text
stringlengths 254
1.16M
|
|---|
---
title: Development and design of the first structured clinic-based program in lower
resource settings to transition emerging adults with type 1 diabetes from pediatric
to adult care
authors:
- Angelica Cristello Sarteau
- Ariba Peerzada
- Alpesh Goyal
- Pradeep A. Praveen
- Nikhil Tandon
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021365
doi: 10.1371/journal.pgph.0000665
license: CC BY 4.0
---
# Development and design of the first structured clinic-based program in lower resource settings to transition emerging adults with type 1 diabetes from pediatric to adult care
## Abstract
### Introduction
Type 1 diabetes (T1D) is increasing in young people worldwide and more children in resource limited settings are living into adulthood. There is a need for rigorous testing and reporting of evidence-based and stakeholder-informed strategies that transition individuals with T1D from pediatric to adult care. We present the development of and design of the first structured transition program in Delhi, India, to inform similar efforts in India and resource limited settings.
### Methods
The intervention development team included clinicians and researchers with expertise in T1D and the implementation context. To select intervention outcomes, establish intervention targets, and design session modules, we drew upon formative research conducted at prospective intervention implementation sites, consensus guidelines, and previous care transition and behavior change research conducted in developed settings. We used the Template for Intervention Description and Replication and GUIDance for the rEporting of intervention Development checklists to report the intervention and development process.
### Results
The 15-month program (“PATHWAY”) includes five quarterly ~30 minute sessions delivered predominantly by diabetes educators at pediatric and adult clinics, which coincide with routine care visits. Primary program components include educational and behavioral sessions that address psychosocial drivers of clinic attendance and self-management, diabetes educators as transition coordinators and counselors, and a one-year “overlap period” of alternating visits between pediatric and adult providers.
### Conclusions
We followed a systematic and transparent process to develop PATHWAY, which facilitated rich description of intervention context, guiding principles, targets, and components. Dependence on previously published program examples to design PATHWAY may have introduced challenges for program feasibility and effectiveness, underscoring the importance of input gathering from prospective intervention actors at multiple points in the development process. This detailed report in combination with future evaluations of PATHWAY support efforts to increase rigorous development and testing of strategies to improve outcomes among emerging adults with T1D.
## Introduction
One of the most common chronic illnesses diagnosed in childhood, type 1 diabetes (T1D) is not only increasing in incidence worldwide, but also in prevalence as treatment advances enable more children to live into adulthood [1–3]. Since the landmark findings put forth by the Diabetes Complications and Control Trial, T1D treatment and management centers about maintaining glycated hemoglobin (HbA1c) below $7\%$ to prevent morbidity and early mortality [4,5]. To support individuals’ achievement of glycemic targets, consensus guidelines recommend a minimum of quarterly visits with a multidisciplinary clinical care team since the complexity of T1D requires ongoing adjustment to treatment and self-management approaches [6,7].
The transition between childhood and adulthood is a physiologically and behaviorally challenging stage during which individuals with T1D all around the world are least likely to meet target glycemic levels, despite being a crucial period for adopting healthy lifelong habits to prevent later adverse health outcomes [8–10]. Although maintaining contact with clinical care during this stage is associated with better glycemic management and fewer acute care visits, long gaps in care are especially common during this period and clinic attendance often declines after transfer to adult care [11–15].
To improve care engagement and glycemic management during emerging adulthood, consensus guidelines promote beginning an intentional transition process from early adolescence that includes gradual practical and psychological preparation to assume self-management tasks independently and navigate the adult clinical care setting and treatment encounters [6,16–20]. In developed settings, there is a maturing evidence base of behavior change and care transition strategies that may improve care engagement and glycemic management during this stage [21–23]. However, weaknesses in this evidence base include few strategies informed by theory, multiple stakeholder perspectives, or rigorously tested via randomized control trials [15,23]. Further, to-date there is no published information about formal care transition programs in lower resource settings, and even more broadly, there is a paucity of evidence about barriers to or strategies for behavior change among young people with T1D in these contexts.
Our formative qualitative work among providers, patients, and parents across a sample of private and public clinics in Delhi, India indicated differences in system, clinic, and patient-level factors relevant to the transition process and post-transfer outcomes as compared to the higher resource settings that dominate the existing evidence base on emerging adults with T1D [24,25]. Informants indicated heterogeneous management of transition processes in the clinical landscape across India (e.g., transfer age, counseling practices) but no formal protocols or programs in place. Providers across private and public settings estimated only about 10–$50\%$ of their T1D patient population followed up with them at least every three months as advised. Among the emerging adults we planned to target with the intervention, informants explained that a visit frequency of once every two to even three years was not uncommon due to geographic displacement, sub-optimal provider and patient rapport, and low self-management readiness and motivation due to various social and economic factors and priorities [24].
Health facilities in India also vary widely within and across public and private settings as regards provider training, treatment and prescribing practices, patient to provider ratios, patient socioeconomic status and out-of-pocket costs. Whereas a visit with an endocrinologist at a public facility could be as low as 0–10 Indian rupees and include a supply of basic insulin (e.g., regular and Neutral Protamine Hagedorn [NPH]), the cost of a consultation at a private facility could reach up to 2000 Indian rupees, exclusive of insulin. As such, because most health care costs are borne by the patient in India, patients of public facilities are more likely to be low or middle income, uninsured, have low literacy, and be unable to afford insulin, blood glucose test strips, or advanced diabetes self-management technology; in contrast, patients of private facilities are more likely to be insured or have a level of income that affords private clinic consultation fees, analog insulin, and routine use of continuous glucose monitoring systems and insulin pumps. Because pediatric endocrinology is not a widely prevalent specialty, adult physicians may treat children from T1D diagnosis and pediatric physicians may treat children from T1D diagnosis into adulthood. Gaps in knowledge therefore persist regarding how to adapt and implement care strategies in lower resource contexts such as India that were developed and evaluated in developed settings, as well as the effectiveness of such strategies in these contexts.
Additionally, detailed reports about the development processes that underpin interventions are extremely limited, which poses obstacles for intervention replication and improvement efforts [21,26]. More widespread publication of these reports and use of consensus-based reporting guidelines to structure them could not only promote learning within and across fields of study, but also promote scientific rigor in the intervention development process by promoting consideration of theory and empirical evidence before undertaking intervention design and implementation. Such transparent reports of programs in developing settings are notably absent despite the arguably greater importance of shared learning to expedite development of evidence-based interventions in the context of greater resource constraints and absolute burden of disease.
The objective of this article is to report the process and outcomes of developing a formal program to improve frequency of routine clinical care attendance and self-management among emerging adults with T1D in Delhi, India after transfer from pediatric to adult care. A systematically developed, theoretically and empirically sound program is foundational to our subsequent aims to rigorously test the program via a randomized controlled trial (CTRI/$\frac{2020}{10}$/028379), modify the program for additional and larger scale evaluations, and ultimately, inform clinical practice guidelines for managing T1D care transitions in India. Through standardized, detailed reporting of the intervention development process and resulting program, we also contribute to efforts to increase rapid development, testing, and sharing of theory, evidence, and stakeholder-informed transition of care interventions for emerging adults with T1D in India and other resource limited settings.
## Methods
The central intervention development team was mainly comprised of endocrinologists, diabetes educators, and researchers with training in epidemiology and behavior change who had context-specific expertise in pediatric and adult T1D and who were employed by a public research hospital in Delhi, India that was also an intervention implementation site. We followed the GUIDance for the rEporting of intervention Development framework to report each step of the intervention development process (S1 Appendix) [26]. The intervention is reported using the Template for Intervention Description and Replication (TIDieR) checklist (S2 Appendix) [27]. Both the formative research study that involved human participants referenced in this report as well as the future randomized controlled trial that will evaluate the program were granted ethical clearance (REF: IEC-$\frac{82}{01.02.2019}$, RP-$\frac{27}{2019}$) by the institutional ethics committee of the All India Institute of Medical Sciences, New Delhi.
A guiding principle of the intervention development process was to promote likelihood that the program could be integrated into existing clinical care practice and widely disseminated across more clinical settings in India if proven effective. As such, a priority was for intervention sessions to coincide with quarterly clinic visits–the minimum recommended frequency of clinical care visits according to consensus guidelines [6,7]. A related priority was for sessions to be delivered as much as possible by existing staff across clinics of different sizes, pay models, counseling practices, staff training backgrounds, and patient socioeconomic status. Because our formative work indicated clinician buy-in was a foundational step for widespread adoption of changes in clinical practice and eventual change in administrative policy, clinicians were asked for input at multiple points of the intervention development process, as described below [24].
In developing the Pediatric to Adult Transition Care for the Health and Wellness of Adolescents with Young Diabetes in India program (PATHWAY), we used a “combination” approach by integrating three formal approaches to intervention development [28]. Starting with a “theory and evidence-based” approach, we identified a list of potential intervention components through an iterative literature review of published research evidence and theories both specific to improving clinical engagement and health outcomes among emerging adults with T1D in the transition between pediatric to adult care as well as general to emerging adults with youth-onset chronic conditions [21–23,29]. Subsequent steps were informed by both a “target population-centred” approach (i.e., attention to views and actions of those using or benefiting from the intervention) and “implementation-based” approach (i.e., attention to the intervention being used in the real world). A stakeholder mapping exercise identified key stakeholder groups involved in T1D care transitions, after which a formative research and analysis phase was undertaken over a 9-month period (May 2019 -March 2020) to refine important intervention targets and elicit suggestions about intervention components, which included in-depth interviews with 38 patients, parents, and providers across public and private clinical settings. Detail about the design and results of the in-depth interviews conducted with key stakeholders is included in a report of our formative research [24].
In December 2019, via a 3 hour in-person workshop with 40 providers from prospective public and private intervention sites, the central research team presented and elicited feedback on the idea of a formal transition program as well as potential intervention and program design options identified through literature review and preliminary findings from the formative research.
A series of bi-weekly one hour working sessions with the central research team were subsequently held over a one-year period to integrate evidence from the formative research, published literature, and clinician feedback, in order to first draft the structure of the intervention at a high level (i.e., intervention length, frequency and objectives of sessions, roles and responsibilities), and then to draft session content and material in detail.
To prioritize and select the constructs that the intervention would target at provider and patient level (i.e., interpersonal and psychosocial factors), we combined context-specific insights from the formative research together with evidence gathered through our literature review, as well as the expanded Social-Ecological Model of Adolescents and Young Adult Readiness for Transition (SMART), an empirically developed, stakeholder and social-ecological theory informed model of the social (e.g., economic status), interpersonal (e.g., provider relationship), and individual-level (e.g., motivation) factors that shape transition readiness and related health outcomes [6,16,19–22,30–34].
Once the high-level design of the intervention was established, the development of specific session activities and materials began with aggregating strategies and content from published health care transition and behavior change studies in emerging adults with T1D, or recommended by T1D or health care transition consensus guidelines to address the priority constructs [21,22]. These strategies and materials were then tailored to be relevant to the context of Delhi, India in which they were to be implemented based on insights from the central research team, formative research. The intervention structure, session activities, and materials were then presented in detail to pediatric and adult providers from prospective implementation sites to elicit feedback. Due to the need to hold the workshops virtually because of the ongoing COVID pandemic, two 1–2 hour online conference sessions were held with 43 pediatric and adult providers in October 2020.
## Findings from stakeholder engagement and influence on PATHWAY program design
Reinforcing findings from our qualitative formative interviews, clinician participants in the stakeholder workshop held in December 2019 expressed unanimous endorsement of a formal transition program [24]. Due to the reality of already time-strapped and insufficient staff at most clinics, participants further emphasized the infeasibility of full reliance on existing staff to take on the additional tasks involved in a transition program, as well as the infeasibility of a joint clinic that would involve time-consuming geographic displacement of providers. This feedback influenced both design and implementation through the decision to incorporate an ‘overlap phase’ into the program (i.e., time period during which the patient sees both adult and pediatric provider) as well as to hire additional staff to support program delivery.
At the follow-up workshops held in October 2020, participants were particularly concerned about successfully convincing patients to switch providers and thus emphasized the importance of incorporating repeated efforts from multiple transition program actors to persuade the patient about the rationale for transfer and manage any misunderstandings that might discourage them from seeking care. This feedback resulted in developing consistent messaging about the rationale for transfer and designing sessions so that this messaging was repeated by adult and pediatric physicians and diabetes educators throughout the program. Workshop participants also suggested that the control group be provided with slightly more support than just receiving notice of the deadline for transfer to adult care in order to motivate their participation and so that the control condition approximately mimicked the most supportive way transfer is currently managed in the existing clinical landscape. This feedback influenced the decision for the control group, and thus the PATHWAY transition program comparator condition, to represent a type of ‘bare bones’ or ‘minimalist’ one-session transition program.
As previously described, a comprehensive report of the results from the qualitative formative interviews has been published elsewhere [24]. The way stakeholder feedback obtained throughout the intervention development process underpinned specific transition program design elements, is further detailed in the subsequent section that describes the PATHWAY program.
## Description of the PATHWAY program design: Logic model, program structure and content
The PATHWAY transition program that will be tested in a future randomized controlled trial is described below. The logic model in Fig 1 depicts how the program components achieve the intervention target outcomes and objectives [35–37]. The 15-month structured transition program has three primary defining features: diabetes educators as the central coordinators and counselors of the transition program, a one year “overlap period” during which time the emerging adult with T1D alternates between adult and pediatric provider team visits, and quarterly counseling sessions focused on targeting theoretically and empirically supported psychosocial factors associated with clinical care engagement and other positive outcomes during health care transitions among emerging adults with T1D [21–23].
**Fig 1:** *The PATHWAY intervention logic model.*
A growing evidence base suggests post-transfer clinical care engagement and glycemic management are improved through structured transition interventions that include both a “transition clinic” to build rapport with the adult diabetes providers (i.e., emerging adults visit with both pediatric and adult providers at a jointly staffed clinic before permanent transfer) and “transition coordinators” who counsel emerging adults on practical and psychosocial factors that impact clinical visit attendance and self-management adherence in adulthood [22,23]. Our context-specific formative work and feedback from workshops with clinicians indicated that joint visits would be infeasible due to the common occurrence of patients switching health care institutions when transferring between pediatric and adult providers. Further, high patient-to-provider ratios challenge the coordination of pediatric and adult physician schedules even when both providers work within the same institution. Thus, in order to facilitate rapport-building and adult physician familiarity with the emerging adult’s diabetes, PATHWAY involves an “overlap period” during which time participants alternate quarterly clinic visits between pediatric and adult provider teams for a yearlong period before official and permanent transfer to the adult care team. These visits combine meetings with the physician and supplemental psychosocial counseling sessions–described below—by the diabetes educators at each site. This quarterly intervention session frequency throughout the 15-month transition program intervention aligns with the minimum recommended frequency of routine clinical care visits for individuals with T1D, consistent with our objective to develop an intervention that could be more readily built into routine care if proven beneficial. In addition to facilitating gradual rapport building with the adult physician and diabetes educator, this program format addresses the practical and psychosocial factors that influence clinical care engagement and increases convenience of the transition program for patients–many of whom travel long distances and may be deterred by the requirement to make additional trips to the clinic outside of those made as part of routine care.
Our formative research also highlighted several context-specific reasons to position diabetes educators at the center of the structured transition program as transition coordinator and counselor. One of the facilitators of care engagement and diabetes self-management most commonly cited in our formative research included trust that the adult provider not only understands the emerging adults’ diabetes treatment regimen, but also empathizes with and helps address the psychosocial factors that shape an emerging adult’s self-management difficulties [24]. Given the case loads of adult endocrinologists in Delhi, India, which make short encounters and case sharing among physicians common features of adult diabetes care visits, informants described the importance of enlisting providers outside of the adult endocrinologist to provide psychosocial support during the transition process to reduce likelihood that care engagement would hinge on establishing what was perceived as an often infeasible, idealistic relationship with the adult physician. Compared to other providers in diabetes care teams across the clinical context in which PATHWAY was designed to be implemented, diabetes educators were described as having the most skills and time to establish rapport with emerging adults and address psychosocial and practical barriers to care engagement and self-management issues during the turbulent period of early adulthood. Our design of PATHWAY to maximize delivery of the educational and behavioral intervention sessions by the diabetes educators employed at private and public pediatric and adult clinics also aligned with a key development objective to produce a program that could be implemented and sustained within the existing clinical landscape. However, given the multiple clinical responsibilities that site diabetes educators manage outside their delivery of PATHWAY, 4 additional diabetes educators (2 nurse educators and 2 dieticians experienced with T1D) were hired in order to coordinate program scheduling, oversee assessments requiring completion as part of the clinical trial, and deliver sessions when site-specific diabetes educators could not provide a session on a date and time that overlapped with participants’ quarterly visits.
Our review of published studies and expert consensus guidelines about care transitions and behavior change among adolescents and young adults with type 1 diabetes also underscored the importance of proactively managing transition-related psychosocial factors in order to improve care engagement and health outcomes after transfer: attachment to and reluctance to leave pediatric provider, lack of awareness of transition timeline, unfamiliarity with differences in adult provider approaches, expectations of patients, and visit procedures [21,23]. Additionally, the evidence base suggests promoting transition readiness, care engagement, and other outcomes by targeting additional factors that influence self-management and care seeking behaviors: self-management knowledge, problem-solving and goal-setting skills, motivation, self-efficacy, social support, emotional regulation, and conflict management [21–23].
Although our formative work reinforced the importance of all these factors to varying degrees, when designing the intervention sessions and their comprising educational and behavioral activities, we targeted the factors that were especially salient according to our context-specific investigations (Table 1). The materials used to facilitate session activities were also informed by our literature review and tailored for the context of intervention implementation based on our formative work that elicited perspectives from patients, parents, and clinicians, as well as the expertise of clinician, behavior change, and public health members of the central research team. A session-by-session description of intervention design and activities can be found in Fig 2 and corresponding materials used can be found in S3 Appendix.
**Fig 2:** *Detailed description of PATHWAY intervention program timeline, sessions, materials, sites, and staff.* TABLE_PLACEHOLDER:Table 1 We decided upon several main changes in transition program content and structure during our bi-weekly intervention development team meetings. First, findings from our formative research and first stakeholder engagement workshop led us to revise our plans to completely rely on the clinic employed diabetes educators to coordinate and deliver the program. Four diabetes educators were employed by the intervention development team to support session delivery during the randomized controlled trial so as to prevent against over-extending existing resources at participating sites, which could dilute fidelity in intervention delivery, lower the rigor with which the strategy was tested through the randomized controlled trial, and reduce provider buy-in for the program. Although this design decision will challenge immediate translation of the program into existing clinical practice if trial results indicate program effectiveness, information gathered from the planned RE-AIM evaluation (i.e., monitoring checklists and interviews with providers and patients) of the randomized controlled trial will generate insights about the staff time and monetary resources that may be required by practices to implement the strategy as well as generating suggestions about the components of the program that may be particularly important for driving target outcomes. These insights could inform a strategy to pare-down the program design to increase feasibility and uptake into existing settings, which could then be assessed for effectiveness through a future evaluation. Second, after the second stakeholder engagement workshop, we made adjustments to deliver a session (Session 1, Fig 2) to all study participants before randomization (i.e., both control and intervention arms) in order to compare the 15-month transition program to a minimalist, one-session transition program, Third, the intervention sessions were initially conceived to be group sessions to reduce staff implementation burden, promote normalization of adherence barriers, and facilitate instrumental and non-instrumental peer support that have been demonstrated to enhance self-management during this developmental stage [21,23]. We introduced flexibility to deliver these sessions in either group or individual format due to challenges aligning patient schedules and because we anticipated more gradual trial recruitment due to COVID-19.
## Discussion
This report provides an example of systematic, evidence-based, and stakeholder-informed development of a context-specific care transition program to promote care engagement and self-management in emerging adults with T1D in Delhi, India (“PATHWAY”), which we plan to rigorously test in a randomized controlled trial in future steps. Strengths include integrating the evidence base from high resource settings together with context-specific stakeholder perspectives from our formative work in order to identify the intervention targets and strategies that define the program. Such an approach may increase the potential that the program promotes intended outcomes among emerging adults with T1D in Delhi, India. Further, our use of GUIDED and TiDIER reporting frameworks (S1 Appendix and S2 Appendix) to transparently report the intervention development process. Employment of these reporting frameworks facilitates comprehension of the program design and may inform development of theoretical and evidence-based care transition strategies for emerging adults with T1D in other resource limited settings, which are currently lacking despite growing incidence and prevalence of this disease in these settings.
Limitations and persisting uncertainties include that the care transition evidence base and our formative research that informed the PATHWAY program were undertaken before COVID-19, which introduce the possibility that unexplored intervention targets and strategies could be relevant to our target population and setting, but are not incorporated into program design. Although we will not systematically collect information to address this point of uncertainty before the program is tested through a clinical trial, a planned mixed methods Reach, Effectiveness, Adoption, Implementation, and Maintenance (RE-AIM) evaluation of the trial will help to identify these factors and enable future refinement of the program. We purposively designed the program to rely on the diabetes educators employed by adult and pediatric clinics to deliver as much of the intervention as possible in order to minimize program expense and maximize potential that the program would be more widely adopted and tested if proven effective through our initial randomized controlled trial evaluation. However, this choice also introduces risk of low and variable fidelity across implementation sites. To promote fidelity and prevent drift, we designed scripted counseling materials to facilitate consistent implementation across sites and we will also monitor a random subset of intervention sessions throughout implementation as part of the trial. The aforementioned RE-AIM evaluation will enable us to understand differences in fidelity across sites and time, identify factors that potentially underly these differences, as well as explore the influence of these factors on the trial outcomes observed.
Detailed reporting of interventions and their underlying processes are limited, especially in the context of resource limited settings and T1D care transitions. Our systematic development and reporting of the PATHWAY program is an important foundational step towards identifying strategies that may benefit the self-management of emerging adults with T1D in India and similarly resourced settings.
## References
1. Miller RG, Secrest AM, Sharma RK, Songer TJ, Orchard TJ. **Improvements in the life expectancy of type 1 diabetes: the Pittsburgh Epidemiology of Diabetes Complications study cohort**. *Diabetes* (2012.0) **61** 2987-92. DOI: 10.2337/db11-1625
2. Marshall SM. *Celebrating 100 years of insulin* (2021.0)
3. Dabelea D, Mayer-Davis EJ, Saydah S, Imperatore G, Linder B, Divers J. **Prevalence of type 1 and type 2 diabetes among children and adolescents from 2001 to 2009**. *Jama* (2014.0) **311** 1778-86. DOI: 10.1001/jama.2014.3201
4. Cleary PA, Dahms W, Goldstein D, Malone J, Tamborlane WV. **Beneficial effects of intensive therapy of diabetes during adolescence: outcomes after the conclusion of the Diabetes Control and Complications Trial (DCCT).**. *J Pediatr* (2001.0) **139** 804-12. DOI: 10.1067/mpd.2001.118887
5. Control D, Trial C. **Intensive diabetes treatment and cardiovascular outcomes in type 1 diabetes: the DCCT/EDIC study 30-year follow-up**. *Diabetes care* (2016.0) **39** 686-93. DOI: 10.2337/dc15-1990
6. Chiang JL, Kirkman MS, Laffel LM, Peters AL. **Type 1 diabetes through the life span: a position statement of the American Diabetes Association**. *Diabetes care* (2014.0) **37** 2034-54. DOI: 10.2337/dc14-1140
7. Pihoker C, Forsander G, Fantahun B, Virmani A, Corathers S, Benitez‐Aguirre P. **ISPAD Clinical Practice Consensus Guidelines 2018: The delivery of ambulatory diabetes care to children and adolescents with diabetes**. *Pediatric diabetes* (2018.0) **19** 84-104. DOI: 10.1111/pedi.12757
8. Foster NC, Beck RW, Miller KM, Clements MA, Rickels MR, DiMeglio LA. **State of type 1 diabetes management and outcomes from the T1D exchange in 2016–2018.**. *Diabetes technology & therapeutics* (2019.0) **21** 66-72. DOI: 10.1089/dia.2018.0384
9. Chen X, Pei Z, Zhang M, Xu Z, Zhao Z, Lu W. **Glycated hemoglobin (HbA1c) concentrations among children and adolescents with diabetes in middle-and low-income countries, 2010–2019: A retrospective chart review and systematic review of literature.**. *Frontiers in endocrinology* (2021.0) **12**
10. Prigge R, McKnight JA, Wild SH, Haynes A, Jones T. **International comparison of glycaemic control in people with type 1 diabetes: an update and extension**. *Diabetic Medicine* (2021.0) e14766. DOI: 10.1111/dme.14766
11. Lotstein DS, Seid M, Klingensmith G, Case D, Lawrence JM, Pihoker C. **Transition from pediatric to adult care for youth diagnosed with type 1 diabetes in adolescence**. *Pediatrics* (2013.0) **131** e1062-e70. DOI: 10.1542/peds.2012-1450
12. Kipps S, Bahu T, Ong K, Ackland F, Brown R, Fox C. **Current methods of transfer of young people with type 1 diabetes to adult services**. *Diabetic Medicine* (2002.0) **19** 649-54. DOI: 10.1046/j.1464-5491.2002.00757.x
13. Nakhla M, Daneman D, To T, Paradis G, Guttmann A. **Transition to adult care for youths with diabetes mellitus: findings from a Universal Health Care System**. *Pediatrics* (2009.0) **124** e1134-e41. DOI: 10.1542/peds.2009-0041
14. Farrell K, Fernandez R, Salamonson Y, Griffiths R, Holmes-Walker D. **Health outcomes for youth with type 1 diabetes at 18 months and 30 months post transition from pediatric to adult care**. *Diabetes research and clinical practice* (2018.0) **139** 163-9. DOI: 10.1016/j.diabres.2018.03.013
15. Hynes L, Byrne M, Dinneen SF, McGuire BE, O’Donnell M, Mc Sharry J. **Barriers and facilitators associated with attendance at hospital diabetes clinics among young adults (15–30 years) with type 1 diabetes mellitus: a systematic review.**. *Pediatric diabetes* (2016.0) **17** 509-18. DOI: 10.1111/pedi.12198
16. Peters A, Laffel L. **Diabetes care for emerging adults: recommendations for transition from pediatric to adult diabetes care systems: a position statement of the American diabetes association, with representation by the American College of osteopathic family physicians, the American Academy of pediatrics, the American association of clinical endocrinologists, the American osteopathic association, the centers for disease control and prevention, children with diabetes, the endocrine Society, the International Society for pediatric and adolescent diabetes, juvenile diabetes research Foundation international, the National diabetes education program, and the pediatric endocrine Society (formerly Lawson Wilkins pediatric endocrine Society).**. *Diabetes care* (2011.0) **34** 2477-85. DOI: 10.2337/dc11-1723
17. Pediatrics AAo Physicians AAoF. **Supporting the health care transition from adolescence to adulthood in the medical home**. *Am Acad Pediatrics* (2011.0)
18. 18Health NAtAA. Six core elements of health care transition 3.0: Transitioning youth to an adult health care clinician. 2021; Available from: http://gottransition.org/resources/index.cfm.
19. Rosen DS, Blum RW, Britto M, Sawyer SM, Siegel DM. **Transition to adult health care for adolescents and young adults with chronic conditions: position paper of the Society for Adolescent Medicine**. *Journal of Adolescent Health* (2003.0) **33** 309-11. DOI: 10.1016/s1054-139x(03)00208-8
20. Sherr JL, Tauschmann M, Battelino T, de Bock M, Forlenza G, Roman R. **ISPAD Clinical Practice Consensus Guidelines 2018: Diabetes technologies.**. *Pediatric diabetes* (2018.0) **19** 302-25. DOI: 10.1111/pedi.12731
21. Schmidt A, Ilango SM, McManus MA, Rogers KK, White PH. **Outcomes of pediatric to adult health care transition interventions: An updated systematic review**. *Journal of pediatric nursing* (2020.0) **51** 92-107. DOI: 10.1016/j.pedn.2020.01.002
22. Schultz AT, Smaldone A. **Components of interventions that improve transitions to adult care for adolescents with type 1 diabetes**. *Journal of Adolescent Health* (2017.0) **60** 133-46. DOI: 10.1016/j.jadohealth.2016.10.002
23. O’Hara M, Hynes L, O’donnell M, Nery N, Byrne M, Heller S. **A systematic review of interventions to improve outcomes for young adults with type 1 diabetes**. *Diabetic Medicine* (2017.0) **34** 753-69. DOI: 10.1111/dme.13276
24. Cristello Sarteau A, Peerzada A, Goyal A, Praveen PA, Tandon N. **Improving care transitions among emerging adults with type 1 diabetes: a qualitative study to inform the first formal program in low-resource settings**. *Manuscript submitted for publication* (2021.0)
25. Gittelsohn J, Steckler A, Johnson CC, Pratt C, Grieser M, Pickrel J. **Formative research in school and community-based health programs and studies:“State of the art” and the TAAG approach.**. *Health education & behavior* (2006.0) **33** 25-39. PMID: 16397157
26. Duncan E, O’Cathain A, Rousseau N, Croot L, Sworn K, Turner KM. **Guidance for reporting intervention development studies in health research (GUIDED): an evidence-based consensus study.**. *BMJ open* (2020.0) **10** e033516. DOI: 10.1136/bmjopen-2019-033516
27. Hoffmann TC, Glasziou PP, Boutron I, Milne R, Perera R, Moher D. **Better reporting of interventions: template for intervention description and replication (TIDieR) checklist and guide.**. *Bmj* (2014.0) 348. DOI: 10.1136/bmj.g1687
28. O’Cathain A, Croot L, Sworn K, Duncan E, Rousseau N, Turner K. **Taxonomy of approaches to developing interventions to improve health: a systematic methods overview.**. *Pilot and feasibility studies* (2019.0) **5** 1-27. DOI: 10.1186/s40814-019-0425-6
29. Craig P, Dieppe P, Macintyre S, Michie S, Nazareth I, Petticrew M. **Developing and evaluating complex interventions: the new Medical Research Council guidance**. *Bmj* (2008.0) 337. DOI: 10.1136/bmj.a1655
30. Schwartz L, Tuchman L, Hobbie W, Ginsberg J. **A social‐ecological model of readiness for transition to adult‐oriented care for adolescents and young adults with chronic health conditions. Child: care**. *health and development* (2011.0) **37** 883-95. DOI: 10.1111/j.1365-2214.2011.01282.x
31. Schwartz LA, Brumley LD, Tuchman LK, Barakat LP, Hobbie WL, Ginsberg JP. **Stakeholder validation of a model of readiness for transition to adult care**. *JAMA pediatrics* (2013.0) **167** 939-46. DOI: 10.1001/jamapediatrics.2013.2223
32. White PH, Cooley WC. **Transitions Clinical Report Authoring Group; American Academy of Pediatrics; American Academy of Family Physicians; American College of Physicians. Supporting the Health Care Transition From Adolescence to Adulthood in the Medical Home (vol 142, e20182587, 2018).**. *Pediatrics* (2019.0) **143**
33. Pierce JS, Aroian K, Schifano E, Milkes A, Schwindt T, Gannon A. **Health Care Transition for Young Adults With Type 1 Diabetes: Stakeholder Engagement for Defining Optimal Outcomes.**. *Journal of Pediatric Psychology* (2017.0) **42** 970-82. DOI: 10.1093/jpepsy/jsx076
34. Pierce JS, Wysocki T. **Topical Review: Advancing Research on the Transition to Adult Care for Type 1 Diabetes.**. *Journal of Pediatric Psychology* (2015.0) **40** 1041-7. DOI: 10.1093/jpepsy/jsv064
35. Goeschel CA, Weiss WM, Pronovost PJ. **Using a logic model to design and evaluate quality and patient safety improvement programs.**. *International Journal for Quality in Health Care* (2012.0) **24** 330-7. DOI: 10.1093/intqhc/mzs029
36. Davidoff F, Dixon-Woods M, Leviton L, Michie S. **Demystifying theory and its use in improvement**. *BMJ quality & safety* (2015.0) **24** 228-38. DOI: 10.1136/bmjqs-2014-003627
37. 37Foundation WKK. Logic Model Development Guide. Available from: http://www.wkkf.org/resource-directory.
|
---
title: Spatial changes in park visitation at the onset of the pandemic
authors:
- Kelsey Linnell
- Mikaela Irene Fudolig
- Aaron Schwartz
- Taylor H. Ricketts
- Jarlath P. M. O’Neil-Dunne
- Peter Sheridan Dodds
- Christopher M. Danforth
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021367
doi: 10.1371/journal.pgph.0000766
license: CC BY 4.0
---
# Spatial changes in park visitation at the onset of the pandemic
## Abstract
The COVID-19 pandemic disrupted the mobility patterns of a majority of Americans beginning in March 2020. Despite the beneficial, socially distanced activity offered by outdoor recreation, confusing and contradictory public health messaging complicated access to natural spaces. Working with a dataset comprising the locations of roughly 50 million distinct mobile devices in 2019 and 2020, we analyze weekly visitation patterns for 8,135 parks across the United States. Using Bayesian inference, we identify regions that experienced a substantial change in visitation in the first few weeks of the pandemic. We find that regions that did not exhibit a change were likely to have smaller populations, and to have voted more republican than democrat in the 2020 elections. Our study contributes to a growing body of literature using passive observations to explore who benefits from access to nature.
## Introduction
Parks are important public infrastructure that provide a venue for interaction with nature, socialization, and exercise. Park access and use has been found to offer both mental and physical health benefits [1–6]. Among the many benefits of exposure to nature are faster healing, decreased stress and increased ability to manage life’s challenges [7, 8]. During the COVID–19 pandemic, access to parks may have been important for mitigating and managing the secondary impacts of the virus. Recent publications indicate that access to parks during the pandemic is important for a variety of reasons including providing a venue for exercise, increasing happiness, and improving social cohesion [9–14].
While park visitation may have provided significant support to personal and public health at the time, it is unclear whether park visitation changed, to what extent, and for whom in the United States. In March of 2020, stay at home orders were issued in most states, and many non-essential workplaces and public spaces were closed. Following these events, overall mobility decreased dramatically for most Americans, reaching a maximum reduction by 34 to $69\%$ depending on the state [15, 16]. While Americans were visiting fewer locations in general, some research suggests that park visitation may not have been subject to this decline. An early study of parks on the West Coast determined changes in visitation at the onset of the pandemic to be primarily motivated by seasonal change, while a study of parks in New Jersey found that early pandemic visitation was higher than the baseline [17, 18]. Together these results indicate that visits to parks may have differed from other points of interest at the onset of the pandemic.
Preliminary examination of trends suggest that changes in park visitation were not universal. In the United States, partisanship, even at the regional level, is associated with behavioral differences. Researchers have found that Thanksgiving dinners were 30 to 50 minutes shorter when the guests and hosts resided in voting precincts that had been in opposition in 2016 [19]. Mobility studies of Americans during the pandemic have found differences along partisan lines as well. The American political system is largely dominated by two political parties: Democrats, and Republicans. This divide in political ideology has been found to be indicative of differing identities and behaviors. This is particularly true of COVID-19 policy response and preferences [20]. Republicans have been found to have lower vaccination rates, have a smaller decrease in mobility during the pandemic, and to be less compliant with non-pharmaceutical interventions [15, 21–25]. Counties with more Republicans also had less severe mobility restrictions, and were less responsive to their governor’s recommendation to stay home [15, 26]. Given these partisan differences in general mobility, we seek to determine whether changes in local park visitation at the onset of the pandemic also differed by partisanship, or whether park visitation uniquely transcended these differences.
Studies of park usage in March and April of 2020 have thus far relied on survey data, or have been geographically limited, and neglected to establish a baseline of seasonality of park usage [17, 18, 27, 28]. Here we utilize mobile device data from across the United States to explore abrupt non-seasonal changes in park visitation at the regional level. We use data from 2019 to discern seasonal visitation patterns, and employ a change-point detection algorithm to diagnose sudden changes in behavior at the onset of the pandemic. By classifying regions by whether or not an abrupt change in park visitation took place, we are able to discern whether or not these abrupt changes occurred along partisan lines. We conduct further comparisons across population, income, and share of employment by industry to provide insight into other factors that may have influenced whether or not an abrupt change occurred. In the Data section we introduce the data used to classify regions, and make these comparisons. The Methods section then gives a detailed explanation of the data aggregation for each region, and the classification procedure and methods of comparison applied to the aggregated data. The results of the comparisons are described, and then discussed.
## Data
To determine whether a partisan effect is observed in park visitation we used park visitation data from across the United States, and voting share data from the 2020 Presidential Election. Differences in regions with and without abrupt changes in visitation were further analyzed using population estimates, and income and employment share data from the US Census and the Bureau of Economic Affairs. Details for these data sources are provided below.
## Park visitation data
Our park visitation dataset was acquired from UberMedia (now part of Near), and consists of daily visitation counts for non-commercial parks for each day of 2019 and 2020. There are 8,135 parks in the data set, including municipal, neighborhood, and city parks. National and State Parks were specifically excluded as predominantly travel destinations. Parks are located in each of the 50 states, and Washington DC. A total of 1,033 counties, roughly a third of all counties, contain at least one park from our dataset.
Daily visitation counts were determined using location data from mobile devices. Each unique device appearing within a park’s bounds on a single day was counted as a visit. A device’s location was reported when an individual used one of over 400 apps utilizing a GPS Software Development Kit (SDK) in partnership with UberMedia ($90\%$ of data by volume), or when a user interacted with an advertisement through real time bidding on one of over 250,000 apps ($10\%$ of data by volume). GPS location and an accompanying timestamp were determined from the device’s operating system.
The number of devices reporting activity in at least one location in the US on a given day is referred to as the Daily Active Users (DAUs). This number refers to all locations, not simply parks. In 2019 and 2020 the monthly DAUs varied between 38 and 60 million, and represented roughly $10\%$ of the adult population in the United States.
In mid December 2019 the set of SDK’s in partnership with UberMedia was updated. This change in data collection corresponded to a large increase in observations throughout the US, and was not spatially uniform. Thus, while the raw 2019 and 2020 park visitation data are not directly comparable, we analyze their relationship where possible.
## Voting and economic data
Voting data at the state and county levels from the 2020 election was retrieved from MIT Election Data and Science Lab, and is available at https://electionlab.mit.edu/data.
The BEA publishes data on employment by industry (using the North American Industry Classification System (NAICS)) for each county in table “CAEMP25N”, which can be found at https://apps.bea.gov/regional/downloadzip.cfm. In this table “Farming” and “Forestry” are considered as separate, though they appear as one sector in the NAICS classification. For this study they were considered separately, as they appear in the table. County level population, income, and economic data from the 2019 American Community Survey were obtained from US census API.
## Methods
In this study we compare regions in the United States where abrupt non-seasonal changes in park visitation did and did not occur at the onset of the pandemic. In order to make these comparisons we begin by classifying regions as having or not having an abrupt change point by applying a change point detection algorithm to aggregated time series of park visitation in each of the regions (Fig 1). Once regions have been classified, the distributions of the regions across population, income, share of employment by industry, and voteshare in the 2020 presidential election are compared.
**Fig 1:** *A heat map of the population of the contiguous United States in log scale overlaid with the locations of the parks used in the study, with each park demarcated with a black point.Observation of this map indicates that the parks in this data set have an urban bias, and that the parks are roughly distributed according to population distribution in the United States. Maps of park locations are shown for Massachusetts (top) and Oklahoma (bottom), with the color of each map representing the political party receiving the most votes in the 2020 Presidential election (Democrats for Massachusetts, and Republicans in Oklahoma). Normalized weekly park visitation for each state is plotted to the right. Visitation for 2019 is plotted in blue, while visitation for 2020 is plotted in orange. For Massachusetts there is a significant dip in visitation bottoming out the week of March 25, 2020, and the visitation plots for 2019 and 2020 diverge. For Oklahoma visitation does not drop off in March, and does not diverge from 2019 visitation patterns. Generated with GeoPandas using Census data [29].*
## Aggregation
Daily visits to a park were defined as the unique number of mobile devices reporting GPS coordinates found inside the park polygon bound on a day. This daily visit count was then normalized by the average Daily Active Users for the month in which it was found, approximating the percentage of devices observed in parks relative to all observed devices. The normalized visitation was then summed over each week in order to minimize noise. Weekly visitation was summed for parks contained in a county, or a state, and thus a time series of weekly park visitation between 2019 and 2020 was created for each county and state containing at least one park from our data set.
## Change point detection
To determine whether a substantive change in visitation is observed in each time series, we use the Bayesian Estimator of Abrupt Change, Seasonality, and Trend (BEAST) [30]. This method decomposes a time series into a seasonal (harmonic) component, and trend (linear) component, and uses Bayesian Inference to fit a model which estimates the location of change points in either of the components. BEAST was chosen because the underlying model acknowledges the seasonal nature of most park visitation time series (more visits in summer). By specifying a 52 week season length, we were able to train the model to the annual cycle shape of the data.
Parametric methods applied without the seasonal decomposition are susceptible to under estimating change points in these particular time series because of the combination of seasonality and the proximity in the series of the data collection change in December 2019 to the onset of the pandemic in March 2020. The initial event represents a sharp increase in visitation volume (roughly 150 pct), while the second appears, for most regions, as a sharp decline. When fit with a single model, these two features appear together as a change in variance, and a parametric model can be nicely fit using a single change point in December 2019.
By decomposing the time series and forcing a decoupling of the two events by specification of seasonal, length we make each event visible as a unique discontinuity in the linear component.
The December 2019 discontinuity could then be accommodated with a trend change point, which incorporates a discontinuity into the linear component. In this way the model was fit while accounting for seasonality, and the abrupt change in data volume.
Allowing a trend change point to be used as described above, the model was effectively limited to selecting a single trend change point, which enabled it to identify the most likely change point in the data. It is possible for the algorithm to detect no change point, reducing concern that one would be identified artificially.
Regions which had a change point occurring in between mid March and mid April 2020 were considered to have had an abrupt change in park visitation coinciding with the onset of the pandemic and social distancing measures. If a region was found not to have had a change point in this window, it can be assumed that either no change point was found in the time series, or any change occurring in the specified window was not as significant as a change at another time.
Changes induced by seasonality are in most cases more gradual than those that occur in the window of interest, and these changes are accounted for by the harmonic component of the model. The harmonic component is fit using both 2019 and 2020 data, which informs the model of the expected seasonal shape. Since these changes are accounted for in the model fitting, it is unlikely that change points identified in the window of interest are due only to seasonal variation. Because the length of the time series only included two seasons (park visitation demonstrates a yearly cycle), it was not pertinent to search for changes in the seasonal structure.
BEAST is less effective in identifying change points in time series with high variance. The recorded park visits in some of the counties were low enough that the behavior of only a few individuals could have large impacts on the time series itself. To ensure that BEAST was only considering counties for which there was enough data we used a mean normalized visitation threshold of 10−5.5 (this corresponds to about 120 visits per week in the month with the least DAUs) in 2020. A total of 322 counties did not meet this criteria and were excluded from further analysis. The remaining 711 counties that contain parks in our dataset met this criteria. The counties included in the analysis are roughly $21\%$ of all the counties in the United States, and span all of the states. Details on the selection of the visitation threshold can be found in the Supplementary Materials (See S1 Fig).
## Comparison
Regions were binned according to whether a change point in mid March 2020 was identified or not, and comparisons of the populations of the regions in each category were made. Using data from the 2020 election, states and counties were assigned a percent of the population having voted either Republican (Trump and Pence) or Democrat (Biden and Harris) in the 2020 election. Counties were assigned personal incomes, and population counts using Census estimates from 2019. Voting records and census data were combined to determine the votes cast per capita for each county. Finally, a fraction of employment (employment share) for each industry in the North American Industry Classification System (NAICS) was assigned to each county in the study using data from the BEA. Counties which had no available employment data for an industry were exclcuded from the analysis of that particular industry. Counties with and without detected change points were compared across vote share, population, votes per capita, personal income, and industry employment using Kolomogorov-Smirnov two sample tests. This test was chosen for its ubiquity in the literature and ability to compare distributions with different sample sizes. The means of the distributions are compared using Welch’s t-test, which also accommodates different sample sizes.
## Partisanship in abrupt changes
With the parameters discussed in the Methods section, BEAST found a change point in the window of interest for 21 states, while the remaining 29 states did not exhibit an abrupt change in visitation. Comparison of the 2020 presidential election results for states where visitation did and did not change abruptly is shown in the top row of Fig 2. The distributions across vote share for the two sets of states were not significantly different for either the Democratic or Republican parties (KS statistic = 0.2, $$p \leq 0.63$$ and KS statistic = 0.2, $$p \leq 0.63$$ respectively). The distributions for each party are neither similar, nor mirrored. The difference is accounted for by third party votes, most notably Libertarian votes.
**Fig 2:** *Top: Distributions of states where a pandemic response change point was detected (pink), and not (green), across proportion of votes cast for the Democratic (left) and Republican (right) parties in the 2020 Presidential Election. The distributions are not significantly different across voting proportion for either party. The apparent outlier in each figure is Washington DC, which did not exhibit a change point, and for which more than 80% of the votes cast were for the Democratic party. Exclusion of DC from the analysis did not change the results. Bottom: Distributions of counties with (pink) and without(green) detected change points across percent voting for the Democratic (left) and Republican (right) parties in the 2020 Presidential Election. Distributions across the percent of votes cast for the Democratic party were determined to be significantly different by the Kolomogorov Smirnov 2 sample test (KS statistic = 0.33, p = 4.23e-10). The distribution of counties with a change point was shifted to the right of those without a change point, indicating counties with change points had greater proportions of votes for the Democratic party. The bulk of the mass of the two distributions lies on either side of 0.5, meaning that the majority of counties with a change point are majority Democrat counties. The distributions across percent voting for the Republican party are likewise significantly different (KS statistic = 0.33, p = 2.35e-10), and indicate counties without change points had greater proportions of votes for the Republican party, and were more likely to be a majority Republican county.*
Comparison of the distributions across percent voting Libertarian (which accounted for less than $3\%$ of the vote in all states) indicates that the distributions were significantly different (KS statistic = 0.44, $$p \leq 0.015$$), where Libertarians had greater vote share in states without an abrupt change. S2 Fig demonstrates the relative proportion of Democrat, Republican, and third party votes for each state and county. The state appearing as an outlier in the distributions, where Democrats had the highest vote share, and which did not have an abrupt change, is Washington DC, which was treated as a state for this study. Exclusion of Washington DC does not change the results.
Partitioning the data by county led to significantly different distributions across vote share (bottom row of Fig 2). When the BEAST classification procedure was applied to county level aggregations of visitation data, 123 of the 711 counties had abrupt visitation changes at the onset of the pandemic. The distribution across Democratic vote share for counties with abrupt changes is shifted to the right of the distribution for counties without- indicating that Democrats were more likely to have greater vote share in counties with abrupt changes. Kolomogorov Smirnov 2 sample results confirm that these distributions are significantly different (KS statistic = 0.33, $$p \leq 4.23$$e-10). Observation of the same distributions across Republican vote share reveals that Republicans were more likely to have greater vote share in counties without abrupt changes. KS 2 sample test results support that these distributions are also significantly different (KS statistic = 0.33, $$p \leq 2.35$$e-10).
Not only are the distributions significantly different, but across the vote share for each party they are translated across the $x = 0.5$ line (drawn in red). This line represents the dividing point in the majority party support in a county. This reveals that Democrats were not only more likely to have greater vote share in counties with abrupt changes, they were more likely to hold a majority in those counties. Likewise, Republicans were more likely to hold a majority in counties without abrupt changes.
The distributions of the counties across vote share for the Democratic and Republican parties are not mirrored on account of votes going to third parties, meaning that “not Democrat” is not the same as “Republican”. Both distributions taken together support that there is a partisan divide between counties with and without abrupt changes in park visitation at the onset of the pandemic. This is further supported by no significant difference found in the distributions of the counties over percent voting Libertarian (KS statistic = 0.13, $$p \leq 0.087$$).
## Population, income, and employment
Partitioning the data by county allowed further analysis using population, employment, and income data. Differences in distribution across population size, and votes cast per resident, for the counties with and without abrupt changes, are displayed in Fig 3. Counties with an abrupt change had more than twice the mean population of counties exhibiting no change, and fewer votes per resident than counties that did not. The distributions across each of these variables is significantly different (KS statistic = 0.28, $$p \leq 1.08$$e-07 for log 10 scale population, and KS statistic = 0.13, $$p \leq 0.045$$ for votes cast per resident).
**Fig 3:** *Distributions of counties with and without change points across the log base 10 of 2019 county population (left) and the votes cast per capita in the 2020 Presidential Election (right).The distributions over population are roughly log normal, and visibly and significantly different (KS statistic = 0.28, p = 1.08e-07). The mean population of counties with change points was 331,131 people, which is more than twice the mean population of counties without change points, namely 144,544. The counties with the lowest populations were exclusively without change points, while the counties with the greatest populations were exclusively those with change points. The distributions across votes per capita are also visibly and significantly different (KS statistic = 0.13, p = 0.045) with the counties with change points having fewer votes per capita than counties without.*
The incomes of the counties were not significantly different (KS statistic = 0.10, $$p \leq 0.23$$), as seen in the distributions in Fig 4.
**Fig 4:** *Distributions of counties with and without change points across the log base 10 of 2019 personal income as reported by the census.The distributions are visually similar, and not significantly different (statistic = 0.10, p-value = 0.23). There is not a statistically significant difference between the incomes of counties where abrupt park visitation changes occurred and those where it did not.*
Counties were also compared on the basis of percent employment in each of the 20 NAICS sectors. The distributions of the counties with and without change points across percent of employment were significantly different ($p \leq 0.05$) for 14 of the sectors. This includes both Farming Employment, and Forestry, Fishing and Related Activities, which comprise a single sector in the NCAIS, but are considered separately here. Of the 14 sectors with significantly different distributions, Welch’s T-Tests found only 10 had significantly different means. The distributions for these 10 sectors is shown for counties with and without abrupt changes in Fig 5. For each sector, the box plot to the left shows the distribution over the fraction of employment for counties with an abrupt change (pink), and the box plot to the right represents the same distribution for counties without an abrupt change (green).
**Fig 5:** *Box plots showing the distribution across employment share for counties with and without change points in the sectors where the distributions and their means were significantly different.Sectors in the plot to the left were those where counties with a change point had significantly higher means (p < 0.05)), sectors in the plot to the right had significantly greater mean employment share in counties without change points. The distributions across employment share for counties with change points are shown in pink, while the distributions for counties without change points is shown in green. While the differences in mean and distribution for all shown sectors are significant, they are small.*
For the 10 sectors with significantly different distributions and means, 5 had higher mean employment share in counties with abrupt changes: Information, Finance and insurance, Professional, scientific, and technical services, Educational services, and Health care and social assistance. These sectors are primarily comprised of white collar workers, and with the exception of Health care and social assistance, require less onsite work. Farm employment, Mining, quarrying, and oil and gas extraction, Construction, Manufacturing, and Retail trade all had higher mean employment share in counties where abrupt changes in park visitation did not occur.
## Discussion
At the state level, there was no significant difference in the partisanship of regions where an abrupt change in park visitation took place, and those where it had not. There was a significant difference in the vote share of Libertarians, with Libertarians having smaller vote share in states with an abrupt change. However, Libertarian voters account for less than $3\%$ of voters in each state, and are unlikely to be themselves pivotal in deciding overall park visitation behavior for a state. Thus, the practical significance of the difference in Libertarian vote share is doubtful. However, at the county level there is a clear divide in the partisanship of regions where park visitation did and did not undergo abrupt change. Counties with an abrupt change were more likely to be majority Democratic, while counties without a change point were more likely to be Republican. Taken together with the urban bias of the data set, it is possible that the state results are confounded by an over representation of urban park visits.
If abrupt park visitation changes were more associated with Democrat behavior, since urban areas have a Democratic bias, it is possible that the behavior of the urban park goers (who are more likely to be Democrats) may have overshadowed park going behavior in the rural parts of states. This possibility is made further plausible by the observation that the counties with a change point tend to be more populated. If park visitation changes are more likely in areas of greater population, and these areas are also over represented in the data, it stands to reason that aggregation to the state level may obscure behavior of the rural residents in the park visitation data.
Of course there is a second implication of these observations which is that whether or not park visitation exhibited an abrupt change is directly related to population density. If true, this relationship would explain why there is a disparity in population size for counties with and without abrupt changes, and why the counties with the lowest populations did not have abrupt changes, while the counties with the greatest populations did. In this case, differences in party affiliation of the respective areas is possibly unrelated, and only appears due to the confounding correlation between population density and party affiliation [31]. Since there is a connection between small populations and extreme partisanship as well, this would offer a potential explanation for why the span of the distribution across vote share for either party is greater for counties without abrupt changes.
Counties without abrupt changes in park visitation were more likely to have higher proportions of employment in Manufacturing, Construction, Mining, and Farming. Many of the workers in these sectors would have been considered “essential,” and much of the work would be site specific. Meanwhile, counties with abrupt changes were more likely to have greater proportions of jobs in Information, Finance and insurance, Professional, scientific, and technical services, and Educational services; sectors where remote work would have been more widely adopted. It is curious that regions with greater proportions of remote workers, who may have had greater time and opportunity to visit parks at the time, were more likely to experience a drop-off in visits. The difference is interesting and suggests it is possible that reductions in employment related mobility impacted other mobility decisions, such as whether or not to visit parks.
However, while there are differences in employment share by sector, they are small, and their practical significance remains undetermined. The most striking differences found in this study were in population, and partisanship. Recent work [15, 21, 26] suggests that regions with higher Republican vote share exhibited less social distancing at the onset of the pandemic, were slower to adopt stay at home orders, and residents visited more points of interest than residents of regions with higher Democratic vote share, suggesting that overall mobility reduction was greater for Democratic counties than Republican ones. Insight from these new studies suggests that the lack of change in park visitation behavior among Republican regions simply reflects this partisan difference in mobility, and indicates that parks were not necessarily uniquely visited more or less relative to other points of interest.
## Limitations and future directions
This study did not account for differences in local COVID-19 response policies. Incorporation of these differences would be necessary to understand how local governance impacted park access, and how willing residents were to defy local mobility restrictions for parks as opposed to other locations.
The spatial distribution of the parks in our data set roughly corresponds to the spatial distribution of the population, creating a substantial urban bias that we do not control for in this study. Weighing park visitation in such a way to allow for aggregation to the state level without over representing the urban parks would enable more revealing analysis at the state level, and additional insight into the demographic differences between counties with and without change points, with less influence from population density.
Augmenting the current data set with visitation data for more rural parks could also aid in these goals. Greater representation of rural parks would also allow a better investigation into population and park access as it relates specifically to population density and general nature accessibility.
Due to a change in collection methodology at the end of 2019, which led to a spatially non-uniform increase in total visitation counts, we were unable to directly compare 2019 and 2020 data. While there are visibly dramatic dips in behavior for some states and counties at the end of March 2020, it is not possible to clearly quantify how these changes deviate from expected behavior, nor how the magnitude of these changes compare across regions. Future work could investigate other park visitation data, and attempt to use it to normalize and perhaps compare visitation changes. The park visitation estimates used in this study are necessarily underestimates because they rely on mobile devices, and do not account for multiple visits made by a single device in one day. Other measurements of visitation could improve accuracy of visitation estimates.
Comparison of visitation levels across years and regions, especially following the initial pandemic reaction, would be extremely helpful in determining whether or not there were differences in how park visitation was valued in different regions. This could also be achieved by comparing dips in park visitation to dips in visitation to other points of interest. In particular, it would be useful to understand how different areas, and different populations, weigh the benefits and risks of park usage in the pandemic, and how park usage diverted visitation to other destinations. Future work in this direction could benefit from spatio-temporal methods, which would account for any spatial correlation of park visits. Studies indicating which populations had access to parks, which may have been greatly beneficial during 2020, could be used to address potential social inequality, and reduce public health risk in the future.
## References
1. Frumkin H, Bratman GN, Breslow SJ, Cochran B, Kahn PH, Lawler JJ. **Nature contact and human health: A research agenda**. *Environmental health perspectives* (2017.0) **125** 075001. DOI: 10.1289/EHP1663
2. Bedimo-Rung AL, Mowen AJ, Cohen DA. **The significance of parks to physical activity and public health: A conceptual model**. *American Journal of Preventive Medicine* (2005.0) **28** 159-168. DOI: 10.1016/j.amepre.2004.10.024
3. Orsega-Smith E, Mowen AJ, Payne LL, Godbey G. **The Interaction of Stress and Park Use on Psycho-physiological Health in Older Adults**. *Journal of Leisure Research* (2004.0) **36** 232-256. DOI: 10.1080/00222216.2004.11950021
4. Lee KH, Heo J, Jayaraman R, Dawson S. **Proximity to parks and natural areas as an environmental determinant to spatial disparities in obesity prevalence**. *Applied Geography* (2019.0) **112** 102074. DOI: 10.1016/j.apgeog.2019.102074
5. Bojorquez I, Ojeda-Revah L. **Urban public parks and mental health in adult women: Mediating and moderating factors**. *International Journal of Social Psychiatry* (2018.0) **64** 637-646. DOI: 10.1177/0020764018795198
6. Reuben A, Rutherford GW, James J, Razani N. **Association of neighborhood parks with child health in the United States**. *Preventive Medicine* (2020.0) **141** 106265. DOI: 10.1016/j.ypmed.2020.106265
7. Berto R. **The role of nature in coping with psycho-physiological stress: a literature review on restorativeness**. *Behavioral sciences* (2014.0) **4** 394-409. DOI: 10.3390/bs4040394
8. Sturm R, Cohen D. **Proximity to Urban Parks and Mental Health**. *The journal of mental health policy and economics* (2014.0) **17** 19-24. PMID: 24864118
9. Cheng Y, Zhang J, Wei W, Zhao B. **Effects of urban parks on residents’ expressed happiness before and during the COVID-19 pandemic**. *Landscape and Urban Planning* (2021.0) **212** 104118. DOI: 10.1016/j.landurbplan.2021.104118
10. Liu S, Wang X. **Reexamine the value of urban pocket parks under the impact of the COVID-19**. *Urban Forestry & Urban Greening* (2021.0) **64** 127294. DOI: 10.1016/j.ufug.2021.127294
11. Slater SJ, Christiana RW, Gustat J. **Recommendations for Keeping Parks and Green Space Accessible for Mental and Physical Health During COVID-19 and Other Pandemics**. *Preventing Chronic Disease* (2020.0) **17** E59. DOI: 10.5888/pcd17.200204
12. Xie J, Luo S, Furuya K, Sun D. **Urban Parks as Green Buffers During the COVID-19 Pandemic**. *Sustainability* (2020.0) **12** 6751. DOI: 10.3390/su12176751
13. Schwartz AJ, Dodds PS, O’Neil-Dunne JPM, Ricketts TH, Danforth CM. **Gauging the happiness benefit of US urban parks through Twitter**. *PLOS ONE* **17** e0261056. DOI: 10.1371/journal.pone.0261056
14. Schwartz AJ, Dodds PS, O’Neil-Dunne JPM, Danforth CM, Ricketts TH. **Visitors to urban greenspace have higher sentiment and lower negativity on Twitter**. *People and Nature* **1** 476-485. DOI: 10.1002/pan3.10045
15. Hsiehchen D, Espinoza M, Slovic P. **Political partisanship and mobility restriction during the COVID-19 pandemic**. *Public Health* (2020.0) **187** 111-114. DOI: 10.1016/j.puhe.2020.08.009
16. Soucy JPR, Sturrock SL, Berry I, Westwood DJ, Daneman N, MacFadden DR. **Estimating effects of physical distancing on the COVID-19 pandemic using an urban mobility index**. *medRxiv* (2020.0)
17. Rice WL, Pan B. **Understanding changes in park visitation during the COVID-19 pandemic: A spatial application of big data**. *Wellbeing, Space and Society* (2021.0) **2** 100037. DOI: 10.1016/j.wss.2021.100037
18. Volenec ZM, Abraham JO, Becker AD, Dobson AP. **Public parks and the pandemic: How park usage has been affected by COVID-19 policies**. *PLOS ONE* (2021.0) **16** e0251799. DOI: 10.1371/journal.pone.0251799
19. Chen MK, Rohla R. **The effect of partisanship and political advertising on close family ties**. *Science* (2018.0) **360** 1020-1024. DOI: 10.1126/science.aaq1433
20. Gadarian SK, Goodman SW, Pepinsky TB. **Partisanship, health behavior, and policy attitudes in the early stages of the COVID-19 pandemic**. *Plos one* (2021.0) **16** e0249596. DOI: 10.1371/journal.pone.0249596
21. Allcott H, Boxell L, Conway J, Gentzkow M, Thaler M, Yang D. **Polarization and public health: Partisan differences in social distancing during the coronavirus pandemic**. *Journal of Public Economics* (2020.0) **191** 104254. DOI: 10.1016/j.jpubeco.2020.104254
22. Clinton J, Cohen J, Lapinski J, Trussler M. **Partisan pandemic: How partisanship and public health concerns affect individuals’ social mobility during COVID-19**. *Science Advances* (2021.0) **7** eabd7204. DOI: 10.1126/sciadv.abd7204
23. Gollwitzer A, Martel C, Brady WJ, Pärnamets P, Freedman IG, Knowles ED. **Partisan differences in physical distancing are linked to health outcomes during the COVID-19 pandemic**. *Nature Human Behaviour* (2020.0) **4** 1186-1197. DOI: 10.1038/s41562-020-00977-7
24. Ye X. **Exploring the relationship between political partisanship and COVID-19 vaccination rate**. *Journal of Public Health* (2021.0) fdab364. DOI: 10.1093/pubmed/fdab364
25. Engle S, Stromme J, Zhou A. **Staying at Home: Mobility Effects of COVID-19**. *SSRN Electronic Journal* (2020.0). DOI: 10.2139/ssrn.3565703
26. Grossman G, Kim S, Rexer JM, Thirumurthy H. **Political partisanship influences behavioral responses to governors’ recommendations for COVID-19 prevention in the United States**. *Proceedings of the National Academy of Sciences* (2020.0) **117** 24144-24153. DOI: 10.1073/pnas.2007835117
27. Geng DC, Innes J, Wu W, Wang G. **Impacts of COVID-19 pandemic on urban park visitation: a global analysis**. *Journal of Forestry Research* (2021.0) **32** 553-567. DOI: 10.1007/s11676-020-01249-w
28. 28Lopez B, Kennedy C, McPhearson T. Parks are Critical Urban Infrastructure: Perception and Use of Urban Green Spaces in NYC During COVID-19; 2020. Available from: https://www.preprints.org/manuscript/202008.0620/v2.
29. 29Jordahl K, den Bossche JV, Fleischmann M, Wasserman J, McBride J, Gerard J, et al. geopandas/geopandas: v0.8.1; 2020. Available from: 10.5281/zenodo.3946761.. DOI: 10.5281/zenodo.3946761
30. Zhao K, Wulder MA, Hu T, Bright R, Wu Q, Qin H. **Detecting change-point, trend, and seasonality in satellite time series data to track abrupt changes and nonlinear dynamics: A Bayesian ensemble algorithm**. *Remote Sensing of Environment* (2019.0) **232** 111181. DOI: 10.1016/j.rse.2019.04.034
31. Gimpel JG, Lovin N, Moy B, Reeves A. **The Urban–Rural Gulf in American Political Behavior**. *Political Behavior* (2020.0) **42** 1343-1368. DOI: 10.1007/s11109-020-09601-w
|
---
title: 'Prevalence, patterns and predictors of metabolic abnormalities in Nigerian
hypertensives with hypertriglyceridemic waist phenotype: A cross sectional study'
authors:
- Casmir E. Amadi
- Amam C. Mbakwem
- Dolapo C. Duro
- Ifeoma C. Udenze
- Clement M. Akinsola
- Jayne N. Ajuluchukwu
- David A. Wale Oke
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021371
doi: 10.1371/journal.pgph.0001203
license: CC BY 4.0
---
# Prevalence, patterns and predictors of metabolic abnormalities in Nigerian hypertensives with hypertriglyceridemic waist phenotype: A cross sectional study
## Abstract
### Background
Simultaneous presence of elevated waist circumference and hypertriglyceridemia (HTGW) is a simple and low-cost measure of visceral obesity, and it is associated with a plethora of cardio-metabolic abnormalities that can increase the risk of cardiovascular diseases and incident Type 2 diabetes mellitus. We decided to study the prevalence, patterns, and predictors of metabolic abnormalities in Nigerian hypertensives with the HTGW phenotype.
### Methods
The medical records of 582 hypertensives with complete data of interest were retrieved and analyzed for the study. Their socio-demographic data, anthropometric data, and booking blood pressure values were retrieved. The results of their fasting plasma glucose, lipid profile, uric acid and serum creatinine were also retrieved for analysis.
### Results
The mean age of the study population was 56.2 ±13.6, with $53.1\%$ being males. The prevalence of smoking and use of alcohol was $4.3\%$ and $26.5\%$ respectively. The prevalence of the HTGW phenotype was $23.4\%$ and were predominantly males ($61\%$). Subjects with the HTGW phenotype were more obese assessed by waist circumference (WC) and body mass index (BMI). Mean serum total cholesterol, triglyceride, very low-density lipoprotein, uric acid, and creatinine were significantly higher in the HTGW phenotype ($$p \leq 0.003$$; <0.001; <0.001; 0.002 and <0.001 respectively). The prevalence of newly diagnosed Type 2 diabetes was $28.7\%$. There was also a preponderance of cardio-metabolic abnormalities (obesity, dyslipidaemia, hyperuricemia) in the HTGW phenotype. In both males and females, the HGTW phenotype was significantly associated with elevated Tc, TG, VLDL, hyperuricemia and atherogenic index of plasma.
### Conclusion
The HTGW phenotype is common amongst Nigerian hypertensives, and it is associated with metabolic abnormalities.
## Introduction
Cardiovascular disease (CVD) remains the number one cause of preventable and premature death globally, accounting for 17.9 million deaths in 2019 [1]. The Low-Middle Income Countries (LMICs) bear about $80\%$ of this burden due to their enormous population, rapid economic development, lifestyle changes, and a high prevalence of co-occurring CVD risk factors [2, 3]. Obesity (diagnosed either as elevated Body Mass Index or elevated waist circumference) is a traditional risk factor for CVD. Body mass index (BMI) is commonly used to assess general obesity. However, waist circumference (WC), as a measure of abdominal obesity, has been proven to be a better predictor of CVD than BMI [4–6]. However, use of WC alone is limited by the fact that it is not able to discriminate between intra-abdominal/visceral adiposity and subcutaneous abdominal adiposity. Several studies have shown that visceral adiposity rather subcutaneous abdominal adiposity is plays a crucial role in glucose metabolism, blood pressure homeostasis, lipid metabolism and inflammation and has positive associations with metabolic abnormalities, including hypertension, type 2 diabetes, dyslipidemia, and hyperinsulinemia [7, 8]. Measurements of visceral adiposity require complex and expensive imaging techniques such as computed tomography (CT) or magnetic resonance imaging (MRI) and are thus not suitable and cost effective for the general population and for routine clinical evaluation of visceral adiposity. Lemieux et al. [ 9] demonstrated that the hypertriglyceridemic waist (HTGW) phenotype (defined as co-occurring hypertriglyceridemia and elevated WC) was a simple and inexpensive tool to identify cardiometabolic abnormalities (especially the metabolic triad of hyperinsulinemia, elevated Apolipo B and small dense LDL-c) in individuals with visceral obesity in routine clinical encounters with patients [10, 11]. HTGW is also associated with increased risk of angiography confirmed coronary artery disease and sub-clinical atherosclerosis [12, 13].
Hypertensives have been reported to be at increased of risk having HTGW phenotype [14–16]. In these hypertensives this phenotype is associated with a plethora of metabolic abnormalities, which increases the risk of incident CVD [17, 18]. In Nigeria the prevalence of hypertension is about $30.8\%$ [19]. Often hypertension aggregates other CVD risk factors especially visceral obesity and dyslipidemia, which expectedly increases risk of incident CVD. The relationship between HTGW and other metabolic abnormalities in Nigerian hypertensives has not been extensively documented. Our study was aimed at describing the prevalence, patterns, and predictors of metabolic abnormalities in Nigerian hypertensives with the HTGW phenotype.
## Study population
This was a cross sectional retrospective study involving the review of the hospital records of adult hypertensives who presented at a Cardiac Centre in Lagos, Nigeria over a 12-year period (2009 to 2021). A total of 4,412 patients presented at this hospital within the study period and 1,430($32\%$) were hypertensives, approximately the reported prevalence of hypertension in Nigeria.
## Data collection/measurements
All hypertensives presenting for the first time at the Cardiac Centre are usually extensively evaluated by cardiologists and trained nurses. This evaluation includes their socio-demographic characteristics, lifestyle risk factors, medical history and thorough physical examination. The nurses measure and record their anthropometric indices as well as their blood pressure using standard protocols as described below. These nurses are trained on how to measure these parameters as part of their pre-employment training. After clinical evaluation these patients are requested to undergo laboratory evaluation.
## Anthropometry
Weight and height were measured to the nearest 0.5 kg and 0.1 cm, respectively, with the patients standing and wearing light clothing, no head gear or footwear. Body Mass Index (BMI) was derived as weight in kilograms divided by height2. Waist circumstance (WC) was measured with an inelastic tape (to the nearest 0.1 cm), at the level midway between the lower rib margin and the iliac crest.
## Blood pressure measurement
The blood pressure (BP) of the patients was measured by the nurses after five minutes of rest, with the patients seated comfortably, feet on the floor, arm at the level of the heart and free of any constricting clothing. Appropriate–sized cuffs connected to an Omron HEM7233 (Osaka, Japan) digital sphygmomanometer were used in measuring the BP, which was taken initially on both arms, and the arm with the higher value was used in subsequent measurements. Three BP readings were usually taken at two-to three-minutes intervals. The average of three readings was taken for analysis.
## Biochemical parameters
Venous blood sample of all hypertensives were collected after an overnight fasting and stored in the appropriate vacutainer specimen bottles. The samples were transported to the laboratory in thermo-coolers for processing and analysis. Serum concentrations of total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), fasting plasma glucose (FPG), serum uric acid, and serum creatinine were measured. From the lipid profile Atherogenic Index of Plasma (AIP) was determined as the log transformation of TG/HDL-C. Non-HDL-C was derived as TC minus HDL.
Relevant patient information required for analysis were retrieved from their case notes by trained research assistants. These included: These were entered into a spread sheet by the research assistants.
## Ethical approval
Ethics approval for the research was obtained from Health Research Ethics Committee of the Lagos University Teaching Hospital.
## Definition of terms
Triglyceridemic and waist phenotypes:
## Statistical analysis
Continuous variables were expressed as mean ± standard deviation (SD) or median and interquartile range when skewed, while categorical variables expressed as percentages. Analysis of variance (ANOVA) and Kruskal Wallis were used to compare differences in the 4 phenotypes for the continuous variables while chi square test was used to compare differences between categorical variables. Logistic regression analyses with odds ratios (ORs) at $95\%$ confidence intervals (CIs) were performed to estimate the association between HTGW phenotype and metabolic abnormalities. All statistical analyses were performed using SPSS version 26.0 software (SPSS Inc, Chicago, IL), and a $p \leq 0.05$ was considered as statistically significant. Charts was used for data presentation where appropriate.
## General characteristics
The records of 1,430 hypertensives were studied but only 582($40.7\%$) had the complete dataset of interest. Table 1 shows the general characteristics of the population.
**Table 1**
| Variable | Total | Male | Female | p-value |
| --- | --- | --- | --- | --- |
| Variable | (n = 582) | (n = 309) | (n = 273) | p-value |
| Age (year) | 56.2 ±13.6 | 54.9 ±13.6 | 57.6 ±13.4 | 0.015* |
| Smoking status (%) | 25 (4.3) | 23 (7.4) | 2 (0.7) | <0.001* |
| Drinking status (%) | 154 (26.5) | 124 (40.1) | 30 (11.0) | <0.001* |
| Clinical data | | | | |
| WC (cm) | 105.4 ± 15.6 | 104.3 ± 16.4 | 106.6 ± 14.6 | 0.078 |
| BMI (kg/m2) | 31.7 ± 6.6 | 30.3 ± 5.8 | 33.2 ± 7.1 | <0.001* |
| Overweight (%) | 173 (29.7) | 104 (33.7) | 69 (25.3) | 0.027* |
| Obesity (%) | 338 (58.1) | 160 (51.8) | 178 (65.2) | 0.001* |
| SBP (mmHg) | 154.3 ± 24.3 | 153.0 ± 24.3 | 155.8 ± 24.2 | 0.157 |
| DBP (mmHg) | 88.9 ± 14.9 | 89.7 ± 14.6 | 87.9 ± 15.2 | 0.156 |
| Duration of HTN (yrs) | 9.0(5–16) | 9.0(5–15) | 9.0(5–17) | 0.847 |
| Laboratory data | | | | |
| Tc (mmol/l) | 5.17 ± 1.2 | 5.05 ± 1.2 | 5.30 ± 1.1 | 0.011* |
| TG (mmol/l)) | 1.48 ± 0.5 | 1.59 ± 0.5 | 1.34 ± 0.5 | <0.001* |
| LDL (mmol/l) | 3.20 ± 1.1 | 3.10 ± 1.2 | 3.33 ± 0.9 | 0.012* |
| VLDL (mmol/l) | 0.64 ± 0.2 | 0.69 ± 0.2 | 0.59 ± 0.2 | <0.001* |
| HDL (mmol/l) | 1.34 ± 0.3 | 1.29 ± 0.3 | 1.39 ± 0.3 | <0.001* |
| Non-HDL ((mmol) | 3.83 ± 1.1 | 3.77 ± 1.2 | 3.90 ± 1.0 | 0.127 |
| Uric acid (μmol/l) | 429.27 ± 132.5 | 458.48±140.9 | 396.20±113.9 | <0.001* |
| HbA1C (%) | 6.28 ± 1.4 | 6.30 ± 1.5 | 6.30 ± 1.3 | 0.813 |
| FBS (mmol/l) | 6.05 ± 2.1 | 6.01 ± 2.19 | 6.09 ± 2.1 | 0.666 |
| Creatinine (mmol/l) | 97.26 (81–115) | 103.45 (88–124) | 88.42 (77–108) | <0.001* |
| Atherogenic index | 0.9(0.5–1.2) | 1.0(0.6–1.3) | 0.7(0.4–1.0) | <0.001* |
There were more males, 309($53.1\%$) but females were relatively older, $$p \leq 0.015.$$ Twenty-five ($4.3\%$) of the hypertensives were smokers while 154($26.5\%$) used alcohol and were predominantly males, $p \leq 0.001$ respectively. The mean duration of hypertension was 9.0 (5–16) years. Females tended to have significantly higher mean BMI, were more obese, and had higher mean Tc and LDL-c, while males had significantly higher mean values of TG, VLDL, creatinine, uric acid, and AIP. Blood pressure phenotypes were comparable in both genders.
Majority 484 ($83.2\%$) of hypertensives were on 2 or more medications for BP control. The top three prescribed antihypertensives were Renin Angiotensin System blockers, diuretics, and calcium channel blockers.
## Prevalence and patterns of triglyceridemic-waist phenotype
Fig 1 shows the distribution of the triglyceridemic-waist phenotype. The prevalence of HTGW phenotype was $23.4\%$.
**Fig 1:** *Distribution of the triglyceridemic-waist phenotype in the study population.*
Table 2 shows the characteristics of the phenotypes. Subjects with the HTGW phenotype were relatively younger, predominantly males ($61.0\%$) and less of smokers compared to NWNT phenotype. The EWNT and HTGW phenotypes had significantly higher mean WC and BMI compared to NWNT and NWET phenotypes, $p \leq 0.001$ respectively. The prevalence of obesity was comparable in the EWNT and HTGW phenotypes ($65.5\%$ vs $63.2\%$) but significantly higher than in the NWNT and NWET phenotypes, $p \leq 0.001.$ Mean SBP and DBP values were comparable in all 4 phenotypes, $$p \leq 0.33$$ and 0.92 respectively. The mean value of all biochemical parameters except LDL-c were significantly higher in EWNT, HTGW and NWET phenotypes compared to NWNT phenotypes.
**Table 2**
| Variable | NWNT | EWNT | NWET | HTGW | p-value |
| --- | --- | --- | --- | --- | --- |
| Variable | (n = 49) | (n = 377) | (n = 20) | (n = 136) | p-value |
| Age(yr)± SD | 56.8 ±17.0 | 57.2 ±13.2 | 49.1 ±13.1 | 54.1±12.9 | 0.012* |
| Sex (%) | | | | | <0.001** |
| Male | 43 (87.8) | 165 (43.8) | 18 (90.0) | 83 (61.0) | |
| Female | 6 (12.2) | 212 (56.2) | 2 (10.0) | 53 (39.0) | |
| Smoking status (%) | 3 (6.1) | 13 (3.4) | 3 (15.0) | 6 (4.4) | 0.004** |
| Drinking status (%) | 11 (22.4) | 88 (23.3) | 11 (55.0) | 44 (32.4) | 0.084** |
| Clinical data | | | | | |
| WC (cm) | 83.9 ± 11.8 | 108.6 ± 13.6 | 79.2 ± 17.6 | 108.4 ± 11.4 | <0.001* |
| BMI (kg/m2) | 24.1 ± 3.3 | 32.7 ± 6.5 | 25.9 ± 2.9 | 32.3 ± 6.1 | <0.001* |
| Overweight (%) | 14 (28.6) | 107 (28.4 | 9 (45.0) | 43 (31.6) | 0.420 |
| Obesity (%) | 2 (4.1) | 248 (65.8) | 2 (10.0) | 86 (63.2) | <0.001** |
| SBP (mmHg) | 158.3 ± 20.3 | 154.8 ± 23.9 | 149.2 ± 21.9 | 152.2 ± 23.4 | 0.332 |
| DBP (mmHg) | 88.6 ± 16.0 | 88.9 ± 15.0 | 91.1 ± 12.3 | 88.6 ± 14.8 | 0.922 |
| Duration of HTN (year) | 8.0 (5.0–15.0) | 9.0(6.0–17.0) | 8.0(4.0–15.0) | 7.0(5.0–14.0) | 0.053 |
| Laboratory data | | | | | |
| Tc (mmol/l) | 4.89 ± 1.1 | 5.08 ± 1.1 | 5.39 ± 1.5 | 5.46 ± 1.1 | 0.003* |
| TG (mmol/l) | 1.12 ± 0.3 | 1.15 ± 0.3 | 2.46 ± 0.7 | 2.36 ± 0.8 | <0.001* |
| LDL (mmol/l) | 3.15 ± 0.9 | 3.15 ± 0.9 | 3.09 ± 0.7 | 3.12 ± 1.1 | 0.592 |
| VLDL (mmol/l) | 0.50 ± 0.1 | 0.50 ± 0.1 | 1.05 ± 0.3 | 1.01 ± 0.3 | <0.001* |
| HDL (mmol/l) | 1.26 ± 0.3 | 1.34 ± 0.3 | 1.25 ± 0.4 | 1.36 ± 0.3 | 0.239 |
| Non-HDL (mmol/l) | 3.63 ± 1.0 | 3.74 ± 1.0 | 4.14 ± 1.7 | 4.11 ± 1.1 | 0.003* |
| Uric acid (μmol/L) | 435.75 ± 114 | 417.10 ± 121 | 514.77 ± 147 | 448.10 ± 109 | 0.002* |
| HbA1C (%) | 6.16 ± 1.3 | 6.20 ±1.4 | 6.74 ± 1.7 | 6.49 ±1.5 | 0.971 |
| FBS (mmol/l) | 6.13 ± 2.0 | 6.00 ± 2.2 | 5.57 ± 2.3 | 6.22 ± 2.1 | 0.599 |
| Creatinine (mmol/l) | 97.26(80–108) | 96.38(80–114) | 110.53(93–213) | 99.03(86–122) | 0.006# |
| Atherogenic index | 0.7(0.4–1.1) | 0.7(0.4–0.9) | 1.5(1.0–1.8) | 1.3(1.1–1.6) | <0.001# |
## Prevalence of metabolic perturbations across the triglyceridemic-waist phenotypes
Table 3 shows the metabolic abnormalities across the phenotypes. The HTGW phenotype had higher prevalence of elevated Tc ($61.8\%$), TG ($100.0\%$), VLDL ($91.2\%$), $p \leq 0.001$; serum uric acid ($67.6\%$) and AIP ($100.0\%$), $$p \leq 0.002$$ and $p \leq 0.001$ respectively compared to NWNT. Males with HTGW phenotype had significantly higher prevalence of elevated Tc ($60.2\%$), TG ($100.0\%$), VLDL ($96.4\%$) and serum uric acid ($63.9\%$), $$p \leq 0.003$$, and $p \leq 0.001$ respectively while in the females this phenotype was associated with higher prevalence of diabetes mellitus ($39.6\%$; newly diagnosed), elevated TG ($100.0\%$) and VLDL ($83\%$) levels; $$p \leq 0.01$$ and $p \leq 0.001$ respectively (Table 3).
**Table 3**
| Variable | Total | NWNT | EWNT | NWET | HTGW | p-value |
| --- | --- | --- | --- | --- | --- | --- |
| Total | (n = 582) | (n = 49) | (n = 377) | (n = 20) | (n = 136) | |
| High Tc | 268(46.0) | 18 (36.7) | 158 (41.9) | 8 (40.0) | 84 (61.8) | <0.001* |
| High TG | 156 (26.8) | 0 (0) | 0 (0) | 20 (100) | 136 (100) | <0.001* |
| High LDL | 185 (31.8) | 14 (28.6) | 123 (32.6) | 3 (15.0) | 45 (33.1) | 0.382 |
| High VLDL | 144 (24.7) | 0 (0) | 0 (0) | 20 (100) | 124 (91.2) | <0.001* |
| Low HDL | 97 (16.7) | 13 (26.5) | 61 (16.2) | 6 (30.0) | 17 (12.5) | 0.051 |
| High Non-HDL | 383 (65.8) | 29 (59.2) | 240 (63.7) | 13 (65.0) | 101 (74.3) | 0.109 |
| High Uric acid | 346 (59.5) | 22 (44.9) | 215 (57.0) | 17 (85.0) | 92 (67.6) | 0.002* |
| High Creatinine | 268 (46.0) | 22 (44.9) | 161 (42.7) | 12 (60.)) | 73 (53.7) | 0.091 |
| IGT | 149 (25.6) | 18 (36.7) | 99 (26.3) | 3 (15.0) | 29 (21.3) | 0.124 |
| DM | 123 (21.1) | 9 (18.4) | 72 (19.1) | 3 (15.0) | 39 (28.7) | 0.100 |
| High HbA1C | 198 (33.3) | 17 (34.7) | 119 (31.6) | 8 (40.0) | 50 (36.8) | 0.639 |
| High AI | 554 (95.2) | 45 (91.8) | 353 (93.6) | 20 (100) | 136 (100) | 0.011* |
| Men | (n = 309) | (n = 43) | (n = 165) | (n = 18) | (n = 83) | |
| High Tc | 133 (43.0) | 16 (37.2) | 60 (36.4) | 7 (38.9) | 50 (60.2) | 0.003* |
| High TG | 101 (32.9) | 0 (0) | 0 (0) | 18 (100 | 83 (100) | <0.001* |
| High LDL | 86 (77.8) | 12 (27.9) | 48 (29.1) | 2 (11.1) | 24 (28.9) | 0.443 |
| High VLDL | 98 (31.7) | 0 (0) | 0 (0) | 18 (100) | 80 (96.4) | <0.001* |
| Low HDL | 65 (21.0) | 10 (23.3) | 38 (23.0) | 5 (27.8) | 12 (14.5) | 0.365 |
| High Non-HDL | 191 (61.8) | 25 (58.1) | 95 (57.6) | 18 (100) | 60 (72.3) | 0.147 |
| High Uric acid | 179 (57.9) | 19 (44.2) | 91 (55.2) | 16 (88.9) | 53 (63.9) | 0.007* |
| High Creatinine | 168 (54.4) | 18 (41.9) | 97 (58.8) | 11 (61.1) | 42 (50.6) | 0.186 |
| IGT | 81 (26.2) | 15 (34.9) | 45 (37.3) | 2 (11.1) | 19 (22.9) | 0.225 |
| DM | 60 (19.4) | 7 (16.3) | 32 (19.4) | 3 (16.7) | 18 (21.7) | 0.889 |
| High HbA1C | 102 (33.0) | 14 (32.6) | 54 (32.7) | 7 (38.9) | 27 (32.5) | 0.960 |
| High AI | 295 (95.6) | 40 (93.0) | 154 (93.3) | 18 (100) | 83 (100) | 0.068 |
| Women | (n = 273) | (n = 6) | (n = 212) | (n = 2) | (n = 53) | |
| High Tc | 135 (49.5) | 2 (33.3) | 98 (46.2) | 1 (50.0) | 34 (64.2) | 0.107 |
| High TG | 55 (20.2) | 0 (0) | 0 (0) | 2 (100) | 53 (100) | <0.001* |
| High LDL | 99 (36.3) | 2 (33.3) | 75 (35.4) | 1 (50.0) | 32 (60.4) | 0.915 |
| High VLDL | 46 (16.8) | 0 (0) | 0 (0) | 2 (100) | 44 (83.0) | <0.001* |
| Low HDL | 32 (11.7) | 3 (50.0) | 23 (10.8) | 1 (50.0) | 5 (9.4) | 0.008* |
| High Non-HDL | 192 (70.3) | 4 (66.7) | 145 (68.4) | 2 (100) | 41 (77.4) | 0.472 |
| High Uric acid | 167 (61.2) | 3 (50.0) | 124 (58.5) | 1 (50.0) | 39 (73.6) | 0.212 |
| High Creatinine | 100 (36.6) | 4 (66.7) | 64 (30.2) | 1 (50.0) | 31 (58.5) | 0.001* |
| IGT | 68 (24.9) | 3 (50.0) | 54 (25.5) | 1 (50.0) | 10 (18.9) | 0.288 |
| DM | 63 (23.1) | 2 (33.3) | 40 (18.9) | 0 (0) | 21 (39.6) | 0.010* |
| High HbA1C | 92 (33.7) | 3 (50.0) | 65 (30.7) | 1 (50.0) | 23 (43.4) | 0.255 |
| High AI | 259 (94.9) | 5 (83.3) | 199 (93.9) | 2 (100) | 53 (100) | 0.168 |
## Patterns of metabolic perturbations across the triglyceridemic-waist phenotypes
Fig 2 shows the frequency of the 5 common CVD risk factors found in the study population. Obesity assessed with BMI and WC, dyslipidemia and hyperuricemia were the top three CVD risk factors.
**Fig 2:** *Prevalence of cardiovascular disease risk factors in the study population.*
Fig 3 shows the distribution of the number of CVD risk factors in the study population. More than $90\%$ of the study population had more than two risk factors.
**Fig 3:** *Distribution of the number of cardiovascular disease risk factors in the study population.*
Fig 4 shows the frequency of the number of CVD risk factors present in each of the 4 triglyceridemic-waist phenotypes. The HTGW phenotype had the highest frequency of co-occurring CVD risk factors of $84.6\%$.
**Fig 4:** *The frequency of the CVD risk factors in the triglyceridemic-waist phenotypes.*
Fig 5 shows the most frequent CVD risk factor combinations in the study population. The most frequent combinations were obesity, hyperuricemia ($25.2\%$) and obesity, abnormal glucose profile, hyperuricemia, and dyslipidemia ($18.5\%$).
**Fig 5:** *Some common cardiovascular risk factor combinations in the study population.*
## Predictors of triglyceridemic phenotypes and metabolic abnormalities
Table 4 shows the adjusted odds ratios of the 4 phenotypes and their association with metabolic abnormalities. Compared with the NWNT phenotype the HTGW phenotype had significantly strong associations with elevated Tc, TG, VLDL, uric acid, and AIP in the total population. Subjects (males and females) with the HTGW phenotype were 2.7 times ($95\%$ CI, 1.42–5.49; $$p \leq 0.003$$) more likely to have elevated Tc; 3.24 times ($95\%$CI, 1.05–10.39; $p \leq 0.001$) more likely to have elevated TG; 4.54 times ($95\%$ CI, 3.59–7.21; $p \leq 0.001$) more likely to have elevated VLDL; 2.57 times ($95\%$ CI, 1.32–5.00; $$p \leq 0.006$$) for hyperuricemia and 8.8 times ($95\%$ CI, 0.9–86.6; $$p \leq 0.03$$) of having elevated AIP (Table 4). With respect to elevated VLDL its association was stronger in females than in males; 10.15 ($95\%$ CI, 2.70–17.93; $p \leq 0.001$) vs 4.54 ($95\%$ CI, 3.59–7.21; $p \leq 0.001$ respectively (Table 4).
**Table 4**
| Metabolic abnormality | Predictor | Total (n = 582) | Total (n = 582).1 | Men (n = 309) | Men (n = 309).1 | Women (n = 273) | Women (n = 273).1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Metabolic abnormality | Predictor | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value |
| High Tc | NWNT | Reference | | Reference | | Reference | |
| High Tc | EWNT | 1.24 (9.67–2.30) | 0.489 | 0.96 (0.48–1.93) | 0.918 | 1.72 (0.31–9.59) | 0.537 |
| High Tc | NWET | 1.14 (0.40–3.34) | 0.800 | 1.07 (0.35–3.33) | 0.902 | 2.00 (0.08–51.59) | 0.676 |
| High Tc | HTGW | 2.78 (1.42–5.49) | 0.003* | 2.26 (1.20–5.46) | 0.015* | 3.58 (0.60–21.39) | 0.162 |
| High TG | NWNT | Reference | | Reference | | Reference | |
| High TG | EWNT | 0.13 (0.01–2.04) | 0.087 | 0.26 (0.02–4.09) | 0.303 | 0.02 (0.00–0.38) | <0.001* |
| High TG | NWET | 9.12 (5.7–14.52) | <0.001* | 3.14 (2.6–11.41) | <0.001* | 0.39 (0.12–2.69) | 0.346 |
| High TG | HTGW | 32.4 (4.05–10.39) | <0.001* | 3.93 (2.15–5.71) | <0.001* | 2.60 (1.7–7.93) | <0.001* |
| High LDL | NWNT | Reference | | Reference | | Reference | |
| High LDL | EWNT | 1.21 (0.63–2.33) | 0.568 | 1.06 (0.50–2.24) | 0.874 | 1.10 (0.20–6.12) | 0.918 |
| High LDL | NWET | 0.44 (0.11–1.75) | 0.243 | 0.32 (0.06–1.62) | 0.170 | 2.00 (0.08–51.59) | 0.676 |
| High LDL | HTGW | 1.24 (0.61–2.53) | 0.561 | 1.05 (0.46–2.38) | 0.905 | 1.31 (0.22–7.82) | 0.765 |
| High VLDL | NWNT | Reference | | Reference | | Reference | |
| High VLDL | EWNT | 0.13 (0.01–2.04) | 0.087 | 0.26 (0.02–4.09) | 0.303 | 0.20 (0.13–0.38) | <0.001* |
| High VLDL | NWET | 912 (57–14581) | <0.001* | 7.14 (4.51–11.93) | <0.001* | 0.38 (0.2- | 0.346 |
| High VLDL | HTGW | 4.54 (3.59–7.21) | <0.001* | 4.54 (3.59–7.21) | <0.001* | 10.15(2.76–17.93) | 0.001* |
| Low HDL | NWNT | Reference | | Reference | | Reference | |
| Low HDL | EWNT | 0.54 (0.27–1.07) | 0.076 | 0.99 (0.45–2.19) | 0.975 | 0.12 (0.02–0.64) | 0.013* |
| Low HDL | NWET | 1.19 (0.38–3.74) | 0.770 | 1.27 (0.36–4.45) | 0.709 | 1.00 (0.04–24.55) | 1.000 |
| Low HDL | HTGW | 0.40 (0.18–0.89) | 0.025* | 0.56 (0.22–1.42) | 0.221 | 0.10 (0.02–0.66) | 0.016* |
| High Non-HDL | NWNT | Reference | | Reference | | Reference | |
| High Non-HDL | EWNT | 1.21 (0.66–2.22) | 0.542 | 0.98 (0.50–1.93) | 0.947 | 1.08 (0.19–6.05) | 0.928 |
| High Non-HDL | NWET | 1.28 (0.43–3.78) | 0.654 | 1.13 (0.37–3.48) | 0.830 | 0.28 (0.22–0.47) | 0.999 |
| High Non-HDL | HTGW | 1.99 (1.00–4.00) | 0.050* | 1.88 (0.87–4.07) | 0.110 | 1.71 (0.28–10.49) | 0.563 |
| Hyperuricemia | NWNT | Reference | | Reference | | Reference | |
| Hyperuricemia | EWNT | 1.63 (0.90–2.96) | 0.110 | 1.55 (0.79–3.05) | 0.201 | 1.41 (0.28–7.15) | 0.679 |
| Hyperuricemia | NWET | 7.00 (1.80–26.84) | 0.005* | 10.11 (2.06–49.48) | 0.004* | 1.00 (0.04–24.55) | 1.000 |
| Hyperuricemia | HTGW | 2.57 (1.32–5.00) | 0.006* | 2.23 (1.05–4.73) | 0.036* | 2.79 (0.50–15.45) | 0.241 |
| High HbA1C | NWNT | Reference | | Reference | | Reference | |
| High HbA1C | EWNT | 0.87 (0.46–1.63) | 0.639 | 1.01 (0.49–2.06) | 0.983 | 0.44 (0.09–2.25) | 0.325 |
| High HbA1C | NWET | 1.26 (0.43–3.66) | 0.678 | 1.32 (0.42–4.13) | 0.636 | 1.00 (0.04–24.55) | 1.000 |
| High HbA1C | HTGW | 1.09 (0.55–2.17) | 0.796 | 0.99 (0.46–2.19) | 0.997 | 0.77 (0.14–4.16) | 0.758 |
| Highatherogenic index | NWNT | Reference | | Reference | | Reference | |
| Highatherogenic index | EWNT | 0.98 (0.702–4.90) | 0.892 | 0.76 (0.16–3.53) | 0.724 | 3.65 (0.47–28.43) | 0.224 |
| Highatherogenic index | NWET | 1.24 (0.13–12.06) | 0.856 | 2.94 (1.94–4.09) | <0.001* | 0.41 (0.01–1.93) | 0.346 |
| Highatherogenic index | HTGW | 8.80 (0.90–85.69) | 0.026* | 4.01 (3.11–5.79) | <0.01* | 10.40 (0.65–16.23) | 0.058 |
## Discussion
Although there are many published studies on metabolic syndrome in Nigerians ours is the first (according to the knowledge of the authors) to have explored the relationship between simultaneous presence hypertriglyceridemia and elevated waist circumference (HTGW) and metabolic abnormalities in hypertensives. Our study showed that the prevalence of the HTGW phenotype in our hypertensive cohort was $23.4\%$; $26.8\%$ and $19.4\%$ in males and females respectively. In the Chinese hypertensive population Chen et al. [ 15] reported a prevalence of $15.8\%$ and $13.5\%$ and $18.1\%$ in male and female hypertensives respectively. In ELSA-Brazil study the prevalence of HTGW in hypertensives was reported to range between $13.3\%$ and $24.7\%$ with higher prevalence in females [20]. Our prevalence is similar to the $21.4\%$ and $21.5\%$ reported by two other Brazilian studies [14, 21]. The varying prevalence might be related to the different cut-off points used for elevated WC in these studies and the type of population studied. The HTGW phenotype can be likened as a surrogate of metabolic syndrome [22–24]. They both share common criteria especially hypertriglyceridemia and elevated WC and the same cut-off points. In Nigeria the prevalence of metabolic syndrome is $28\%$ using the IDF criteria [25].
Our study reported a higher prevalence of HTGW in males. The impact of gender on the HTGW phenotype in published literature is varied. While some studies reported no gender differences in prevalence [26, 27], other studies reported either male preponderance [22, 28] or female preponderance [14, 15]. These differences might be methodological (gender composition of the study populations and in populations with varying cardiovascular risk factors). In our study we had more males, and this might explain the higher prevalence of HTGW in males.
Our study also demonstrated greater aggregation of metabolic perturbations in the HTGW phenotype, which has been demonstrated by other studies [14–16]. The direct implication of this perturbations is a heightened risk of CVD, prospectively [29–31]. The EPIC-Norfolk study, the largest prospective study to date to explore the relationship between HTGW phenotype and CVD risk, reported of increased risk of CVD and poor outcomes in men and women with the HTGW phenotype compared with those with NWNT phenotype [11]. Other prospective studies in other parts of Europe arrived at this same conclusion [32, 33]. In the Framingham Offspring Study, Wilson et al. [ 34] reported that hypertriglyceridemic waist contributed to elevated CVD risk and even more to associated with incident Type 2 diabetes mellitus.
The major metabolic abnormalities and the predictors of the found HTGW phenotype in our study were elevated Tc, VLDL, TG, uric acid, and AIP. In males the predictors of the HTGW phenotype were elevated Tc, VLDL-c, uric acid, and AIP. In females only elevated VLDL-c and uric acid were associated with the HTGW phenotype. However, VLDL was a stronger predictor of the HTGW phenotype in females than in males; 10.15 ($95\%$ CI, 2.70–17.93; $p \leq 0.001$) vs 4.54 ($95\%$ CI, 3.59–7.48; $p \leq 0.001$ respectively.
Although the prevalence of elevated high LDL-c was high in both males and females with HTGW phenotype ($28.9\%$ vs $60.4\%$ respectively), it was not a predictor of this phenotype. Some studies have reported elevated LDL-c as a predictor of the HTGW phenotype [14–18, 20]. Our HTGW cohort might have been on treatment with statins at the time of enrolment into the study, and this might explain the observed trend in their LDL-c levels. The clinical significance of elevated LDL-c in atherosclerotic CVD is well established. However, serum level of LDL-c does completely and carefully account for atherosclerotic CVD (ASCVD) even as the biological relationship between CVD and dyslipidemia continues to evolve. Very Low-Density Lipoprotein Cholesterol (VLDL-c), a component of non-HDL and triglyceride rich lipid fraction, is an identified risk factor for ASCVD and its importance in the prevention of CVD is widely recognized [34, 35]. Elevated VLDL-c is a major type of dyslipidemia, especially in China, and recent epidemiological studies have demonstrated the superiority of VLDL-c over low-density lipoprotein cholesterol (LDL-c) in terms of the population-attributable risk proportion for ASCVD [36]. Even in the presence of on-target LDL-c levels and in the absence of the traditional risk factors for CVD, elevated VLDL-c levels is associated with 2.19–3.36-fold increased risk of coronary heart disease [36]. Secondly, in the presence of obesity and diabetes (as is the case with our HTGW phenotype) LDL-c is not an accurate marker of CVD risk [37]. Cardiovascular disease events occur in the scenario of on-target or even low levels of LDL-c and/or statin therapy in some individuals, a phenomenon referred to as residual cardiovascular risk [37]. Triglyceride-rich lipoproteins (especially VLDL) and their cholesterol content could partially explain this residual risk [38]. These lipoproteins are novel biomarkers driving residual cardiovascular risk in our contemporary era of high burden of obesity, diabetes, and metabolic syndrome [38, 39]. The JUPITER trial demonstrated that VLDL-c, particularly the smallest remnant subclass, was associated with cardiovascular disease risk when LDL was low [40]. VLDL remnants (as well as LDL particles) have been shown to migrate across the endothelium where they are entrapped by macrophages, forming foam cells, promoting low-grade inflammation, and facilitating atheromatous plaque growth [41].
Hyperuricemia is strongly associated with CVD and CVD risk [42, 43]. In addition, this relationship has a J-curve phenomenon [44]. It is positively associated with obesity, hypertension, and dyslipidemia, and hyperuricemic subjects tend to have a clustering of these cardiovascular risk factors which also are components of the metabolic syndrome [45]. Viscerally obese people are known to produce high levels of uric acid. Our HTGW subjects had a significantly higher prevalence of hyperuricemia ($67.6\%$) compared to those with NWNT phenotype ($44.9\%$). This corroborates findings from similar studies [16, 27, 28]. The high prevalence of hyperuricemia in the HTGW phenotype agrees with the reported finding that elevated serum uric acid is associated with the metabolic syndrome, of which the HTGW phenotype is a surrogate of [46–48]. This clustering of hyperuricemia with metabolic syndrome or its components is associated with increased CVD risk. Our study also reported prevalence of $59.5\%$ for hyperuricemia in all subjects irrespective of their phenotypes. In Nigeria the prevalence of hyperuricemia is higher in hypertensives with prevalence rates ranging between $36.7\%$ and $60.2\%$ compared to $17.2\%$ to $20.5\%$ in the general population [49–51]. Hyperuricemia and hypertension have a bi-directional relationship. The former is more prevalent in hypertensives and is also a risk factor for incident hypertension. The association between serum uric acid levels and high blood pressure in humans is well established. For example, cross-sectional and longitudinal studies have shown that there is a $13\%$ increase in the risk of incident hypertension for each 1 mg/dL increase in serum uric acid in a general normotensive population not treated for hyperuricemia [52, 53]. This association is linear and commoner in the younger population and in females [54]. Moreover, hyperuricemia also contributes to the development of hypertension from prehypertension although no causality has been established between hyperuricemia and hypertension [55–57]. Possible mechanisms for the hyperuricemia and hypertension relationship include activation of the intra-renal renin-angiotensin system, urate deposition in the lumen of the nephrons and inflammation [56, 58, 59]. In our study hyperuricemia was a strong predictor of the HTGW phenotype especially in females.
Our study was able to demonstrate high prevalence of the hypertriglyceridemic waist phenotype in a cross section of Nigerian hypertensives and this phenotype is associated with multiple co-occurring metabolic perturbations that are likely to drive the risk of incident CVD. The triglyceridemic-waist is an inexpensive and simple to measure clinical parameter that can add a lot of value to assessing the cardiovascular risk of the hypertensives.
Our study has a few limitations. First it is a cross sectional study and will not be able to prove causality. Secondly the strength of association from this cross-sectional study cannot be as strong as a longitudinal study.
## Transfer Alert
This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.
## References
1. 1World Health Organization Factsheet: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds). Accessed May 24th, 2022.
2. Joseph P, Leong D, McKee M, Anand S, Schwalm J, Teo K. **Reducing the Global Burden of Cardiovascular Diseases, Part 1**. *The Epidemiology and Risk Factors:* (2017.0) **121** 677-694
3. Roth G, Forouzanfar M, Moran A, Barber R, Nguyen G, Feigin V. **Demographic and epidemiologic drivers of global cardiovascular mortality**. (2015.0) **372** 1333-1341. DOI: 10.1056/NEJMoa1406656
4. Brown P. **Waist circumference in primary care**. *Prim Care Diabetes* (2009.0) **3** 259-261. 2. DOI: 10.1016/j.pcd.2009.09.006
5. Klein S, Allison D, Heymsfield S. **Waist circumference and cardiometabolic risk: a consensus statement from shaping America’s health**. (2007.0) **30** 1647-1652. 3. PMID: 17360974
6. Seidell J.. **Waist circumference and waist/hip ratio in relation to all-cause mortality, cancer, and sleep apnea**. (2010.0) **64** 35-41. 4. PMID: 19639001
7. Després J, Lemieux I. **Abdominal obesity, and metabolic syndrome**. (2006.0) **444** 881-887. DOI: 10.1038/nature05488
8. Mathieu P, Poirier P, Pibarot P, Lemieux I, Després J. **Visceral obesity; the link among inflammation, hypertension, and cardiovascular disease**. (2009.0) **53** 577-584. DOI: 10.1161/HYPERTENSIONAHA.108.110320
9. Lemieux I, Pascot A, Couillard C, Lamarche B, Tchernof A, Almeras N. **Hypertriglyceridemic waist: a marker of the atherogenic metabolic triad (hyperinsulinemia; hyperapolipoprotein B; small, dense LDL) in men**. (2000.0) **102** 179-84. DOI: 10.1161/01.cir.102.2.179
10. Sam S, Haffner S, Davidson M, D’Agostino R, Feinstein S, Kondos G. **Hypertriglyceridemic waist phenotype predicts increased visceral fat in subjects with type 2 diabetes**. (2009.0) **32** 1916-20. DOI: 10.2337/dc09-0412
11. Arsenault B, Lemieux I, Despres J, Wareham N, Kastelein J, Khaw K. **The hypertriglyceridemic waist phenotype and the risk of coronary artery disease: results from the EPIC-Norfolk prospective population study**. (2010.0) **182** 1427-32. DOI: 10.1503/cmaj.091276
12. Lemieux I, Poirier P, Bergeron J, Alméras N, Lamarche B, Cantin B. **Hypertriglyceridemic waist: a useful screening phenotype in preventive cardiology?**. (2009.0) **25** 140
13. Namdarimoghaddam P, Fowokan A, Humphries K, John Machini G, Scott L. **Association of “hypertriglyceridemic waist” with increased 5-year risk of subclinical atherosclerosis in a multi-ethnic population: a prospective cohort study**. (2021.0) **21** 63. DOI: 10.1186/s12872-021-01882-1
14. Borges L, Comini L, de Oliveira L, Dias H, Ferreira E, Batistelli C. **Hypertriglyceridemic waist phenotype and associated factors in individuals with arterial hypertension and/or diabetes mellitus**. (2021.0) **10** e74. DOI: 10.1017/jns.2021.71
15. Chen S, Guo X, Yu S, Yang H, Sun G, Li Z. **Hypertriglyceridemic waist phenotype and metabolic abnormalities in hypertensive adults: A STROBE compliant study**. (2016.0) **95** e5613. DOI: 10.1097/MD.0000000000005613
16. Xuan Y, Shen Y, Wang S, Gao P, Gu X, Tang D. **The association of hypertriglyceridemic waist phenotype with hypertension: A cross-sectional study in a Chinese middle aged-old population**. (2022.0) **24** 191-199
17. Cabral N, Ribeiro V, França A, Salgado J, Santos A, Salgado Filho N. **Hypertriglyceridemic waist and cardiometabolic risk in hypertensive women**. (2012.0) **58** 568-73. DOI: 10.1590/s0104-42302012000500014
18. Cabral Rocha A, Feliciano Pereira P, Cristine Pessoa M, Gonçalves Alfenas Rde C, Segheto W, da Silva D. **Hypertriglyceridemic waist phenotype and cardiometabolic alterations in Brazilian adults**. (2015.0) **32** 1099-106. DOI: 10.3305/nh.2015.32.3.9305
19. Adeloye D, Owolabi E, Ojji D, Auta A, Dewan M, Olanrewaju T. **Prevalence, awareness, treatment, and control of hypertension in Nigeria in 1995 and 2020: A systematic analysis of current evidence**. *J Clin Hypertens (Greenwich)* (2021.0) 963-977. DOI: 10.1111/jch.14220
20. Freitas R, Fonseca M, Schmidt M, Molina M, Almeida M. **Hypertriglyceridemic waist phenotype: associated factors and comparison with other cardiovascular and metabolic risk indicators in the ELSA-Brasil study**. (2018.0) **34** e00067617. PMID: 29617485
21. 21Mendes M. Cintura hipertrigliceridêmica e sua associação com fatores de risco metabólicos [MSc dissertation]. Belo Horizonte, Brasil, Univer-sidade Federal de Minas Gerais; 2009.
22. Gomez-Huelgas R, Bernal-López M, Villalobos A, Mancera-Romero J, Baca-Osorio A, Jansen S. **Hypertriglyceridemic waist: an alternative the metabolic syndrome? Results ofthe IMAPStudy (multidisciplinary intervention in primary care)**. *Int J Obes (Lond)* (2011.0) **35** 292-299. PMID: 20548300
23. Nawabzad R, Champin B. **Concordance between three definitions for metabolic syndrome (Hypertriglyceridemic waist, National Cholesterol Education Program, International Diabetes Federation), and prevalence of the syndrome in a French population**. (2010.0) **60** 15-23. PMID: 22530271
24. Blackburn P, Lemieux I, Alméras N, Bergeron J, Côté M, Tremblay A. **The hypertriglyceridemic waist phenotype versus the National Cholesterol Education Program-Adult Treatment Panel III and International Diabetes Federation clinical criteria to identify high-risk men with an altered cardiometabolic risk profile**. (2009.0) **58** 1123-1130. DOI: 10.1016/j.metabol.2009.03.012
25. Oguoma V, Nwose E, Richards R. **Prevalence of Cardio-metabolic syndrome in Nigeria**. *A systematic review:* (2015.0) **125** 413-423
26. Ma C, Wang R, Liu X, Lu N, Lu Q, Yin F. **The Relationship between Hypertriglyceridemic Waist Phenotype and Early Diabetic Nephropathy in Type 2 Diabetes**. (2017.0) **4** 295-300. DOI: 10.1159/000477828
27. Cunha de Oliveira C, Carneiro Roriz A, Eickemberg M, Barreto Medeiros J, Barbosa Ramos L. **Hypertriglyceridemic waist phenotype: association with metabolic disorders and visceral fat in adults**. **30** 25-31. DOI: 10.3305/nh.2014.30.1.7411
28. Solati M, Ghanbarian A, Rahmani M, Sarbazi N, Allahverdian S, Aziz F. **Cardiovascular risk factors in males with hypertriglycemic waist (Tehran Lipid and Glucose Study)**. (2004.0) **28** 706-709. DOI: 10.1038/sj.ijo.0802582
29. Poirier J, Kubow S, Noël M, Dupont C, Egeland G. **The hypertriglyceridemic-waist phenotype is associated with the Framingham risk score and subclinical atherosclerosis in Canadian**. (2015.0) **25** 1050-1055
30. Wang A, Li Z, Zhou Y, Wang C, Luo Y, Liu X. **Hypertriglyceridemic waist phenotype and risk of cardiovascular diseases in China: results from the Kailuan Study**. (2014.0) **174** 106-9. DOI: 10.1016/j.ijcard.2014.03.177
31. Samadi S, Bozorgmanesh M, Khalili D, Momenan A, Sheikholeslami F, Azizi F. **Hypertriglyceridemic waist: the point of divergence for prediction of CVD vs. mortality: Tehran Lipid and Glucose Study**. (2013.0) **165** 260-5. DOI: 10.1016/j.ijcard.2011.08.049
32. Czernichow S, Bruckert E, Bertrais S, Galan P, Hercberg S, Oppert J. **Hypertriglyceridemic waist and 7.5-year prospective risk of cardiovascular disease in asymptomatic middle-aged men**. **31** 791-6. DOI: 10.1038/sj.ijo.0803477
33. Tanko L, Bagger Y, Qin G, Alexandersen P, Larsen P, Christiansen C. **Enlarged waist combined with elevated triglycerides is a strong predictor of accelerated atherogenesis and related cardiovascular mortality in postmenopausal women**. (2005.0) **111** 1883-1890. DOI: 10.1161/01.CIR.0000161801.65408.8D
34. Wilson P, D’Agostino R, Parise H, Sullivan L, Meigs J. **Metabolic syndrome as a precursor of cardiovascular disease and Type 2 diabetes mellitus**. (2005.0) **112** 3066-3072. DOI: 10.1161/CIRCULATIONAHA.105.539528
35. Prenner S, Mulvey C, Ferguson J, Rickels M, Bhatt A, Reilly M. **Very low-density lipoprotein cholesterol associates with coronary artery calcification in type 2 diabetes beyond circulating levels of triglycerides**. *Atherosclerosis* (2014.0) **236** 244-250. DOI: 10.1016/j.atherosclerosis.2014.07.008
36. Catapano A, Graham I, De Backer G, Wiklund O, Chapman M, Drexel H. **ESC/EAS guidelines for the management of dyslipidaemias: the task force for the management of dyslipidaemias of the European Society of Cardiology (ESC) and European Atherosclerosis Society (EAS)developed with the special contribution of the European Association for Cardiovascular Prevention & Rehabilitation (EACPR)**. (2016.0) **2016**
37. Ren J, Grundy S, Liu J, Wang W, Wang M, Sun J. **Long term coronary heart disease risk associated with very-low-density lipoprotein cholesterol in Chinese: the results of a 15-year Chinese Multi-Provincial Cohort Study (CMCS)**. (2010.0) **211** 327-332. DOI: 10.1016/j.atherosclerosis.2010.02.020
38. Gentile M, Iannuzzi A, Giallauria F, D’Andrea A, Venturini E, Pacileo M. **Association between Very Low-Density Lipoprotein Cholesterol (VLDL-C) and Carotid Intima-Media Thickness in Postmenopausal Women Without Overt Cardiovascular Disease and on LDL-C Target Levels**. (2020.0) **9** 1422. DOI: 10.3390/jcm9051422
39. Varbo A, Nordestgaard B. **Remnant lipoproteins**. (2017.0) **28** 300-307. DOI: 10.1097/MOL.0000000000000429
40. Abi-Ayad M, Abbou A, Abi-Ayad F, Behadada O, Benyoucef M. **HDL-C, ApoA1 and VLDL-TG as biomarkers for the carotid plaque presence in patients with metabolic syndrome**. (2018.0) **12** 175-179. DOI: 10.1016/j.dsx.2017.12.017
41. Lawler P, Akinkuolie A, Harada P, Glynn R, Chasman D, Ridker P. **Residual risk of atherosclerotic cardiovascular events in relation to reductions in very-low-density lipoproteins**. (2017.0) **6** e007402. DOI: 10.1161/JAHA.117.007402
42. Generoso G, Janovsky C, Bittencourt M. **S. Triglycerides and triglyceride-rich lipoproteins in the development and progression of atherosclerosis**. (2019.0) **26** 109-116. PMID: 30694827
43. Zhang S, Wang Y, Cheng J, Huangfu N, Zhao R, Xu Z. **Hyperuricemia and Cardiovascular Disease**. (2019.0) **25** 700-709. DOI: 10.2174/1381612825666190408122557
44. Shahin L, Patel K, Heydari M, Heydari M, Kesselman M. **Hyperuricemia and Cardiovascular Risk**. (2021.0) **13** e14855. DOI: 10.7759/cureus.14855
45. Kuo C, See L, Yu K, Chou I, Chiou M, Luo S. **Significance of serum uric acid levels on the risk of all-cause and cardiovascular mortality**. (2013.0) **52** 127-34. DOI: 10.1093/rheumatology/kes223
46. Li F, Chen S, Qiu X, Wu J, Tan M, Wang M. **Serum Uric Acid Levels and Metabolic Indices in an Obese Population: A Cross-Sectional Study**. (2021.0) **14** 627-635. DOI: 10.2147/DMSO.S286299
47. Dong H, Xu Y, Zhang X, Tian S. **Visceral adiposity index is strongly associated with hyperuricemia independently of metabolic health and obesity phenotypes**. (2017.0) **7** 8822. DOI: 10.1038/s41598-017-09455-z
48. Huang X, Jiang X, Wang L, Chen L, Wu Y, Gao P. **Visceral adipose accumulation increased the risk of hyperuricemia among middle-aged and elderly adults: a population-based study**. (2019.0) **17** 341. DOI: 10.1186/s12967-019-2074-1
49. Abidemi J, Olusegun A, Rotimi O, Musa Y, Oladipo G, Ayodele B. **Hyperuricemia and Its Correlation with Target Organ Damage and Electrocardiographic Changes in Newly Diagnosed Adult Nigerian Hypertensive Patients**. (2018.0) **7** 1-7
50. Abiodun M, Aliyu A. **Serum Uric Acid Levels among Nigerians with Essential Hypertension**. (2013.0) **28** 41-44. PMID: 23955405
51. Alikor C, Emem-Chioma P, Odia O. **Prevalence of hyperuricaemia in a rural population of Nigerian Niger Delta region**. (2013.0) **22** 187-92. PMID: 24180145
52. Wang J, Qin T, Chen J, Li Y, Wang L, Huang H. **Hyperuricemia and Risk of Incident Hypertension: A Systematic Review and Meta-Analysis of Observational Studies**. (2014.0) **9** e114259. DOI: 10.1371/journal.pone.0114259
53. Grayson P, Kim S, LaValley M, Choi H. **Hyperuricemia and incident hypertension: a systematic review and meta-analysis**. (2011.0) **63** 102-10. DOI: 10.1002/acr.20344
54. Kuwabara M, Hisatome I, Niwa K, Hara S, Roncal-Jimenez C, Bjornstad P. **Uric acid is a strong risk marker for developing hypertension from prehypertension: a 5-year Japanese cohort study**. (2018.0) **71** 78-86. DOI: 10.1161/HYPERTENSIONAHA.117.10370
55. Liu L, Gu Y, Li C, Xhang C, Meng G, Wu H. **Serum uric acid is an independent predictor for developing prehypertension: a population-based prospective cohort study**. (2017.0) **31** 116-120. DOI: 10.1038/jhh.2016.48
56. Feig D, Madero M, Jalal D, Sanchez-Lozada L, Johnson R. **Uric acid and the origins of hypertension**. (2013.0) **162** 896-902. DOI: 10.1016/j.jpeds.2012.12.078
57. Lanaspa M, Andres-Hernando A, Kuwabara M. **Uric acid and hypertension**. (2020.0) **43** 832-834. DOI: 10.1038/s41440-020-0481-6
58. Sanchez-Lozada L, Lanaspa M, Cristobal-Garcia M, Garcia-Arroyo F, Soto V, Cruz-Robles D. **Uric acid-induced endothelial dysfunction is associated with mitochondrial alterations and decreased intracellular ATP concentrations**. (2012.0) **121** e71-78. DOI: 10.1159/000345509
59. Wilson C, Lee MD, Heathcote HR, Zhang X, Buckley C, Girkin JM. **Mitochondrial ATP production provides long-range control of endothelial inositol trisphosphate-evoked calcium signaling**. (2019.0) **294** 737-758. DOI: 10.1074/jbc.RA118.005913
|
---
title: Community-based referral for tuberculosis preventive therapy is effective for
treatment completion
authors:
- Sheela V. Shenoi
- Tassos C. Kyriakides
- Emily Kainne Dokubo
- Vijayanand Guddera
- Peter Vranken
- Mitesh Desai
- Gerald Friedland
- Anthony P. Moll
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021376
doi: 10.1371/journal.pgph.0001269
license: CC0 1.0
---
# Community-based referral for tuberculosis preventive therapy is effective for treatment completion
## Abstract
Expansion of tuberculous preventive therapy (TPT) is essential to curb TB incidence and mortality among people with HIV (PWH), yet implementation has been slow. Innovative strategies to operationalize TPT are urgently needed. Here we present an evaluation of community-based identification and referral of PWH on completion of a six-month course of isoniazid in a highly prevalent region in rural South Africa. Using a community-based TB/HIV intensive case finding strategy, a team of nurses and lay workers identified community members with HIV who were without fever, night sweats, weight loss, or cough and referred them to the government primary care clinics for daily oral isoniazid, the only available TPT regimen. We measured monthly adherence and six-month treatment completion in the community-based identification and referral (CBR) group compared to those already engaged in HIV care. Adherence was measured by self-report and urine isoniazid metabolite testing. A multivariable analysis was performed to identify independent predictors of TPT completion. Among 240 participants, $81.7\%$ were female, median age 35 years (IQR 30–44), and $24.6\%$ had previously been treated for TB. The median CD4 count in the CBR group was 457 (IQR 301–648), significantly higher than the clinic-based comparison group median CD4 of 344 (IQR 186–495, $p \leq 0.001$). Independent predictors of treatment completion included being a woman (aOR 2.41, $95\%$ 1.02–5.72) and community-based identification and referral for TPT (aOR 2.495, $95\%$ 1.13–5.53). Among the CBR group, treatment completion was $90.0\%$, an absolute $10.8\%$ higher than the clinic-based comparison group ($79.2\%$, $$p \leq 0.02$$). Adherence was significantly greater in the CBR group than the clinic-based comparison group, as measured by self-report ($$p \leq 0.02$$) and urine isoniazid testing ($$p \leq 0.01$$). Among those not on ART at baseline, $10\%$ of eligible PWH subsequently initiated ART. Community members living with HIV in TB endemic regions identified and referred for TPT demonstrated higher treatment completion and adherence compared to PWH engaged for TPT while receiving clinic-based care. Community-based identification and referral is an innovative adjunctive strategy to facilitate implementation of TB preventive therapy in people living with HIV.
## Introduction
Tuberculosis (TB) is the leading cause of infectious death worldwide and among people living with HIV (PWH) [1]. South Africa has the highest global prevalence of TB/HIV coinfection, particularly KwaZulu-Natal province, with 65–$85\%$ of patients with newly diagnosed TB also HIV coinfected [2, 3]. World Health Organization-endorsed TB preventive therapy (TPT) reduces TB incidence by 60–$70\%$ and significantly reduces mortality among PWH, independent of antiretroviral therapy (ART) [4–7]. South Africa incorporated TPT into national guidelines in 2010, accounting for the majority of global TPT uptake [8, 9]. Widespread implementation lags in resource-limited settings, requiring innovative strategies to operationalize TPT to reduce morbidity, mortality, and the global TB reservoir [10, 11].
TPT implementation faces multiple patient-level and systemic barriers along the cascade of care; initiation, adherence, and completion rates are suboptimal [11–17]. Community-based strategies for engaging individuals in HIV and TB services have proven successful and valuable [18–29], while modeling and cost effectiveness studies have demonstrated favorable impact on HIV and TB incidence and mortality [30–32]. Transmission in community settings contributes substantially to TB propagation, suggesting that interventions such as active TB case finding and TPT eligibility determined beyond the historic foci of household contacts and health care facilities are important to epidemic control [33–36]. We sought to determine the benefit of an innovative community-based approach for identification and referral (CBR) on TPT completion and adherence, compared to traditional HIV clinical care in resource-limited high TB/HIV burden areas.
## Setting
This study was conducted in the rural sub-district of Msinga in KwaZulu-Natal province, an area covering 2000 square kilometers, and a population of 180,000 traditional Zulu people. Msinga is among the most impoverished regions in the country [37, 38]. Residents live in isolated family compounds of traditional Zulu huts, often without ventilation, low levels of electricity ($61\%$) and clean water ($69\%$), and with high levels of unemployment ($85\%$) [37]. Adult HIV prevalence ($27\%$) and drug-susceptible TB ($\frac{1100}{100}$,000) incidence are extremely high, the latter with $70\%$ HIV coinfection [39]. Transportation is challenging due to the rugged terrain and unpaved roads requiring vehicles with high ground clearance. The provincial district hospital, 16 primary care clinics and three mobile clinics serve the region. TB specimens may be collected at primary care clinics, but all diagnostics, including GeneXpert MTB/RIF® (Cepheid, Sunnyvale, CA, USA) and chest radiography, are only available at the district hospital [18].
## Procedures
From 2013–2016, a team of health educators, nurses, and HIV counselors attended community-based congregate settings, such as bus stations, municipality events, and pension pay points (social grant distribution sites), to provide integrated TB/HIV screening services [18]. The team provided health education on a variety of topics, and offered HIV and TB screening services consisting of rapid fingerstick HIV testing with confirmatory rapid testing if positive following national guidelines [40], WHO-endorsed four symptom screen, and if symptoms reported, sputum collection for GeneXpert testing [18, 41, 42]. Phlebotomy for CD4 cell count was offered, testing was performed on whole blood on by National Health Laboratory Services at the government district hospital, and results were used to determine eligibility for ART according to South African guidelines, which changed during the course of the study [43]. Guidelines expanded access to ART from those with CD4<350 cells/mm3 to those with CD4<500 cells/mm3 in January 2015 [40]. TPT consisted of oral isoniazid daily for six months. Tuberculin skin testing was recommended to determine TPT duration but had not been implemented [44]. PWH ≥18 years with negative TB symptom screen [40, 42] were informed about TPT including the purpose, dosing, duration, adverse effects, availability at local facilities, and referred to their local primary care clinic.
PWH who were referred and subsequently linked to care, defined as a clinic visit for TPT, at one of five study clinics in Msinga were offered written informed consent. For every enrollee in this CBR group, a PWH ≥18 years initiating TPT in routine HIV care during the same week and clinic, was offered written informed consent to enroll into a clinic-based comparison (CBC) group. After providing consent, all participants received TPT education and medications were subsequently prescribed and dispensed by Department of Health personnel to be taken daily for six months, without routine pyridoxine supplementation or hepatic function monitoring, and integrated with HIV services [44]. Enrollment began July 2013 and all follow up visits and data collection were completed in August 2016. Tuberculin skin testing (TST) was recommended in the South African national guidelines at that time to determine duration of TPT but had not been implemented by the facilities at the time of this study [44]. Study visits were scheduled concurrently with each monthly visit for medication refills. Sociodemographic data, as well as history of prior tuberculosis treatment, cost of transport to the clinic, and alcohol use were elicited. Alcohol consumption was assessed using the AUDIT tool; harmful levels of drinking were defined as a score ≥8 in men and ≥6 in women [45, 46]. Adherence was assessed monthly using two measures—a validated 7-day recall tool (i.e. none, few, half, most, all pills in the last week) [47, 48] and a point of care urine test for isoniazid metabolites (IsoScreen GFC Diagnostics, Oxfordshire, UK), based on the Arkansas colorimetric assay [49–51]. Urine metabolite testing was scored as ‘1’–yellow: no INH, ‘2’–green: INH present, or ‘3’–blue-purple: INH ingestion within the last 24–48 hours. All participants in the CBR and clinic-based comparison group received TPT adherence counseling by the same staff at each visit.
The primary outcome was six month TPT completion [44], defined as pickup of 6 months of medications. TPT completion was compared in the community-based referred (CBR) group and the clinic-initiated comparison group using a non-inferiority design, where the study assesses whether the CBR group is not less successful than the clinic-initiated group. The sample size was calculated based on TPT treatment completion, with a $15\%$ non-inferiority margin, anticipated $85\%$ retention in care, and two-sided alpha of 0.05. With a minimum sample size of 120 participants per group, the study had $95\%$ power to detect non-inferiority.
## Analysis
Descriptive statistics were used to characterize the population and bivariate analyses were performed to identify factors associated with completing TPT. Comparisons between the two groups were made using parametric (t-test) or non-parametric equivalent (for continuous variables) or by chi-square (categorical variables). Variables that were significant at the $p \leq 0$·2 level in univariate logistic regression models predicting the primary outcome (TPT completion) were entered into a stepwise multivariable logistic regression analysis. Unadjusted and adjusted odds ratios and $95\%$ confidence intervals were calculated. Self-reported adherence was defined as a participant reporting taking their pills. The five categories of self-reported adherence were assigned a score of 0 (’No/Few/Half TPT pills per day’) or 1 (’Most/All TPT pills per day’). Using urine metabolite testing, adherence was assessed as strong if the urine result was blue-purple, indicating isoniazid intake in the last 24-48hours. Strict adherence was defined as blue-purple urine at all of the six follow-up visits. Self-report was also correlated with urine testing scores. For all analyses, a p-value <0.05 was considered statistically significant. All statistical analyses were performed using SAS 9·3 (SAS Institute, Cary NC).
Ethical approval was obtained from the institutional review boards at the South African Medical Association and Yale University School of Medicine. The protocol was also reviewed in accordance with the U.S. Centers for Disease Control and Prevention (CDC) human research protection procedures and was determined to be research, but CDC investigators did not interact with human subjects or have access to identifiable data or specimens for research purposes.
## Results
During 334 community-based congregate site visits, 563 individuals were identified as HIV positive. Of those, 411 ($73\%$) had a negative TB symptom screen, deemed eligible, and referred for TPT care (Fig 1). Among PWH referred for TPT, 285 ($69\%$) sought care at one of five study clinics; 120 were enrolled and 120 patients already engaged in HIV care and initiating TPT at the same clinic during the same week were enrolled in the comparison group. Participants were predominantly ($81.7\%$) female (Table 1), with a median age of 35 years (Interquartile Range, (IQR) 30–44). Among 209 ($87\%$) with available CD4 values at baseline, the median CD4 count was 417 cells/mm3(IQR 238–588). Women had a median CD4 count of 439 (IQR 255–619), while men had a median CD4 count of 393 (IQR 205–514), with no significant difference in CD4 based on gender ($$p \leq 0.47$$). Among the CBR group, the median CD4 count was 457 (IQR 301–648), significantly higher than the clinic-initiated comparison group whose median CD4 cell count at baseline was 344 (IQR 186–495, $p \leq 0.001$*). Approximately one quarter of participants in the CBR group and comparison group had been previously diagnosed with TB ($$p \leq 0.65$$), and the proportion with harmful drinking was higher in those identified in the community ($8.3\%$) as compared to persons identified at the clinic ($3.2\%$, $$p \leq 0.09$$).
**Fig 1:** *Flow chart depicting community-based TB/HIV screening to enrollment.* TABLE_PLACEHOLDER:Table 1 The majority of clinic TPT initiators ($70.8\%$) were already on ART, and had been on treatment for a median of 442 days (IQR53-1223) days. Significantly fewer CBR participants ($50\%$) were taking ART at baseline ($$p \leq 0.001$$), but were on treatment for a median 1398 days (IQR 970–2350).
A comparison of the primary outcome of TPT completion was performed between the two groups (Table 2). Among the CBR group, treatment completion was $90.0\%$, an absolute $10.8\%$ higher rate than in the clinic-initiated comparison group, where completion was $79.2\%$ ($$p \leq 0.03$$). Overall, 37 ($15.4\%$) participants did not complete TPT, disproportionately higher among clinic initiators ($$n = 25$$/120, $20.8\%$) compared to the CBR group ($$n = 12$$/120, $10\%$), who exhibited significantly ($$p \leq 0.03$$) less loss-to follow up. Among those who did not complete the TPT course ($$n = 37$$), 3 ($8.1\%$) were due to side effects and 2 ($5.4\%$) participants died during the course of the study; both were in the CBR arm and their causes of death (stroke, gunshot wounds) were not associated with study participation.
**Table 2**
| Unnamed: 0 | Comparison Group (n = 120) | CBR (n = 120) | p-value |
| --- | --- | --- | --- |
| TPT completion | 95 (79.2%) | 108 (90%) | 0.02& |
| Non-completion | 25 (20.8%) | 12 (10%) | |
| Proportion with strict adherence across 6 visits | Proportion with strict adherence across 6 visits | Proportion with strict adherence across 6 visits | Proportion with strict adherence across 6 visits |
| Self-report | 65 (54.2%) | 84 (70%) | 0.02& |
| Urine Isoniazid | 49 (40.8%) | 68 (56.7%) | 0.01& |
Study participants reported taking all or most of their pills during $97.9\%$ (1228 of 1254) patient visits. Of the CBR group, $70\%$ self-reported strict adherence, as compared to $54.2\%$ of the comparison group ($$p \leq 0.02$$). Urine metabolite testing was performed in $97\%$ (1216 of 1254) study visits, and of these, $95.8\%$ (1165 of 1216) were blue-purple, demonstrating strong adherence to TPT. Strict adherence over the TPT course in the CBR group was demonstrated in $56.7\%$ by urine testing, significantly greater than the $40.8\%$ in the comparison group ($$p \leq 0.01$$). Self-reported adherence was positively correlated to urine testing (correlation coefficient of 0.88, $p \leq 0.0001$).
Among 60 CBR participants not receiving ART at baseline, 28 were eligible or became eligible during the course of the study, and among these, 12 ($42.9\%$) participants initiated ART. Among 35 participants in the comparison group not on ART at enrollment, 14 were eligible or became eligible and 8 ($57.1\%$) initiated ART prior to the end of the study. There was no significant difference in ART initiation between these groups.
Bivariate analyses demonstrated that gender, CD4 count ≤ 350 cells/mm3, and community-based referral were associated with TPT completion, whereas concurrent ART, either at baseline ($$p \leq 0.90$$) or initiated over the course of the study ($$p \leq 0.34$$), was not associated with TPT completion. Multivariable stepwise regression analysis identified female gender and community-based referral as independent predictors of TPT completion while CD4 count ≤ 350 cells/mm3 dropped out of the model (Table 3).
**Table 3**
| Unnamed: 0 | Unadjusted OR (95% CI) | Adjusted OR (95% CI) | p-value |
| --- | --- | --- | --- |
| Age | 1.04 (0.97–1.04) | | |
| Female Gender | 1.84 (0.82–4.15) | 2.41(1.02–5.72) | 0.042 |
| Transport Cost to Clinic | 1.01(0.97–1.05) | | |
| Elevated AUDIT Score | 0.99(0.89–1.10) | | |
| Household TB Contact | 0.53(0.06–4.99) | | |
| Prior TB | 0.86(0.39–1.90) | | |
| Baseline CD4 cells/mm3 | 1.0(0.99–1.0) | | |
| Baseline CD4 count ≤200 cells/mm3 | 0.65(0.28–1.51) | | |
| Baseline CD4 count ≤350 cells/mm3 | 0.54(0.26–1.13) | | |
| Baseline CD4 count ≤500 cells/mm3 | 0.82(0.37–1.83) | | |
| On ART at baseline | 1.05(0.51–2.14) | | |
| On ART at end of study | 1.44(0.68–3.05) | | |
| Community based referral for TPT | 2.37(1.13–4.97) | 2.495(1.13–5.53) | 0.036 |
## Discussion
We compared two strategies for initiating TPT in a high HIV and TB prevalence community in rural South Africa. We demonstrate that a community-based approach to identifying and referring TPT-eligible PWH resulted in a $10\%$ absolute higher completion rate and significantly better adherence compared to those PWH who were already engaged in care. This observed effect was independent of CD4 count on entry into the study. This is among the first studies to successfully leverage a community-based approach for implementing TPT for PWH. The difference in TPT treatment adherence and completion between groups may be due to a number of factors. PWH identified in the community and referred for TPT, often not yet taking ART, may have different perspectives and motivations on their health status and how best to improve their health compared to PWH in the comparison group who were already engaged in HIV care and were largely already taking ART. A higher proportion of PWH engaged in clinic-based care were taking ART and may not prioritize TPT as highly as ART or were impacted by pill burden, resulting in selective adherence to ART over TPT [13, 52]. Thus, PWH in the community may have been particularly motivated to obtain TPT, particularly if not yet eligible for ART. While health care workers may not support or be knowledgeable about the benefit of TPT, this would not differentially affect the study groups [15]. Increased awareness of the effectiveness of TPT in preventing TB incidence and reducing mortality and recommendations for TPT provision to PWH is needed to facilitate widespread implementation. We speculate that community-based engagement results in a potentially different relationship with health workers and subsequent perspective on health care, possibly leading to different motivations to engage in care [53, 54]. Furthermore, individuals identified in the community engaged specifically about TPT may have received more education about TPT than patients in the clinic. However, these are speculative; the study was not designed to answer this question and requires further rigorous evaluation.
Community-based strategies have been successful in identifying individuals with previously unknown or known communicable and non-communicable diseases, and linking them to care [18, 22, 25, 26, 55, 56]. HIV services, particularly in resource-limited settings by NGO partners as well as government programs, have long incorporated community approaches while rigorous evaluations have demonstrated success in improving diagnosis, linkage to care, and initiation of ART [18, 25, 27, 57]. Prevention services to date have primarily emphasized implementing universal HIV test and treat, with a focus on treatment as prevention, as well as medical male circumcision [58–62]. Modeling strategies have demonstrated the benefit of such approaches on HIV incidence and mortality and have demonstrated cost effectiveness [30–32]. Such alternative or differentiated care models need to be employed to improve implementation of preventive strategies, including TPT [63, 64]. Innovative studies are now underway evaluating the benefit of community-based approaches for TPT [62, 65].
Gender had a large impact on outcomes in this study. Women were more successful in completing the TPT course, compared to men, independent of CD4 count. This is not unexpected given that women are more likely to undergo testing, link to care, initiate ART, and overall experience better TB and HIV outcomes than men, particularly in resource limited settings [66–69]. This study contributes to the sparse literature on gender disparities in TB preventive therapy completion [70, 71], suggesting that like HIV retention and TB treatment completion, the TB prevention cascade may differ by gender. Similar to antiretroviral therapy [72], men lag in initiating TB preventive therapy as well. Prevention efforts may be enhanced with gender-specific interventions, such as community case finding targeting men, or adherence clubs targeting men receiving TPT, incorporating gender and cultural norms.
Community-based approaches may also likely support medication adherence. In contrast to other studies, adherence was excellent among all enrollees, potentially attributable to the education and counseling about TPT in both arms. National guidelines [44] do not specify counseling or education, and our overall high rates of completion even among clinic patients may be attributable to increased understanding about the importance of this preventive strategy [13]. The participants in the community-based referral group were significantly more likely to report higher adherence and significantly more likely to have higher isoniazid metabolites in urine. Both groups received the same literacy sessions from the clinic staff and regular counseling at each study visit from study staff. The reasons for better adherence in one group could be related to concurrent ART, however, in our study, concurrent ART use was not an independent predictor of TPT adherence. Though ART use at enrollment was significantly lower in the CBR group, newly prescribed TPT in addition to more recent ART initiation in the comparison group may have contributed to difficulties with TPT adherence.
Among those not receiving ART at enrollment, nearly $10\%$ subsequently initiated ART during the study follow up period. In this instance, engaging community members for preventive therapy also contributed to ART initiation. While the proportion subsequently initiating ART was not statistically significant between both groups, the proportion of those referred from the community for TPT and then subsequently initiating ART was substantial. With the removal of CD4 criteria for ART eligibility and implementation of differentiated care strategies such as community-based ART provision, ART initiation would likely be higher. Another notable finding was the significantly higher CD4 count at enrollment among the CBR group, demonstrating entry into HIV care at an earlier stage of disease accompanied by the potential benefit of longer disease-free survival. This type of community-based ‘hook’ into the health care system can be a valuable tool to engaging individuals into clinical care. In the era of ‘test and treat’, the emphasis remains on rapid ART initiation, but TPT and ART are necessary companions in HIV care, and patients will benefit from bundled care [5, 73]. Laudably, South African guidelines have recently integrated TPT with ART initiation. CBR may facilitate linkage and initiation of either or both TPT and ART, especially if coupled with community-based provision. Though new regimens are becoming available which will facilitate TB preventive therapy adherence and completion, the first step remains initiation. Novel strategies that can effectively engage individuals for integrated HIV and TB care are essential.
There are a number of limitations in this study. Community members referred for TPT were not traced to other than the five study sites, limiting ascertainment of TPT outcomes for referred community members who may have gone to a different facility. Study participants may also represent those particularly motivated to obtain TPT. Secondly, some PWH who were lost to follow-up due to moving out of the area may have continued TPT after relocating; it is possible that if all completed their TPT course, the completion rates would have been similar. Next, urine metabolite testing is limited by detection only over the prior 48–72 hours, and enrollees expecting to be tested may have altered their behavior just prior to an upcoming appointment [49, 50]. Regardless, such a point of care tool is useful to measure adherence, provides rapid patient feedback, and may facilitate adherence monitoring in routine care [74]. Next, the South African government has made progress on implementing TPT, and the clinic environment and knowledge among clinic staff may be different currently than during the study period. Lastly, despite the availability of new shorter regimens, isoniazid remains the current TPT regimen in South Africa and much of the world [1] and the results of this study remains relevant to current global practice. Community-based approaches are valuable regardless of TPT regimen used and offer evidence for improving PWH engagement and strengthening TPT and ART implementation in the current global context.
## Conclusion
TB preventive therapy is an effective strategy to reduce TB incidence in PWH and a WHO-recommended approach to reduce the global burden of TB. Despite being recommended for more than a decade, implementation has not reached optimal levels. Novel strategies such as community-based case finding and referral, designed here to engage individuals for TPT, can be implemented more broadly to engage individuals in care. With updated WHO recommendations on who should receive TPT, including HIV negative adult household contacts of TB patients and other immunosuppressed patients, community-based strategies can offer an innovative complementary approach to traditional TB contact tracing by identifying additional at risk individuals with and without HIV to expand implementation [6, 27, 75]. Community-based identification and referral is an innovative adjunctive strategy for TB prevention to facilitate TB epidemic control.
## References
1. 1World Health Organization. Global Tuberculosis Report. Geneva: World Health Organization; 2020.. *Global Tuberculosis Report* (2020.0)
2. 2World Health Organization. WHO Global Tuberculosis Report. 2018.. *WHO Global Tuberculosis Report* (2018.0)
3. Jacobson K, Moll A, Friedland G, Shenoi S. **Successful Tuberculosis Treatment Outcomes among HIV/TB Coinfected Patients Down-Referred from a District Hospital to Primary Health Clinics in Rural South Africa**. *PLoS One* (2015.0) **10**. DOI: 10.1371/journal.pone.0127024
4. Golub JE, Pronyk P, Mohapi L, Thsabangu N, Moshabela M, Struthers H. **Isoniazid preventive therapy, HAART and tuberculosis risk in HIV-infected adults in South Africa: a prospective cohort**. *AIDS* (2011.0) **23** 631-6
5. Badje A, Moh R, Gabillard D, Guehi C, Kabran M, Ntakpe JB. **Effect of isoniazid preventive therapy on risk of death in west African, HIV-infected adults with high CD4 cell counts: long-term follow-up of the Temprano ANRS 12136 trial**. *The Lancet Global health* (2017.0) **5** e1080-e9. DOI: 10.1016/S2214-109X(17)30372-8
6. 6World Health Organization. Latent TB Infection: Updated and consolidated guidelines for programmatic management. Geneva: WHO; 2018.. *Latent TB Infection: Updated and consolidated guidelines for programmatic management* (2018.0)
7. **A Trial of Early Antiretrovirals and Isoniazid Preventive Therapy in Africa**. *New England Journal of Medicine* (2015.0) **373** 808-22. DOI: 10.1056/NEJMoa1507198
8. 8Republic of South Africa Department of Health. Guidelines for Tuberculosis Preventive Therapy Among HIV Infected Individuals in South Africa. 2010.. *Guidelines for Tuberculosis Preventive Therapy Among HIV Infected Individuals in South Africa* (2010.0)
9. 9World Health Organization. Global Tuberculosis Report
2017. Available from: http://www.who.int/tb/publications/global_report/en/.. *Global Tuberculosis Report* (2017.0)
10. Alsdurf H, Hill PC, Matteelli A, Getahun H, Menzies D. **The cascade of care in diagnosis and treatment of latent tuberculosis infection: a systematic review and meta-analysis**. *The Lancet Infectious Diseases* (2016.0) **16** 1269-78. DOI: 10.1016/S1473-3099(16)30216-X
11. Rangaka MX, Cavalcante SC, Marais BJ, Thim S, Martinson NA, Swaminathan S. **Controlling the seedbeds of tuberculosis: diagnosis and treatment of tuberculosis infection**. *Lancet* (2015.0) **386** 2344-53. DOI: 10.1016/S0140-6736(15)00323-2
12. Getahun H, Granich R, Sculier D, Gunneberg C, Blanc L, Nunn P. **Implementation of isoniazid preventive therapy for people living with HIV worldwide: barriers and solutions**. *AIDS* (2010.0) S57-65. DOI: 10.1097/01.aids.0000391023.03037.1f
13. Jacobson KB, Niccolai L, Mtungwa N, Moll AP, Shenoi SV. **"It’s about my life": facilitators of and barriers to isoniazid preventive therapy completion among people living with HIV in rural South Africa**. *AIDS Care* (2017.0) **29** 936-42. DOI: 10.1080/09540121.2017.1283390
14. Kufa T, Chihota VN, Charalambous S, Churchyard GJ. **Isoniazid preventive therapy use among patients on antiretroviral therapy: a missed opportunity**. *International Journal of Tuberculosis and Lung Disease* (2014.0) **18** 312-4(3). DOI: 10.5588/ijtld.13.0505
15. Lester R, Hamilton R, Charalambous S, Dwadwa T, Chandler C, Churchyard GJ. **Barriers to implementation of isoniazid preventive therapy in HIV clinics: a qualitative study**. *AIDS* (2010.0) S45-8. DOI: 10.1097/01.aids.0000391021.18284.12
16. Maharaj B, Gengiah TN, Yende-Zuma N, Gengiah S, Naidoo A, Naidoo K. **Implementing isoniazid preventive therapy in a tuberculosis treatment-experienced cohort on ART**. *Int J Tuberc Lung Dis* (2017.0) **21** 537-43. DOI: 10.5588/ijtld.16.0775
17. Sutton BS, Arias MS, Chheng P, Eang MT, Kimerling ME. **The cost of intensified case finding and isoniazid preventive therapy for HIV-infected patients in Battambang, Cambodia**. *Int J Tuberc Lung Dis* (2009.0) **13** 713-8. PMID: 19460246
18. Shenoi SV, Moll AP, Brooks RP, Kyriakides T, Andrews L, Kompala T. **Integrated Tuberculosis/Human Immunodeficiency Virus Community-Based Case Finding in Rural South Africa: Implications for Tuberculosis Control Efforts**. *Open Forum Infect Dis* (2017.0) **4** ofx092. DOI: 10.1093/ofid/ofx092
19. Ayles H, Muyoyeta M, Du Toit E, Schaap A, Floyd S, Simwinga M. **Effect of household and community interventions on the burden of tuberculosis in southern Africa: the ZAMSTAR community-randomised trial**. *The Lancet* (2013.0) **382** 1183-94. DOI: 10.1016/S0140-6736(13)61131-9
20. Corbett E, Bandason T, Duong T, Dauya E, Makamure B, Churchyard G. **Comparison of two active case-finding strategies for community-based diagnosis of symptomatic smear-positive tuberculosis and control of infectious tuberculosis in Harare, Zimbabwe (DETECTB): a cluster-randomised trial**. *The Lancet* (2010.0) **376** 1244-53. DOI: 10.1016/S0140-6736(10)61425-0
21. Churchyard GJ, Fielding KL, Lewis JJ, Coetzee L, Corbett EL, Godfrey-Faussett P. **A trial of mass isoniazid preventive therapy for tuberculosis control**. *N Engl J Med* (2014.0) **370** 301-10. DOI: 10.1056/NEJMoa1214289
22. Chamie G, Kwarisiima D, Clark TD, Kabami J, Jain V, Geng E. **Uptake of Community-Based HIV Testing during a Multi-Disease Health Campaign in Rural Uganda**. *PLoS One* (2014.0) **9** e84317. DOI: 10.1371/journal.pone.0084317
23. Chamie G, Kwarisiima D, Clark TD, Kabami J, Jain V, Geng E. **Leveraging rapid community-based HIV testing campaigns for non-communicable diseases in rural Uganda**. *PLoS One* (2012.0) **7** e43400. DOI: 10.1371/journal.pone.0043400
24. Uwimana J, Zarowsky C, Hausler H, Swanevelder S, Tabana H, Jackson D. **Community-based intervention to enhance provision of integrated TB-HIV and PMTCT services in South Africa**. *Int J Tuberc Lung Dis* (2013.0) **17** 48-55. DOI: 10.5588/ijtld.13.0173
25. Barnabas RV, van Rooyen H, Tumwesigye E, Brantley J, Baeten JM, van Heerden A. **Uptake of antiretroviral therapy and male circumcision after community-based HIV testing and strategies for linkage to care versus standard clinic referral: a multisite, open-label, randomised controlled trial in South Africa and Uganda**. *The Lancet HIV* (2016.0) **3** e212-20. DOI: 10.1016/S2352-3018(16)00020-5
26. Shapiro AE, van Heerden A, Schaafsma TT, Hughes JP, Baeten JM, van Rooyen H. **Completion of the tuberculosis care cascade in a community-based HIV linkage-to-care study in South Africa and Uganda**. *Journal of the International AIDS Society* (2018.0) **21**
27. Sharma M, Ying R, Tarr G, Barnabas R. **Systematic review and meta-analysis of community and facility-based HIV testing to address linkage to care gaps in sub-Saharan Africa**. *Nature* (2015.0) **528** S77-85. DOI: 10.1038/nature16044
28. Upadhya D, Moll AP, Brooks RP, Friedland G, Shenoi SV. **What motivates use of community-based human immunodeficiency virus testing in rural South Africa?**. *Int J STD AIDS* (2016.0) **27** 662-71. DOI: 10.1177/0956462415592789
29. Calligaro GL, Zijenah LS, Peter JG, Theron G, Buser V, McNerney R. **Effect of new tuberculosis diagnostic technologies on community-based intensified case finding: a multicentre randomised controlled trial**. *The Lancet Infectious Diseases* (2017.0) **17** 441-50. DOI: 10.1016/S1473-3099(16)30384-X
30. Gilbert JA, Long EF, Brooks RP, Friedland GH, Moll AP, Townsend JP. **Integrating Community-Based Interventions to Reverse the Convergent TB/HIV Epidemics in Rural South Africa**. *PLoS ONE* (2015.0) **10** e0126267. DOI: 10.1371/journal.pone.0126267
31. Gilbert JA, Shenoi SV, Moll AP, Friedland GH, Paltiel AD, Galvani AP. **Cost-Effectiveness of Community-Based TB/HIV Screening and Linkage to Care in Rural South Africa**. *PLoS ONE* (2016.0) **11** e0165614. DOI: 10.1371/journal.pone.0165614
32. Smith JA, Sharma M, Levin C, Baeten JM, van Rooyen H, Celum C. **Cost-effectiveness of community-based strategies to strengthen the continuum of HIV care in rural South Africa: a health economic modelling analysis**. *The Lancet HIV* (2015.0) **2** e159-68. DOI: 10.1016/S2352-3018(15)00016-8
33. Shah NS, Auld SC, Brust JC, Mathema B, Ismail N, Moodley P. **Transmission of Extensively Drug-Resistant Tuberculosis in South Africa**. *N Engl J Med* (2017.0) **376** 243-53. DOI: 10.1056/NEJMoa1604544
34. Dowdy DW, Azman AS, Kendall EA, Mathema B. **Transforming the fight against tuberculosis: targeting catalysts of transmission**. *Clin Infect Dis* (2014.0) **59** 1123-9. DOI: 10.1093/cid/ciu506
35. Churchyard G, Kim P, Shah NS, Rustomjee R, Gandhi N, Mathema B. **What We Know About Tuberculosis Transmission: An Overview**. *The Journal of Infectious Diseases* (2017.0) **216** S629-S35. DOI: 10.1093/infdis/jix362
36. 36World Health Organization. WHO Consolidated Guidelines on TB. Module 2: Screening—Systematic screening for tuberculosis disease. Geneva: WHO; 2021.. *WHO Consolidated Guidelines on TB. Module 2: Screening—Systematic screening for tuberculosis disease* (2021.0)
37. **Umzinyathi District Profile**. *KwaZulu Natal* (2012.0)
38. 38District Health Barometer 2010/2011 [Internet]. 2011. Available from: http://www.hst.org.za/sites/default/files/DHB_Datafile_19Dec2011.xlsx.
39. Gandhi NR, Moll A, Sturm AW, Pawinski R, Govender T, Lalloo U. **Extensively drug-resistant tuberculosis as a cause of death in patients co-infected with tuberculosis and HIV in a rural area of South Africa**. *The Lancet* (2006.0) **368** 1575-80
40. **National consolidated guidelines for the prevention of mother-to-child transmission of HIV (PMTCT) and the management of HIV in children, adolescents and adults**. *Pretoria* (2014.0)
41. 41Republic of South Africa Department of Health. National Tuberculosis Management Guidelines. 2014.. *National Tuberculosis Management Guidelines* (2014.0)
42. **Guidelines for intensified tuberculosis case-finding and isoniazid preventive therapy for people living with HIV in resource constrained settings**. *Geneva* (2010.0)
43. **South African Antiretroviral Treatment Guidelines**. (2010.0)
44. **The South African Antiretroviral Treatment Guidelines. Pretoria**. (2013.0)
45. 45World Health Organization. The Alcohol Use Disorders Identification Test, Guidelines for Use in Primary Care. 2nd edition ed. Geneva: World Health Organization; 2001.. *The Alcohol Use Disorders Identification Test, Guidelines for Use in Primary Care* (2001.0)
46. Morojele NK, Nkosi S, Kekwaletswe CT, Shuper PA, Manda SO, Myers B. **Utility of Brief Versions of the Alcohol Use Disorders Identification Test (AUDIT) to Identify Excessive Drinking Among Patients in HIV Care in South Africa**. *Journal of Studies on Alcohol and Drugs* (2017.0) **78** 88-96. DOI: 10.15288/jsad.2017.78.88
47. Mannheimer S, Friedland G, Matts J, Child C, Chesney M. **The consistency of adherence to antiretroviral therapy predicts biologic outcomes for human immunodeficiency virus-infected persons in clinical trials**. *Clinical Infectious Diseases* (2002.0) **34** 1115-21. DOI: 10.1086/339074
48. Mannheimer S, Thackeray L, Hullsiek KH, Chesney M, Gardner EM, Wu AW. **A randomized comparison of two instruments for measuring self-reported antiretroviral adherence**. *AIDS Care* (2008.0) **20** 161-9. DOI: 10.1080/09540120701534699
49. Guerra R, Conde M, Efron A, Loredo C, Bastos G, Chaisson R. **Point-of-care Arkansas method for measuring adherence to treatment with isoniazid**. *Respiratory Medicine* (2010.0) **104** 754-7. DOI: 10.1016/j.rmed.2010.02.001
50. Hanifa Y, Mngadi K, Lewis J, Fielding K, Churchyard G, Grant AD. **Evaluation of the Arkansas method of urine testing for isoniazid in South Africa**. *IJTLD* (2007.0) **11** 1232-6. PMID: 17958987
51. Schraufnagel DE, Stoner R, Whiting E, Snukst-Torbeck G, Werhane MJ. **Testing for isoniazid. An evaluation of the Arkansas method**. *Chest* (1990.0) **98** 314-6. DOI: 10.1378/chest.98.2.314
52. Makanjuola T, Taddese HB, Booth A. **Factors associated with adherence to treatment with isoniazid for the prevention of tuberculosis amongst people living with HIV/AIDS: a systematic review of qualitative data**. *PLoS One* (2014.0) **9** e87166. DOI: 10.1371/journal.pone.0087166
53. Mwai GW, Mburu G, Torpey K, Frost P, Ford N, Seeley J. **Role and outcomes of community health workers in HIV care in sub-Saharan Africa: a systematic review**. *Journal of the International AIDS Society* (2013.0) **16** 18586. DOI: 10.7448/IAS.16.1.18586
54. Sinha P, Shenoi SV, Friedland GH. **Opportunities for community health workers to contribute to global efforts to end tuberculosis**. *Global public health* (2019.0) **15** 474-84. DOI: 10.1080/17441692.2019.1663361
55. Suthar AB, Klinkenberg E, Ramsay A, Garg N, Bennett R, Towle M. **Community-based multi-disease prevention campaigns for controlling human immunodeficiency virus-associated tuberculosis**. *Int J Tuberc Lung Dis* (2012.0) **16** 430-6. DOI: 10.5588/ijtld.11.0480
56. Sinha P, Moll AP, Brooks RP, Deng YH, Shenoi SV. **Synergism between diabetes and human immunodeficiency virus in increasing the risk of tuberculosis**. *Int J Tuberc Lung Dis* (2018.0) **22** 793-9. DOI: 10.5588/ijtld.17.0936
57. Desai MA, Okal DO, Rose CE, Ndivo R, Oyaro B, Otieno FO. **Effect of point-of-care CD4 cell count results on linkage to care and antiretroviral initiation during a home-based HIV testing campaign: a non-blinded, cluster-randomised trial**. *The Lancet HIV* (2017.0) **4** e393-e401. DOI: 10.1016/S2352-3018(17)30091-7
58. Sgaier SK, Reed JB, Thomas A, Njeuhmeli E. **Achieving the HIV prevention impact of voluntary medical male circumcision: lessons and challenges for managing programs**. *PLoS Med* (2014.0) **11** e1001641. DOI: 10.1371/journal.pmed.1001641
59. Simwinga M, Bond V, Makola N, Hoddinott G, Belemu S, White R. **Implementing Community Engagement for Combination Prevention: Lessons Learnt From the First Year of the HPTN 071 (PopART) Community-Randomized Study**. *Current HIV/AIDS Reports* (2016.0) **13** 194-201. DOI: 10.1007/s11904-016-0322-z
60. Grabowski MK, Serwadda DM, Gray RH, Nakigozi G, Kigozi G, Kagaayi J. **Combination HIV Prevention and HIV Incidence in Uganda**. *N Engl J Med* (2017.0) **377** 2154-66. PMID: 29171817
61. Gaolathe T, Wirth KE, Holme MP, Makhema J, Moyo S, Chakalisa U. **Botswana’s progress toward achieving the 2020 UNAIDS 90-90-90 antiretroviral therapy and virological suppression goals: a population-based survey**. *The Lancet HIV* (2016.0) **3** e221-30. DOI: 10.1016/S2352-3018(16)00037-0
62. Howard AA, Hirsch-Moverman Y, Saito S, Gadisa T, Daftary A, Melaku Z. **The ENRICH Study to evaluate the effectiveness of a combination intervention package to improve isoniazid preventive therapy initiation, adherence and completion among people living with HIV in Ethiopia: rationale and design of a mixed methods cluster randomized trial**. *Contemporary Clinical Trials Communications* (2017.0) **6** 46-54. DOI: 10.1016/j.conctc.2017.03.001
63. Pathmanathan I, Ahmedov S, Pevzner E, Anyalechi G, Modi S, Kirking H. **TB preventive therapy for people living with HIV: key considerations for scale-up in resource-limited settings**. *Int J Tuberc Lung Dis* (2018.0) **22** 596-605. DOI: 10.5588/ijtld.17.0758
64. Pathmanathan I, Pevzner E, Cavanaugh J, Nelson L. **Addressing tuberculosis in differentiated care provision for people living with HIV**. *Bull World Health Organ* (2017.0) **95** 3. DOI: 10.2471/BLT.16.187021
65. Hirsch-Moverman Y, Howard AA, Frederix K, Lebelo L, Hesseling A, Nachman S. **The PREVENT study to evaluate the effectiveness and acceptability of a community-based intervention to prevent childhood tuberculosis in Lesotho: study protocol for a cluster randomized controlled trial**. *Trials* (2017.0) **18** 552. DOI: 10.1186/s13063-017-2184-0
66. Hensen B, Taoka S, Lewis JJ, Weiss HA, Hargreaves J. **Systematic review of strategies to increase men’s HIV-testing in sub-Saharan Africa**. *AIDS* (2014.0) **28** 2133-45. DOI: 10.1097/QAD.0000000000000395
67. Bor J, Rosen S, Chimbindi N, Haber N, Herbst K, Mutevedzi T. **Mass HIV Treatment and Sex Disparities in Life Expectancy: Demographic Surveillance in Rural South Africa**. *PLoS Med* (2015.0) **12** e1001905. DOI: 10.1371/journal.pmed.1001905
68. Sharma M, Barnabas RV, Celum C. **Community-based strategies to strengthen men’s engagement in the HIV care cascade in sub-Saharan Africa**. *PLoS Med* (2017.0) **14** e1002262. DOI: 10.1371/journal.pmed.1002262
69. Arnesen R, Moll AP, Shenoi SV. **Predictors of loss to follow-up among patients on ART at a rural hospital in KwaZulu-Natal, South Africa**. *PLoS One* (2017.0) **12** e0177168. DOI: 10.1371/journal.pone.0177168
70. Munseri PJ, Talbot EA, Mtei L, Fordham von Reyn C. **Completion of isoniazid preventive therapy among HIV-infected patients in Tanzania**. *Int J Tuberc Lung Dis* (2008.0) **12** 1037-41. PMID: 18713501
71. Ngamvithayapong J, Uthaivoravit W, Yanai H, Akarasewi P, Sawanpanyalert P. **Adherence to tuberculosis preventive therapy among HIV-infected persons in Chiang Rai, Thailand**. *AIDS* (1997.0) **11** 107-12. DOI: 10.1097/00002030-199701000-00016
72. Tsai AC, Siedner MJ. **The missing men: HIV treatment scale-up and life expectancy in Sub-Saharan Africa**. *Plos Med* (2015.0) **12** e1001906. DOI: 10.1371/journal.pmed.1001906
73. Hakim J, Musiime V, Szubert AJ, Mallewa J, Siika A, Agutu C. **Enhanced Prophylaxis plus Antiretroviral Therapy for Advanced HIV Infection in Africa**. *New England Journal of Medicine* (2017.0) **377** 233-45. DOI: 10.1056/NEJMoa1615822
74. Eidlitz-Markus T, Zeharia A, Baum G, Mimouni M, Amir J. **Use of the urine color test to monitor compliance with isoniazid treatment of latent tuberculosis infection**. *Chest* (2003.0) **123** 736-9. DOI: 10.1378/chest.123.3.736
75. Shapiro AE, Variava E, Rakgokong MH, Moodley N, Luke B, Salimi S. **Community-based targeted case-finding for tuberculosis and HIV in household contacts of tuberculosis patients in South Africa**. *American Journal of Respiratory and Critical Care Medicine* (2012.0) **185** 1110-6. PMID: 22427532
|
---
title: 'Factors associated with depression among heart failure patients at selected
public hospitals in Addis Ababa, Ethiopia: A cross sectional study'
authors:
- Kassahun Alemayehu
- Yohannes Ayalew Bekele
- Teshome Habte Wurjine
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021377
doi: 10.1371/journal.pgph.0000853
license: CC BY 4.0
---
# Factors associated with depression among heart failure patients at selected public hospitals in Addis Ababa, Ethiopia: A cross sectional study
## Abstract
This study aimed to Assess Factors associated with depression among heart failure patients at cardiac follow-up clinics in a government teaching hospital of Addis Ababa. A cross-sectional study design was employed to assess Factors associated with depression among 424 heart failure patients at selected public hospitals of Addis Ababa who were selected by using a systematic random sampling method from January 1 to 30, 2021 at four public hospitals. Sample was proportionally allocated for each study hospital and then data were collected by using structured-interview questionnaires. Bivariate and Multivariate logistic regression analysis was done to examine the possible predictors and variables with the statistical association of P-value of < 0.05 and a $95\%$ confidence interval were considered. Data were gathered from heart failure patients in cardiac follow clinic with $100\%$ response rate. Among the 424 respondents [mean age: 52.7 (SD) 17.5 years; $56.1\%$ women], prevalence of depression was $56.1\%$. Among the 424 respondents [mean age: 52.7 (SD) 17.5 years; $56.1\%$ women], prevalence of depression was $56.1\%$. New York Heart Association class III and IV was highly associated with depression respectively. Furthermore, poor self-care behaviours alcohol use, poor social support, poor knowledge level, were associated with depression and statistically significant. The findings from this study showed that depression is highly prevalent among heart failure patients and age of respondent, alcohol intake, self-care behaviour, social support, knowledge level, and co-morbidity were independently associated with depression. Therefore, all institutions of cardiac centre should work on screening for depression in heart failure patients, and consult psychiatrists and psychologists for early detection and measures.
## 1. Introduction
According to Global Burden of disease report, 62 million people suffered from problem related to heart failure and more than half of the patients were in severe stages. Depression is one of the most co-morbid disorders in patients with cardiovascular disease and it is one of the main public health problems across the world [1,2].
A study indicates that depression has strong association with heart failure (HF). Depressed patients have less functionality with increased physiological activity of the heart, heart failure symptoms and impaired health-related quality of life. Moreover, HF patients with depression are at risk for re-hospitalization [3]. Therefore, HF with depression, highly influence on their quality of life and they had 2-fold increased risk of death or cardiac events. mortality rates at five years for HF patients is $50\%$ [4]. Among heart failure patients with depressive symptoms there will be an increased risk of mortality 2–3 times higher than in patients without symptoms of depression [5]. Furthermore, hospitalized heart failure patients with depression are at particularly at high risk for mortality. Median survival is 1.7 years for men and 3.2 years for women, with only $25\%$ of men and $38\%$ of women surviving for the last 5 years. This mortality rate is 4–8 times greater than that of the general population with similar age [6,7]. A study done in New York Hospitalized Heart Failure (HF) patients with depression the rate from $13\%$ to $77.5\%$ and out- patients are from $13\%$ to $42\%$ and depression is five times more prevalent in HF patients compared to the whole population [8,9]. Another study conducted in west Amhara region in Ethiopia, the incidence of depression among hear failure patients is $49\%$, and Dessie were $50\%$ [1,10].
The presence of social support has been associated with lower incidence of depression and faster remission of depressive symptoms. In contrast, the lack of social support, conflict relation-ship have been linked to the presence of depression [11]. A systematic and meta-analysis conducted in Ethiopia indicates that depression is a common co-morbid illness among patients with diabetes [12]. And another meta-data analysis done in Greece, Patients classified in New York Heart Association (NYHA) in class “I to IV” of them class III and IV were more likely to be depressed than class I and II Patients [13].
According to the studies done in Nigeria and Greece, depression was associated with poor health care behaviors and additional risks factors, such as smoking, sedentary life, unhealthy dietary habits, lack of regular exercise, uncontrolled weight gain may lead to worsening of depression [14,15]. Another study done in Brazil indicates that selfcare behaviour was significantly associated with depression of the heart failure participants [16].
A cross sectional study conducted in west Amhara region in 2019 showed that poor knowledge of HF patients has strong association with depression [1,17]. But factors associated with depression in heart failure are not assessed adequately in developing countries including Ethiopia. Therefore, this study aimed to assess Factors associated with depression among heart failure patients at selected public hospitals in Addis Ababa, Ethiopia. This study finding may support for health care professionals to focuses their interventional strategies on the management of depression in heart failure patients, policy makers and responsible offices at various level of health care interventions to take appropriate measures and serve as a base line information for other researchers who are interested to conduct similar studies.
## Conceptual frame work
Conceptual framework for this study was established after reviewing and adapted from different literatures related to similar sociodemographic characteristics of the study population and identified variables as illustrated in the “Fig 1” below, the socio-demographic factors (marital status, age, gender of the participants, educational level and occupation), self- care behavior such as smoking habit and alcohol use, psycho-social factor like cognitive, perception and social support and co-morbidity like heart failure patients with diabetic, hypertension and chronic kidney disease and stage of heart failure and associated factors of depression. The direction of the relationship between outcome variable (Depression) and explanatory variables are illustrated [3,12,18].
**Fig 1:** *Conceptual framework for assessment of prevalence and associated factors of depression among heart failure patients at cardiac follow-up clinic.*
Government Hospitals with cardiac centre found in Addis Ababa are: X1, X2, X3 and X4 hospitals X1 [1667]X3 [200] n1=246n4=25n total =n1+n2 + n3 + n4 =246+123+30+25=424Lottery methodProportional allocation
## Ethical approval and consent to participate
Ethical clearance and approval were obtained from Institutional Review Board (IRB) of Addis Ababa University, College of Health Sciences and approved by Department of Nursing and Midwifery. Permission was obtained from clinical director of each study hospitals, Matron and heads of the respective ward. After explaining the purpose of the study, possible benefit of the study and time to complete the questionnaire and why the participants are chosen, oral and written informed consent was obtained from each participant before proceeding the procedure. The participants were fully explained that they have the right not to participate in the study, to stop at any time in between or not to answer any questions they were not willing to answer. Confidentiality was maintained; no unauthorized person had access to the information and names or other identifiers were not recorded. All methods and subjects provided written consent, and the study was conducted in accordance with Helsinki declarations.
## 2.1. Study setting
The study was conducted at selected public hospitals of Addis Ababa city administration, Ethiopia. Addis *Ababa is* the largest and the most populated capital city of Ethiopia. It is a metropolitan area with a population of estimated to be around 5,006,000 people in 2021. The capital city holds 527 square kilometers of area in Ethiopia. The population density is estimated to be near 5,165 individuals per square kilometer available. Based on the 2020 population enumeration annual growth rate is $4.42\%$. The city has 15 public hospitals from this four of them randomly selected to conduct this research study namely: Tikur Anbesa specialized hospital, St. Paulo’s hospital, Yekatit 12 Hospital and Armed force specialized hospital.
## 2.2. Study design and period
An institution-based cross-sectional study design was conducted from January 1 to 30, 2021. 2.3. 2.3. Study population All heart failure patients those who had follow- up at selected four government hospitals of cardiac centre in Addis Ababa during the study period.
## 2.4. Sampling procedure and technique
Those patients who were clinically confirmed as having heart failure assessed by reviewing their clinical charts with a check list that incorporates the PHQ-9 tool for depression measurement used. And patients assessed using structured-interview questionnaires and to select a total of 424 heart failure patients sampled from 2,867 heart failure patients after proportionally allocated patients from four selected public hospitals as stated below.
The total number of heart failure patients in the four selected study hospitals were 2,867. Hence, after the sample size was determined by using simple population proportion formula (n-424). The study sample was proportionally allocated to each study hospitals in line with: proportional allocation formula.
The sample size was proportionally allocated for each hospital as depicted in “Fig 2” below. Study participants were selected from proportionally allocated study subject in each hospital using a random sampling technique from eligible patients visiting the cardiac clinic during the data collection period was interviewed.
**Fig 2:** *Schematic presentation of sampling technique and sample size proportional allocation.*
## 2.5. Procedure of Data Collection and tool used
Data was collected by four trained BSc nurses using $5\%$ pretested interviewer administered questionnaire and supervised by two MSc nurses. The data collection instrument includes the following components.
PHQ-9: by using a check-list that was developed on the basis of prior similar studies, the data was collected by using the Patients Health Questionnaire (PHQ-9), the questionnaire has nine items, the total score ranges from 0 to27 a score 5,10,15,20 represent cut point for mild, moderate, moderate-severe and patient e depression respectively. In PHQ-9 tool there are four options (0 = none at all, 1 = several days, 2 = more than half of the days, and3 = nearly every day). Which have been used to screen depression symptoms from the study participants 21. and it has $88\%$ specificity and sensitivity. HF patients score between 1–4 categorized as having “no depression”. HF patients score between 5–9 was “mild depression”, HF patients score between 10–14 was “moderate depression”, between 15–19 categorized as having “moderate to severe depression” and score >20 was categorized as having “severe depression”. PHQ-9 tool *Cronbach alpha* value was.904 [21,22].
European HF Self-care behaviours scale-9 (EHFScBC-9): The EHFScBS-9 had supportive psychometric properties of validity, reliability and precision, and it’s used to measure self-care behaviours in clinical practice and research. The EHFScBS-9 has nine items of questions each item uses a 5-point Likert scale from 1 (“completely agree”) to 5 (“completely disagree”). The possible score was 9 to 45, with the level of self-care behaviour score described as the following: A score <or = 2 on each item and total score < or = 18 suggest high self-care behaviours, a score >2 on each item and total score >18 suggest low self-care behaviours) [22,23]. Alpha Cronbach was 0.806.
Oslo social support scale (OSSS-3): Evidence supports reliability and validity of the OSSS-3 as a measure of social determinants of health in the general population. The OSSS-3 consists of three items assessing the level of social support the sum score ranges from 3 to 14, with high values (>8) representing strong levels and low values (3–8) representing poor levels of social support [14]. Alpha Cronbach in this study was 0.869.
The Dutch HF knowledge scale has 15 multiple choices for each item patients can choose from the three options, with one of the options being the correct answer. When a person gives the correct answer, 1 point is given whereas the answer is wrong the person receive 0 point for that question. The possible total score for knowledge of heart failure ranges from 0 to 15 and interpreted as Study subject’s knowledge on HF was found to be a score of DHFKS above Median was used as a cut off for good knowledge and poor knowledge as the score is below Median [24].
## 2.6. Data quality assurance
Structured and pre-tested questionnaire was used and also training was given for data collectors and supervisors on the objective of the study, method, contents and also how to maintain confidentiality and privacy of the study subject was strictly conducted. Data was collected by four experienced staff nurses with BSc degree and above. Pre-test was conducted on $5\%$ of heart failure patients at Zewditu hospital before the actual data collection period and based on the finding the necessary correction made on the method and materials.
## 2.7. Data processing and analysis
The collected data was entered in to epi data processed and analysed by using SPSS version 25. Descriptive statistics was employed to describe the percentages and frequency distributions of the variables in the study. Adjusted odd ratio with $95\%$ confidence interval was estimated to measure the strength of association a P- Value of ≤ 0.05 was used for statistical significance.
## 2.8 Operational definitions
Depressive symptom-The Patients Health Questionnaire 9 (PHQ9) is a self-questionnaire designed to help diagnosis and detect mental disorders commonly encountered in the primarily clinical setting. The individual participant had interviewed depressive symptoms when he/she has at least four positive symptoms in the PHQ-9 interview. including Q1(Little interest or pleasure in doing things) and Q2 (Feeling down, depressed or hopeless) [25,26].
Major depressive disorder: A score of ≥10 on the PHQ-9 scale was considered as a major depressive disorder [27].
Poor self-care-The individual participant had poor self-care when scored with from 9 to 45 with a higher score above >18 indicating poor, and lower <18 is good self-care of European Heart Failure Self-Care Behaviour Score-9 [17].
Poor social support-The individual participant had poor social support when she/he scored between the 3–8 out of 14 [23].
Knowledge: A score of DHFKS above Median was used as a cut off for good knowledge and poor knowledge as the score is < Median [28].
Knowledge of HF patients defines as ability to recognize and interpreted HF symptoms [29].
Low knowledge- A score of DHFKS “≥10” was used as a cut off for good knowledge and poor knowledge as the score is “<10” [28].
Self–care behaviour: is defined as the activity of HF patients performs to take care of their health in terms of exercise habit, abstain from cigarette smoking and alcohol intake [17].
Psycho-social support: defines as help maintain a continuum of social support during or after a problem and prevent from long term mental disorder and make him to developed good perception [23].
Co-morbidity-The individual heart failure patients has additional chronic disease like DM, HTN, CKD [17].
Diabetes mellitus: -a disease in which the body’s ability to produce or respond to the hormone insulin is impaired, resulting in abnormal metabolism of carbohydrate and elevated level of glucose in the blood [30].
Hypertension: -abnormal high blood pressure which is systolic blood pressure >130 and diastolic blood pressure >85mmHg [25].
Chronic kidney disease: is abnormality of kidney structure or function, present for >3 month, with implications for health [30].
## 3.1 Socio-demographic characteristics
A total of 424 study participants were included with $100\%$ response rate and about 238 ($55.2\%$) of the study participants were female. The mean and standard deviation of age of the respondents were 52.68 and17.471years respectively. A high proportion of 134 ($31.6\%$) of the respondents were within the age group of 37–54 years and about 166 ($39.1\%$) the study reveals participants were married. The study found that one out of four participants did not attend formal education or illiterate and two-third of participants were employed in government and non-governmental institutions. As shown in Table 1 below. About $36.3\%$ participants are having a monthly income in the range of 601–1650 Ethiopian birr.
**Table 1**
| Socio-demographic factor | Category | Frequency | Percentage |
| --- | --- | --- | --- |
| Gender | Male | 186 | 43.9% |
| Gender | Female | 238 | 56.1% |
| Age | 18–36 | 100 | 23.6% |
| Age | 37–54 | 134 | 31.6% |
| Age | 55–72 | 114 | 26.9% |
| Age | >72 | 76 | 17.9% |
| Marital status | Married | 166 | 39.1% |
| Marital status | Divorced | 117 | 27.6% |
| Marital status | Single | 61 | 14.4% |
| Marital status | Widow | 80 | 18.9% |
| Educational level | Illiterate | 110 | 25.9% |
| Educational level | Elementary | 104 | 24.5% |
| Educational level | Secondary | 115 | 27.1% |
| Educational level | Preparatory | 40 | 9.4% |
| Educational level | University level | 55 | 23.8% |
| Profession | Governmental employ | 101 | 23.8% |
| Profession | Self-employ | 175 | 41.3% |
| | Pension | 18 | 4.2% |
| | Student | 54 | 12.7% |
| | House wife | 76 | 17.9 |
| Monthly income per month | 0–600 | 9 | 2.1% |
| Monthly income per month | 601–1650 | 154 | 36.3% |
| Monthly income per month | 1651–3200 | 67 | 15.8% |
| Monthly income per month | 3201–5250 | 69 | 16.3% |
| Monthly income per month | 5251–7300 | 84 | 19.8% |
| Monthly income per month | 7301–10899 | 29 | 6.8% |
| Monthly income per month | >10899 | 12 | 2.8% |
## 3.2 Clinical characteristics of respondents
From the total of 424 Heart failure patients about 415 ($97.9\%$) of them found to have comorbidity. According to the New York Heart Association guideline more than one third of the participants 140($33\%$) were categorized in class III, patients experienced hypertensive disorder 147($34.7\%$), diabetes mellites 109($25.7\%$), and chronic kidney disease accounts 19 ($4.5\%$).
## 3.3 Self-care behaviour, social support and knowledge of respondents
Majority of participants 257($60.6\%$) were having poor knowledge and more than half 230 ($54.2\%$) of study participants had poor social support. The mean of heart failure patients with self-care behaviour score was 19.29, with SD of 4.91. From the total study participants,235($55.4\%$) had poor self-care behavior with cigarette smoking 120 ($28.3\%$). More than half 235($54.4\%$) of them were alcohol users.
## 3.4 prevalence of depression
The study indicates that depression is common problem as shown in Fig 3 below, the incidence of depression was $59\%$ ($95\%$ CI 54.5–63.7). Being in NYHA class III and IV [(AOR:12.8(2.2–71.6),$95\%$CI, $$P \leq 0.004$$, (AOR:19.2(1.9–189.9), $95\%$ CI, $$P \leq 0.011$$)] respectively, having poor self-care behaviours[(AOR: 9.1(2.4–34.6), $95\%$CI, $$P \leq 0.001$$)], having alcohol use[(AOR: 17.7(4.14–35.65),$95\%$ CI,$$P \leq 0.001$$)], having poor social support [(AOR: 4.6(1.2–16.7),$95\%$CI,$$P \leq 0.020$$)], having poor knowledge [(AOR:5.1(1.3–20.6), $95\%$CI, $$P \leq 0.020$$), and being single or unmarried [(0.108(0.03–0.47), $95\%$CI, $$P \leq 0.001$$)] were independently associated with depression. Low educational attainment or being illiterate was also associated with depression.
**Fig 3:** *Prevalence of depression among heart failure patients attending cardiac follow-up clinic at selected government specialized teaching hospitals in Addis Ababa, Ethiopia (n = 424).*
## 3.5 Factor associated with depression among heart failure patients attending cardiac follow-up clinic
The bi-variate logistic regression analysis indicates that, age of respondent, marital status, educational status, profession, NYHA classification, self-care behaviour, high alcohol intake, social support, knowledge level of the disease prognosis and co-morbidity were associated with depression among heart failure patients as shown in Table 2 below.
**Table 2**
| Variable | Category | Depressive symptom | Depressive symptom.1 | COR | P -Value |
| --- | --- | --- | --- | --- | --- |
| Variable | Category | Nondepression | depression | COR | P -Value |
| Marital status | Single or un-married | 53(30.5%) | 205(82%) | 10.4(6.5–16.4) | <0.001 |
| Marital status | Married | 121(69.5%) | 45(18%) | 1 | |
| Age | 18–36 | 76(43.7%) | 24(9.6%) | 1 | |
| Age | 37–54 | 86(49.4%) | 48(19.2%) | 1.7(0.99–3.1) | 0.054 |
| Age | 55–72 | 8(4.6%) | 106(42.4%) | 41.9((17–98.4) | 0.001 |
| Age | >72 | 4(2.3%) | 72(28.8%) | 57(18.8–172.3) | <0.001 |
| Educational level | Illiterate | 13(7.5%) | 97(38.8%) | 4.8(1.-12.) | <0.001 |
| Educational level | Higher school | 125(71.8%) | 134(53.6%) | 0.874(0.41–1.86) | 0.728 |
| Educational level | University level | 36(20.6%) | 19(7.6%) | 1 | |
| Profession | Governmental employ | 77(44.3%) | 24(9.6%) | 7.4(4.4–12.5) | <0.001 |
| Profession | Non-governmental employ | 97(55.7%) | 226(90.4%) | 1 | |
| NYHA Class | I | 75(43.1%) | 17(6.8%) | 1 | |
| NYHA Class | II | 69(39.7%) | 46(18.4%) | .019(0.007–0.051) | <0.001 |
| NYHA Class | III | 24(13.8%) | 116(46.4%) | .056(.023-.140) | <0.001 |
| NYHA Class | IV | 6(3.4%) | 71(28.4%) | .408(.159–1.048) | 0.068 |
| Co-morbidity | Diabetes mellites | 26(14.9%) | 83(33.2%) | 5.9(3.4–10.3) | <0.001 |
| Co-morbidity | Hypertension | 38(21.8%) | 109(43.6%) | 5.3(3.2–8.8) | <0.001 |
| Co-morbidity | CKD | 13(7.5%) | 6(2.4%) | 0.81((0.3–2.3) | 0.775 |
| Co-morbidity | Non-co-morbid | 97(55.7%) | 52(20.8%) | 1 | |
| Alcohol intake | Yes | 24(13.8%) | 211(84.4%) | 33.8(19.5–58.6) | <0.001 |
| Alcohol intake | No | 150(86.2%) | 39(15.6%) | 1 | |
| Self-care behaviour | Good | 163(93.7%) | 26(10.4%) | 1 | |
| Self-care behaviour | Poor | 11(6.3) | 224(89.6%) | 127.6(61.3–265.7) | <0.001 |
| Social-support | Good | 158(90.8%) | 36(14.4%) | 1 | |
| Social-support | Poor | 16(9.2%) | 214(85.6%) | 58(31.4–109.5) | <0.001 |
| Knowledge | Good | 143(82.5%) | 24(9.6%) | 1 | |
| Knowledge | Poor | 31(17.8%) | 226(90.4%) | 43.3(24.5–77) | <0.001 |
The multivariate logistic regression analysis also shows that, NYHA class, self-care behavior, high alcohol intake, social support, knowledge level and marital status were found to be associated with depression and statistical significance with a P-value of less than 0.05.
Those heart failure patients who had NYHA Class category of III and IV were more likely to be depressed than class I and II[(AOR:12.8(2.2–71.6), $95\%$ CI, $$P \leq 0.004$$, (AOR:19.2(1.9–189.9), $95\%$ CI, $$P \leq 0.011$$)] respectively. Those heart failure patients who had poor self-care behaviour are 9-fold at risk of depressive disorder as compared to those who had good self-care behaviour [(AOR: 9.1(2.4–34.6),$95\%$ CI, $$P \leq 0.001$$)]. Those heart failure patients they were alcohol users are 18 times more likely to be depressed than those who have not taken alcohol [(AOR: 17.7(4.14–35.65),$95\%$ CI, $$P \leq 0.001$$)]. Those heart failure patients who were poor social support are 5 times more likely to be depressed when compared to good social support [(AOR: 4.6(1.2–16.7),$95\%$CI, $$P \leq 0.020$$)] as depicted in Table 3 below. In this study those heart failure patients who had poor knowledge were more likely 5 times to be depressed as compared to good knowledge[(AOR:5.1(1.3–20.6),$95\%$CI, $$P \leq 0.020$$)]. Those heart failure patients who are unmarried 9.25 times more likely to be depressed than married [(0.108(0.03–0.47),$95\%$CI, $$P \leq 0.001$$)].
**Table 3**
| Variable | Category | Depressive symptom | Depressive symptom.1 | AOR | P Value |
| --- | --- | --- | --- | --- | --- |
| Variable | Category | NonDepression | Depression | AOR | P Value |
| Marital status | Single or not Married | 53(30.5%) | 205(82%) | 0.108(0.03–0.47) | 0.001* |
| Marital status | Married | 121(69.5%) | 45(18%) | 1 | |
| NYHA Class | I | 75(43.1%) | 17(6.8%) | 1 | |
| NYHA Class | II | 69(39.7%) | 46(18.4%) | 3.8(0.81–17.9) | 0.09 |
| NYHA Class | III | 24(13.8%) | 116(46.4%) | 12.8(2.2–31.6) | 0.004* |
| NYHA Class | IV | 6(3.4%) | 71(28.4%) | 19.2(9–89.9) | 0.011 |
| Alcohol intake | Yes | 24(13.8%) | 211(84.4%) | 17.7(4.14–35.65) | <0.001* |
| Alcohol intake | No | 150(86.2%) | 39(15.6%) | 1 | |
| Self-care behaviour | Good | 163(93.7%) | 26(10.4%) | 1 | |
| Self-care behaviour | Poor | 11(6.3) | 224(89.6%) | 9.1(2.4–34.6) | 0.001* |
| Social-support | Good | 158(90.8%) | 36(14.4%) | 1 | |
| Social-support | Poor | 16(9.2%) | 214(85.6%) | 4.6(1.2–16.7) | 0.020* |
| Knowledge | Good | 143(82.5%) | 24(9.6%) | 1 | |
| Knowledge | Poor | 31(17.8%) | 226(90.4%) | 5.1(1.3–20.6) | 0.020* |
## 4. Discussions
The aim of this study was to assess Factors associated with depression among heart failure patients at selected public hospitals in Addis Ababa, Ethiopia. The incidence of depression in this study reveals that $59\%$ ($95\%$ CI,54.5–63.7), this indicates that depression is highly associated with heart failure. It is obviously clear that depression is highly affecting the quality of life and efficacy of care.
A study conducted in Australia shows the incidence of depression was $52\%$ and South Africa in Johannesburg indicates $50\%$ [8,21]. This similarity might be due to problem of depression in heart failure patients spreading across in different nations globally. The current study reveals that higher incidence compared to study done in Greece (20–$40\%$), Japanese ($5.8\%$), United States of America ($42.1\%$) [20,27,28]. The difference might be due to different screening strategy, study design and sociodemographic characteristics of the study population. And incidence of depression observed in the present study was higher than a study conducted in west Amhara region of Dessie city administration were $50\%$ [1], this variation might be different educational background, their life style and sociodemographic characters.
This study indicated that those heart failure patients with NYHA class III and IV are more likely at risk of depressive disorders compare to those who had NYHA class I and II. The possible explanation could be individual heart failure patients with advanced stage might be worry about their worsening symptom, illness-related complication, dietary restriction and un able to do any activity and they are always dependent on others. This might be directly or indirectly lead to depression. This finding supported by a study done in Greece 2020, and study done in Ethiopia [1,15]. This indicate that advanced heart failure patients need to be early evaluation of depression and cardiac clinic work in collaboration with psychiatry department to screen and therapeutic interventions. This study found that heart failure patients who had poor self-care behaviour is positively associated with depression when compared to good self-care behaviour, this might be due to poor self-care behaviour, prone to depression and potential to develop bad habit, like cigarette smoking, the use of shisha, chat chewing and lack of regular exercise. The other co-relation between psychological factors and disease outcomes, such as poor quality of life, effect of poor practice of self-care behaviour as the result heart failure patients may be potential to develop depression. This finding was in line with the study done in New-York, Brazil and Gondar [10,16,17]. This similarity might be due to prevention strategy of the country and pathological nature of disease process.
In this study, those heart failure patients who experienced high alcohol intake shows strongly positive association with depressive disorder than those who have not experienced alcohol use [(AOR: 17.7(4.14–35.65),$95\%$ CI, $$P \leq 0.001$$)] this might be due to the fact that alcohol use exposes to depression. Most alcohol users might be affected by physiological, psychological, social and economic behaviour that can alters metabolic condition of an individual life and as the result of conflict to family and society at large. This finding supported by research done in England, [24] and this finding contradict to research done in United Kingdom showed that there is no significant association alcohol intake and depression. The difference might be due to difference in socio-demographic characteristics of the study population. And also, this study indicates that heart failure patients who had poor social support is more likely to be depressed when compare to good social support [(AOR: 4.6(1.2–16.7),$95\%$CI, $$P \leq 0.020$$)]. The reason might be patients who have poor social support may not share their own stressor. and also plays great role in the coping strategies, so that this situation might be directly or indirectly expose to depression. This finding supported by research done in Pakistan, USA and Greece [25,26]. This reveals that and contact with support group for those HF patients who had poor social support. In this study heart failure patients who had poor knowledge about their disease were having positive associated with depression [(AOR:5.1(1.3–20.6),$95\%$CI, $$P \leq 0.020$$)]. This finding supported by research done in Ethiopia [1,30]. This indicate that health care institution should be focused on health education specially for heart failure patients who had poor knowledge. *In* general health care professionals should focus on education about their disease process and associated factors.
This study reveal that unmarried patients have positively associated with depression compared to married ones. [( AOR:0.108(0.03–0.47),$95\%$CI, $$P \leq 0.001$$)] This might be heart failure patients who were unmarried did not share their own stressor to life partner. This finding was in line with the study conducted in Pakistan and Ethiopia [11,27].
## 5.1 Conclusion
This study reveals that Factors associated with depression among heart failure patients at selected public hospitals found to be very high and advanced stage of heart failure patients were more depressive, poor self-care behaviour, alcohol users and poor social support were more likely at risk of depressive disorder. And also, heart failure patients who had good knowledge are at lower risk of depressive disorder and having single or unmarried patients were at risk of depressive disorder and those variables were associated with the odds of depression among heart failure patients in Addis Ababa.
## 5.2 Recommendations
All health institution of cardiac units should work on screening of heart failure patients for depression and consult psychiatrist and Psychologist for early detection and possible measure. In addition to this health care workers should focus to teach heart failure patients about disease prognosis and associated risk factors and patients’ education should be a part of heart failure patients management guideline.
## References
1. Yazew K, Beshah D. **Factors Associated with Depression among Heart Failure Patients at Cardiac Follow-Up Clinics in Northwest Ethiopia, 2017: A Cross-Sectional Study.**. *Hindawi Psychiatry J* (2019.0) 1-8. DOI: 10.1155/2019/6892623
2. Dimos AK, Stougiannos PN, Kakkavas AT, Trikas AG. **Depression and heart failure**. *Hellenic J Cardiol* (2009.0) **50** 410-7. PMID: 19767283
3. York K., Sheps D.. **Psychobiology of depression/distress in congestive heart failure. Heart Failure.**. *Europe PMC* (2009.0) **14** 35-50. DOI: 10.1007/s10741-008-9091-0
4. Djarv A., Lagergren P.. **“Number and burden of cardiovascular diseases in relation to health-related quality of life in a cross-sectional population-based cohort study**. *BMJ* (2012.0) **2** 1-5. DOI: 10.1136/bmjopen-2012-001554
5. Abedi H, Abdeyazdan G. **Quality of Life in heart failure patients referred to the Kerman outpatients’ centers**. *J Shahrekord Univ Med Sci* (2011.0) **13** 55-63. DOI: 10.1186/s12955-021-01861-2
6. Polikandrioti M., Koutelekos L., Vasilopoulos G., Gerogianni G., Gourni M., Zyga S.. **Factors Associated with Depression and Anxiety of Hospitalized Patients with Heart Failure Europe PMC**. (2015.0) **56** 26-35
7. Celano CM, Villegas AC, Albanese AM, Gaggin HK, Huffman JC. **Depression and Anxiety in Heart Failure: A Review.**. *Harvard review of psychiatry* (2018.0) **26** 175-84. DOI: 10.1097/HRP.0000000000000162
8. Tsabedze N, Kinsey J-L, Mpanya D, Mogashoa V, Klug E, Manga P. **The Prevalence of depression in outpatients attending a chronic heart failure with reduced ejection fraction clinic in a tertiary academic center in Johannesburg, South Africa**. *international jornal of mental health* (2020.0) **1** 24137. DOI: 10.21203/rs.3.rs-24137/v1
9. Mbakwem A., Amadi C.. **Depression in Patients with Heart Failure: Is Enough Being Done Cardiac Failure.**. *Cardiac Failure Review* (2016.0) **2** 110-2. DOI: 10.15420/cfr
10. Edmealem A, Olis C. **Factors Associated with Anxiety and Depression among Diabetes, Hypertension, and Heart Failure Patients at Dessie Referral Hospital, Northeast Ethiopia.**. *Behavioral Neurology.* (2020.0) **2020** 3609873. DOI: 10.1155/2020/3609873
11. Abdul S., Ruqia S., Mureed H., Muhammad T., Shagufta B., Arslan K. **Hospital Anxiety and Depression of Patients with Heart Failure in South Punjab Pakistan: A Sectional Survey Study Transaction Journal of Engineering**. *Management Applied-Sciences& Technologies* (2020.0) **11** 1-10. DOI: 10.14456/itjemast.2020.103
12. Henok M., Getenet D., Fasil W., Cheru T., and Dube J.. **The prevalence of depression among diabetic patients in Ethiopia: a systematic review and meta-analysis**. *Depression Research and Treatment* (2018.0) **2018** 1-8. DOI: 10.1155/2018/6135460
13. Rutledge VAR S T.. **Depression in heart failure a meta-analytic review of prevalence, intervention effects, and associations with clinical outcomes**. *J Am Coll Cardiol* (2006.0) **48** 1527-37. DOI: 10.1016/j.jacc.2006.06.055
14. Iloh G, Aguocha G, Amadi A, Chukwuonye M. **Depression among ambulatory adult patients in a primary care clinic in southeastern Nigeria**. *Nigerian Postgraduate Medical Journal* (2018.0) **25** 204-12. DOI: 10.4103/npmj.npmj_107_18
15. Polikandrioti M., Morou Z., Kotronoulas G., Evagelou H., Kyritsi H.. **Evaluation of depression in patients with congestive heart failure**. *Health Science Journal* (2010.0) **4** 37-47
16. Viviane M., Luma N., Rejane K., André S., Lídia A., Rosana A.. **Self-care, sense of coherence and depression in patients hospitalized for decompensated heart failure**. *Journal of School of Nursing USP* (2015.0) **49** 387-393. DOI: 10.1590/s0080-623420150000300005
17. Tsegu H., Kalayou K., Weyzer T., Haftea H., Kbrom G.. **Self-Care Behavior and Associated Factors among Heart Failure Patients**. *Clinical Nursing Research* (2020.0) **1** 1-8. DOI: 10.1177/1054773820961243
18. Beker TB J., Mekonin A., Hailu E.. **Predictors of adherence to self-care behavior among patients with chronic heart failure attending Jimma University Specialized Hospital Chronic Follow up Clinic, South West Ethiopia**. *Journal of Cardiovascular Diseases & Diagnosis* (2014.0) **2** 1-8. DOI: 10.4172/2329-9517.1000180
19. 19Addis Ababa, Ethiopia metroArea-population-1950-2020, 10–25. Available from: https://www.macrotrends.net/cities/20921/addis-ababa/population.
20. Hailu A., Berhe H., Aregay A.. **Assessment of depression prevalence and its Determinants among adult patients admitted in governmental hospitals**. *A Cross-sectional Study Stress* (2012.0) **4** 1882-92. DOI: 10.13040/ijPSR.0975-8232
21. Kroenke K, Spitzer RL, Williams JB. **The PHQ-9: validity of a brief depression severity measure**. *J Gen Intern Med* (2001.0) **16** 606-13. DOI: 10.1046/j.1525-1497.2001.016009606.x
22. Friedmann H., Thomas A., Chapa D., Lee H.. **Poor social support is associated with increases in depression but not anxiety over 2 years in heart failure out patients**. *Journal of Cardio-vascular Nursing* (2014.0) **29** 20-8. DOI: 10.1097/JCN.0b013e318276fa07
23. Kocalevent R., Berg L., Beutel M., Hinz A., Zenger M., Härter M. **social support in the general population: standardization of the Oslo social support scale (OSSS-3).**. *BMC Psychology* (2018.0) **6** 31. DOI: 10.1186/s40359-018-0249-9
24. van der Wal M., Jaarsma T, Moser D, van Veldhuisen D. **Development and Testing of the Dutch Heart Failure Knowledge Scale**. *European Journal of Cardiovascular Nursing* (2005.0) **4** 273-7. DOI: 10.1016/j.ejcnurse.2005.07.003
25. Simon G., Rutter C., Peterson D, Oliver M, Whiteside U, Operskalski B. **Does response on the PHQ-9 Depression Questionnaire predict subsequent suicide attempt or suicide death Psychiatric services**. (2013.0) **64** 1195-202. DOI: 10.1176/appi.ps.201200587
26. Sewagegn N, Fekadu S, Chanie T. **Adherence to self-care behaviours and knowledge on treatment among heart in Ethiopia**. *Journal of Pharmaceutical Care & Health Systems* (2015.0) **17** 1-7. DOI: 10.4172/2376.0419S4-001
27. Suzuki T., Omori H., Tatsumi F., Nishimura K., Hagiwara N.. **Depression and outcomes in Japanese outpatients with cardiovascular disease. a prospective observational study**. (2016.0) **80** 2482-8. DOI: 10.1253/circj.CJ-16-0829
28. Bhautesh D., Frances S., Véronique L., Susan A., Ruoxiang J., Alanna M.. **“Comorbid-depression and heart failure: a community cohort study.**. *PLoS ONE* (2016.0) **11** e0158570. DOI: 10.1371/journal.pone.0158570
29. Graven L., Martorella G., Gordon G., Grant Keltner J., Higgins M.. **Predictors of Depression in Outpatients with Heart Failure.**. *International journal of Nursing study* (2017.0) **69** 57-65. DOI: 10.1016/j.ijnurstu.2017.01.014
30. Khan M., Monaghan M., Klein N., Ruiz G., John A.. **Associations among Depression Symptoms with Alcohol and Smoking Tobacco Use in Adult Patients with Congenital Heart isease**. *Congenital heart disease* (2015.0) **10** 243-9. DOI: 10.1111/chd.12282
|
---
title: 'Menstrual health challenges in the workplace and consequences for women’s
work and wellbeing: A cross-sectional survey in Mukono, Uganda'
authors:
- Julie Hennegan
- Justine N. Bukenya
- Fredrick E. Makumbi
- Petranilla Nakamya
- Natalie G. Exum
- Kellogg J. Schwab
- Simon P. S. Kibira
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021399
doi: 10.1371/journal.pgph.0000589
license: CC BY 4.0
---
# Menstrual health challenges in the workplace and consequences for women’s work and wellbeing: A cross-sectional survey in Mukono, Uganda
## Abstract
This study describes women’s menstrual health needs at work in Uganda and explores the associations between unmet needs and women’s work and wellbeing. We undertook a cross-sectional survey of women working in marketplaces, public primary schools, and health care facilities in Mukono district, central Uganda. Survey questions were designed to capture women’s experiences of managing menstrual bleeding, pain, social support, and the social environment. A total 435 women working in markets, 45 teachers and 45 health care facility workers participated. Of these, $15\%$ missed work due to their last period, and $41\%$ would prefer not to work during menstruation. Unmet menstrual health needs were associated with consequences for women’s work and psychological wellbeing. Experiencing menstrual pain (aPR 3.65 $95\%$CI 1.48–9.00), along with the use of improvised menstrual materials (aPR 1.41 $95\%$CI 1.08–1.83), not feeling comfortable to discuss menstruation at work (aPR 1.54 $95\%$CI 1.01–2.34) and the expectation that women should stay home when menstruating (aPR 2.44 $95\%$CI 1.30–4.60) were associated with absenteeism due to menstruation. In contrast, not having menstrual management needs met (aPR 1.45 $95\%$CI 1.17–1.79) and the attitude that menstruating women are dirty (aPR 1.94 $95\%$CI 1.50–2.51), along with pain (aPR 1.59 $95\%$CI 1.12–2.24) and norms around absenteeism were associated with wanting to miss work. After adjustment for age and poverty, unmet menstrual management needs (b = -5.97, $95\%$CI -8.89, -2.97), pain (b = -3.89, $95\%$CI -7.71, -0.08) and poor social support (b = -5.40, $95\%$CI -9.22, -1.57) were associated with lower wellbeing measured using the WHO-5. Attitudes that menstruation should be kept secret ($b = 4.48$, $95\%$CI 0.79, 8.17) and is dirty ($b = 4.59$, $95\%$CI 0.79, 8.40) were associated with higher wellbeing. Findings suggest that supporting care for menstrual pain, addressing secrecy surrounding menstruation and the perception of menstruation as dirty, and improving access to materials and facilities for managing menstrual bleeding are avenues for programs and policies to support working women.
## Introduction
Women’s participation in decent work is essential for sustainable development, reducing poverty, and improving the health of women and their families [1–4]. The many hours spent at work also makes workplaces important sites which can support, or undermine, health [5]. The average female menstruates 65 days of the year, yet women’s menstrual health needs in the workplace are frequently overlooked [6].
Menstrual health has been recognised as an essential part of sexual and reproductive health, and a core consideration for gender sensitive water, sanitation, and hygiene (WASH) service provision [7–11]. However, research to understand menstrual experiences and develop policy and practice responses in low- and-middle-income countries (LMICs) has focused almost exclusively on adolescent girls [12]. While adolescence represents a window of opportunity to safeguard menstrual health [13], menstrual health needs continue into adulthood [14,15]. Calls for greater attention to menstruation over the life-course have emphasised the need for research to understand women’s experiences at work and identify levers for improvement [6,16].
## Consequences of unmet menstrual health needs
Nationally representative surveys have found that many women report missing work or other daily activities due to menstruation. Performance Monitoring for Action surveys in Burkina Faso and Nigeria found $19\%$ and $17\%$ respectively, and almost one in four women in the lowest wealth tertile in both countries, missed work due to their period [15]. Multiple Indicator Cluster Surveys across countries have found up to $35\%$ of women missed participating in school, work or other social events while menstruating [17]. While this nationally representative data highlights the importance of menstruation for women’s work, the surveys did not assess the reasons for absenteeism to inform policy and practice responses [18]. Further, a focus on attendance does not acknowledge other impacts on women’s lives [12,19,20]. Research with adolescent girls has been criticised for failing to capture impacts on participation at school, confidence or wellbeing, alongside attendance [21,22].
Prior qualitative interviews reported in Hennegan et al [14] undertaken with women working in markets, schools, and health care facilities in Mukono, Uganda found women experienced numerous consequences for their lives due to menstruation. These included: missed days at work, feeling uncomfortable or needing to persevere through distress or discomfort to remain at work, and anxiety and distress [14]. Thus, in this study we describe the prevalence of each of these consequences among our sample. Further, we investigate the extent to which different unmet menstrual health needs are associated with these outcomes. Women also reported implications of menstruation for their financial wellbeing, however this is outside the scope of this investigation.
## Unmet menstrual health needs
Past studies undertaken with adolescent girls have identified a wide range of unmet menstrual health needs across contexts [12]. These include: difficulties accessing sufficient materials to absorb or catch menstrual bleeding [20], poor availability of menstrual-friendly infrastructure to change materials or for cleaning the body and laundering materials [23], insufficient information about the menstrual cycle [24], inadequate support for reducing menstrual pain and discomforts [25], as well as social norms and attitudes surrounding menstruation which restrict behaviour and participation [12,20,26]. These challenges have been identified as core requirements for achieving menstrual health [7], and pillars for intervention design summarised briefly as: knowledge, materials and infrastructure for menstrual management, support for discomforts and disorders, and a supportive social environment [10]. Our interviews with women working in markets, schools and health care facilities echoed this body of evidence [14]. We found that strong social expectations to keep menstruation secret, and disgust surrounding menstruation resulting in strict hygiene requirements dictated women’s experiences. Menstrual pain and heavy bleeding were significant challenges, as was inadequately supportive infrastructure to change menstrual materials or wash the body. Difficulties affording sufficient menstrual materials for some, along with taboos held by market customers and varied social support from others at work also influenced women’s experiences of menstruation at work. In contrast to past research with adolescents [16], women working in markets, schools, and health care facilities did not emphasise menstrual knowledge gaps during our qualitative interviews [14].
Despite the breadth of menstrual health challenges identified in past qualitative research, there is scarce quantitative research to estimate the prevalence of unmet needs across different populations or to estimate the associations between these challenges and impacts on women’s lives [27]. Research that has been undertaken has often focused on the use of menstrual pads rather than the broad array of challenges identified through qualitative research [28,29]. Positively, more recent research with adolescents has investigated the contribution of a wider range of determinants including biological, menstrual management, and sociocultural considerations [30–32] but studies have not explored these associations in adult populations. It is plausible that different unmet menstrual health needs are more important contributors to different outcomes such as absenteeism or wellbeing. Understanding the associations between unmet menstrual health needs and consequences for women’s lives can help to prioritize program and policy responses and identify the outcomes that should be monitored in their evaluation.
## The present study
To address multiple gaps in the research evidence, the present study sought to address two aims. First, we aimed to describe working women’s unmet menstrual health needs and the prevalence of self-reported consequences for their work. Second, we aimed to test the associations between unmet menstrual health needs identified through our qualitative research and consequences for women’s lives, including (a) work absenteeism, (b) discomfort at work and (c) wellbeing, to inform future intervention approaches.
## Methods
This work is reported in accordance with the STROBE statement (included as S1 File) [33].
## Study population
Our research program included women working in public markets, government primary schools and public health care facilities (HCFs) in Mukono district, in the central region of Uganda. Mukono district was selected as an emerging industrial setting with a mix of rural and urban characteristics in which the study team had strong relationships with local governments and organizations. Market, school, and health care facility worker groups were selected as priority populations with government mandate for sanitation service provision [34]. Women working in marketplaces were the primary focus. In Uganda, informal employment has been estimated to account for $85\%$ of non-agriculture employment [35], with many women in informal employment working as market vendors selling retail and food [36].
Markets operating in the district for at least 8 hours per day for 3 days per week were eligible. This excluded weekend-only markets, anticipating workers may have other week-day employment, and markets with restricted hours such as those only open for evening mealtimes. In collaboration with the local government, we identified 10 markets. Government primary schools and public HCFs in closest proximity to each market were then sampled. We recruited five teachers and five HCF workers for each of the 10 markets, sampled based on their availability with sampling extended to the next closest facility if there were less than five.
Women working in markets were sampled proportionally to the total population of female workers estimated from site visits and advice from market leaders. We sampled $50\%$ of the population in each market, except for the largest market in the municipality which was under sampled ($20\%$) to achieve sufficient representation from smaller markets. Enumerators mapped each market and systematically sampled female workers by selecting every second or fifth working woman. Women aged 18–45 who had worked at least 3 days per week over the past month and had not participated in the qualitative portion of the study were eligible. Ineligible women were replaced by the neighbouring female worker.
We sought a sample of 500 women in markets, alongside the 50 teachers and 50 HCF workers to explore sanitation needs and menstruation. The sample of teachers and HCF workers was limited by cost and feasibility. In our qualitative investigation [14] we found consistent consequences for women’s lives and the same set of contributing unmet menstrual health needs reported across worker groups. Thus, we hypothesised that associations between unmet menstrual health needs and consequences for work and wellbeing may be consistent across groups and all three groups were included in this quantitative study. Sensitivity analyses describe associations among the market sample alone. Our sample size was calculated to achieve $80\%$ power to detect modest correlations between unmet needs and outcomes 0.20 ($p \leq 0.01$) such as work absenteeism, while allowing up to $30\%$ of the sample not to have had a menstrual period [37] and answer questions about menstrual health. Power calculations were undertaken prior to the qualitative study, and as such we did not specify a single outcome for these analyses, nor did we have any past research reporting on the prevalence of key variables such as discomfort at work or unmet menstrual health needs to draw on for this analysis.
## Data collection
Data were collected in March 2020. Surveys were programmed onto smartphones using Open Data Kit (ODK) and administered verbally, with data uploaded to a secure cloud server at the end of each day, and downloaded for analysis. Ten experienced female, Ugandan enumerators received five days of training on the survey tool, sampling, and informed consent process. Surveys were conducted in Luganda or English based on participant preference, with auditory privacy. Participants were informed of their right to decline to answer any questions and provided written informed consent for participation. Enumerators worked around participant schedules, pausing interviews for workplace tasks, or returning later as needed. Surveys lasted approximately 45–60 minutes, and participants received a bar of soap (approx. US$1) in appreciation.
## Measures
Survey questions were developed in English and designed to capture menstrual health needs, experiences, and consequences for women’s lives described across the integrated model of menstrual experience [12], with specific questions informed by past research [38,39] and findings from our qualitative investigation reported in Hennegan et al [14]. Questions were translated and back translated by bilingual research team members (JNB, SPSK, PN). Cognitive interviews with seven women were undertaken to assess question acceptability and comprehension, with modifications made as indicated. Questions were further workshopped during enumerator training. The full survey is available on the project page: www.osf.io/nzjtq. Broadly, survey topics included: demographic information, psychological wellbeing, life at work, biological menstrual characteristics such as pain, menstrual management practices and experience, the social environment, consequences for work and social participation, and women’s experiences of their workplace sanitation infrastructure.
## Women’s work and wellbeing outcomes
We aimed to describe the prevalence of consequences for women’s work and test the associations between unmet menstrual health needs (outlined below) and three consequences for women’s lives identified through our qualitative study: [1] absence from work, [2] discomfort at work, and [3] wellbeing.
## Menstrual health needs (exposures)
Unmet menstrual health needs hypothesised to contribute to consequences for work and wellbeing were selected based on our qualitative findings reported in Hennegan et al. and grouped according to the categories identified through that analysis [14]. Table 1 displays the category reported in the qualitative study findings, identified unmet menstrual health need, and the survey question used. Further details on the measurement of each concept are provided below.
**Table 1**
| Category from qualitative findings [14] | Menstrual health need | Survey measure |
| --- | --- | --- |
| Managing menses and cleaning the body | Menstrual management needs | Score on the Menstrual Practice Needs Scale. |
| | Use of improvised menstrual materials | The use of improvised materials (rather than commercial disposable or reusable pads) at work |
| Menstrual cycle characteristics | Pain | “Do you experience cramping or pain in the abdomen, back or legs during your period?” |
| | Pain severity | “How would you rate the severity of your pain from 0–10, where 0 is no pain and 10 is the worst possible pain?” |
| Workplace environment | Social support | “How comfortable do/would you feel discussing menstruation with someone in your workplace?” Very uncomfortable, uncomfortable, comfortable, very comfortable |
| | Social support | “If your period started unexpectedly in the workplace, is there someone you could ask to help you?” Agree/Disagree |
| Keeping clean | Supportive sociocultural environment | Attitude: “Women should avoid working during menstruation for workplace hygiene” Agree/Disagree |
| | | Injunctive norm: “Women working here are expected to stay at home when they are menstruating” Agree/Disagree |
| Keeping menstruation secret | Supportive sociocultural environment | Attitude: “Women should not discuss menstruation with others in the workplace, it is a private matter” Agree/Disagree |
| | | Injunctive norm: “Most women working here expect others not to discuss menstruation” Agree/Disagree |
| Modern knowledge and restrictions | Supportive sociocultural environment | Injunctive norm: “Most people shopping in this marketplace would avoid purchasing food from a woman if they knew she was menstruating.” Agree/Disagree |
Menstrual management needs. Women’s experiences of managing menstrual bleeding were assessed through the Menstrual Practice Needs Scale (MPNS) [42]. This self-report scale assesses the extent to which women had access to resources and environments to care for their body which supported their preferences, comfort, privacy and safety during their last period [7]. Women reported whether needs were met on a four-point response scale from ‘never’ to ‘always’. The total mean score was calculated across all items applicable to the respondent. The MPNS performance in this population was assessed and is reported elsewhere [43]. The revised scale for adults was used for analysis. In contrast to past research, the scale was scored such that higher scores represent more negative experiences to support more easily interpreted prevalence ratios, with 0 representing the lowest level of unmet menstrual management needs, and 3 the highest. For description we also report differences across a categorical variable grouping respondents with a total score between 0 and 0.5 (few unmet needs), 0.51 and 1.49 (some unmet needs), and 1.5 to 2.49 (many unmet needs).
Use of improvised menstrual materials. The use of improvised materials (rather than commercial disposable or reusable pads) at work was included as a dichotomous independent variable. For 10 women who reported not attending any work during the last menstrual period, their menstrual material used at home was incorporated to avoid missing data.
Pain. Women were asked to report if they experienced menstrual pain and those experiencing pain were asked to rate the usual severity from 0 to 10. Those experiencing no pain were considered to have a severity of ‘0’.
Social support. Support in the workplace was assessed through two items, reported in Table 1, assessing women’s comfort discussing menstruation with someone at work and access to someone who could help them if their period started unexpectedly. Women’s comfort discussing menstruation with someone in their workplace was dichotomized as “uncomfortable” and “comfortable”.
Sociocultural environment. In qualitative interviews (see Table 1) women’s attitudes and the norm that menstruation should be kept secret, and that menstruation was dirty, so women needed to ‘keep clean’ were key determinants of menstrual experiences. In our survey we assessed women’s own endorsement of this attitude, the presence of the corresponding descriptive norm (what the respondent believes others do) and injunctive norm (what the respondent believes women are expected to do) [44]. Our qualitative findings highlighted the importance of the expectations of others, thus the injunctive norm was included for quantitative analysis. Women working in markets were also asked about customer behaviour highlighted during qualitative interviews, reporting if they felt shoppers would avoid menstruating women. Responses were included in sensitivity analysis undertaken with only women working in markets.
## Demographic characteristics
Questions captured respondents’ age, marital status, level of education and other workplace details such as job type and days worked. Poverty was assessed through a 5-item lived poverty index [45] which asked how often over the past year the participants’ household had gone without food, water, medical treatment, fuel for cooking or cash income. A total score (0–20) was calculated.
The practices undertaken to manage menstrual bleeding, such as the type of absorbent used and disposal mechanisms were captured using questions from the Menstrual Practices Questionnaire to describe the sample [46].
## Analysis
Analyses were conducted using Stata 17. To address the first study aim we use descriptive statistics to report the prevalence of unmet menstrual health needs, and the prevalence of consequences for women’s lives including self-reported absenteeism, discomfort at work, and wellbeing.
To assess the associations between unmet menstrual health needs and absenteeism, discomfort at work and wellbeing (Aim 2) we tested the bivariate and multivariable associations between menstrual health needs and these outcomes. For dichotomous work consequences (absenteeism and discomfort) we used Poisson regressions with a robust variance estimator to provide prevalence ratios [47]. This method was selected as neither outcome was rare and thus odds ratios would represent a poor approximation of risk ratios [48]. To account for clustering at the level of the workplace we used generalized estimating equations with exchangeable correlation structure (assuming observations within the cluster are equally correlated) to provide a population-averaged effect [49]. Due to the small number of clusters ($$n = 29$$) we computed bias-corrected standard errors using the Kauermann and Carroll correction for the full sample [50]. Needs associated with consequences at $p \leq .10$ in bivariate analyses were included in the multivariable comparisons.
To test the associations between unmet menstrual health needs and wellbeing we undertook ordinary least-squares regression, with standard errors adjusted for clustering within workplaces. *As* generalised wellbeing was assessed, we adjusted for demographic factors (age and poverty) to assess the association between each menstrual health need and wellbeing individually (model 1). Associations with $p \leq .10$ were carried through to a multivariable model (model 2) to assess the relative contribution of different menstrual health needs.
Given the limited number of teachers and HCF workers included in the study, we undertook sensitivity analysis to explore the associations reported exclusively among women working in markets.
## Ethical approvals
Ethical approval was provided by Makerere University School of Public Health Higher Degrees, Research and Ethics Committee (HDREC: 739) and Johns Hopkins Bloomberg School of Public Health Institutional Review Board (IRB: 00010015). The Uganda National Council for Science and Technology (UNCST) approved the study (ref: SS 5143). Workplace administrators (Headteachers, Health Care Facility Administrators and Market Chairpersons) permitted recruitment of participants from their workplaces. Approval for the study in the area was also provided by the Mukono district chief administrator’s office and the Mukono Municipality Town clerk’s office.
## Respondents
Of the 600 women who participated in the quantitative survey, $87.5\%$ had menstruated in the past six months and were asked questions about their menstrual experiences ($$n = 525$$). A total of 435 women working in markets, 45 teachers, and 45 HCF workers are thus included in this study.
Table 2 describes the characteristics of the sample. A total $83.0\%$ of the sample reported having gone without food, water, fuel, medicines, or income within the past year. The mean and median days worked was 6, and almost half the sample ($42\%$) worked 7 days per week. Most of the sample spent 9 to 12 hours in the workplace on a typical workday. Of those working in markets, $70.3\%$ selected their own market hours, with a further $14.7\%$ reporting that hours were dictated by the number of customers. Just $14.0\%$ had a supervisor who determined their work hours.
**Table 2**
| Unnamed: 0 | n | % |
| --- | --- | --- |
| Age | | |
| 18–25 | 155.0 | 29.5 |
| 26–30 | 139.0 | 26.5 |
| 31–35 | 77.0 | 14.7 |
| 36–40 | 95.0 | 18.1 |
| 41–45 | 59.0 | 11.2 |
| Religion | | |
| Christian | 424.0 | 80.8 |
| Muslim | 101.0 | 19.2 |
| Highest education level attended | | |
| None or primary school | 188.0 | 35.8 |
| Secondary school | 244.0 | 46.5 |
| Post-secondary school | 93.0 | 17.7 |
| Usual days worked | | |
| 3–4 | 74.0 | 14.1 |
| 5 | 83.0 | 15.8 |
| 6 | 150.0 | 28.6 |
| 7 | 218.0 | 41.5 |
| Hours worked in a typical day | | |
| <9 | 61.0 | 11.6 |
| 9–12 | 313.0 | 63.4 |
| 13+ | 151.0 | 28.8 |
| Menstrual material used most often at work during the last period (n = 514) | | |
| Cloth | 85.0 | 16.5 |
| Disposable pad | 363.0 | 70.6 |
| Reusable pad | 42.0 | 8.2 |
| Other: toilet paper, cotton wool or underwear alone | 24.0 | 4.7 |
| Washed and reused any materials during the last period | | |
| Yes | 145.0 | 27.6 |
| No | 380.0 | 72.4 |
| Changed menstrual materials at work during the last period | | |
| Never | 28.0 | 5.3 |
| One or some days | 142.0 | 27.1 |
| Every day | 354.0 | 67.6 |
## Consequences for work and wellbeing
Table 3 displays the prevalence of work consequences due to menstruation and wellbeing reported for each worker group. A total of $19.3\%$ of respondents reported usually missing work due to their period, $15.1\%$ due to the last menstrual period, and $40.6\%$ reported they would avoid scheduling work, if possible, during their period. Pain was the most common reason reported for absenteeism, along with other physical symptoms such as fatigue. A total $43\%$ of those missing work included concerns about menstrual management or facilities as a reason for absenteeism.
**Table 3**
| Unnamed: 0 | Marketsn = 435n (%) | HCFn = 45n (%) | Teachersn = 45n (%) |
| --- | --- | --- | --- |
| Work | | | |
| Usually misses work due to menstruation | 94 (21.7) | 6 (13.3) | 1 (2.2) |
| Missed work due to last menstrual period | 72 (16.6) | 6 (13.3) | 1 (2.2) |
| Time missed during the last menstrual period (n = 79) | | | |
| < 1 day | 6 (8.3) | 0 | 0 |
| 1 day | 31 (43.1) | 4 (66.7) | 1 (100.0) |
| 2 days | 18 (25.0) | 0 | 0 |
| 3+ days | 23.6 (17) | 2 (33.3) | 0 |
| Reasons for missing work (n = 79)1 | | | |
| Pain | 54 (75.0) | 5 (83.3) | 1 (100.0) |
| Other physical symptoms: fatigue, heavy bleeding, gastrological symptoms | 17 (23.6) | 1 (16.7) | 0 |
| Containment fears or inadequate materials | 27 (37.5) | 3 (50.0) | 0 |
| Inadequate sanitation facilities | 4 (5.6) | 1 (16.7) | 0 |
| Other | 2 (2.8) | 0 | 0 |
| Discomfort at work | | | |
| Would prefer not to work during menstruation | 199 (45.8) | 9 (20.0) | 5 (11.1) |
| Wellbeing | Mean (SD) | Mean (SD) | Mean (SD) |
| WHO5 | 47.1 (22.06) | 51.75 (20.03) | 50.93 (19.68) |
## Associations between menstrual health needs and consequences for work
Table 4 displays the total prevalence of menstrual health needs among the sample. Scores on the MPNS ranged from 0 to 1.89 with a mean of 0.53 (SD = 0.40). Most of the sample reported few unmet needs ($56.6\%$), with $41.3\%$ reporting some unmet needs and $2.1\%$ many unmet needs. Most women reported experiencing menstrual pain. Half of participants reported they would feel comfortable discussing menstruation with someone in the workplace, and $46.6\%$ had someone they could ask for help. Over half of the sample believed menstruation should be kept secret, a greater $68.4\%$ reported perceiving a norm that women should not discuss menstruation, and $15.6\%$ agreed that women were expected to stay home from work while menstruating.
**Table 4**
| Unnamed: 0 | Total% (mean) | Missed work n (%) / M (SD) | Did not miss workn (%) / M (SD) | PR (95%CI) | aPR (95%CI) | Would prefer to miss workn (%) | Would not prefer to miss n (%) | PR (95%CI).1 | aPR (95%CI).1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Poverty | (4.33) | 5.48 (3.95) | 4.11 (3.66) | 1.07 (0.99–1.17) | 1.01 (0.92–1.10) | 5.04 (3.95) | 3.85 (3.51) | 1.04 (1.02–1.07) | 1.00 (0.99–1.02) |
| Managing menses | | | | | | | | | |
| Menstrual Practice Needs total score (0–3) | (0.53) | 0.67 (0.47) | 0.51 (0.38) | 2.00 (1.25–3.19) | 1.24 (0.85–1.79) | 0.64 (0.43) | 0.45 (0.36) | 1.85 (1.37–2.50) | 1.45 (1.17–1.79) |
| Few unmet needs | 56.6 | 40 (13.5) | 257 (86.5) | | | 102 (34.3) | 195 (65.7) | | |
| Some unmet needs | 41.3 | 35 (16.2) | 181 (83.8) | | | 104 (47.9) | 113 (52.1) | | |
| Many unmet needs | 2.1 | 4 (36.4) | 7 (63.6) | | | 7 (63.6) | 4 (36.4) | | |
| Uses improvised materials at work (yes) | 21.8 | 26 (22.8) | 88 (77.2) | 1.67 (1.30–2.16) | 1.41 (1.08–1.83) | 57 (50.0) | 57 (50.0) | 1.18 (0.85–1.66) | |
| Uses commercial materials at work | 78.2 | 52 (12.7) | 357 (87.3) | 1.00 | | 155 (37.8) | 255 (62.2) | 1.00 | |
| Pain | | | | | | | | | |
| Experiences pain (yes) | 77.7 | 74 (18.2) | 333 (81.8) | 3.97 (0.98–16.16) | 3.65 (1.48–9.00) | 184 (45.1) | 224 (54.9) | 1.71 (1.21–2.42) | 1.59 (1.12–2.24) |
| Does not experience pain | 22.3 | 5 (4.3) | 112 (95.7) | 1.00 | 1.00 | 29 (24.8) | 88 (75.2) | 1.00 | 1.00 |
| Pain severity (1–10) | (4.45) | 5.97 (2.86) | 4.18 (3.36) | 1.15 (1.09–1.21) | - | 5.30 (3.27) | 3.87 (3.28) | 1.07 (1.03–1.12) | - |
| Social support at work | | | | | | | | | |
| Comfortable talking to someone (yes) | 50.0 | 30 (11.5) | 232 (88.6) | 1.00 | 1.00 | 89 (33.8) | 174 (66.2) | 1.00 | 1.00 |
| Not comfortable talking to someone | 50.0 | 49 (18.7) | 213 (81.3) | 1.59 (1.12–2.26) | 1.54 (1.01–2.34) | 124 (47.3) | 138 (52.7) | 1.36 (1.09–1.69) | 1.19 (0.99–1.43) |
| Has someone she could ask for help (yes) | 46.6 | 40 (16.4) | 204 (83.6) | 1.00 | - | 93 (38.0) | 152 (62.0) | 1.00 | - |
| Does not have someone | 53.4 | 39 (13.9) | 241 (86.1) | 0.84 (0.52–1.36) | - | 120 (42.9) | 160 (57.1) | 1.04 (0.90–1.20) | - |
| Sociocultural environment: Attitudes & norms | | | | | - | | | | |
| Attitude: Menstruation should be kept secret | | | | | | | | | |
| Agree | 56.8 | 49 (16.5) | 248 (83.5) | 1.22 (0.58–2.56) | - | 140 (47.1) | 157 (52.9) | 1.35 (1.13–1.62) | 1.11 (0.98–1.26) |
| Disagree | 43.2 | 30 (13.3) | 196 (86.7) | 1.00 | - | 73 (32.2) | 154 (67.8) | 1.00 | 1.00 |
| Attitude: Women should avoid work during menstruation for hygiene | | | | | | | | | |
| Agree | 33.5 | 43 (24.6) | 132 (75.4) | 2.31 (1.48–3.61) | 1.56 (0.90–2.69) | 115 (65.3) | 61 (34.7) | 2.33 (1.74–3.13) | 1.94 (1.50–2.51) |
| Disagree | 66.5 | 36 (10.4) | 331 (89.6) | 1.00 | 1.00 | 97 (28.0) | 250 (72.0) | 1.00 | 1.00 |
| Injunctive norm: Women are expected to keep menstruation secret | | | | | | | | | |
| Agree | 68.4 | 61 (17.2) | 294 (82.8) | 1.66 (1.01–2.71) | 1.08 (0.71–1.66) | 161 (45.2) | 195 (54.8) | 1.45 (0.95–2.22) | 1.10 (0.63–1.94) |
| Disagree | 31.6 | 17 (10.4) | 147 (89.6) | 1.00 | 1.00 | 47 (28.7) | 117 (71.3) | 1.00 | 1.00 |
| Injunctive norm: Women are expected to stay home when menstruating | | | | | | | | | |
| Agree | 15.6 | 28 (34.6) | 65.43 (53) | 3.02 (2.13–4.30) | 2.44 (1.30–4.60) | 57 (69.5) | 25 (30.5) | 1.97 (1.52–2.55) | 1.49 (1.02–2.18) |
| Disagree | 84.5 | 49 (11.1) | 391 (88.9) | 1.00 | 1.00 | 153 (34.8) | 287 (65.2) | 1.00 | 1.00 |
Table 4 also presents women’s self-reported work absenteeism and discomfort according to unmet menstrual health needs, and the bivariate and multivariable associations between these unmet needs and work consequences.
Absenteeism. In the multivariable model, experiencing pain was associated with a much greater prevalence of absenteeism. Using an improvised (rather than commercially produced) menstrual material was associated with a greater prevalence of missing work. Not feeling comfortable to discuss menstruation and believing that women are expected to stay home during menstruation was also associated with absenteeism. The level of pain (rated from 0 to 10) was associated with absenteeism in bivariate comparisons, but due to high collinearity with reporting pain and 0-inflation we used the experience of pain as the predictor variable in the multivariable model.
Sensitivity analysis undertaken with only market women are reported in S2 File. The pattern of results remained the same as for the full sample. Agreeing that shoppers would avoid purchasing from a menstruating woman was associated with absenteeism in the bivariate comparison (PR = 1.50, $95\%$CI 1.01–2.19) but no longer statistically significant in the multivariable analysis (PR = 1.17, $95\%$CI 0.77–1.77).
Discomfort at work. Menstrual practice needs were significantly associated with wanting to miss work during menstruation, with an increase in 1 point on the MPNS associated with 1.45 times higher prevalence of wanting to avoid work during menstruation. Use of improvised materials was not associated with the desire to miss work. Pain remained associated, but social support was no longer statistically significant. Agreeing that women should avoid work during menstruation for hygiene was associated with preferring not to work during menstruation, along with the norm that women are expected to stay home. The attitude and norm that menstruation should be kept secret were associated with discomfort in bivariate comparisons but not in the final multivariable model.
In sensitivity analysis with market women, the pattern of results was similar, however pain was no longer significantly associated with wanting to miss work in the multivariable model (aPR = 1.36 $95\%$CI 0.99–1.94), although this broader confidence interval may request the reduced sample. Not feeling comfortable to talk to someone at work was associated with a higher prevalence of desiring to miss work (aPR = 1.15 $95\%$CI 1.02–1.30) with a similar effect size to that reported for the full sample (aPR = 1.19). For market women, believing that shoppers would avoid purchases from a menstruating woman was associated with wanting to miss work in the multivariable model (aPR = 1.39 $95\%$CI 1.07–1.80).
## Associations between menstrual health needs and wellbeing
Both age and poverty were significantly associated with wellbeing scores and were included as covariates in assessing the multivariable associations between menstrual health needs and wellbeing (see Table 5). Having unmet menstrual practice needs was associated with poorer wellbeing in the individual adjusted model (model 1) and the full multivariable model with a 1-point increase on the MPNS associated with a 6-point decrease in wellbeing (WHO-5) score. The use of improvised materials at work was not significantly associated with wellbeing.
**Table 5**
| Unnamed: 0 | Model 1 (predictor with adjustment for age & poverty) | Model 1 (predictor with adjustment for age & poverty).1 | Model 2 (full multivariable model)(n = 520) | Model 2 (full multivariable model)(n = 520).1 |
| --- | --- | --- | --- | --- |
| Predictor | b (Std. Error) | 95%CI | b (Std. Error) | 95%CI |
| Age | -0.35 (0.13) | -0.63, -0.08 | -0.34 (0.13) | -0.59, -0.08 |
| Poverty | -1.93 (0.22) | -2.38, -1.47 | -1.75 (0.22) | -2.20, -1.30 |
| Managing menses | | | | |
| Menstrual Practice Needs total score | -7.50 (1.38) | -10.33, -4.66 | -5.97 (1.47) | -8.98, -2.97 |
| Uses improvised materials at work (Yes) | 1.96 (1.48) | -1.06, 4.99 | - | |
| Pain | | | | |
| Experiences pain (Yes) | -4.51 (1.92) | -8.45, -0.56 | -3.89 (1.86) | -7.71, -0.08 |
| Pain severity (1–10) | -0.53 (0.28) | -1.10, 0.05 | - | |
| Social support at work | | | | |
| Comfortable talking to someone (No) | -5.02 (1.66) | -8.42, -1.62 | -5.40 (1.87) | -9.22, -1.57 |
| Has someone she could ask for help (No) | -2.15 (2.14) | -6.54, 2.24 | - | |
| Attitudes & norms | | | | |
| Attitude: Menstruation should be kept secret (Agree) | 3.13 (1.58) | -0.12, 6.38 | 4.48 (1.80) | 0.79, 8.17 |
| Attitude: Women should avoid work during menstruation for hygiene (Agree) | 4.55 (1.81) | 0.84, 8.26 | 4.59 (1.86) | 0.79, 8.40 |
| Injunctive norm: Women are expected to keep menstruation secret (Agree) | -1.18 (1.65) | -4.57, 2.21 | - | |
| Injunctive norm: Women are expected to stay home when menstruating (Agree) | 4.12 (1.30) | 1.45, 6.79 | 1.79 (1.33) | -0.93, 4.51 |
| Intercept | | | 70.27 (5.15) | |
| Adj R2 | | | 0.18 | |
Experiencing pain during menstruation was associated with lower wellbeing as was not being comfortable to talk to someone at work about menstruation, with results presented in Table 5. Endorsing attitudes that menstruation should be kept secret, and that work should be avoided for workplace hygiene, suggesting a view of menstruation as dirty, was associated with greater wellbeing in the full multivariable model. Believing that others expected secrecy, or women to remain at home during menstruation, were not associated with wellbeing in the multivariable model. Together, age, poverty and menstrual health needs accounted for $18\%$ of the variance in WHO-5 index scores.
The pattern of results was broadly consistent for sensitivity analysis including only women working in markets (see S2 File), however for this group pain was not significantly associated with wellbeing in the individual model (model 1). Wide confidence intervals meant attitudes that menstruation should be kept secret and that women should avoid work were no longer statistically significant in this more restricted sample ($$p \leq 0.067$$ and 0.090, respectively). Reporting that shoppers would avoid purchases from a menstruating woman was not significantly associated with wellbeing in the individual adjusted model (model 1).
## Discussion
Our study aimed to [1] describe the menstrual health needs and menstrual-related consequences for women’s lives in Mukono, Uganda, and [2] explore the associations between unmet menstrual health needs and consequences for work and wellbeing. We found a high prevalence of absenteeism and discomfort due to menstruation. Three in every 20 participants reported missing work due to their last period, and two in five reported that they would avoid scheduling work during menstruation if it were possible. Women reported many unmet menstrual health needs, including challenges with pain and caring for their body during menstruation, and difficulties in receiving social support and the social environment. In exploring the associations between these needs and work and wellbeing outcomes, we found that pain and social support were associated with absenteeism, while difficulties managing menstrual bleeding and social attitudes surrounding menstrual hygiene were associated with discomfort at work. Our findings provide the first quantitative evidence that menstrual health challenges may negatively impact adult women’s mental health. Results highlight that support for pain management, materials, and facilities for managing menstrual bleeding, and addressing stigma and silence surrounding menstruation are all important avenues for intervention. Addressing different unmet menstrual health needs may help to alleviate different consequences, with pain particularly critical for absenteeism and menstrual management essential for supporting comfort at work.
Consistent with research among adolescents [51], and surveys from high-income countries [52,53], menstrual pain was associated with work absenteeism. Of those who missed work, three quarters mentioned pain as a reason for absenteeism. Of those experiencing pain during menstruation, $18\%$ missed work because of their last period, compared to $4\%$ who did not report pain. Counter to hypotheses, reporting unmet menstrual management needs was not associated with absenteeism in multivariable comparisons despite many of those who missed work reporting management concerns as one of their reasons. However, use of an improvised menstrual material such as reusable cloth was associated with higher absenteeism. This was consistent with studies testing the associations between pad use and absenteeism that did not include other menstrual health needs [29]. Women reported varied menstrual material preferences in our qualitative interviews, but these quantitative results suggest that improvised materials may not perform as well as commercial products. It is possible that a greater risk of leakage or more time required to change these materials contributed to absences [54]. Not feeling comfortable to talk to someone at work about menstruation was associated with a higher prevalence of absenteeism, although reporting having someone to ask for help was not. This may suggest that the degree of comfort more validly captured women’s openness about menstruation and the support received. The expectation that women should stay home when menstruating was associated with a higher prevalence of absenteeism. This may reflect a negative expectation that menstruating women should not be present at work but could also capture those receiving greater permission from supervisors or co-workers to stay home if needed. In qualitative interviews many women reported that their supervisor would not be supportive if they needed to miss work due to menstruation [14], so it is possible this may be viewed as a positive norm.
This study advances evidence on the impacts of unmet menstrual health needs by exploring consequences for discomfort and wellbeing, not only absenteeism. Many women in our study reported desiring to miss work, indicating discomfort. Having unmet menstrual management needs was an important predictor in binary and multivariable comparisons. A total $48\%$ of women with some unmet management needs reported they would rather miss work, compared to $34\%$ of those with few unmet needs. Reflecting our qualitative finding that menstruating women were viewed as dirty, the attitude that women should avoid work during menstruation for hygiene, and the social expectation that women should stay home during menses was associated with the desire to miss work. A total $65\%$ of women who thought menstruating women should avoid work for workplace hygiene would rather miss work, compared to $28\%$ of those who did not agree with this belief. These findings suggest that while struggles related to managing menstrual bleeding may be less important for absenteeism, they are crucial for discomfort at work. Pain remained significantly associated with discomfort among the full sample, however this association was not significant in analyses including only women working in markets.
We found a statistically significant association between unmet menstrual health needs and women’s mental health. There are few studies against which to compare average WHO-5 scores for this sample, although scores suggested poor wellbeing. Mean scores observed in our study were poorer than those recorded for health care workers in Malawi [55] and similar to a sample of HIV-positive adults in Tanzania [56] as well as women in a study in India assessing the association between sanitation insecurity and mental health [57]. After adjustment for age and poverty, we found that unmet menstrual management needs, pain, and social support were all negatively associated with wellbeing in individual and multivariable comparisons. In contrast, reporting an attitude that menstruation should be kept secret, and that women should avoid work during menstruation for workplace hygiene were associated with better wellbeing in the full sample. While these findings may seem surprising when these attitudes were associated with absenteeism and a desire to avoid work, they are consistent with the findings from the qualitative interviews. In our qualitative analysis we found that women expressed pride in successfully enacting social expectations to keep menstruation secret and to keep clean [14]. It is plausible that for those endorsing and adhering to these expectations, this had a positive effect on wellbeing.
## Strengths and limitations
Survey questions and quantitative analyses were informed by in-depth qualitative investigation undertaken with the study population, along with past research. We also used our qualitative findings to aid the interpretation of results reported here, providing triangulation, and strengthening conclusions. While we were unable to take a full census and random sample of the workers, our proportional systematic sampling in markets offered a feasible and rigorous approach for this population. Only a small number of teachers and HCF workers were included due to feasibility constraints. These workers were included in the main quantitative analysis as similar consequences and unmet menstrual health needs were identified in the qualitative study and we hypothesised the same associations [14]. Quantitative sensitivity analysis including only women working in markets showed some differences to comparisons in the full sample. Future studies should investigate these effects in a larger sample of different worker groups. To assess menstrual management needs we used the newly validated Menstrual Practice Needs Scale to offer a comprehensive assessment [42,43]. We used the total score, rather than sub-scales to assess needs across the broad spectrum of blood management practices. However, this approach did include items capturing women’s experiences at home, not only those applicable to the workplace. Our findings are from cross-sectional data and as such we cannot infer causality or directionality. We elected to include a dichotomous variable indicating the presence of pain in multivariable analyses due to zero-inflation of pain reported on a rating scale. Questions capturing whether women were able to successfully reduce their menstrual pain should be explored in future studies to better understand the potential for interventions supporting pain mitigation. Our study relied on women’s self-reported unmet menstrual health needs and consequences for work. Such self-report is open to social desirability bias. Particularly in the interview format where women may have felt embarrassed to report difficulties surrounding menstruation to enumerators. Participants may have underreported unmet menstrual health needs or consequences of menstruation for their work.
We did not include any assessment of menstrual-related knowledge in our analysis. Inadequate knowledge about the menstrual cycle is often highlighted as a need among adolescents [58]. In our qualitative interviews, women did not report a high level of knowledge needs [14], although they desired more detailed information about menstrual products to inform their purchasing. Knowledge was not prioritized within the limited length of the quantitative survey.
## Implications for research and practice
Women spend many of their waking hours at work. In our sample, most worked six or seven days per week and between nine and 12 hours each day. Achieving sustainable development goals of decent work for all means ensuring work in safe environments, with equal opportunities for women. Our findings highlight that menstruation is an important contributor to women’s lives at work and must be considered in infrastructure provision and workplace policies [59,60].
Our findings demonstrate the importance of taking a holistic approach to menstrual health [4,7,12,16,61,62] which acknowledges the contribution of self-care challenges, pain, social support, attitudes, and norms. This range of drivers must be considered when designing programs to support women. We also found that different menstrual health needs may be more influential for some consequences. Comprehensive outcome assessment should be used in intervention trials and program evaluations to ensure the many consequences of menstruation for women’s lives are considered.
Menstrual pain was particularly important for work absenteeism, associated with a more than three-fold increase in missing work. Other correlates of absenteeism, being comfortable to talk to someone and expected to stay home during menstruation, may both serve to support women in managing their pain or symptoms at work. In contrast, unmet menstrual management needs were associated with a significantly increased desire to avoid work during menstruation, as did viewing menstruating women as dirty. Interventions focused on improving women’s access to infrastructure and materials may be more effective at reducing discomfort at work than absenteeism, while pain-focused interventions may be best placed to improve attendance. Use of improvised menstrual materials was associated with missing work, suggesting these may offer less protection against leakage despite women’s preferences. In contrast women’s own perspectives on their needs were associated with discomfort and wellbeing, while commercial product use was not.
Unmet menstrual health needs are associated with women’s wellbeing, with blood management needs and social support particularly important. At the same time, we found that attitudes constructing menstruation as dirty and something that should be kept secret had a positive association with wellbeing suggesting norm-change interventions must be navigated carefully.
## References
1. Brinda EM, Rajkumar AP, Enemark U. **Association between gender inequality index and child mortality rates: a cross-national study of 138 countries**. *BMC public health* (2015.0) **15** 1-6. DOI: 10.1186/s12889-015-1449-3
2. Moyo T, Dhliwayo R. **Achieving Gender Equality and Women’s Empowerment in Sub-Saharan Africa: Lessons from the Experience of Selected Countries**. *Journal of Developing Societies* (2019.0) **35** 256-81
3. 3United Nations Development Programme (UNDP). Gender Equality Strategy 2018–2021. New York, USA: United Nations Development Programme, 2018.. *Gender Equality Strategy 2018–2021* (2018.0)
4. Sommer M, Torondel B, Hennegan J, Phillips-Howard PA, Mahon T, Motivans A. **How addressing menstrual health and hygiene may enable progress across the Sustainable Development Goals**. *Global Health Action* (2021.0) **14** 1920315. DOI: 10.1080/16549716.2021.1920315
5. Asselberg K, Stecher-Rasmussen S. *How can social protection systems, interlinked with the access to public services and sustainable infrastructure, contribute to achieve gender equality and the empowerment of women and girls?* (2018.0)
6. Sommer M, Chandraratna S, Cavill S, Mahon T, Phillips-Howard PA. **Managing menstruation in the workplace: an overlooked issue in low- and middle income countries**. *International Journal for Equity in Health* (2016.0) **15** 86. DOI: 10.1186/s12939-016-0379-8
7. Hennegan J, Winkler IT, Bobel C, Keiser D, Hampton J, Larsson G. **Menstrual Health: A Definition for Policy, Practice, and Research**. *Sexual and Reproductive Health Matters* (2021.0) **29** 1-8. DOI: 10.1080/26410397.2021.1911618
8. Hennegan J, Tsui AO, Sommer M. **Missed Opportunities: Menstruation Matters for Family Planning**. *International perspectives on sexual and reproductive health* (2019.0) **45** 55-9. DOI: 10.1363/45e7919
9. 9UNFPA. Technical brief on the integration of menstrual health into sexual and reproductive heatlh and rights policies and programmes. Johannesburg, South Africa: United Nations Population Fund,, 2021.. *Technical brief on the integration of menstrual health into sexual and reproductive heatlh and rights policies and programmes* (2021.0)
10. 10UNICEF. Guidance on Menstrual Health and Hygiene. New York, USA: UNICEF. Available from https://www.unicef.org/wash/files/UNICEF-Guidance-menstrual-health-hygiene-2019.pdf [Accessed July 2019], 2019.. *Guidance on Menstrual Health and Hygiene* (2019.0)
11. Schmitt M, Clatworthy D, Ogello T, Sommer M. **Making the Case for a Female-Friendly Toilet**. *Water* (2018.0) **10** 1193
12. Hennegan J, Shannon AK, Rubli J, Schwab KJ, Melendez-Torres GJ. **Women’s and girls’ experiences of menstruation in low- and middle-income countries: a systematic review and qualitative metasynthesis**. *PLOS Medicine* (2019.0) **16** e1002803. DOI: 10.1371/journal.pmed.1002803
13. Sommer M.. **An Early Window of Opportunity for Promoting Girls’ Health: Policy Implications of the Girl’s Puberty Book Project in Tanzania**. *International Electronic Journal of Health Education* (2011.0) **14** 77-92
14. Hennegan J, Kibira SP, Exum NG, Schwab KJ, Makumbi FE, Bukenya J. **‘I do what a woman should do’: a grounded theory study of women’s menstrual experiences at work in Mukono District, Uganda**. *BMJ global health* (2020.0) **5** e003433. DOI: 10.1136/bmjgh-2020-003433
15. Hennegan J, OlaOlorun FM, Oumarou S, Alzouma S, Guiella G, Omoluabi E. **School and work absenteeism due to menstruation in three West African countries: findings from PMA2020 surveys**. *Sexual and Reproductive Health Matters* (2021.0) **29** 1915940. DOI: 10.1080/26410397.2021.1915940
16. Tellier S, Hyttel M. *Menstrual Health Management in East and Southern Africa: a Review Paper* (2018.0)
17. 17WHO, UNICEF. Progress on household drinking water, sanitation and hygiene 2000–2020: Five years into the SDGs. Geneva: World Health Organization (WHO) and the United Nations Children’s Fund (UNICEF), 2021.. *Progress on household drinking water, sanitation and hygiene 2000–2020: Five years into the SDGs* (2021.0)
18. Larson E, Turke S, Miko NH, Oumarou S, Alzouma S, Rogers A. **Capturing menstrual health and hygiene in national surveys: insights from performance monitoring and accountability 2020 resident enumerators in Niamey, Niger**. *Journal of Water, Sanitation and Hygiene for Development* (2021.0) **11** 295-303
19. Rheinländer T, Gyapong M, Akpakli DE, Konradsen F. **Secrets, Shame and Discipline: School Girls’ Experiences of Sanitation and Menstrual Hygiene Management in a Peri-Urban Community in Ghana**. *Health Care for Women International* (2019.0) **40** 13-32. DOI: 10.1080/07399332.2018.1444041
20. Scorgie F, Foster J, Stadler J, Phiri T, Hoppenjans L, Rees H. **"Bitten By Shyness": Menstrual Hygiene Management, Sanitation, and the Quest for Privacy in South Africa**. *Medical anthropology* (2016.0) **35** 161-76. DOI: 10.1080/01459740.2015.1094067
21. Hennegan J.. **Menstrual hygiene management and human rights: the case for an evidence-based approach**. *Women’s Reproductive Health* (2017.0) **4** 212-31
22. Geertz A, Iyer L, Kasen P, Mazzola F, Peterson K. *An Opportunity to Address Menstrual Health and Gender Equity*
23. Girod C, Ellis A, Andes KL, Freeman MC, Caruso BA. **Physical, Social, and Political Inequities Constraining Girls’ Menstrual Management at Schools in Informal Settlements of Nairobi, Kenya**. *Journal Of Urban Health: Bulletin Of The New York Academy Of Medicine* (2017.0) **94** 835-46. DOI: 10.1007/s11524-017-0189-3
24. Mason L, Nyothach E, Alexander K, Odhiambo FO, Eleveld A, Vulule J. **’We keep it secret so no one should know’—A qualitative study to explore young schoolgirls attitudes and experiences with menstruation in rural Western Kenya**. *PLoS ONE* (2013.0) **8**. PMID: 24244435
25. Wong LP. **Premenstrual syndrome and dysmenorrhea: urban-rural and multiethnic differences in perception, impacts, and treatment seeking**. *Journal of pediatric and adolescent gynecology* (2011.0) **24** 272-7. DOI: 10.1016/j.jpag.2011.03.009
26. Mohamed Y, Durrant K, Huggett C, Davis J, Macintyre A, Menu S. **A qualitative exploration of menstruation-related restrictive practices in Fiji, Solomon Islands and Papua New Guinea**. *PLOS ONE* (2018.0) **13** e0208224. DOI: 10.1371/journal.pone.0208224
27. Phillips-Howard PA, Caruso B, Torondel B, Zulaika G, Sahin M, Sommer M. **Menstrual hygiene management among adolescent schoolgirls in low-and middle-income countries: research priorities**. *Global health action* (2016.0) **9** 33032. DOI: 10.3402/gha.v9.33032
28. van Eijk AM, Zulaika G, Lenchner M, Mason L, Sivakami M, Nyothach E. **Exploring menstrual products: a systematic review and meta-analysis of menstrual cups for public health internationally**. *The Lancet Public Health* (2019.0)
29. Krenz A, Strulik H. **Menstruation hygiene management and work attendance in a developing country**. *CEGE Discussion Papers, Number* (2019.0)
30. Hennegan J, Sol L. **Confidence to manage menstruation at home and at school: findings from a cross-sectional survey of schoolgirls in rural Bangladesh**. *Culture, health & sexuality* (2019.0) 1-20. DOI: 10.1080/13691058.2019.1580768
31. Stoilova D, Cai R, Aguilar S, Batzer N, Nyanza E, Benshaul-Tolonen A. **Biological, material and socio-cultural constraints to effective menstrual hygiene management among secondary school students in Tanzania**. *PLOS Global Public Health. in press*
32. Davis J, Macintyre A, Odagiri M, Suriastini W, Cordova A, Huggett C. **Menstrual hygiene management and school absenteeism among adolescent students in Indonesia: evidence from a cross‐sectional school‐based survey**. *Tropical Medicine & International Health* (2018.0). DOI: 10.1111/tmi.13159
33. Von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. **The Strengthening the Reporting of Observational Studies in Epidemiology [STROBE] statement: guidelines for reporting observational studies**. *Gaceta Sanitaria* (2008.0) **22** 144-50. DOI: 10.1157/13119325
34. Burr P. *The State of WASH Financing in Eastern and Southern Africa: Uganda Country Level Assessment* (2019.0)
35. 35Uganda Bureau of Statistics. National Labour Force Survey 2016/17. Kampala, Uganda: Uganda Bureau of Statistics, 2018.. *National Labour Force Survey 2016/17* (2018.0)
36. Mugoda S, Esaku S, Nakimu RK, Bbaale E. **The portrait of Uganda’s informal sector: What main obstacles do the sector face?**. *Cogent Economics & Finance* (2020.0) **8** 1843255
37. Hennegan J, Shannon AK, Schwab KJ, investigators P. **Wealthy, urban, educated. Who is represented in population surveys of women’s menstrual hygiene management?**. *Reproductive health matters* (2018.0) **26** 1484220. DOI: 10.1080/09688080.2018.1484220
38. 38UNICEF. Guidance for Monitoring Menstrual Health and Hygiene.
New York: UNICEF, 2020.. *Guidance for Monitoring Menstrual Health and Hygiene.* (2020.0)
39. Hennegan J, Brooks DJ, Schwab KJ, Melendez-Torres G. **Measurement in the study of menstrual health and hygiene: A systematic review and audit**. *Plos One* (2020.0) **15** e0232935. DOI: 10.1371/journal.pone.0232935
40. Topp CW, Østergaard SD, Søndergaard S, Bech P. **The WHO-5 Well-Being Index: a systematic review of the literature**. *Psychotherapy and psychosomatics* (2015.0) **84** 167-76. DOI: 10.1159/000376585
41. Bech P, Gudex C, Johansen KS. **The WHO (Ten) well-being index: validation in diabetes**. *Psychotherapy and psychosomatics* (1996.0) **65** 183-90. DOI: 10.1159/000289073
42. Hennegan J, Nansubuga A, Smith C, Redshaw M, Akullo A, Schwab KJ. **Measuring menstrual hygiene experience: development and validation of the Menstrual Practice Needs Scale (MPNS-36) in Soroti, Uganda**. *BMJ open* (2020.0) **10**
43. Hennegan J, Bukenya JN, Kibira SP, Nakamya P, Makumbi FE, Exum N. **Re-validation and adaptation of the Menstrual Practice Needs Scale to measure the menstrual experiences of adult women working in Mukono District, Uganda**. *SocArXiv* (2021.0)
44. Bicchieri C.. *Norms in the wild: How to diagnose, measure and change social norms* (2017.0)
45. 45Afrobarometer. Surveys and methods: Afrobarometer; 2018 [cited http://www.afrobarometer.org/surveys-and-methods Accessed, September 2018].. *Surveys and methods: Afrobarometer*
46. Hennegan J, Nansubuga A, Akullo A, Smith C, Schwab KJ. **The Menstrual Practices Questionnaire (MPQ): development, elaboration, and implications for future research**. *Global health action* (2020.0) **13** 1829402. DOI: 10.1080/16549716.2020.1829402
47. Holmberg MJ, Andersen LW. **Estimating risk ratios and risk differences: alternatives to odds ratios**. *Jama* (2020.0) **324** 1098-9. DOI: 10.1001/jama.2020.12698
48. Cummings P.. **The relative merits of risk ratios and odds ratios**. *Archives of pediatrics & adolescent medicine* (2009.0) **163** 438-45. DOI: 10.1001/archpediatrics.2009.31
49. Gallis JA, Li F, Turner EL. *XTGEEBCV: Stata module to compute bias-corrected (small-sample) standard errors for generalized estimating equations* (2021.0)
50. Kauermann G, Carroll RJ. **A Note on the Efficiency of Sandwich Covariance Matrix Estimation**. *Journal of the American Statistical Association* (2001.0) **96** 1387-96. DOI: 10.1198/016214501753382309
51. Miiro G, Rutakumwa R, Nakiyingi-Miiro J, Nakuya K, Musoke S, Namakula J. **Menstrual health and school absenteeism among adolescent girls in Uganda (MENISCUS): a feasibility study**. *BMC women’s health* (2018.0) **18** 4. DOI: 10.1186/s12905-017-0502-z
52. Schoep ME, Adang EM, Maas JW, De Bie B, Aarts JW, Nieboer TE. **Productivity loss due to menstruation-related symptoms: a nationwide cross-sectional survey among 32 748 women**. *BMJ open* (2019.0) **9** e026186. DOI: 10.1136/bmjopen-2018-026186
53. Armour M, Parry K, Manohar N, Holmes K, Ferfolja T, Curry C. **The prevalence and academic impact of dysmenorrhea in 21,573 young women: a systematic review and meta-analysis**. *Journal of women’s health* (2019.0) **28** 1161-71. DOI: 10.1089/jwh.2018.7615
54. Hennegan J, Dolan C, Wu M, Scott L, Montgomery P. **Schoolgirls’ experience and appraisal of menstrual absorbents in rural Uganda: a cross-sectional evaluation of reusable sanitary pads**. *Reproductive Health* (2016.0) **13** 143. DOI: 10.1186/s12978-016-0260-7
55. Lohmann J, Shulenbayev O, Wilhelm D, Muula AS, De Allegri M. **Psychological wellbeing in a resource-limited work environment: examining levels and determinants among health workers in rural Malawi**. *Human resources for health* (2019.0) **17** 1-11. PMID: 30606232
56. Nolan C, O’Donnell P, Desderius B, Mzombwe M, McNairy M, Peck R. **Depression screening in HIV-positive Tanzanian adults: comparing the PHQ-2, PHQ-9 and WHO-5 questionnaires**. *Global Mental Health* (2018.0) 5. DOI: 10.1017/gmh.2018.31
57. Caruso BA, Cooper HL, Haardörfer R, Yount KM, Routray P, Torondel B. **The association between women’s sanitation experiences and mental health: A cross-sectional study in Rural, Odisha India**. *SSM-population health* (2018.0) **5** 257-66. DOI: 10.1016/j.ssmph.2018.06.005
58. Alam M-U, Luby SP, Halder AK, Islam K, Opel A, Shoab AK. **Menstrual hygiene management among Bangladeshi adolescent schoolgirls and risk factors affecting school absence: results from a cross-sectional survey**. *BMJ open* (2017.0) **7** e015508. DOI: 10.1136/bmjopen-2016-015508
59. Rajaraman D, Travasso SM, Heymann SJ. **A qualitative study of access to sanitation amongst low-income working women in Bangalore, India**. *Journal of Water Sanitation and Hygiene for Development* (2013.0) **3** 432-40. DOI: 10.2166/washdev.2013.114
60. Kamau A, Alfers L, Sverdlik A. *Revealing and strengthening the links between WASH, productivity, and well-being for informal vendors in Durban, South Africa, and Nakuru, Kenya* (2019.0)
61. Crichton J, Okal J, Kabiru CW, Zulu EM. **Emotional and Psychosocial Aspects of Menstrual Poverty in Resource-Poor Settings: A Qualitative Study of the Experiences of Adolescent Girls in an Informal Settlement in Nairobi**. *Health Care for Women International* (2013.0) **34** 891-916. DOI: 10.1080/07399332.2012.740112
62. Kansiime C, Hytti L, Nalugya R, Nakuya K, Namirembe P, Nakalema S. **Menstrual health intervention and school attendance in Uganda (MENISCUS-2): a pilot intervention study**. *BMJ open* (2020.0) **10**. DOI: 10.1136/bmjopen-2019-031182
|
---
title: 'Variation in the incidence of type 1 diabetes mellitus in children and adolescents
by world region and country income group: A scoping review'
authors:
- Apoorva Gomber
- Zachary J. Ward
- Carlo Ross
- Maira Owais
- Carol Mita
- Jennifer M. Yeh
- Ché L. Reddy
- Rifat Atun
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021400
doi: 10.1371/journal.pgph.0001099
license: CC BY 4.0
---
# Variation in the incidence of type 1 diabetes mellitus in children and adolescents by world region and country income group: A scoping review
## Abstract
### Introduction
Around 18.7 million of the 537 million people with diabetes worldwide live in low-income and middle-income countries (LMIC), where there is also an increase in the number of children, adolescents, and young adults diagnosed with type 1 diabetes (T1D). There are substantial gaps in data in the current understanding of the epidemiological patterns and trends in incidence rates of T1D at the global level.
### Methods
We performed a scoping review of published studies that established the incidence of T1D in children, adolescents, and young adults aged 0–25 years at national and sub-national levels using PubMed, Embase and Global Health. Data was analyzed using R programming.
### Results
The scoping review identified 237 studies which included T1D incidence estimates from 92 countries, revealing substantial variability in the annual incidence of T1D by age, geographic region, and country-income classification. Highest rates were reported in the 5–9 and 10–14 year age groups than in the 0–4 and 15–19 year age groups, respectively. In the 0–14 year age group, the highest incidence was reported in Northern Europe (23.96 per 100,000), Australia/New Zealand (22.8 per 100,000), and Northern America (18.02 per 100,000), while the lowest was observed in Melanesia, Western Africa, and South America (all < 1 per 100,000). For the 0–19 year age group, the highest incidence was reported in Northern Europe (39.0 per 100,000), Northern America (20.07 per 100,000), and Northern Africa (10.1 per 100,000), while the lowest was observed in Eastern and Western Africa (< 2 per 100,000). Higher incidence rates were observed in high-income countries compared to LMICs. There was a paucity of published studies focusing on determining the incidence of T1D in LMICs.
### Conclusion
The review reveals substantial variability in incidence rates of T1D by geographic region, country income group, and age. There is a dearth of information on T1D in LMICs, particularly in sub-Saharan Africa, where incidence remains largely unknown. Investment in population-based registries and longitudinal cohort studies could help improve the current understanding of the epidemiological trends and help inform health policy, resource allocation, and targeted interventions to enhance access to effective, efficient, equitable, and responsive healthcare services.
## Introduction
Diabetes is a major threat to health systems globally. An estimated 537 million people are living with diabetes worldwide in 2021, with around 18.7 million people living in low-income and middle-income countries (LMICs) [1]. The number of children, adolescents, and young adults diagnosed with type 1 diabetes (T1D) in LMICs is rising with over 1.2 million estimated in 2021, with several countries reporting higher incidence rates than current available best estimates resulting in more than half ($54\%$) of them <15 years of age [1].
The number of people living with diabetes is projected to increase to 643 million by 2030 and 783 million by 2045, with adverse consequences for the health and economic wellbeing of individuals, households, and countries [1–3]. An estimated 240 million people are presently living with undiagnosed diabetes worldwide with the global economic burden of diabetes estimated to increase from US$1.3 trillion in 2015 to $2.5 trillion by 2030 due to premature disability and mortality [4].
However, there are critical gaps in the current understanding of diabetes in children, adolescents, and young adults, with recent studies suggesting substantial variation in incidence at the global and sub-national levels, with higher incidence rates observed in several LMICs than earlier estimates indicated [5–11]. For many countries like Nigeria and South Africa, data are not available and are extrapolated from a nearby country with similar characteristics which may not be accurately presented. Less is known about the natural history, etiology, likelihood of complications, and the capability of health systems to deliver effective healthcare services to patients living with T1D in varied settings. It is crucial to understand the disease burden especially in LMICs where data is scant—and how it affects Early Child Development (ECD) and influences the active participation of adolescents and young adults in their economic and social pursuits. Such knowledge could inform health policies, resource allocation, and clinical practice to improve health outcomes for children, adolescents, and young adults in LMICs such as the need to secure adequate sustainable supplies of insulin.
There have been several notable initiatives to estimate the incidence of T1D documenting geographic and temporal trends, the impacts of health system barriers to diagnosis have not been quantified when estimating incidence rates, which may explain in part the large variation and increasing trends in observed incidence. In 1990, the World Health Organization launched the Multinational Project for Childhood Diabetes (WHO DIAMOND Project) to determine the global incidence of T1D over ten years, from 1990 to 1999. The goals of this project were threefold, namely to: I) gather standard information on incidence, risk factors, complications, and mortality associated with type 1 diabetes; II) assess the efficiency and effectiveness of health care and the economics of diabetes; and III) establish national and international training programs in diabetes epidemiology [6]. The DIAMOND project demonstrated more than 350-fold variation in the incidence of T1D among countries, although incidence data from several LMICs were only available for the first five years of the study period. However, the WHO DIAMOND project played a significant catalytic role to inspire subsequent initiatives to develop regional, national and multi-country registries to monitor the incidence and prevalence of T1D worldwide [6, 9–12]. The International Diabetes Federation (IDF) Atlas 10th Edition provides the most current estimates for incidence rates of T1D from 215 countries and territories, grouped into seven IDF Regions: Africa (AFR), Europe (EUR), Middle East, and North Africa (MENA), North America and Caribbean (NAC), South and Central America (SACA), South-East Asia (SEA) and the Western Pacific (WP) which include newly available data from AFR since 2019. Previous estimates in the 9th Atlas from 2019 included 138 out of the 211 countries worldwide. Incidence rates were available for approximately $6\%$ ($\frac{3}{47}$) of countries in sub-Saharan Africa, $31\%$ in Western Pacific ($\frac{11}{36}$), $33\%$ in North America and the Caribbean ($\frac{8}{24}$), $57\%$ in the Middle East and North Africa ($\frac{12}{21}$), $57\%$ in South-East Asia ($\frac{4}{7}$), $63\%$ in South and Central America ($\frac{12}{19}$) and $77\%$ ($\frac{44}{57}$) in Europe [8]. In 2021 IDF Atlas revision, only 97 of the 215 countries and territories have their own incidence data estimating a total of 1,211,900 children and adolescents under 20 years with T1D and 651,700 children under 15 years age. It is estimated that each year around 108,200 children and adolescents under 15 years are diagnosed with T1D with EUR and NAC regions reporting the highest number of prevalent cases and high incidence rates [1].
An initial search for systematic and scoping reviews on the global incidence of T1D in children, adolescents, and young adults was conducted, but no studies were identified. This study reviews the literature to examine the incidence of T1D in children, adolescents, and young adults (0–25 years) worldwide and explains the wide heterogeneity in the incidence rates that differ by age, sex, world region, and country income classification.
## Search strategy
In designing and conducting this scoping review, we adopted the methodological framework proposed by Arksey and O’Malley [13]. A systematic scoping review was conducted according to the PRISMA recommendations retrieved from original papers published in English up to April 12, 2021, in peer-reviewed journals reporting the incidence of T1D among individuals aged <25 years of age in population-based studies and reporting the diagnostic criteria used to define T1D. The databases used for the literature search were Medline / PubMed (National Library of Medicine, NCBI), Embase (Elsevier, embase.com), and Global Health (C.A.B. International, Ebsco). The protocol for the search was registered in the International Prospective Register of Systematic Reviews (PROSPERO). Controlled vocabulary terms (i.e. MeSH, Emtree, CAB Thesaurus) were included when available and appropriate. The search strategies were designed and executed by a librarian (CM) from Harvard University in discussion with the authors. Tables A and B in S1 Text outline both the concepts and controlled vocabulary used for the search strategy and summary. Fig 1 presents the flow diagram of the bibliographic search using the PRISMA 2020 checklist (PRISMA-ScR checklist shown in S1 Text). Ethics approval for performing the study was not required as no primary data were included.
**Fig 1:** *PRISMA flow chart of study selection.(Source: Authors).*
## Inclusion and exclusion criteria
We included studies published between 1990–2021 if they contained information about the incidence of T1D in children, adolescents, and young adults, under 25 years of age. Data from national diabetes registries or databases that explicitly tracked a pediatric or adolescent population were included. Studies were excluded if they did not report incidence rates of T1D or reported incidence rates for other diseases. Studies reporting incidence rates for populations over 25 years of age were excluded. Studies published in languages other than English were excluded from data extraction eligibility. Abstracts without associated full text, including abstracts published from conference presentations, were not included. Abstracts, where full text could not be located, were also excluded. All full-text manuscripts where the incidence rates were not reported were excluded from the review. Efforts were made to obtain the value of incidence rates for each country at the national level. If more than one study was available for a country, we applied the following four criteria to select the most suitable study: more recent studies, covering a large part of the country, including the age, ranges 0–14 and 15–19 years, providing age/sex-specific rates for 0–4, 5–9, 10–14 and 15–19-year age-groups.
## Screening, data extraction and analysis
A total of 7,647 results were retrieved across the three search databases (4,527 results from PubMed/ Medline; 2,484 results from Embase and 636 results from Global Health), resulting in 5320 unique references. Duplicate records were removed using EndNote and 22 additional duplicates were identified and removed during the import of references into the Covidence software for a title and abstract screening. 60 additional duplicates were discovered and excluded during full-text screening or data extraction. All abstracts were independently screened by three authors (AG, CR and MO) to identify articles meeting inclusion criteria. A process of consensus resolved disagreements, and inter-rater reliability was determined using a $20\%$ random sample to calculate the Cohen’s Kappa coefficient ($k = 0.785$). 538 full text articles were reviewed by two authors for eligibility, with final inclusion of the 237 studies determined by 2–3 author consensus. Data were extracted from the 237 studies by AG using a data collection template developed by ZW.
For each article, the following information was extracted and tabulated in the template (Table C in S1 Text): 1) Identification of the study with the year of publication: Authors, title, DOI and PMID; 2) Country and study location; 3) Geographical coverage of the study: Nationwide (when the study was conducted across the whole nation) or regional (when the study was restricted to a region, city or single-center) or Global (when the study involved rates reported globally); 4) Incidence rates expressed as new cases per 100,000 people (for both sexes) per year in the following 5-year age/sex subgroups (males/females) 0–4, 5–9, 10–14, 15–19 years and also in 0–14 and 0–19 years, the age standard stratifications used by IDF, to ensure comparability of our results. The rates were retrieved from either the tables or graphs, with the study periods recorded for each study. Supplementary information was also reviewed to include information on incidence rates from the literature for reported study periods.
Data extraction was performed using Microsoft Excel 16.51, and statistical analysis was performed using R v3.3.61. Data were graphically represented on maps that contained information from countries at the national level. We compared the incidence of T1D for countries that had information for the age groups 0–19 years and 0–14 years. Regional means and $95\%$ CIs were estimated using an inverse-variance weighting of the country-specific log incidence rates, re-transformed using Duan’s smearing estimator [14]. Secular trends were estimated for each region by regressing the log incidence rate on the calendar year, with inverse-variance weighting used to weight each estimate. We performed subgroup analysis by assigning countries based on the World Health Organisation (WHO) regions and by country income groups based on gross national income (GNI) per capita in 2018, as published in the June 2019 World Bank Income Classification: low-income country (LIC) with per capita GNI of $1025, lower-middle-income country (LMIC) $1026 to $3995, upper-middle-income country (UMIC) $3996 to $12,735, and high-income country (HIC) >$12,735 respectively [15].
## Study characteristics
We identified and extracted data from 237 studies, which provided 1852 estimates reporting T1D incidence covering 92 countries (Fig A in S1 Text). Of these studies, 79 reported T1D incidence in age groups 0–14 years, while 13 reported incidence in those aged 0–19 years. The majority of published studies ($$n = 1498$$, $81\%$) reported incidence estimates with accompanying $95\%$ confidence intervals. For a minority of published studies ($$n = 354$$, $19\%$), estimated incidence rates of T1D were calculated by dividing the number of patients diagnosed with T1D (numerator) by the corresponding person-years at risk (denominator). Of the 1852 T1D incidence estimates presented, $50\%$ ($$n = 920$$) were nationwide, $43\%$ ($$n = 804$$) were regional or multi-center, and $7\%$ ($$n = 128$$) were global (Fig B in S1 Text). Among the studies reporting incidence estimates from primary data sources, $64\%$ ($$n = 1189$$) were reported from population-based registries (national or regional), and $36\%$ ($$n = 663$$) were facility-based (single-center, hospital, or medical record) (Fig B in S1 Text).
## Variability in type 1 diabetes incidence rates
We observed substantial variation in the incidence of T1D worldwide and how it differed by age, sex, world region, and country income classification. We retrieved and compared national and subnational-level data from 92 countries in individuals 0–14 years and 0–19 years age, with sub-analysis for the IDF standard five-year age groups 0–4, 5–9, 10–14, and 15–19 years.
T1D incidence varied substantially between the 0–4, 5–9, 10–14, 15–19, 0–14 and 0–19 age groups across regions (Figs 2 and 3, Tables 1 and 2). Overall, the incidence was highest in the 10–14 year age group [18.02 per 100,000 ($95\%$ CI [17.54;21.49])] and lowest in the 15–19 year age group [6.71 per 100,000 ($95\%$ CI [4.54;7.91])]. Among those diagnosed in the 0–4 age category (Table 2), incidence was highest in Northern Africa [31.11 per 100,000 ($95\%$ CI [31.11;31.11])], Northern Europe [21.54 per 100,000 ($95\%$ CI [19.05;24.35]) and Western Europe [15.21 per 100,000 ($95\%$ CI [13.84;16.72]); it was lowest in Southern Asia, Eastern Asia, Western Asia and Eastern African countries (< 5 per 100,000), with a mean incidence of 10.27 per 100,000 ($95\%$ CI [8.77,12.97]). Within the 5–9 years age category (Table 2), incidence was highest in Northern Africa [44.78 per 100,000 ($95\%$ CI [NA])], Northern Europe [37.17 per 100,000 ($95\%$ CI [33.60,41.12])] and Northern America ([26.31 per 100,000 ($95\%$ CI [23.82,29.06])], whereas it was lowest in Southern Asia [0.92 per 100,000 ($95\%$ CI [0.65,1.31])], Eastern Asia [1.93 per 100,000 ($95\%$ CI [1.64,2.27])] and South America [4.47 per 100,000 ($95\%$ CI [1.51,13.25])], with a mean incidence of [17.19 per 100,000 ($95\%$ CI [15.67,19.41])]. Among those diagnosed within the 10–14 age category, incidence was highest in Northern Europe [41.48 per 100,000 ($95\%$ CI [37.42,45.97])], Northern Africa [40.92 per 100,000 ($95\%$ CI [NA])] and Northern America [33.50 per 100,000 ($95\%$ CI [29.54,38.00])], but lowest in Southern Asia [1.99 per 100,000 ($95\%$ CI [1.45,2.76])], Eastern Asia [2.78 per 100,000 ($95\%$ CI [2.41,3.21])] and South America [3.62 per 100,000 ($95\%$ CI [NA])] with a mean incidence of [18.02 per 100,000 ($95\%$ CI [17.54, 21.48])]. Finally, in the 15–19 age category (Table 2), incidence rates were highest in Northern America [17.68 per 100,000 ($95\%$ CI [13.31,23.48])] and Southern Europe [9.71 per 100,000 ($95\%$ CI [NA])], whereas they were lowest in East Asia [1.43 per 100,000 ($95\%$ CI [NA])] and African countries [1.07 per 100,000 ($95\%$ CI [0.58,1.96])], with a mean incidence of [6.71per 100,000 ($95\%$ CI [4.54, 7.91])].
**Fig 2:** *Variation in TlD incidence rates for age groups 0–14 and 0–19 years by region.[Point sizes are proportional to their weight (i.e. inverse variance). The lines indicate means and shaded areas indicate $95\%$ Cls].* **Fig 3:** *Variation in TlD incidence rates for age groups 0–4, 5–9, 10–14, 15–19 years by region.[Point sizes are proportional to their weight (i.e. inverse variance). The lines indicate means and shaded areas indicate $95\%$ Cls].* TABLE_PLACEHOLDER:Table 1 TABLE_PLACEHOLDER:Table 2 There was substantial variation in T1D incidence rates by country income group (Fig 4, Table 1). Higher incidence rates were reported in high-income countries [7.89 per 100,000 ($95\%$CI [7.24;8.59])], followed by upper-middle-income [0.87 per 100,000 ($95\%$ CI [0.60;1.25])], lower-middle-income [0.57 per 100,000 ($95\%$ CI [0.34;0.98])] and low-income countries [0.19 per 100,000 ($95\%$ CI [0.13;0.28])]. Sub-analysis of incidence rates by sex (Table 3) shown as Fig C in S1 Text revealed a mean male-to-female ratio of 1.04 ($95\%$ CI [1.02, 1.06]).
**Fig 4:** *Geographical representation of incidence rates of TlD by income.[Point sizes are proportional to their weight (i.e. inverse variance). The lines indicate means and shaded areas indicate $95\%$ Cls].* TABLE_PLACEHOLDER:Table 3 Fig D in S1 Text depicts secular trends of T1D in different regions, with each circle proportionate to the weight of published literature reporting incidence rates. The overall incidence of T1D appears to be rising in some regions such as Australia, Northern America, and Europe, with relatively stable estimates reported from Asia. We noted very few older, and less reliable, estimates reported from Africa and Latin America dating back to the 1980s and 1990s, which hindered estimation of meaningful secular trends.
## Discussion
The results reveal substantial variability in T1D incidence worldwide by geographic region, country income classification, and age, which ranges from over $95\%$ of cases in some regions (such as North America, Australia and Europe), to less than $35\%$ in areas of Africa and Asia. However, there is a paucity of country-level studies reporting data on incidence, especially in LMICs.
Worldwide, the incidence of T1D is increasing [16–20]. A recent study published global incidence and prevalence data of T1D using sensitivity analysis and estimated 234,710 and 9,004,610, respectively, in 2017 [21]. However, our review reveals more rigorous estimates and secular trends for each region supporting that increase in incidence and the rate of change is not uniform, with several notable nuances.
First, T1D incidence may follow a bimodal distribution pattern as regional differences are noted in peak incidence by age group. While we observed an increase in the incidence of T1D in the age categories of 5–9 years and 10–14 years, there appears to be substantial variation when peaks occur in different populations. In Sweden, for example, the highest incidence was noted in the 0–9 years [27.1 (25.6–27.4)] [22], whereas, in a 14-year longitudinal study conducted in the United States, the highest incidence occurred in the 10–14 years (45.5/ 100,000 person-years) [23]. Interestingly, in the 0–4 years category, there appears to be less variability in incidence across world regions. One explanation for this observation could be the temporal relationship between T1D triggers (environmental, seasonal, socioeconomic background, and other contextual factors) and the onset of symptoms in at-risk individuals, in addition to differences in predisposition that vary within-country populations [16, 24–26].
Second, the incidence of T1D appears to be increasing with time. A recent study analyzing three distinct age categories in 15 countries over two time periods (1975–1999 and 2000–2017), reported an increased incidence in the 0–4 year age group (1.9 times), followed by the 5–9 year age group (1.8 times) and the 10–14 year age group (1.4 times) [27]. Epidemiological, environmental, and health system factors may explain this observation. More frequent and intense triggering events, reductions in competing childhood mortality with demographic transitions, and improved health system diagnostic capabilities in certain regions, among other factors, may all be contributory.
Third, there appears to be no difference in incidence of T1D by sex. However, there may be differences in the age of peak incidence by sex. The male to female incidence ratio was 1.04 ($95\%$ CI [1.02, 1.06]), which shows a slight predominance in males. Another study reported predominance in males by age ten and persisted throughout adulthood with the male to female incidence ratio of 1.32 ($95\%$ CI [1.30–1.35]) [23]. These sex differences may be explained with exposure to certain gendered behavioral practices and susceptibility to T1D environmental triggers or innate genetic predisposition and hormonal variance. Studies of sex differences in hormonal fluctuations at the time of adrenarche in genetically predisposed individuals may provide clues to the underlying pathogenesis of T1D [28, 29].
Fourth, context matters: there is a higher incidence of T1D observed in high-income settings. The incidence and prevalence are the highest in high-income countries which constitute only $10\%$ of the world population in both the 0–14 and 0–19 year age groups [8]. The observed high incidence in these countries may be explained by health system factors and those relating to the social determinants of health. In high-income countries, more robust health systems enable more effective diagnosis of T1D and/or referral to specialists when the diagnosis is uncertain. In addition, patients that present with complications, for example, diabetes ketoacidosis (DKA), are more likely to receive timely and effective management needed for survival and diagnosis. Studies have also found that for some children an accurate T1D diagnosis is initially missed by clinicians even in some high-income settings [16, 30–33]. For example, a patient survey in the United States found that $25\%$ of all patients reported being initially misdiagnosed with another condition, which was associated with $18\%$ increased risk of DKA than those who were correctly diagnosed [34]. The potential barriers to receiving timely diagnosis as explored in the survey, included lack of a primary care provider, lack of time, lack of insurance, having a high deductible or copay, difficulty getting an appointment with a physician, and lack of transportation. In many LMICs, levels of misdiagnosis are likely higher, with studies finding that in sub-Saharan Africa most healthcare workers are not able to recognize symptoms and signs of DKA in a timely manner, which leads to high mortality [35, 36]. Indeed, many children presenting with diabetic coma in Africa are likely to be treated for more common diseases such as cerebral malaria or meningitis before the correct diagnosis is suspected [37]. It has therefore long been suspected that many children in such settings die before being correctly diagnosed [38, 39]. This helps to explain the paucity of data from many LMICs and emphasizes on the importance of accurate and timely diagnosis in these settings.
Higher incidence in high-income countries may also be explained by environmental and lifestyle or behavioral transitions in diet and physical activity associated with increases in GDP per capita. A recent study analyzing two study periods (1975–1999 and 2000–2017) found a positive correlation between T1D incidence and GDP per capita. By using a linear regression model, they suggested that for the year 1991, the country-to-country variation in GDP explained $9\%$ of the country-to-country variation in incidence rates, which was $17\%$ for the year 2006 [27]. This relationship with GDP could be explained by differences in behavior, lifestyle, and nutrition that are influenced by one’s economic position. A positive relationship exists between a country’s level of economic development as measured by GDP and obesity [40]. An elevated BMI and a sedentary lifestyle may exacerbate insulin resistance, which could lead to β-cells fatigue and triggering of an autoimmune response with resultant β-cell apoptosis [41, 42]. There exists a complex relationship between environmental factors and genetic risk for T1D or type 2 diabetes which likely plays a role in autoimmunity pathogenesis and presentation of clinical disease.
Globally, we find wide variability in incidence of T1D and higher estimates than recent estimates from the IDF Diabetes Atlas [1]. Several factors such as environmental triggers, ethnic differences, genetic susceptibility, and the ability of country health systems to diagnose new cases, are all logical explanations [6, 9, 12, 43, 44]. An association with the seasonality of onset has been reported in numerous studies in Northern Europe [45–47], among other countries [44, 48, 49]. The cyclic, sinusoidal model of T1D "seasonality" has been reported, with a peak occurring in winter in both sexes in all age groups, but is more profound in regions with more significant temperature fluctuations [50]. In Norway and Finland, the incidence of T1D has decreased among young Finnish children, with findings implying that environmental factors driving the immune system toward islet autoimmunity are changing in young children [51, 52]. Differences in HLA association may alter predisposition to T1D, despite individuals being exposed to the same trigger. Health systems differ widely in their ability to effectively screen and diagnose individuals with diabetes, in addition to their capability to address other causes of competing mortality and prolong life that might otherwise have prevented the expression of T1D.
There is insufficient data available for Latin America and the Caribbean to estimate stable trends. While the literature reports low incidence rates in these regions, the frequency of diabetes in Latin *America is* expected to increase by $38\%$ over the next ten years, compared with an estimated $14\%$ increase in their total population, with overall numbers exceeding the number of cases in the US, Canada, and Europe by 2025 [53]. Recent evidence from Brazil demonstrates a marked increase in incidence rates of T1D ($3.1\%$ annually, with an absolute crude increase of 2.5-fold for the 0–14 year age group), making up almost three-quarters of total incidence in Latin America and the Caribbean [53].
The studies from Africa reported notably higher incidence rates in the 15–19 year age group compared to the 0–14 year age group, which may reveal different patterns of incidence and drivers of T1D by age-group and region or simply that weak health systems and access to healthcare matter even more. Simulation-based analysis estimates that by 2050, the total number of new cases of T1D are projected to increase especially in Africa which will account for $51\%$ of global new cases of T1D per year and highest DKA admissions at diagnosis [54].
Further reliable and recent research from the sub-Saharan Africa region will help to refine this ratio. Africa has a rapidly growing population of children, adolescents, and young adults, and knowledge of the recent trends is crucial to develop appropriate policies and healthcare services that improve health outcomes.
Our review reveals the paucity of data available on the incidence of T1D among children and adolescents in most world regions and the need to redouble efforts on developing efficient data systems to collect, pool, and store reliable healthcare data, particularly in LMICs. While the overall impression is wide heterogeneity in the incidence rates of T1D and plateauing in the Scandinavian countries, as data systems are strengthened, the true nature of the incidence dynamics will be revealed. National population-based prospective registries provide a platform to obtain estimated T1D incidence rates. However, such initiatives are typically only sustained in high-income countries leading to a discrepancy in data and incidence estimates at the global level. Investing in data systems to capture data more efficiently, especially in Africa, where investments are needed to strengthen health systems that could inform policy and practices related to resource allocation and the development of targeted interventions to improve the effectiveness, equity, efficiency, and responsiveness of healthcare services for children and adolescents with T1D. One such example could be diagonally integrating diabetes care into the broader health system in order to improve the quality of care for T1D and also other chronic illnesses [55–57]. This could include policy prioritization, innovative financing, task-shifting, secure and sustainable access to essential medicines and diagnostic tools, and comprehensive service delivery.
There were several limitations to this study. First, we were not able to obtain information for crucial associations including socioeconomic status (for example, education level and occupational status, among other proxy measures), health-seeking behavior, ethnicity, and seasonal variability due to the incompleteness of data from all 237 included studies. When reporting the regional incidence rates, we did not control for these crucial associations. Second, inclusion of both population-based and cohort studies could serve as a bias and may not truly represent the population characteristics in each country. In addition, studies that were published in non-English languages such as Spanish, Russian and French were also not included in our investigation due to time constraints and the non-availability of an English translation. Third, the low accuracy of decrease in incidence trends with age beyond 14 years was not extensively explored due to the paucity of data available for the age category of 15–19 years. Finally, this study was unable to obtain data on the trends of DKA and mortality rates which warrants further investigations in future reviews.
Notwithstanding these limitations, this first global scoping review reveals the substantial variations in the incidence rates of T1D worldwide by region, country income group, and age category. There is a substantial paucity of data from LMICs and comparatively overwhelming evidence from high-income countries. A more accurate and holistic picture of the global burden of T1D is critical to inform health policies to strengthen health systems and improve access to effective, efficient, equitable and responsive healthcare services for children and adolescents and improve health outcomes.
## References
1. 1IDF Diabetes Atlas 2021 | IDF Diabetes Atlas. International Diabetes Federation. IDF Diabetes Atlas, 10th edn. Brussels, Belgium: 2021. https://www.diabetesatlas.org. Accessed January 3, 2022. https://diabetesatlas.org/atlas/tenth-edition/
2. Saeedi P, Petersohn I, Salpea P. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from the International Diabetes Federation Diabetes Atlas, 9th edition**. (2019.0) **157** 107843. DOI: 10.1016/j.diabres.2019.107843
3. Bommer C, Heesemann E, Sagalova V. **The global economic burden of diabetes in adults aged 20–79 years: a cost-of-illness study**. (2017.0) **5** 423-430. DOI: 10.1016/S2213-8587(17)30097-9
4. 4Global Economic Burden of Diabetes in Adults: Projections From 2015 to 2030 | Diabetes Care. Accessed August 26, 2021. https://care.diabetesjournals.org/content/41/5/963
5. Ogle GD, Middlehurst AC, Silink M. **The IDF Life for a Child Program Index of diabetes care for children and youth**. (2016.0) **17** 374-384. DOI: 10.1111/pedi.12296
6. **Incidence and trends of childhood Type 1 diabetes worldwide 1990–1999**. (2006.0) **23** 857-866. DOI: 10.1111/j.1464-5491.2006.01925.x
7. Gale E a. M.. **Spring harvest? Reflections on the rise of type 1 diabetes**. (2005.0) **48** 2445-2450. DOI: 10.1007/s00125-005-0028-z
8. Patterson CC, Karuranga S, Salpea P. **Worldwide estimates of incidence, prevalence and mortality of type 1 diabetes in children and adolescents: Results from the International Diabetes Federation Diabetes Atlas, 9th edition**. (2019.0) **157** 107842. DOI: 10.1016/j.diabres.2019.107842
9. Tuomilehto J, Ogle GD, Lund-Blix NA, Stene LC. **Update on Worldwide Trends in Occurrence of Childhood Type 1 Diabetes in 2020**. (2020.0) **17** 198-209. DOI: 10.17458/per.vol17.2020.tol.epidemiologychildtype1diabetes
10. Soltesz G, Patterson CC, Dahlquist G. **Worldwide childhood type 1 diabetes incidence—what can we learn from epidemiology?**. *Pediatric Diabetes* (2007.0) **8** 6-14. DOI: 10.1111/j.1399-5448.2007.00280.x
11. **Variation and trends in incidence of childhood diabetes in Europe**. (2000.0) **355** 873-876. PMID: 10752702
12. Green A, Gale EA, Patterson CC. **Incidence of childhood-onset insulin-dependent diabetes mellitus: the EURODIAB ACE Study**. (1992.0) **339** 905-909. DOI: 10.1016/0140-6736(92)90938-y
13. Levac D, Colquhoun H, O’Brien KK. **Scoping studies: advancing the methodology**. (2010.0) **5** 69. DOI: 10.1186/1748-5908-5-69
14. Duan N.. **Smearing Estimate: A Nonparametric Retransformation Method**. (1983.0) **78** 605-610. DOI: 10.2307/2288126
15. 15World Bank Country and Lending Groups–World Bank Data Help Desk. Accessed August 25, 2021. https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups
16. Maahs DM, West NA, Lawrence JM, Mayer-Davis EJ. **Chapter 1: Epidemiology of Type 1 Diabetes**. (2010.0) **39** 481-497. DOI: 10.1016/j.ecl.2010.05.011
17. Abdullah MA. **Epidemiology of type I diabetes mellitus among Arab children**. (2005.0) **26** 911-917. PMID: 15983673
18. Barat P, Lévy-Marchal C. **Epidemiology of diabetes mellitus in childhood**. (2013.0) **20** S110-6. DOI: 10.1016/s0929-693x(13)71424-6
19. Vandewalle CL, Coeckelberghs MI, De Leeuw IH. **Epidemiology, clinical aspects, and biology of IDDM patients under age 40 years. Comparison of data from Antwerp with complete ascertainment with data from Belgium with 40% ascertainment**. *The Belgian Diabetes Registry.* (1997.0) **20** 1556-1561. DOI: 10.2337/diacare.20.10.1556
20. Weets I, De Leeuw IH, Du Caju MV. **The incidence of type 1 diabetes in the age group 0–39 years has not increased in Antwerp (Belgium) between 1989 and 2000: evidence for earlier disease manifestation**. (2002.0) **25** 840-846. DOI: 10.2337/diacare.25.5.840
21. Green A, Hede SM, Patterson CC. **Type 1 diabetes in 2017: global estimates of incident and prevalent cases in children and adults**. (2021.0) **64** 2741-2750. DOI: 10.1007/s00125-021-05571-8
22. Thunander M, Petersson C, Jonzon K. **Incidence of type 1 and type 2 diabetes in adults and children in Kronoberg, Sweden**. (2008.0) **82** 247-255. DOI: 10.1016/j.diabres.2008.07.022
23. Rogers MAM, Kim C, Banerjee T, Lee JM. **Fluctuations in the incidence of type 1 diabetes in the United States from 2001 to 2015: a longitudinal study**. (2017.0) **15** 199. DOI: 10.1186/s12916-017-0958-6
24. Alghamdi RAI, Al-Agha AE, Banasser AM. **Environmental factors predisposing to type 1 diabetes mellitus in children: A descriptive study of various pediatric endocrine clinics in Jeddah, Saudi Arabia**. (2018.0) **22** 201-206
25. Rewers M, Ludvigsson J. **Environmental risk factors for type 1 diabetes**. (2016.0) **387** 2340-2348. DOI: 10.1016/S0140-6736(16)30507-4
26. Chiang JL, Maahs DM, Garvey KC. **Type 1 Diabetes in Children and Adolescents: A Position Statement by the American Diabetes Association**. (2018.0) **41** 2026-2044. DOI: 10.2337/dci18-0023
27. Gomez-Lopera N, Pineda-Trujillo N, Diaz-Valencia PA. **Correlating the global increase in type 1 diabetes incidence across age groups with national economic prosperity: A systematic review**. (2019.0) **10** 560-580. DOI: 10.4239/wjd.v10.i12.560
28. Tramunt B, Smati S, Grandgeorge N. **Sex differences in metabolic regulation and diabetes susceptibility**. (2020.0) **63** 453-461. DOI: 10.1007/s00125-019-05040-3
29. Benitez-Aguirre P, Craig ME, Cass HG. **Sex differences in retinal microvasculature through puberty in type 1 diabetes: are girls at greater risk of diabetic microvascular complications?**. (2014.0) **56** 571-577. DOI: 10.1167/iovs.14-15147
30. Lee JJ, Thompson MJ, Usher-Smith JA, Koshiaris C, Van den Bruel A. **Opportunities for earlier diagnosis of type 1 diabetes in children: A case-control study using routinely collected primary care records**. (2018.0) **12** 254-264. DOI: 10.1016/j.pcd.2018.02.002
31. Baldelli L, Flitter B, Pyle L. **A Survey of Youth with New Onset Type 1 Diabetes: Opportunities to Reduce Diabetic Ketoacidosis**. (2017.0) **18** 547-552. DOI: 10.1111/pedi.12455
32. Lokulo-Sodipe K, Moon RJ, Edge JA, Davies JH. **Identifying targets to reduce the incidence of diabetic ketoacidosis at diagnosis of type 1 diabetes in the UK**. (2014.0) **99** 438-442. DOI: 10.1136/archdischild-2013-304818
33. Townson J, Cannings-John R, Francis N, Thayer D, Gregory JW. **Presentation to primary care during the prodrome of type 1 diabetes in childhood: A case-control study using record data linkage**. (2019.0) **20** 330-338. DOI: 10.1111/pedi.12829
34. Muñoz C, Floreen A, Garey C. **Misdiagnosis and Diabetic Ketoacidosis at Diagnosis of Type 1 Diabetes: Patient and Caregiver Perspectives**. (2019.0) **37** 276-281. DOI: 10.2337/cd18-0088
35. Majaliwa EJ, Mohn A, Chiavaroli V, Ramaiya KK, Swai ABM, Chiarelli F. **Managment of diabetic ketoacidosis in children and adolescents in sub-Saharan Africa: A review**. (2010.0) **87** 167-173. DOI: 10.4314/eamj.v87i4.62207
36. Monabeka HG, Mbika-Cardorelle A, Moyen G. **Ketoacidosis in children and teenagers in Congo**. (2003.0) **13** 139-141. PMID: 14693472
37. Rwiza HT, Swai AB, McLarty DG. **Failure to diagnose diabetic ketoacidosis in Tanzania**. (1986.0) **3** 181-183. DOI: 10.1111/j.1464-5491.1986.tb00738.x
38. Ward ZJ, Yeh JM, Bhakta N, Frazier AL, Atun R. **Estimating the total incidence of global childhood cancer: a simulation-based analysis**. (2019.0) **20** 483-493. DOI: 10.1016/S1470-2045(18)30909-4
39. Atun R, Bhakta N, Denburg A. **Sustainable care for children with cancer: a Lancet Oncology Commission**. (2020.0) **21** e185-e224. DOI: 10.1016/S1470-2045(20)30022-X
40. Egger G, Swinburn B, Islam FMA. **Economic growth and obesity: an interesting relationship with world-wide implications**. (2012.0) **10** 147-153. DOI: 10.1016/j.ehb.2012.01.002
41. Cerf ME. **Beta Cell Dysfunction and Insulin Resistance**. *Front Endocrinol (Lausanne)* (2013.0) **4** 37. DOI: 10.3389/fendo.2013.00037
42. March CA, Becker DJ, Libman IM. **Nutrition and Obesity in the Pathogenesis of Youth-Onset Type 1 Diabetes and Its Complications**. *Front Endocrinol (Lausanne)* (2021.0) **12** 622901. DOI: 10.3389/fendo.2021.622901
43. Abellana R, Ascaso C, Carrasco JL, Castell C, Tresserras R. **Geographical variability of the incidence of Type 1 diabetes in subjects younger than 30 years in Catalonia, Spain**. *Medicina Clínica* (2009.0) **132** 454-458. DOI: 10.1016/j.medcli.2008.10.042
44. Vlad A, Serban V, Green A. **Time Trends, Regional Variability and Seasonality Regarding the Incidence of Type 1 Diabetes Mellitus in Romanian Children Aged 0–14 Years, Between 1996 and 2015**. (2018.0) **10** 92-99. DOI: 10.4274/jcrpe.5456
45. Karvonen M, Tuomilehto J, Virtala E. **Seasonality in the clinical onset of insulin-dependent diabetes mellitus in Finnish children. Childhood Diabetes in Finland (DiMe) Study Group**. (1996.0) **143** 167-176. DOI: 10.1093/oxfordjournals.aje.a008726
46. Fichera G, Arpi ML, Squatrito S, Purrello F, Ashkenazi I, Laron Z. **Seasonality of month of birth of children (0–14 years old) with type 1 diabetes mellitus in the district of Catania, Sicily**. *Journal of Pediatric Endocrinology and Metabolism* (2001.0) **14** 95-96. DOI: 10.1515/jpem.2001.14.1.95
47. Karvonen M, Jäntti V, Muntoni S. **Comparison of the seasonal pattern in the clinical onset of IDDM in Finland and Sardinia**. (1998.0) **21** 1101-1109. DOI: 10.2337/diacare.21.7.1101
48. Adar A, Shalitin S, Eyal O. **Birth during the moderate weather seasons is associated with early onset of type 1 diabetes in the**. *Mediterranean area.* (2020.0) **36**. DOI: 10.1002/dmrr.3318
49. Moltchanova EV, Schreier N, Lammi N, Karvonen M. **Seasonal variation of diagnosis of Type 1 diabetes mellitus in children worldwide**. (2009.0) **26** 673-678. DOI: 10.1111/j.1464-5491.2009.02743.x
50. Lévy-Marchal C, Patterson CC, Green A. **Geographical variation of presentation at diagnosis of type I diabetes in children: The EURODIAB study**. (2001.0) **44** B75-B80. DOI: 10.1007/pl00002958
51. Parviainen A, But A, Siljander H, Knip M. **Decreased Incidence of Type 1 Diabetes in Young Finnish Children**. (2020.0) **43** 2953-2958. DOI: 10.2337/dc20-0604
52. Aamodt G, Stene LC, Njølstad PR, Søvik O, Joner G. **Spatiotemporal trends and age-period-cohort modeling of the incidence of type 1 diabetes among children aged <15 years in Norway 1973–1982 and 1989–2003**. (2007.0) **30** 884-889. DOI: 10.2337/dc06-1568
53. Aschner P.. **Diabetes trends in Latin America**. (2002.0) **18** S27-31. DOI: 10.1002/dmrr.280
54. Ward ZJ. **Estimating the total incidence of type 1 diabetes in children and adolescents 0–19 from 1990–2050: a global simulation-based analysis**
55. Atun R, Davies JI, Gale EAM. **Diabetes in sub-Saharan Africa: from clinical care to health policy**. (2017.0) **5** 622-667. DOI: 10.1016/S2213-8587(17)30181-X
56. Knaul FM, Bhadelia A, Atun R, Frenk J. **Achieving Effective Universal Health Coverage And Diagonal Approaches To Care For Chronic Illnesses**. (2015.0) **34** 1514-1522. DOI: 10.1377/hlthaff.2015.0514
57. Atun R, Jaffar S, Nishtar S. **Improving responsiveness of health systems to non-communicable diseases**. (2013.0) **381** 690-697. DOI: 10.1016/S0140-6736(13)60063-X
|
---
title: Non-adherence to the World Health Organization’s physical activity recommendations
and associated factors among healthy adults in urban centers of Southwest Ethiopia
authors:
- Sabit Zenu
- Endegena Abebe
- Mohammed Reshad
- Yohannes Dessie
- Rukiya Debalke
- Tsegaye Berkessa
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021407
doi: 10.1371/journal.pgph.0001451
license: CC BY 4.0
---
# Non-adherence to the World Health Organization’s physical activity recommendations and associated factors among healthy adults in urban centers of Southwest Ethiopia
## Abstract
Physical inactivity is a major risk-factor of non-communicable diseases. The World Health Organization has set physical activity recommendations for adults to reduce physical inactivity and its consequences. However, 1.4 billion adults are non-adherent to the recommendation worldwide. The prevalence of non-adherence to this recommendation and its predictors has not been assessed in urban Ethiopia. This study aimed to determine the prevalence of non-adherence to physical activity recommendations and identify its associated factors among healthy adults in urban centers of Southwest Ethiopia. A community-based cross-sectional study was employed from May to June 2021, involving 1191 adults in urban centers of Southwest Ethiopia. Data was collected using Global Physical Activity Questionnaire. Multivariable logistic regression was used to identify factors associated with non-adherence to physical activity recommendations using $95\%$ confidence interval of adjusted odds ratio at P-value of < 0.05.Overall, $61.2\%$ of participants were non-adherent to physical activity recommendations. Older age (AOR = 6.6; $95\%$CI (2.3–19)), female sex (AOR = 6.1; $95\%$CI (3.5–10.5)), lower educational status (AOR = 0.5; $95\%$CI (.28–0.93)), less community engagement (AOR = 2.7;$95\%$ CI (1.3–5.5)), lower level of happiness (AOR = 4.7; $95\%$CI (1.3–16.8)) and physical inactivity of family members (AOR = 2.5; $95\%$CI (1.4–4.3)) were associated with non-adherence. The prevalence of non-adherence to physical activity recommendations in the study area is high. Age, sex, educational status, community engagement, level of happiness and physical inactivity of family members were predictors of non-adherence to the recommendations. Interventions have to target females and older adults. Community participation and family based physical activity have to be advocated to avert the consequences of physical inactivity.
## Introduction
Physical inactivity (PI) is a major behavioral risk factor of non-communicable diseases (NCDs), and it is the fourth leading cause of death worldwide. PI is raising around the world posing serious threat for people’s health by contributing to the increasing prevalence of NCDs. PI is thought to be the primary cause of 21–25 percent of breast and colon cancer, 27 percent of diabetes, and 30 percent of ischemic heart disease [1, 2]. Regular physical activity (PA) helps to minimize the risk of NCDs [1, 3]. In addition, it reduces symptoms of depression and anxiety; enhances thinking, learning, and judgment skills. Furthermore, PA improves energy balance, and aids in weight management, which helps to combat the global NCD pandemic [2].
The World Health Organization (WHO) has issued global physical activity recommendation (PAR) in 2010 and updated it with further emphasis in 2020. The organization recommends healthy adults to perform at least 150 minutes of moderate, or 75 minutes of vigorous intensity aerobic physical activity throughout a week; or to do an equivalent combination of moderate and vigorous activities. The recommendation can also be met if adults accumulate at least 600 metabolic equivalents in a week. The organization has called on member states to put special emphasis on creating suitable and healthy transportation, and suitable environments that promote physical activity. In 2018, WHO launched a new Global Action Plan on Physical Activity which outlined four policy action areas, 20 specific policy recommendations and actions for various stakeholders, to increase PA worldwide [3–5]. The organization has also introduced ACTIVE toolkit in 2019 that provides specific technical guidance on how to start and implement the 20 policy recommendations outlined in the global action plan [6].
The PAR and global action plan are meant to promote health, individual wellbeing and prevention of NCDs. PI, or performing less than the recommended level of PA leads to several unwanted health consequences including the increased risk of NCDs [1, 2, 7, 8]. Despite this recommendation and initiatives, 1.4 billion adults are doing less than the recommended levels of PA. Worldwide, one in three women and one in four men do not do enough PA to stay healthy. Levels of PA are reported to be twice as high in high-income countries compared to low-income countries. Furthermore, the global levels of PA didn’t show any improvement over the last two decades. Contrarily, insufficient activity increased by $5\%$ (from $31.6\%$ to $36.8\%$) in high-income countries between 2001 and 2016 [3, 5].
The level of non-adherence to PAR among adults is also high in developing countries. In Malaysia, more than four in ten of adult population is categorized as non-adherent to PAR [9]. The magnitude stands at $34\%$ in Papua New Guinea [10] and much higher in Kerala, India where the level of non-adherence to PAR is as high as $90.3\%$ [11]. In African countries, three quarters of adult population are doing less than recommended level of PA. In Nigeria, the prevalence is as high as $77.8\%$ among university employees [12] and $69.7\%$ among civil servants [13]. The Kenyan national STEPS survey included 4066 study participants in 2015; among these, $80.3\%$ are categorized as physically inactive and are not meeting current PAR [14].
In Ethiopia, most studies assessed the problem of non-adherence to health guidelines, including PAR, among the already diseased populations. These studies assessed the level of non-adherence to PAR mainly among hypertensive and diabetic patients [15–17]. National STEPS survey in 2015 reported the level of non-adherence to PAR to be as low as $6\%$. In addition, a study conducted to assess the level of non-adherence to PAR among public servants in Northern Ethiopia reported prevalence of the problem to be $41\%$ [18, 19]. Non-adherence to PAR has negative impacts on individuals, health systems, economic development, community well-being and quality of life. Despite this, little is known on the prevalence of non-adherence to the current recommendations in urban centers of Ethiopia. In addition, advent of COVID-19 pandemic has resulted in a unique challenge to health of population including physical exercise. This study aimed to determine level of non-adherence to WHO’s physical activity recommendations and identify its associated factors among adult urban residents in Southwest Ethiopia.
## Study area and period
The study was conducted among adults in three major towns of Southwest Ethiopia; Bedelle, Mettu and Gambella. Bedelle and Mettu towns are capitals of Ilu Aba Bor and Buno Bedelle Zones in Oromia region respectively. Gambella town is the regional capital of Gambella region.
Community based cross sectional study was conducted from May to June, 2021.
## Sample size determination and sampling technique
Sample size was calculated for both objectives by using STATCAL of Epi Info-7, and the larger sample size was used. The larger sample size was obtained\ by using the prevalence of non-adherence to PAR among adults in Northern Ethiopia as $41\%$, $95\%$ CI, margin of error of 0.04, design effect of two and adding $10\%$ for non-response. This calculation resulted in sample size of 1277. Multi-stage sampling was used to reach on respondents. From the three towns, lowest administrative units (Kebeles) were selected by lottery method. After developing list of households in selected kebeles, samples were proportionally allocated to each kebeles and systematic random sampling technique was employed to reach on households depending on house numbers. From selected households, one individual was selected from eligible adults by Kish method.
## Data collection tools and procedures
PA was measured using the second version of Global Physical Activity Questionnaire (GPAQ). The participants were asked about their physical activities by using GPAQ translated to the local language, Afaan Oromo, with the help of language experts. Information on socio demographic characteristics of participants, family practices and social characteristics were collected using questionnaires from the EDHS 2015 [20]. Social capital and related variables were measured by using the integrated social capital measurement tool validated for use in developing countries [21]. Trained nurses collected the data. Show-cards were used as per the recommendation of GPAQ to indicate examples of activities categorized in the three dimensions of PA.
## Data analysis procedures
After collection, data were entered to epidata version 3.1 and exported to SPSS version 20 for analysis. The GPAQ questionnaire assessed PA in three dimensions. These dimensions are: work related PA, recreational PA and travel or transportation related PA. All dimensions were converted in to the Metabolic Equivalents. Individuals who achieved less than 600 Metabolic Equivalents in a week were categorized as non-adherers to the WHO recommendation of PA. Proportions and other descriptive statistical analysis were used to describe the data. Multivariable logistic regression was used to identify factors associated with non-adherence to PAR. Associations between non adherence and independent variables were declared by using the $95\%$ CI for the AOR at p-value of 0.05.
## Ethics statement
This research was conducted according to the principles of Declaration of Helsinki. The proposal and conduct of the study was ethically cleared by Mettu University ethics review committee. Written informed consent was taken from selected participants. All information provided by the participants was kept confidential. In addition, any information leading to identification of study participants was not included in data collection tool.
## Socio demographic characteristics of participants
Among 1277 total sample size, 1191 adults volunteered to participate in the study resulting in $93.3\%$ response rate. Most of the participants were in the age group of 30–44 years. Females accounted for more than half of the participants 644($54.0\%$), while more than seven in ten of participants were ethnic Oromo 847($71.1\%$). Orthodox Christianity 433($37.8\%$), Protestant 373($31.7\%$) and Islam 347($28.7\%$) were the predominant religions of the participants (Table 1).
**Table 1**
| S.N | Variables | Categories | Frequency | Percentages |
| --- | --- | --- | --- | --- |
| 1.0 | Age | 18–29 | 375 | 31.4% |
| 1.0 | Age | 30–44 | 452 | 37.9% |
| 1.0 | Age | 45–59 | 206 | 17.3% |
| 1.0 | Age | 60–69 | 105 | 8.8% |
| 1.0 | Age | > = 70 | 53 | 4.5% |
| 2.0 | Sex | Female | 644 | 54.% |
| 2.0 | Sex | Male | 547 | 46% |
| 3.0 | Ethnicity | Oromo | 847 | 71.2% |
| 3.0 | Ethnicity | Amhara | 116 | 9.7% |
| 3.0 | Ethnicity | Agnua | 103 | 8.6% |
| 3.0 | Ethnicity | Nuer | 91 | 7.6% |
| 3.0 | Ethnicity | Others | 34 | 2.9% |
| 4.0 | Religion | Orthodox | 433 | 36.4% |
| 4.0 | Religion | Protestant | 387 | 32.6% |
| 4.0 | Religion | Muslim | 356 | 29.9% |
| 4.0 | Religion | Catholic | 13 | 1.1% |
## Non-adherence to the WHO recommendations of physical activity
From the overall study participants, 729($61.2\%$), $95\%$ CI (56.2–66.0) failed to meet the WHO recommendation for physical activity. The level of non-adherence to PAR is higher among females when compared with males (chi square tests = 37.4; df = 1, $p \leq 0.001$). Non-adherence to PAR steadily increases as age increases from $46.2\%$ among age group 18–29 to $85\%$ among age group of 70 years and above(chi square test = 14.4; df = 4, $$p \leq 0.006$$).
## Factors associated with non-adherence to physical activity recommendations
On the final multivariable model; older age, female sex, lower educational status, not having family members who do physical exercise, not participating in community activities and poor self-reported level of happiness is associated with non-adherence to PAR (Table 2).
**Table 2**
| S.N | Variables | Variable Categories | Adherence to PA recommendations | Adherence to PA recommendations.1 | COR(95%CI) | AOR(95%CI) | p-value |
| --- | --- | --- | --- | --- | --- | --- | --- |
| S.N | Variables | Variable Categories | Non-adherent | Adherent | COR(95%CI) | AOR(95%CI) | p-value |
| 1. | Sex | Male | 240 | 297 | Reference | Reference | |
| 1. | Sex | Female | 489 | 165 | 3.6(2.4–5.6) | 6.1(3.5–10.5) | <0.001* |
| 2. | Age | 18–29 | 219 | 169 | Reference | Reference | |
| 2. | Age | 30–44 | 246 | 210 | 0.95(.6–1.5) | 1.2(0.66–2.3) | 0.499 |
| 2. | Age | 45–59 | 141 | 60 | 1.8(.98–3.4) | 2.5(1.05–5.9) | 0.036* |
| 2. | Age | 60–69 | 69 | 21 | 2.5(1.02–6.4) | 6.6(2.3–19) | 0.001* |
| 2. | Age | > = 70 | 56 | 9 | 4.6(1.3–16.6) | 6.9(1.5–32.0) | 0.014* |
| 3. | Marital Status | Married | 282 | 174 | Reference | Reference | - |
| 3. | Marital Status | Others | 447 | 288 | 0.9(.63–1.4) | - | - |
| 4. | Educational Status | Primary and Below | 318 | 180 | 1.2(0.8–1.83) | 0.51(0.3–0.9) | 0.03* |
| 4. | Educational Status | Secondary and Above | 411 | 282 | Reference | Reference | |
| 5. | Group Membership | Not member of group | 150 | 99 | .95(.57–1.5) | - | - |
| 5. | Group Membership | Member of a group | 579 | 363 | Reference | - | - |
| 6. | House condition (Crowding) | No crowding | 672 | 441 | Reference | - | - |
| 6. | House condition (Crowding) | Over crowding | 57 | 21 | 1.7(.73–4.3) | - | - |
| 7. | Wealth Index | 1st quartile | 159 | 135 | Reference | Reference | |
| 7. | Wealth Index | 2nd Quartile | 201 | 99 | 1.7(.96–3.0) | 1.7(0.8–3.5) | 0.147 |
| 7. | Wealth Index | 3rd Quartile | 204 | 96 | 1.8(1.01–3.2) | 1.24(0.6–2.6) | 0.566 |
| 7. | Wealth Index | 4th Quartile | 165 | 132 | 1.06(.6–1.8) | 0.7(0.33–1.5) | 0.36 |
| 8. | Level of Happiness | Happy | 645 | 429 | Reference | Reference | |
| 8. | Level of Happiness | Neither | 57 | 15 | 2.52(0.9–7.0) | 4.7(1.3–16.8) | 0.016* |
| 8. | Level of Happiness | Unhappy | 27 | 18 | 0.99(.34–2.8) | 0.13(0.01–0.9) | 0.049* |
| 9. | Active Family | No | 462 | 198 | 2.3(1.5–3.5) | 2.5(1.4–4.3) | 0.001* |
| 9. | Active Family | Yes | 267 | 264 | Reference | Reference | |
| 10. | Active Friends | No | 537 | 213 | 3.2(2.1–5.0) | - | - |
| 10. | Active Friends | Yes | 192 | 249 | Reference | - | - |
| 11. | Participation in Community activities | No | 207 | 63 | 2.5(1.5–4.3) | 2.7(1.3–5.5) | 0.007* |
| 11. | Participation in Community activities | Yes | 522 | 399 | Reference | Reference | |
In this study, old aged participants are nearly seven times more likely to be non-adherent to PAR when compared with younger participants; AOR = 6.6: $95\%$ CI [2.3–19]. In addition, female participants are six times as non-adherent as their male counterparts; AOR = 6.1:$95\%$ CI [3.5–10.5]. In addition, participants who are primary and below in their educational statuses have $50\%$ reduction in probability of being non-adherent to PAR; AOR = 0.51: $95\%$ CI [.28–0.93].
Participants whose none of their family members do physical exercises are nearly three times more likely to be non-adherent to PAR when compared with participants who have family members who do physical exercises; AOR = 2.5: $95\%$ CI [1.4–4.3]. Participating in community-wide activities appears to be protective against physical inactivity. In this study participants who did not participate in community activities are nearly three times non-adherent to PAR when compared with participants who participated in such activities; AOR = 2.7:$95\%$CI [1.3–5.5].
In this study, self-reported feeling of happiness is also associated with being non-adherent to PAR. Participants who reported to be unhappy in their daily life are roughly five times more non-adherent to PAR when compared with happy counterparts AOR = 4.7: CI [1.3–16.8].
## Discussion
In the present study, majority of the study participants ($61.2\%$), $95\%$ CI: (56.2–66.0) are non-adherent to PAR. Older adults and females are more non-adherent to the recommendation when compared with their counterparts.
Level of non-adherence to PAR in the study area is very high when compared with the national prevalence, which reported the prevalence of non-adherence to PAR as $6\%$. The higher prevalence from the current study may be explained by the difference in the sampled population and sample size. The current study only included urban residents while the national survey included sample of citizens from both urban and rural settings. The population distribution of *Ethiopia is* predominantly rural with agriculture as a primary source of income. Participation in farm activities demand moderate to vigorous intensity PA which may have lowered the prevalence of PI at the national level. In addition, the relatively large sample size for national survey may also have an impact on the magnitude of non-adherence [18, 22].
The prevalence of no-adherence to PAR in the study area is also higher than reports from Papua New Guinea where $34\%$ of participants are not-adherent to PAR. The current finding is also higher than a report from India among adults in Kani tribe where only $9.7\%$ are non-adherent. The relatively lower prevalence of non-adherence to the recommendation in Papua New Guinea and Kani tribe of India may be due to their lifestyle that involve rural agriculture, which tends to increase the time for physical activity [10, 11].
In contrary, the prevalence of non-adherence to PAR from the current study is lower than reports from Nigeria that showed the prevalence of non-adherence to PAR as $77.8\%$ among students and $69.7\%$ among civil servants. The relatively higher prevalence in Nigeria may be due to the type of participants included in the studies. The current study involved urban residents while the studies in Nigeria were conducted on civil servants and university students [12, 13]. The finding of the current study is also lower than results of STEPS survey in Kenya where more than $80\%$ are not adhering to the recommendation. The relatively higher level of non-adherence to PAR in Kenya may be due to its relative industrialization and digitalization when compared with Ethiopia [14].
Several factors were associated with non-adherence to PAR. These include older age, female sex, and educational status, family history of physical activity; community participation and self-reported level of happiness.
This study found that older adults are nearly seven times more likely to be non-adherent to PAR when compared with younger adults. The relative disadvantage of older adults in the physical activity is well documented in several studies. This may be due to unsuitability of their environment for regular exercise and fear of falling while doing physical exercise [23–25].
In this study, females are six times more likely to be non-adherent to PAR when compared with male participants. Several studies also reported similar findings indicating more likelihood of non-adherence to PAR among females [26–28] This may be due to less involvement of urban women and housewives on works that require moderate to vigorous intensity exercise unlike rural women who are involved in farm activities. Urban women tend to be at home or office, working less vigorous works [29]. Employment pattern of women may also pose females at risk of staying home, which in turn reduce the time used to perform physical activities. According to the Central Statistics Agency of Ethiopia, only $41\%$ of women who are expected to work are employed [22].
This study revealed that participants with lower educational achievement have reduced probability of non-adherence to PAR that reaches up to $50\%$. This finding contradicts with a report from Malaysia that reported up to $20\%$ increase in physical inactivity among participants with lower educational achievement [9]. This may be due to engagement of peoples with lower educational status in informal employment activities. Informal employment activities tend to require more moderate to vigorous intensity physical activities when compared to office based jobs. A recent study in Australia indicated that desk based working environments are not adequately equipped with facilities that encourage movement for office workers [30].
Participants without physically active family member are more likely to be non-adherent to PAR when compared with participants who have physically active family members. Families are essential components of one’s assets that cue peoples to practice healthy behaviors and refrain from something that may harm health. Studies in different parts of the world also acknowledged the importance of families on adopting a regular PA plan among school aged children [31–34]. Frequent encouragement and support from family members has also resulted in reduced leisure time sedentariness [35].
Participants who were involved in community based communal activities have lower probability of being non-adherent to PAR. Similarly, participants who do not participate in such community based activities tend to be more physically inactive. This may be due to the overall effect of social participation on health and wellbeing [36]. Community participation is found to enhance health through several means; among these is enhancement of mental health that in turn causes to take preventive measures against chronic diseases. PA is one of the essential activities to prevent the occurrence of chronic diseases [37].
Participants who self-reported to be unhappy in their current conditions are also more likely to be non-adherent to PAR when compared with happier ones. This unfavorable mental status may result in reduced intention to take proactive measures in maintaining health by performing PA. Such status is also an indicator of poor social capital, and social capital is found to be extremely important in physical activities as shown from many studies [38, 39].
Generally, this study found a high prevalence of non-adherence to PAR among urban residents in Southwest Ethiopia. This could result in consequent increase in the incidence and prevalence of NCDs in the near future, if the current trend continues. Old-aged urban residents, females, relatively educated residents are more likely to be non-adherent to the recommendation. In addition, not participating in community-wide activities and low level of perceived happiness is also associated with non-adherence to PAR.
## Limitations
This study assessed the level of adherence to PAR depending on respondent’s response to set of questions using the GPAQ. The tool fails to objectively measure the level of adherence unlike other advanced tools. In addition, the nature of questionnaire may have introduced recall bias as subjects may forget some of their physical movements.
## Conclusion
Majority of the study participants failed to meet the global recommendations for PA. This high level of physical inactivity poses the risk of future occurrence of NCDs and other health problems. Age, sex, educational status, community engagement, level of happiness and PA of family members were associated with non-adherence to PAR.
Community based interventions to raise the level of PA and boost adherence to the recommendation has to be implemented to protect the health of the community from the looming danger posed by PI. Priority has to be given to old aged adults and females. Community organizations and networks have to be used to encourage PA among members. Family based changes have to be fostered to reduce PI by using family members as role models. Strengthening social capital and social cohesion also helps to tackle the problem in the long run.
## References
1. 1World Health Organization. Global Action Plan for the Prevention and Controlof Non Communicable Diseases 2013–2020. 1st ed. Geneva; 2013.. *Global Action Plan for the Prevention and Controlof Non Communicable Diseases 2013–2020* (2013)
2. 2World Health Organization. Global Recommendations on Physical Activity for Health for age 18–64 years old. 2011. p. 01.. *Global Recommendations on Physical Activity for Health for age 18–64 years old* (2011) 01
3. 3World Health Organization. Global action plan on physical activity 2018–2030: more active people for a healthier world. Geneva; 2018.. *Global action plan on physical activity 2018–2030: more active people for a healthier world* (2018)
4. Musumeci G.. **Physical activity for health—An overview and an update of the physical activity guidelines of the italian ministry of health**. *J Funct Morphol Kinesiol* (2016) **1** 269-275. DOI: 10.3390/JFMK1030269
5. 5World Health Organization. WHO guidelines on physical activity and sedentary behaviour. Geneva: World Health Organization; 2020.. *WHO guidelines on physical activity and sedentary behaviour* (2020)
6. 6World Health Organzation. A technical package for increasing physical activity. 1st ed. Geneva, Switzerland: WHO; 2016.. *A technical package for increasing physical activity* (2016)
7. Abbafati C, Abbas KM, Abbasi-Kangevari M, Abad-Allah F, Abdelalim A. **Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet* (2020) **396** 1223-1249. DOI: 10.1016/S0140-6736(20)30752-2
8. Abbafati C, Abbas KM, Abbasi-Kangevari M, Abd-Allah F, Abdelalim A, Abdollahi M. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet* (2020) **396** 1204-1222. DOI: 10.1016/S0140-6736(20)30925-9
9. Ying C, Kuay LK, Huey TC, Hock LK. **Prevalence and factors associated with physical inactivity among Malaysian adults**. *Southeast asian J troP med Public Heal* (2014) **45** 467-479. PMID: 24968689
10. Rarau P, Vengiau G, Gouda H, Phuanukoonon S, Kevau IH, Bullen C. **Prevalence of non-communicable disease risk factors in three sites across Papua New Guinea: a cross- sectional study**. *BMJ Glob Heal* (2017) **2** 1-14. DOI: 10.1136/bmjgh-2016-000221
11. Sajeev P, Soman B. **Prevalence of noncommunicable disease risk factors among the Kani tribe in Thiruvananthapuram district, Kerala**. *Indian Heart J* (2018) **70** 598-603. DOI: 10.1016/j.ihj.2018.01.022
12. Agaba EI, Akanbi MO, Agaba PA, Ocheke AN, Gimba ZM. **A survey of non-communicable diseases and their risk factors among university employees: a single institutional study**. *Cardiovasc J Afr* (2017) **28** 377-384. DOI: 10.5830/CVJA-2017-021
13. Olawuyi AT, Adeoye IA. **The prevalence and associated factors of non- communicable disease risk factors among civil servants in Ibadan, Nigeria**. *PLoS One* (2018) **13** 1-19. DOI: 10.1371/journal.pone.0203587
14. Wekessah FM, Nyanjau L, Kibachio J, Mutua MK, Mohammed SF, Grobbee DE. **Individual and household level factors associated with presence of multiple non- communicable disease risk factors in Kenyan adults**. *BMC Public Health* (2018) **18** 41-53
15. Bonger Z, Shiferaw S, Tariku EZ. **Adherence to diabetic self-care practices and its associated factors among patients with type 2 diabetes in addis Ababa, Ethiopia**. *Patient Prefer Adherence* (2018) **12** 963-970. DOI: 10.2147/PPA.S156043
16. Edmealem A, Ademe S, Tegegne B. **Level of physical activity and its associated factors among type ii diabetes patients in dessie referral hospital, northeast ethiopia**. *Diabetes, Metab Syndr Obes Targets Ther* (2020) **13** 4067-4075. DOI: 10.2147/DMSO.S279772
17. Negrea ZG ED. **Prevalence and predictors of nonadherence to diet and physical activity recommendations among type 2 diabetes patients in Southwest Ethiopia: a cross-sectional study**. *Int J Endocrinol* (2020) **2020** 1-8. DOI: 10.1155/2020/1512376
18. 18Ethiopian Public Health Institute. Ethiopia STEPS report on risk factors for non- communicable diseases and prevalence of selected NCDs in Ethiopia. 1st ed. Addis Ababa: EPHI; 2016.. *Ethiopia STEPS report on risk factors for non- communicable diseases and prevalence of selected NCDs in Ethiopia* (2016)
19. Gebremariam LW, Chi C, Yatsuya H, Haregot E. **Non-communicable disease risk factor profile among public employees in a regional city in northern Ethiopia**. *Sci Rep* (2018) **8** 1-11. DOI: 10.1038/s41598-018-27519-6
20. 20Central Statistical Agency. Ethiopian Demographic and Health survey. Addis Ababa: CSA; 2016.. *Ethiopian Demographic and Health survey* (2016)
21. Grootaert C, Narayan D, Jones VN, Woolcock M. *Measuring social capital an integrated questionnaire* (2016)
22. 22Central Statstical Agency. The Federal Democratic Republic of Ethiopia, Central Statistical Agency: key findings on the 2012 urban unemployment survey. 1st ed. Addis Ababa: CSA; 2012.. *The Federal Democratic Republic of Ethiopia, Central Statistical Agency: key findings on the 2012 urban unemployment survey* (2012)
23. Tappen R, Vieira ER, Gropper SS, Newman D, Horne C. **Sustaining or declining physical activity: Reports from an ethnically diverse sample of older adults**. *Geriatr* (2021) 6. DOI: 10.3390/geriatrics6020057
24. McPhee JS, French DP, Jackson D, Nazroo J, Pendleton N, Degens H. **Physical activity in older age: perspectives for healthy ageing and frailty**. *Biogerontology* (2016) **17** 567-580. DOI: 10.1007/s10522-016-9641-0
25. Crombie IK, Irvine L, Williams B, McGinnis AR, Slane PW, Alder EM. **Why older people do not participate in leisure time physical activity: a survey of activity levels, beliefs and deterrents**. *Age Ageing* (2004) **33** 287-292. DOI: 10.1093/ageing/afh089
26. Aniza I, Health M, Fairuz MR, Health M. **Factors influencing physical activity level among secondary school adolescents in Petaling district, Selangor**. *Med J Malaysia* (2009) **64** 228-232. PMID: 20527274
27. Wickel EE, Eisenmann JC. **Within- and between-individual variability in estimated energy expenditure and habitual physical activity among young adults**. *Eur J Clin Nutr* (2006) **60** 538-544. DOI: 10.1038/sj.ejcn.1602348
28. Hou X, Liu JM, Tang ZY, Ruan B, Cao XY. **The gender difference in association between home-based environment and different physical behaviors of chinese adolescents**. *Int J Environ Res Public Health* (2020) **17** 1-15. DOI: 10.3390/ijerph17218120
29. Kotikula A, Hill R RW. *What works for working women,understanding female labor force participation in urban Bangladesh* (2019)
30. Hadgraft N, Winkler E, Goode AD, Gunning L, Dunstan DW, Owen N. **How supportive are workplace environments for sitting less and moving more? A descriptive study of Australian workplaces participating in the BeUpstanding program**. *Prev Med Reports* (2021) **24** 101616. DOI: 10.1016/j.pmedr.2021.101616
31. Brown HE, Schiff A, Slujis EM. **Engaging families in physical activity research: a family-based focus group study**. *BMC Public Health* (2015) **15** 1-8. DOI: 10.1186/s12889-015-2497-4
32. Cleland V, Timperio A, Salmon J, Hume C, Telford A, Crawford D. **A longitudinal study of the family physical activity environment and physical activity among youth**. *Am J Heal Promot* (2011) **25** 159-167. DOI: 10.4278/ajhp.090303-QUAN-93
33. Al Yazeedi B, Berry DC, Crandell J, Waly M. **Family influence on children’s nutrition and physical activity patterns in Oman**. *J Pediatr Nurs* (2021) **56** e42-e48. DOI: 10.1016/j.pedn.2020.07.012
34. Xu H, Wen LM, Rissel C. **Associations of parental influences with physical activity and screen time among young children: A systematic review**. *J Obes* (2015) **2015** 1-23. DOI: 10.1155/2015/546925
35. Wang X, Liu QM, Ren YJ, Li LM. **Family influences on physical activity and sedentary behaviours in Chinese junior high school students: a cross-sectional study**. *BMC Public Health* (2015) **15** 1-9. DOI: 10.1186/s12889-015-1593-9
36. Tomioka K, Kurumatani N, Hosoi H. **Positive and negative influences of social participation on physical and mental health among community-dwelling elderly aged 65–70 years: a cross-sectional study in**. *BMC Geriatr* (2017) **17** 1-13. DOI: 10.1186/s12877-017-0502-8
37. Santini ZI, Jose PE, Koyanagi A, Meilstrup C, Nielsen L. **Formal social participation protects physical health through enhanced mental health: A longitudinal mediation analysis using three consecutive waves of the Survey of Health, Ageing and Retirement in Europe (SHARE)**. *Soc Sci Med* (2020) **251** 112906. DOI: 10.1016/j.socscimed.2020.112906
38. Fu C, Wang C, Yang F, Cui D, Wang Q, Mao Z. **Association between social capital and physical activity among community-dwelling elderly in Wuhan, China**. *Int J Gerontoogy* (2018) **12** 155-159
39. Richardson AS, Troxel WM, Ghosh-Dastidar GP, Hunter R, Beckman N. **Pathways through which higher neighborhood crime is longitudinally associated with greater body mass index**. *Int J Behav Nutr Phys Act* (2017) **14** 1-10. DOI: 10.1186/s12966-017-0611-y
|
---
title: 'Ramadan fasting and weight change trajectories: Time-varying association of
weight during and after Ramadan in low-income and refugee populations'
authors:
- Daniel E. Zoughbie
- Tin Lok James Ng
- Jacqueline Y. Thompson
- Kathleen T. Watson
- Rami Farraj
- Eric L. Ding
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021413
doi: 10.1371/journal.pgph.0000371
license: CC BY 4.0
---
# Ramadan fasting and weight change trajectories: Time-varying association of weight during and after Ramadan in low-income and refugee populations
## Abstract
Obesity is a significant driver of the global burden of non-communicable diseases. Fasting is one approach that has been shown to improve health outcomes. However, the effects of Ramadan fasting differ in that the type, frequency, quantity, and time of food consumption vary. This phenomenon requires in-depth evaluation considering that $90\%$ of Muslims (~2 billion people) fast during Ramadan. To address this issue, we evaluated the pattern of weight change during and following Ramadan for a total of 52 weeks. The study was conducted in Amman, Jordan. Between 2012 and 2015, 913 participants were recruited as part of a trial investigating the efficacy of a weight loss intervention among those with or at risk for diabetes. Weight was measured weekly starting at the beginning of Ramadan, and changes were analyzed using discrete and spline models adjusted for age, sex, and trial group. Results show slight weight gain within the first two weeks and weight loss in the subsequent weeks. During the first week of Ramadan, the estimate for a weight reduction was 0·427 kg, ($95\%$ CI: -0·007, 0·861), increasing to 1·567 kg, ($95\%$ CI: 2·547, 3·527) at week 26. There was clear evidence of gradual weight gain from about 4 to 15 weeks and a drop towards the end of the investigation at week 28 (-0·12kg, $95\%$ CI: -0·89, 0·56). Our results show that weight changes occurred during and after Ramadan. Weight fluctuations may affect health risks, and thus, findings from this study can inform interventions. Public health agencies could leverage this period of dietary change to sustain some of the benefits of fasting. The authors (DEZ, EFD) acknowledge the Mulago Foundation, the Horace W. Goldsmith Foundation, Robert Wood Johnson Foundation, and the World Diabetes Foundation. TRIAL REGISTRATION. Clinicaltrials.gov registry identifier: NCT01596244.
## Introduction
As of 2017, approximately 1·8 billion people worldwide observed the Muslim faith [1], and $90\%$ fast during Ramadan [2, 3]. The practise of abstaining from food, water, and other activities is a pillar of the *Islamic religious* praxis with enormous public health implications that are not yet well understood [4, 5]. Therefore, this period of religious observance represents a large-scale dietary and lifestyle intervention that affects approximately $20\%$ of the world’s population each year [3, 4].
Fasting and intermittent fasting, which generally does not include abstention from water, have been shown to improve health outcomes in chronic diseases [6–8]. By contrast, the long-term health benefits or risks of Ramadan fasting are unclear [5, 9–14]. In observant populations, general practices of Ramadan may include an iftar meal to break one’s fast after the sun sets and suhoor, a light pre-dawn meal [15]. Ramadan meals may be calorie-dense, water abstention in warm climates may result in dehydration, overeating may occur before bed, and regular sleep patterns may be disturbed [16]. Therefore, the physiological changes associated with Ramadan fasting differ from general, intermittent fasting [17]; the former may have unidentified effects when body functions are disrupted [5, 10].
One driver of the uncertainty concerning Ramadan fasting is its dynamic nature. Since the Ramadan fast lasts from dawn until dusk, fasting from food and water varies significantly depending on geographical location and daylight hours. Unlike fixed-date holidays like Christmas, the Ramadan fasting period is based on the lunar calendar and may occur during the Fall, Winter, Spring, or Summer seasons [2]. Socio-economic status and culture also affect fasting practices: wealthy populations may have the flexibility to reverse their work schedules, sleeping more during the day and eating late into the night [18], whereas less privileged individuals may experience malnourishment during this period [19].
In addition to the diversity of experiences within the defined month of Ramadan, its long-term effects remain unexplored. The implications of dietary and lifestyle changes that persist after the fasting period are especially salient in places like the Middle East, where the rates of obesity and diabetes are rising [20, 21]. In Jordan, the prevalence of diabetes increased from $14\%$ in 1990 to $16\%$ in 2020 [22] and is expected to rise even further [20, 22, 23]. To address this growing burden of non-communicable diseases, government agencies must understand how best to fine-tune population-level diabetes interventions to account for large numbers of diabetic individuals who fast.
Existing strategies [24] aimed at the Ramadan fasting period, such as structured nutrition for diabetes [25], dietary interventions [26], and micro-clinic social network programs [27–30], have demonstrated promising results. However, these studies have not yet elucidated specific patterns of weight change during and after Ramadan that can be enacted upon by public health institutions. This study aims to address this gap by providing preliminary findings on the association between the practise of Ramadan and patterns of weight change over a longitudinal period.
## Data collection
This study is a secondary analysis using information from all participants who participated in a clinical trial between 2012 and 2015. Briefly, 1025 volunteers were screened using the study eligibility criteria. 913 participants were recruited between 2012–2013 from three community health centers in Amman, Jordan, as part of a multicenter, 3-arm randomized controlled trial (NCT01596244). Participants in Arm A received “Full Microclinic program (MCP)” with curriculum-activated social network interactions. Participants in Arm B received “Basic MCP” educational sessions. While those in Arm C, the Control Arm, received standard monitoring and care. The primary study evaluated the effectiveness of a 6-month Microclinic Program (MCP) on diabetes management, behavioral risk factors, weight, and metabolic outcomes. Participants in the trial were followed for two years. Men and women 18 years or older were eligible to participate in the trial if they had diabetes or were at risk of diabetes. The study intervention was administered after nurses or study coordinators received signed consent forms from participants. Weight was measured in kilograms using a hospital and homecare scale(Health Scale SVR 160) following a standardized study protocol administered by nurses and study coordinators. The protocol was approved by Western IRB (USA) and the Jordanian National Center for Diabetes, Endocrinology, and Genetics. Details of the study participants are in Table 1, and further information on the trial procedure is in S1 Text [Zoughbie et al., Under review] [31].
**Table 1**
| Characteristic, mean (SD) or % | Full MCP | Basic MCP | Control Group |
| --- | --- | --- | --- |
| Characteristic, mean (SD) or % | (n = 540) | (n = 185) | (n = 188) |
| Age, years | 54 · 2 | 56 · 6 | 56 · 2 |
| Women, (%) | 66 · 5 | 67 · 0 | 65 · 0 |
| Weight, kg | 85 · 9 | 85 · 0 | 86 · 0 |
| Height, m | 160 · 3 | 159 · 6 | 160 · 3 |
| BMI, kg/m2 | 33 · 6 | 33 · 5 | 33 · 4 |
| Systolic blood pressure, mm Hg | 129 · 1 | 131 · 7 | 132 · 2 |
| Diastolic blood pressure, mm Hg | 81 · 3 | 81 · 0 | 81 · 3 |
| Mean arterial pressure, mm Hg | 97 · 2 | 97 · 9 | 98 · 3 |
| HbA1c (%) (SD) | 6 · 91 | 6 · 90 | 6 · 91 |
| Fasting plasma glucose*, mg/dL | 147 · 8 | 142 · 8 | 145 · 7 |
## Statistical analysis
Descriptive summaries of participant characteristics and outcome measures are presented using means with standard deviations. We combined data for this secondary data analysis and analyzed participants in all study arms as a single cohort. All statistical analyses were performed using R version 3.6.3 and reported following relevant guidelines [32]. We analyzed the time-varying relationship between Ramadan and weight using two multi-level mixed effect modelling approaches–one discrete and one continuous. For both models, we assumed that the relationship between Ramadan and weight depends on the number of weeks since the start of Ramadan. The discrete version of the multi-level model assumes that Ramadan’s relationship with weight changes stepwise with a jump at the end of each week. In contrast, the continuous version of the multi-level model treats time as a continuous variable and models the relationship between Ramadan fasting with weight using a linear spline. The linear spline model assumes a piecewise linear relationship between two adjacent knots. The resulting spline function is continuous with a change in slope at each knot location which captures the change in weight trajectory. The number of knots and the locations of knots for the spline method were determined based on the results from the discrete model. We further controlled for the following covariates—age, sex and trial arms in the model. Sensitivity analysis was performed to investigate the consistency of the results with respect to small changes in the knot locations.
The discrete multi-level model was further extended to allow the possibility that the Ramadan relationship with weight depends on risk factors. In particular, we incorporated the interaction between sex and the number of weeks since the start of Ramadan in the multi-level model to allow the possibility that the relationship between Ramadan and weight depends on sex. Furthermore, the interaction between diabetes and the number of weeks from the onset of Ramadan was added to the multi-level model, allowing the change in weight to depend on whether a participant has diabetes or is at risk of diabetes. We also controlled for the following covariates—age, sex and trial arms in the model.
## Results
Baseline characteristics of the study participants are presented in Table 1. Further details are in S2 Text. The estimated time-varying effects of Ramadan on weight are shown in Fig 1. We observe a small amount of weight gain in the first two weeks of Ramadan, followed by a gradual decline in weight in the subsequent weeks. By the second week of Ramadan, weight gain was 0·53kg ($95\%$ CI: 0·06, 1·01), compared to a weight loss of 0·55kg ($95\%$ CI: 0·05, 1·05) by the 8th week, or one month after Ramadan ended. There was gradual weight gain from the 8th week, which continued over the next 18 weeks with an estimated weight gain of 2·54kg ($95\%$ CI: 1·57, 3·53) by week 26. A sharp drop of 2·66kg in weight was observed between the 26th (2·54kg, $95\%$ CI: 1·56, 3·53) and the 28th week (-0·12kg, $95\%$ CI: -0·89, 0·56) before it stabilized with average weight returning to similar to baseline values.
**Fig 1:** *Estimated time-varying effects of Ramadan on weight using the discrete model.Effect on Weight (kg). Time (weeks).*
A similar pattern was observed for the continuous model, as shown in Fig 2. Weight decreased during the first eight weeks of the study (estimated slope, ES: -0·104, $95\%$ CI: -0·158, -0·050) followed by a gradual increase in weight until the 26th week (ES: 0·080, $95\%$ CI: 0·044, 0·116). Another significant drop in weight was observed around week 27 (ES: -0·136, $95\%$ CI: -0·426, 0·154), after which weight changes stabilized (ES: -0·011, $95\%$ CI: -0·021, -0·001).
**Fig 2:** *Estimated time-varying effects of Ramadan on weight using the continuous model.Ramadan effect on Weight (kg). Time (weeks).*
The estimated time-varying effects of Ramadan on weight for males and females are shown in Figs 3 and 4, respectively. While the overall weight trajectory is similar for both sexes, there are noticeable differences. We observe a more considerable fluctuation in the weight trajectory for females compared to males. For males, we observe a gradual decline in weight in the first five weeks before a gradual increase. For females, the weight gradually drops in the first ten weeks before an increase is observed. For both sexes, weight fluctuation stabilizes around week 30.
**Fig 3:** *Estimated change in weight (kg) relative to baseline over time in weeks for males.Error bars represent the 95% confidence intervals. Change in Weight (kg) relative to baseline. Time (weeks). Error bars represent the 95% confidence intervals.* **Fig 4:** *Estimated change in weight (kg) relative to baseline over time in weeks for females.Error bars represent the 95% confidence intervals. Change in Weight (kg) relative to baseline. Time (weeks). Error bars represent the 95% confidence intervals.*
The estimated time-varying effects of Ramadan on weight for non-pre-diabetics and non-diabetics (HbA1c < 5·$7\%$), as well as pre-diabetic (HbA1c between 5·$7\%$ and 6·$5\%$) and diabetic (HbA1c > 6·$5\%$) individuals are shown in Figs 5–7, respectively. Overall, we observe similar weight trajectories for the three groups of individuals. However, the weight fluctuation for non-prediabetic and non-diabetic individuals is considerably larger than the other two groups. The largest weight gain is observed at week 27 for both non-prediabetic, non-diabetic, and pre-diabetic individuals. In comparison, weight peaked at week 20 for diabetic individuals.
**Fig 5:** *Estimated change in weight (kg) relative to baseline over time in weeks for non-prediabetic and non-diabetic individuals (HbA1c < 5.7%).Error bars represent the 95% confidence intervals. Change in Weight (kg) relative to baseline. Time (weeks). Error bars represent the 95% confidence intervals.* **Fig 6:** *Estimated change in weight (kg) relative to baseline over time in weeks for pre-diabetic individuals (HbA1c between 5.7% and 6.5%).Error bars represent the 95% confidence intervals. Change in Weight (kg) relative to baseline. Time (weeks). Error bars represent the 95% confidence intervals.* **Fig 7:** *Estimated change in weight (kg) relative to baseline over time in weeks for diabetic individuals (HbA1c > 6.5%).Error bars represent the 95% confidence intervals. Change in Weight (kg) relative to baseline. Time (weeks). Error bars represent the 95% confidence intervals.*
## Discussion
To our knowledge, this is the first study to evaluate the pattern of weight change during and following Ramadan for a total of 52 weeks using data gathered over two years. The study participants consisted of a sample of individuals with diabetes, pre-diabetes, or at-risk adults. Specifically, there was a steady loss in weight until week 8 (0·5 kg), followed by steady weight gain until week 26 (3 kg), then a sharp drop in weight between 26 and 28 (>2·5 kg). From the 28th week onwards, we observe that weight stabilizes. Over time, this fluctuation in weight may be concerning since bodyweight fluctuations have been shown to increase the risk of incident type 2 diabetes, with higher variability increasing risk [33]. The present study can help inform public health organizations of when to intervene and how to tailor interventions [34]. Therefore, an important finding of this study is that it identifies critical turning points associated with fluctuations in weight over a longitudinal period. Our results are similar to previous systemic reviews [11, 13], where Ramadan fasting was shown to reduce a small amount of weight (− 1·022 kg, $95\%$ CI − 1·164 to − 0·880 kg) but only in the short term [35, 36]. However, because of the long-term follow-up in our study, we were also able to show that weight begins to increase shortly after Ramadan ends.
One limitation of our study is that data was collected as part of a clinical trial focused on weight loss, limiting its external validity and comparability with other studies and potentially influencing the weight changes seen. However, despite individuals being enrolled in the trial, a clear pattern of weight change during and following Ramadan was seen regardless of when the participants began the trial, which was on a rolling basis. Also, the study took place in Jordan between 2012 and 2015, with Ramadan overlapping the Summer months of June, July and August. While we would expect similar dynamic patterns to emerge from other Ramadan seasons, our findings are not necessarily generalizable to other times of the year, especially the winter months where days are shortened, and severe weather may alter economic activities. It is also plausible that the patterns of weight change seen in this study were influenced by other factors, such as economic, seasonal, or political factors that change throughout the year.
Another limitation is that as a secondary analysis, we did not explicitly inquire about physical activity levels or fasting adherence during Ramadan among the trial participants. However, the rates of fasting among *Muslims is* known to be high ($90\%$) [2, 3], and although people with diabetes are generally advised not to fast [37], roughly $80\%$ of people with type 2 diabetes still choose to fast during Ramadan [3]. This, together with reports from the study staff and nurses involved in the participants’ care, means we are confident that the vast majority of participants fasted during Ramadan.
Lastly, this study is not randomized. It is almost impossible to randomize participants into fasting and non-fasting groups because *Ramadan is* a religious obligation adhered to by most people in Jordan. Consequently, unknown and known confounders such as sex [2], duration of fasts [11], smoking, nutrition, dietary pattern [11], socio-economic status, cultural practices, or customs [38] may confound the association between weight loss and fasting during Ramadan. Furthermore, our findings are not generalizable to Muslims who do not fast during Ramadan and may not directly translate to high-income countries and other geographic regions.
## Policy implications
In a recent priority evaluation for non-communicable diseases [39], obesity was again identified as a significant public health problem that lacks effective population-level interventions. Ramadan presents a valuable opportunity for leaders to develop and implement complementary but effective policies [24]. Our study demonstrates weight fluctuations during and following Ramadan. While previous studies have shown that fasting during Ramadan may improve weight [13], no study has provided longitudinal evidence highlighting key time points during and after Ramadan fasting when the ebb and flow of weight trajectories can be identified. One way to harness the opportunity around *Ramadan is* to deliver targeted public health interventions, raise awareness within communities [24], and implement Ramadan-focused health monitoring campaigns. Future studies should evaluate the magnitude of the temporal effects of Ramadan on weight and other measures of metabolic health, especially in different geographic, cultural, and seasonal contexts. It would also be helpful for such studies to assess findings across sub-groups of low and high-risk populations to aid in the deployment of population-wide interventions.
## Conclusion
Weight fluctuations were observed during and after the Ramadan fasting period. Public health institutions need to be aware of this pattern to harness possible weight loss effects and protect against weight regain.
## References
1. 1Lipka, M., Muslims and Islam: Key findings in the U.S. and around the world. 2017, Pew Research Center: Washington, DC 20036, USA.
2. Azizi F.. **Research in Islamic Fasting and Health**. *Annals of Saudi Medicine* (2002.0) **22** 186-191. DOI: 10.5144/0256-4947.2002.186
3. Salti I.. **A Population-Based Study of Diabetes and Its Characteristics During the Fasting Month of Ramadan in 13 Countries: Results of the Epidemiology of Diabetes and Ramadan 1422/2001 (EPIDIAR) study**. *Diabetes Care* (2004.0) **27** 2306-2311. DOI: 10.2337/diacare.27.10.2306
4. 4Ghani, F., Most Muslims say they fast during Ramadan. 2013, Pew Research Center: Washington, DC 20036, USA.
5. Tootee A., Larijani B.. **Ramadan fasting and diabetes, latest evidence and technological advancements: 2021 update**. *Journal of Diabetes & Metabolic Disorders* (2021.0) **20** 1003-1009. PMID: 33996651
6. Su J.. **Remodeling of the gut microbiome during Ramadan-associated intermittent fasting**. *The American Journal of Clinical Nutrition* (2021.0) **113** 1332-1342. PMID: 33842951
7. R. De Cabo M.P.M.. **Effects of Intermittent Fasting on Health, Aging, and Disease**. *N Engl J Med* (2020.0) **382** 978
8. Patterson R.E., Sears D.D.. **Metabolic Effects of Intermittent Fasting**. *Annu Rev Nutr* (2017.0) **37** 371-393. PMID: 28715993
9. BaHammam A.S., Almeneessier A.S.. **Recent Evidence on the Impact of Ramadan Diurnal Intermittent Fasting, Mealtime, and Circadian Rhythm on Cardiometabolic Risk: A Review**. *Frontiers in Nutrition* (2020.0) **7**
10. Tootee A., Larijani B.. **Ramadan fasting during Covid-19 pandemic**. *Journal of Diabetes & Metabolic Disorders* (2020.0) **19** 1-4. PMID: 32363167
11. Correia J.M.. **Effects of Ramadan and Non-ramadan Intermittent Fasting on Body Composition: A Systematic Review and Meta-Analysis**. *Frontiers in Nutrition* (2021.0) **7**
12. Al-Rawi N.. **Effect of diurnal intermittent fasting during Ramadan on ghrelin, leptin, melatonin, and cortisol levels among overweight and obese subjects: A prospective observational study**. *PLOS ONE* (2020.0) **15** e0237922. DOI: 10.1371/journal.pone.0237922
13. Fernando H.. **Effect of Ramadan Fasting on Weight and Body Composition in Healthy Non-Athlete Adults: A Systematic Review and Meta-Analysis**. *Nutrients* (2019.0) **11** 478. DOI: 10.3390/nu11020478
14. Faris M.A.E.. **Ramadan intermittent fasting and immunity: An important topic in the era of COVID-19**. *Ann Thorac Med* (2020.0) **15** 125-133. DOI: 10.4103/atm.ATM_151_20
15. Ali S.. **Guidelines for managing diabetes in Ramadan**. *Diabetic Medicine* (2016.0) **33** 1315-1329. PMID: 26802436
16. Abaïdia A.-E., Daab W., Bouzid M.A.. **Effects of Ramadan Fasting on Physical Performance: A Systematic Review with Meta-analysis**. *Sports Medicine* (2020.0) **50** 1009-1026. DOI: 10.1007/s40279-020-01257-0
17. Faris M.e.A.-I.E.. **Impact of diurnal intermittent fasting during Ramadan on inflammatory and oxidative stress markers in healthy people: Systematic review and meta-analysis**. *Journal of Nutrition & Intermediary Metabolism* (2019.0) **15** 18-26
18. Faris M.A.E.. **Effect of diurnal fasting on sleep during Ramadan: a systematic review and meta-analysis**. *Sleep Breath* (2020.0) **24** 771-782. DOI: 10.1007/s11325-019-01986-1
19. Majid F., Behrman J., Mani S.. **Short-term and long-term distributional consequences of prenatal malnutrition and stress: using Ramadan as a natural experiment**. *BMJ global health* (2019.0) **4** e001185-e001185
20. Alghadir A.. **Ten-year Diabetes Risk Forecast in the Capital of Jordan: Arab Diabetes Risk Assessment Questionnaire Perspective-A Strobe-Complaint Article**. *Medicine* (2016.0) **95** e3181-e3181. DOI: 10.1097/MD.0000000000003181
21. Weiderpass E.. **The Prevalence of Overweight and Obesity in an Adult Kuwaiti Population in 2014**. *Frontiers in Endocrinology* (2019.0) **10**
22. Awad S.F.. **Characterizing the type 2 diabetes mellitus epidemic in Jordan up to 2050**. *Scientific Reports* (2020.0) **10** 21001. DOI: 10.1038/s41598-020-77970-7
23. Ajlouni K.. **Time trends in diabetes mellitus in Jordan between 1994 and 2017**. *Diabetic Medicine* (2019.0) **36** 1176-1182. DOI: 10.1111/dme.13894
24. Correia J.C.. **Interventions targeting hypertension and diabetes mellitus at community and primary healthcare level in low- and middle-income countries:a scoping review**. *BMC Public Health* (2019.0) **19** 1542. PMID: 31752801
25. Mohd Yusof B.-N.. **Comparison of Structured Nutrition Therapy for Ramadan with Standard Care in Type 2 Diabetes Patients**. *Nutrients* (2020.0) **12** 813. DOI: 10.3390/nu12030813
26. 26Zoughbie, D.E., Community-based diabetes programme: the micro-clinic project. 2009(1020–3397 (Print)).
27. Shahin Y.M.. **Evaluation of the Microclinic Social Network Model for Palestine Refugees with Diabetes at UNRWA Health Centers**. *Journal of Diabetes Mellitus* (2018.0) **08** 15
28. 28Zoughbie, D.E., et al., A social-network behavioral health program on sustained long-term body weight and glycemic outcomes: 2-year follow-up of a 4-month Microclinic Health Program in Jordan. 2018(2211–3355 (Print)).
29. 29Ding, E.L., et al. Randomized trial of social network lifestyle intervention for obesity: MICROCLINIC intervention results and 16-month follow-up. in Circulation. 2013. LIPPINCOTT WILLIAMS & WILKINS 530 WALNUT ST, PHILADELPHIA, PA 19106–3621 USA.
30. 30Prescott, M., et al. THE MICROCLINIC HEALTH PROGRAM: A SOCIAL NETWORK-BASED INTERVENTION FOR WEIGHT LOSS AND DIABETES RISK MANAGEMENT. in American Journal of Epidemiology. 2013. OXFORD UNIV PRESS INC JOURNALS DEPT, 2001 EVANS RD, CARY, NC 27513 USA.
31. 31Zoughbie, D.E., et al., Social Network Interventions for Obesity and Diabetes in Low-income and Refugee Communities: A Three Arm Randomized Trial of the Jordan Microclinic Social Network Program with 2 Years Follow-up. [Under Review], 2021.
32. Erik von Elm D.G.A., Matthias Egger. **The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies**. *Annals of Internal Medicine* (2007.0) **147** 573-577. DOI: 10.7326/0003-4819-147-8-200710160-00010
33. Park K.-Y.. **Body weight fluctuation as a risk factor for type 2 diabetes: results from a nationwide cohort study**. *Journal of clinical medicine* (2019.0) **8** 950. PMID: 31261984
34. Liao J.. **Experiences and views of people with diabetes during Ramadan fasting: A qualitative meta-synthesis**. *PLOS ONE* (2020.0) **15** e0242111. PMID: 33226993
35. Jahrami H.A.. **A systematic review, meta-analysis, and meta-regression of the impact of diurnal intermittent fasting during Ramadan on body weight in healthy subjects aged 16 years and above**. *Eur J Nutr* (2020.0) **59** 2291-2316. PMID: 32157368
36. Hajek P.. **Weight change during and after Ramadan fasting**. *Journal of Public Health* (2012.0) **34** 377-381. PMID: 22083256
37. Hanif S.. **Managing People with Diabetes Fasting for Ramadan During the COVID‐19 Pandemic: A South Asian Health Foundation Update**. *Diabetic Medicine* (2020.0) **37** 1094-1102. PMID: 32333691
38. Hassanein M.. **Comparison of the dipeptidyl peptidase-4 inhibitor vildagliptin and the sulphonylurea gliclazide in combination with metformin, in Muslim patients with type 2 diabetes mellitus fasting during Ramadan: results of the VECTOR study**. *Current Medical Research and Opinion* (2011.0) **27** 1367-1374. PMID: 21568833
39. Pia Schneider B.P., Shekar Meera, Julia Dayton Eberwein, Charlotte Block, Kyoko Shibata Okamura, Popkin M.S.a.B.. (2020.0)
|
---
title: 'Estimated burden, and associated factors of Urinary Incontinence among Sub-Saharan
African women aged 15–100 years: A systematic review and meta-analysis'
authors:
- Martin Ackah
- Louise Ameyaw
- Mohammed Gazali Salifu
- Cynthia OseiYeboah
- Abena Serwaa Ampomaa Agyemang
- Kow Acquaah
- Yaa Boatema Koranteng
- Asabea Opare-Appiah
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021416
doi: 10.1371/journal.pgph.0000562
license: CC BY 4.0
---
# Estimated burden, and associated factors of Urinary Incontinence among Sub-Saharan African women aged 15–100 years: A systematic review and meta-analysis
## Abstract
Hospital and community based-studies had been conducted for Urinary Incontinence (UI) in Sub-Sahara Africa (SSA) countries. A significant limitation of these studies is likely under-estimation of the burden of UI in SSA. It is therefore, imperative that a well-structured systematic review and meta-analytical models in SSA are required to accurately and reliably estimate the burden of UI. Medline/PubMed, Google Scholar, Africa Journal Online (AJOL) were searched to identified data on burden of UI studies in SSA. Two independent authors performed the initial screening of studies based on the details found in their titles and abstracts. The quality of the retrieved studies was assessed using the Newcastle-Ottawa Quality Assessment instrument. The pooled burden of UI was calculated using a weighted inverse variance random-effects model. A sub-group and meta-regression analyses were performed. Publication bias was checked by the funnel plot and Egger’s test. Of the 25 studies included, 14 were hospital-based, 10 community- based, and 1 university-based studies involving an overall 17863 participants from SSA. The systematic review showed that the prevalence of UI ranged from $0.6\%$ in Sierra Leone to $42.1\%$ in Tanzania. The estimated pooled burden of UI across all studies was $21\%$ [$95\%$ CI: $16\%$-$26\%$, I2 = $91.01\%$]. The estimated pooled prevalence of stress UI was $52\%$ [$95\%$ CI: $42\%$-$62\%$], urgency UI $21\%$ [$95\%$ CI: $15\%$-$26\%$], and mixed UI $27\%$ [$95\%$ CI: $20\%$-$35\%$]. The common significant independent factors were; parity, constipation, overweight/obese, vaginal delivery, chronic cough, gestational age, and aging. One out of every five women in SSA suffers from UI. Parity, constipation, overweight/obesity, vaginal delivery, chronic cough, gestational age, and age were the most important risk variables. As a result, interventions aimed at reducing the burden of UI in SSA women aged 15 to 100 years old in the context of identified determinants could have significant public health implications.
## Introduction
Pelvic Floor Disorders (PFD) affects millions of women worldwide [1–3]. About $10\%$ of women have surgery for Urinary Incontinence (UI), pelvic organ prolapse, or both, according to a regional survey in the United States, and $30\%$ of those women have two or more surgical procedures in their lifetime [3, 4]. Wu and colleagues estimated that $25\%$ of women in affluent countries suffers from one or more PFDs [5] with UI being the most common [6].
The International Continence Society (ICS) defines UI as the involuntary leakage of urine, with three basic subtypes identified: urgency UI (UUI), stress UI (SUI), and mixed UI (MUI; both UUI and SUI) [7, 8]. It is a widespread problem with an estimated global burden of nearly $5.0\%$ to $55\%$ with detrimental consequences on social life, personal relationships, feelings, sleep, and vitality [9, 10]. A comprehensive review and meta-analysis of 54 studies comprising 138722 women aged 10 to 90 years in Low- and Middle-Income Countries estimated the burden of UI to be $26\%$ [1]. In addition, the prevalence of UI ranged from $2.8\%$ in Nigeria to $57.7\%$ in the Islamic Republic of Iran [11]. UI is frequently underestimated and underdiagnose in developing and industrialized countries [11].
In comparison to patients with continence, recent investigations have shown that UI is a predictor of death [12, 13]. As a results, in order to strengthen continence programs, health systems should be able to estimate the burden especially in a region with a frail health system, such as SSA. Hospital and community based-studies had been conducted for UI in SSA countries [14–17]. A significant limitation of these observational studies is likely under-estimation of the burden of UI in SSA [16]. Following a thorough search of the literature, it was revealed that no prior systematic review and meta-analysis addressing the burden of UI and associated factors on the African continent has yet been conducted and published.
Many risk factors, such as multiple pregnancies, positive family history, parity, episiotomy use, body mass index, advanced age, spontaneous perineal tear at delivery and so on, appear to be implicated for UIs in High Income Countries (HIC) [2, 18–20].There are insufficient researches to draw conclusion on the risk factors for UI in SSA. In addition, large, diversified population-based studies have assessed prevalence rates of UI; however, there is limited robust evidence describing the burden of UI amongst women in SSA, where parity on average is higher than those in High-Income Countries [21].
It is therefore, imperative that a well-structured systematic review and meta-analytical models in SSA are required to accurately and reliably estimate the burden, and associated factors for UI. In this context, the current study aims to assess the burden of UI in SSA, as well as the risk factors associated with it. Thus, the review sought to answer the questions; what is the burden of UI in SSA? and what are the factors associated with the burden of UI in SSA?
## Overview
This systematic review was registered in PROSPERO [CRD42021267551]. This systematic review and meta-analysis was conducted and reported according to the guidelines of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) [22] [S1 Checklist].
## Inclusion criteria
Observational studies such as longitudinal, cohort, case control and cross-sectional studies reporting prevalence/or risk factors of UI were incorporated in the current review, as well as conference abstracts with enough information to calculate prevalence UI. Original observational studies published in English and adult SSA woman aged ≥18 years were included. Burden of UI studies that compared both SSA men and women, only information on the women were extracted. Finally, both hospital and community/population-based studies were included and later stratified in the pooled meta-analysis.
## Exclusion criteria
Studies reporting animal studies, reviews, commentaries, letter to editors were excluded. Prevalence of UI articles published in other languages were excluded. Studies that looked at the management and treatment of UI, as well as quality of life, depression, without data on burden of UIs were excluded. Studies from North Africa countries, other LMICs, and HIC were also excluded. UI studies involving children and adolescent females were also excluded.
## Data sources and search strategy
Medline/PubMed, Google Scholar, Africa Journal Online were searched to extract data on burden of UI studies in SSA countries, as well as their respective risk factors’ information. The articles that were considered were published between $\frac{2000}{1}$/1 and $\frac{2021}{9}$/30. There was also a manual search of the reference lists of the studies that were included. Medical Sub-Heading (MeSH) terms and free text were used in the search approach. These terms were coupled with the Boolean operators ‘OR’ and ‘AND’. The keywords included; Burden, prevalence, Pelvic Floor Disorders, Urinary Incontinence, Sub-Saharan Africa. The final search strategy is displayed in (S1 Table).
## Selection process
To ensure a rigorous review strategy, articles were reviewed individually by two independent co-authors [MA, and KA]. The data screening was done in two stages; the title abstract screening, followed by the full-text screening. Both steps were completed independently by two review authors. A third reviewer (ASA) was available to resolved the disagreement between MA and KA. Finally, all the studies were imported into Mendeley desktop reference manager.
## Data collection and management
MA and KA independently extracted data into an excel sheet, and discrepancies were resolved through discussion. Extracted data were; Author’s name, year of publication, country, age, study design, population, sample size, setting, information on the burden of UI. Finally, significant independent risk factors from the individual studies were extracted.
## Quality assessment and risk of bias
The quality of the retrieved studies were assessed using the Newcastle-Ottawa Quality Assessment instrument, which was customized for cross-sectional research [23]. The assessment’s goal was to determine the research’ internal and external validity, as well as to reduce the possibility of bias [23]. The findings of the quality assessment are presented in S2 Table.
## Data synthesis and strategy
The pool burden of UI was calculated using a weighted inverse variance random-effects model. This was visually represented using the forest plot. The presence of heterogeneity among studies was quantified by estimating variance using the I2 statistics [24]. The I2 takes values between 0 and $100\%$, and a value of $0\%$ indicates absence of heterogeneity. I2 was interpreted based on Higgins and Thompson classification, percentages of $25\%$, $50\%$ and $75\%$ was considered as low, moderate and high heterogeneity, respectively [24].
A subgroup analysis was performed to determine the sources of heterogeneity on the study characteristics (year of publication, sample size, setting, and sub-regions). The funnel plot and Egger’s regression test was used to screened for publication bias. Finally, meta-regression was performed to assessed the factors influencing the review’s heterogeneity.
## Study selection process
The study identified one thousand nine hundred and twenty-nine [1929] articles from PubMed, AJOL, and Google scholar, out of which 700 were removed as duplicate. One thousand two hundred and twenty-nine articles were screened, of which 1071 papers were excluded. A total of 148 publications were evaluated for eligibility, with 25 studies [14–17, 25–45] being included in the current review [Fig 1]. This study included a total of 17863 women from SSA.
**Fig 1:** *PRISMA 2020 flow diagram for new systematic reviews which included searches of databases, registers and other sources.*
## Characteristics of the included studies and quality assessment
The characteristics of the included studies are shown in Table 1. The studies were published between 2005 and 2021. Nigeria had the highest number of eligible studies [$$n = 14$$], followed by Ethiopia [$$n = 3$$], followed by Ghana, and South Africa with two studies each. The age of the participants ranged from 15 to 100 years. Of the 25 studies, 14 were hospital-based, 10 community- based, and 1 university-based studies. The sample size ranged from 100 to 5000, with overall 17863 participants from SSA. Seventy-two percent [$$n = 17$$] of the articles had low risk of bias.
**Table 1**
| SN | Study ID | Year of publication | Country | Study Design | Age/year | Sample Size | Burden [%] | Stress UI [%] | Urgency UI [%] | Mixed UI [%] | Setting | Risk of bias assessment |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | Berhe et al. [14] | 2020 | Ethiopia | Cross-sectional | 18–45 | 317 | 4.3 | 58.9 | 10.9 | 30.1 | Hospital | Low |
| 2 | Demissie et al. [15] | 2021 | Ethiopia | Cross-sectional | 19–70 | 542 | 3.3 | | | | Community | Low |
| 3 | Ofori et al. [16] | 2020 | Ghana | Cross-sectional | 19–88 | 400 | 12.0 | 22.9 | 33.3 | 20.8 | Hospital | Low |
| 4 | Adanu et al. [42] | 2005 | Ghana | Not reported | 17–70 | 200 | 42.0 | 100.0 | 0.0 | 0.0 | Hospital | Low |
| 5 | Balde et al. [44] | 2020 | Guinea | Retrospective Cohort | 15–70 | 1770 | 10.2 | | | | Hospital | Low |
| 6 | Bekele et al. [45] | 2016 | Ethiopia | Cross-sectional | 16–40 | 422 | 11.4 | | | | Hospital | Low |
| 7 | Bowling et al. [25] | 2010 | Liberia | Not reported | Not reported | 424 | 1.7 | | | | Community | Moderate |
| 8 | Ojengbede et al. [34] | 2010 | Nigeria | Prospective Cohort | 15–45+ | 5001 | 2.8 | | | | Community | Low |
| 9 | Usifoh et al. [38] | 2012 | Nigeria | Cross-sectional | 15–60+ | 412 | 29.4 | 44.6 | 14.9 | 40.5 | Community | Low |
| 10 | Rabiu et al. [29] | 2015 | Nigeria | Cross-sectional | 15–44 | 257 | 15.2 | 43.6 | 46.2 | 20.2 | Hospital | Moderate |
| 11 | Okunola et al. [35] | 2018 | Nigeria | Cross-sectional | 18–45 | 442 | 28.1 | 62.1 | 24.2 | 19.4 | Hospital | Low |
| 12 | Abiola et al. [40] | 2016 | Nigeria | Cross-sectional | Not reported | 229 | 12.7 | 58.6 | 27.6 | 17.2 | Community | Low |
| 13 | Akinlusi et al. [17] | 2020 | Nigeria | Cross-sectional | 25–67 | 395 | 32.9 | 54.6 | 23.1 | 22.3 | Hospital | Low |
| 14 | Adaji et al. [41] | 2009 | Nigeria | Cross-sectional | 15–42 | 204 | 21.1 | 60.4 | 25.6 | 9.3 | Hospital | Moderate |
| 15 | Yağmur et al. [39] | 2021 | Nigeria | Cross-sectional | 40–69 | 286 | 30.1 | 30.2 | 7.0 | 31.4 | Community | Moderate |
| 16 | Badejoko et al. [43] | 2015 | Nigeria | Cross-sectional | 20–100 | 1250 | 5.2 | 35.4 | 46.2 | 18.6 | Hospital | Low |
| 17 | Bello et al. [36] | 2018 | Nigeria | Cross-sectional | 16–46+ | 500 | 21.4 | 40.2 | 8.4 | 51.4 | Hospital | Low |
| 18 | Njoku et al. [32] | 2020 | Nigeria | Cross-sectional | Not reported | 658 | 16.1 | 73.6 | 16.9 | 9.4 | Hospital | Low |
| 19 | Irshad et al. [28] | 2021 | Nigeria | Cross-sectional | 15–45 | 282 | 26.2 | 56.8 | 8.1 | 33.8 | Hospital | Moderate |
| 20 | Obioha et al. [33] | 2015 | Nigeria | Prospective Cohort | Not reported | 230 | 12.2 | | | | Hospital | Moderate |
| 21 | Gashugi et al. [26] | 2005 | Rwanda | Not reported | 20–64 | 1030 | 41.9 | | | | Community | Low |
| 22 | Patel et al.[27] | 2014 | Sierra Leone | Not reported | Not reported | 1320 | 0.6 | | | | Community | Moderate |
| 23 | Skaal et al. [37] | 2011 | South Africa | Not reported | Not reported | 145 | 31.7 | | | | University | Low |
| 24 | Madombwe et al. [30] | 2010 | South Africa | Not reported | 21–76 | 100 | 35.4 | | | | Community | Low |
| 25 | Masenga et al. [31] | 2019 | Tanzania | Cross-sectional | 18–90 | 1048 | 42.1 | 39.0 | 22.0 | 39.0 | Community | Low |
## Estimated Burden of Urinary Incontinence amongst Sub-Saharan Africa women
The systematic review showed that the prevalence of UI ranged from $0.6\%$ in Sierra Leone to $42.1\%$ in Tanzania. In the meta-analysis, the pooled estimate of the burden of UI across all studies was $21\%$ [$95\%$ CI: $16\%$-$26\%$, I2 = $91.01\%$]. A substantial statistically significant heterogeneity was detected across the studies [Fig 2].
**Fig 2:** *Forest plot of pooled Burden of Urinary Incontinence amongst SSA women.*
With respect to the sub-types of UI, 14 studies were included, and the estimated pooled prevalence of stress UI was $52\%$ [$95\%$ CI: $42\%$-$62\%$, I2 = $70.78\%$], urgency UI $21\%$ [$95\%$ CI: $15\%$-$26\%$, I2 = $0.00\%$], and mixed UI $27\%$ [$95\%$ CI: $20\%$-$35\%$, I2 = $46.37\%$] [Table 2].
**Table 2**
| Sub group | Variable | Dataset | Pooled burden % [95%CI] |
| --- | --- | --- | --- |
| Sample size | | | |
| | ≤400 | 12.0 | 24.0 [18.0–30] |
| | >400 | 13.0 | 19.0 [10.0–27] |
| Types of UI | | | |
| | Stress UI | 14.0 | 52.0 [42.0–62.0] |
| | Urgency UI | 14.0 | 21.0 [15.0–26.0] |
| | Mixed UI | 14.0 | 27.0 [20.0–35.0] |
| Sub-Region | | | |
| | East Africa | 5.0 | 31.0 [19.0–42.0] |
| | West Africa | 18.0 | 16.0 [11.0–21.0] |
| | South Africa | 2.0 | 23.0 [12.0–33.0] |
| Study setting | | | |
| | Hospital based | 14.0 | 18.0 [13.0–22.0] |
| | Community-based | 10.0 | 23.0 [16.0–26.0] |
## Sub-group analysis
Sub-group analysis were performed with regards to sub-region [*East africa* vs.West Africa vs. South Africa], and setting [Hospital-based vs. Community-based studies]. The pooled estimate of UI in East Africa was $31\%$ [$95\%$ CI: $19\%$-$42\%$], West Africa was $16\%$ [$95\%$ CI: $11\%$-$21\%$], and South Africa $35\%$ [$95\%$ CI: $15\%$-$55\%$]. With regards to settings, the estimated burden of UI was $18\%$ [$95\%$ CI: $13\%$-$22\%$] for hospital based studies, and $23\%$ [$95\%$ CI: $12\%$-$33\%$] for community-based studies. The findings are summarised in Table 2.
## Meta-regression
Meta-regression revealed non-sigifcant, decreased trend in the year of publication (coefficient = -0.0033, $$p \leq 0.533$$), and significant decreased trend in the number of sample size (coefficient = -0.0001, $$p \leq 0.048$$) with increasing burden of UI amongst women in SSA [S3 Table].
## Publication bias
There was no evidence of publication bias in both the subjective funnel plot [Fig 3] and the objective Egger’s regression test ($z = 1.74$, $$p \leq 0.0825$$).
**Fig 3:** *Assessment of publication bias using the funnel plot.*
## Systematic review of associated factors for UI amongst Sub-Saharan African women
Systematic review of factors associated with UI amongst SSA women is shown Table 3. The significant independent factors were; parity [14, 31, 32], constipation [14, 17, 45], overweight/obese [17, 35, 45], vaginal delivery [32, 34, 35] chronic cough [16, 45], gestational age [14, 35] and aging [16, 32]. For example, Berhe et al. found that multiparous women had approximately 6 times higher chances of developing UI compared to primigravida women in Ethiopia [14]. A population-based study in rural Tanzania showed that women who had experienced multiple births had 2 times chance of reporting UI [31]. Also, women with history of constipation had 2times chance of reporting UI compared to women without prior constipation in a Nigerian study [17]. Similarly, an Ethiopian study reported history of constipation as an associated factor for UI [45]. The results are presented in Table 3.
**Table 3**
| SN | Study ID | Country of Study | Type of Study | Independent Risk Factors | p-value |
| --- | --- | --- | --- | --- | --- |
| 1.0 | Berhe et al. [14] | Ethiopia | Cross-sectional | Gestational age | ** |
| | | | | Parity | ** |
| | | | | Prior Miscarriage | ** |
| | | | | Constipation | ** |
| | | | | Respiratory Problem | ** |
| | | | | Weak PFM | ** |
| 2.0 | Ofori et al. [16] | Ghana | Cross-sectional | Age>60 | ** |
| | | | | History of chronic cough | ** |
| 3.0 | Bekele et al. [45] | Ethiopia | Cross-sectional | Episiotomy | ** |
| | | | | Constipation | ** |
| | | | | Obese women | ** |
| | | | | Chronic cough/Sneezing | ** |
| | | | | Asthma/Allergies/Sinusitis | ** |
| 4.0 | Okunola et al. [35] | Nigeria | Cross-sectional | overweight/obese | ** |
| | | | | Gestational age | ** |
| | | | | Previous Vaginal/Instrumental delivery | ** |
| 5.0 | Akinlusi et al. [17] | Nigeria | Cross-sectional | Previous Constipation | ** |
| | | | | overweight/obese | ** |
| 6.0 | Njoku et al. [32] | Nigeria | Cross-sectional | Age>40 | ** |
| | | | | Parity>3 | ** |
| | | | | Low educational level | ** |
| | | | | Vaginal/Instrumental Delivery | ** |
| | | | | Carry Heavy Load | ** |
| | | | | Farming | ** |
| 7.0 | Masenga et al. [31] | Tanzania | Cross-sectional | Parity | ** |
| | | | | Delivery at home | ** |
| | | | | Labour>24hrs | ** |
| 8.0 | Ojengbede et al. [34] | Nigeria | Prospective Cohort | Vaginal Delivery | ** |
| | | | | Diabetes | ** |
## Discussion
The systematic review and meta-analysis included 25 studies with 17863 participants from 9 countries across SSA. We present a comprehensive review of the burden of UI and associated variables in SSA women in this study. The systematic review showed that the burden of UI ranged from $0.6\%$ in Sierra Leone to $42.1\%$ in Tanzania. According to the meta-analysis, the sub-region has an estimated pooled burden of UI to be $21\%$ [$95\%$ CI: $16\%$-$26\%$]. The study further revealed that the commonest type of UI was stress UI ($52\%$), mixed UI ($27\%$), and urgency UI ($21\%$). There were a wide range of prevalent estimates among the participants. For example, Patel et al. [ 27] discovered that less than $1\%$ of Sierra Leonean women have UI, but Masenga et al. [ 31] found that $42.1\%$ of Tanzanian women have UI. The current estimated burden is lower than reported by Xue et al in China [46], Mostafaei et al in developing countries [11], and Batmani et al in the global estimate [47]. Similarly, a study in the United States found that $45\%$ of participants aged 30–60 years reported UI [48], which is significantly higher than the current estimate. The comparatively lower burden in the current review could be attributed to variation in the method used, under-reporting, underdiagnosis as a results of low health seeking behavior among SSA women [49]. As a result, initiatives to reduce the burden of UI amongst SSA women aged 15–100 years could have significant public health implications.
The pooled estimate of UI in East Africa was $31\%$, West Africa was $16\%$, and South Africa $35\%$. Tanzania [31] reported the greatest prevalence in East Africa, while Ethiopia [14] recorded the lowest. Ofori et al. [ 16] found the highest prevalence of UI in West Africa, while Patel et al. [ 27] found the lowest. In the southern SSA, both the highest and lowest burden were reported from South Africa [30, 37]. The sub-regional variance could be attributed to the sample size included from individual countries. For example, a trend meta-regression analysis in this review found an inverse relationship between the burden of UI and sampled size.
In addition, our stratified analysis based on study setting showed that community dwelling women in SSA had an estimated burden of $23\%$ compared to hospital-based of $18\%$. Ten and fourteen studies reported on community and hospital-based studies respectively.
Three studies reported that parity was independently associated with UI [14, 31, 32]. Masenga et al. [ 31] adduced that women with at least three children have a two-fold increased risk of contracting UI in Tanzania. Similarly, Njoku et al. [ 32] in Nigeria corroborated with this findings. In Ethiopia, multiparous women had a 6 times higher risk of UI than primigravida women [14]. The current findings are supported meta-analyses [46, 47, 50]. The findings are ascribed to pelvic floor musculature and connective tissue injury that affects normal urine continence function during parturition. [ 50, 51].
Constipation was another commonly reported independent factors associated with UI in SSA [14, 17, 45]. The finding is consistent with systematic reviews conducted in China [46] and worldwide [47]. The mechanism underpinning this association is not well understood, although in an animal research, Chen et al. found that colon distension increased contractility, similar to bladder distention and the vesicovascular reflex, hence it could be inferred that chronic colorectal distension caused by constipation or chronically high abdominal muscle pressure during defecation limits bladder distension which exacerbates irritative bladder symptoms [52, 53].
Another important factor associated with UI among women in SSA was overweight/obese Three studies [17, 35, 45] reported on this factor. For instance, a study in Nigeria reported that an overweight/obese women had $60\%$ increased odds of UI compared to apparently normal body weight women. Bekele et al. [ 45] reported similar magnitude of the event. The current result is in accordance with Batman et al. [ 47]. Excess body weight is thought to increase abdominal pressure, which raises bladder pressure and urethral mobility, causing stress UI and worsening detrusor instability and overactive bladder [54]. In the same vein, obesity/overweight can create chronic strain on the soft tissues, and other pelvic floor structures, hence, straining and weakening these important urethral mobility systems [55].
Furthermore, vaginal delivery [32, 34, 35], chronic cough [16, 45], and gestational age [14, 35] were all found to be factors associated with UI amongst SSA women aged 15–100 years. This is consistent with several observational studies and systematic reviews [46, 47, 50].
Finally, two studies identified aging as a risk factor for UI [16, 32]. In Ghana, women aged 60 were approximately three times more likely than women aged 18–39 to have urinary incontinence [42]. In Nigerian study, Njoku et al. [ 32] estimated that women over the age of 40 had a five-fold increased risk of urine incontinence.
Our findings should be viewed in the context of some caveats. First, there was significant heterogeneity among the studies. Second, studies from Central Africa were scarce, thus no research from this region was included which could affect generalization of our findings. Furthermore, only articles published in English were considered. Finally, the authors identified relevant studies using a small database. However, heterogeneity is common in meta-analyses of observational data, and it does not always invalidate the conclusions. This is the first and largest systematic review and meta-analysis on the burden of UI among SSA women. Our results are more reliable evident by no obvious publication bias.
## Recommendations, research and policy implications
Our research showed that, at least $21\%$ of women in SSA have some form of urinary incontinence. This implies that there is the need to increase public health education and create awareness among people in order to promote health seeking behaviors among women. Healthy practices can be encouraged among women in order to reduce the disease burden in women. Such practices may include; maintaining a healthy body weight through exercise and diet to reduce the incidence of obesity and its complications. Avoiding constipation by taking in a high fiber diet, adequate intake of water, reducing immobility and regular emptying of bowel. Women should be taught how to strengthen their pelvic floor muscles especially during and after pregnancy by performing kegel’s exercises. Family planning methods should be made accessible to all irrespective of their social and economic status so women can effectively exercise their rights to choose the number of children they would want to have while taking into consideration their health. Finally, there are few studies examining the burden of UI, particularly in SSA’s eastern and southern regions. It is therefore critical that more researches be undertaken in these sub-regions in order to obtain a complete, accurate, and consistent picture of this treatable condition.
## Conclusion
This is the first and most comprehensive systematic review and meta-analysis on the burden of UI amongst SSA women aged 15 to 100 years. According to the study, one out of every five women in SSA suffers from UI. Parity, constipation, overweight/obesity, vaginal delivery, chronic cough, gestational age, and age were the most important risk variables. As a result, interventions aimed at reducing the burden of UI in SSA women aged 15 to 100 years old in the context of identified determinants could have significant public health implications.
## References
1. Kenton K, Mueller ER. **The global burden of female pelvic floor disorders**. *BJU Int* (2006.0) **98** 1-5. DOI: 10.1111/j.1464-410X.2006.06299.x
2. Gedefaw G, Demis A. **Burden of pelvic organ prolapse in Ethiopia: A systematic review and meta-analysis.**. *BMC Womens Health* (2020.0) **20** 1-9. PMID: 31898500
3. Nygaard I.. **Prevalence of Symptomatic Pelvic Floor Disorders in US Women**. *JAMA* (2008.0) **23** 1311-1316. DOI: 10.1001/jama.300.11.1311
4. Olsen AL, Smith VJ, Bergstrom JO, Colling JC, Clark AL. **Epidemiology of surgically managed pelvic organ prolapse and urinary incontinence.**. *Obstet Gynecol* (1997.0) **89** 501-6. DOI: 10.1016/S0029-7844(97)00058-6
5. Wu JM, Vaughan CP, Goode PS, Redden DT, Burgio KL, Richter HE. **Prevalence and trends of symptomatic pelvic floor disorders in U.S. women**. *Obstet Gynecol* (2014.0) **123** 141-8. DOI: 10.1097/AOG.0000000000000057
6. Islam RM, Oldroyd J, Rana J, Romero L, Karim MN. **Prevalence of symptomatic pelvic floor disorders in community-dwelling women in low and middle-income countries: a systematic review and meta-analysis.**. *Int Urogynecol J* (2019.0) **30** 2001-11. DOI: 10.1007/s00192-019-03992-z
7. Milsom I, Coyne KS, Nicholson S, Kvasz M, Chen CI, Wein AJ. **Global prevalence and economic burden of urgency urinary incontinence: A systematic review.**. *Eur Urol [Internet]* (2014.0) **65** 79-95. DOI: 10.1016/j.eururo.2013.08.031
8. Haylen BT, De Ridder D, Freeman RM, Swift SE, Berghmans B, Lee J. **An International Urogynecological Association (IUGA)/International Continence Society (ICS) joint report on the terminology for female pelvic floor dysfunction**. *Int Urogynecol J* (2010.0) **21** 5-26. DOI: 10.1007/s00192-009-0976-9
9. Minassian V.A., Drutz H.P. AA-B. **Serum levels of insulin, IGF, BP1 in preeclampsia and eclampsia.**. *Int J Gynaecol Obstet* (2003.0) **81** 151-6. PMID: 12706271
10. Abrams P, Smith AP, Cotterill N. **The impact of urinary incontinence on health-related quality of life (HRQoL) in a real-world population of women aged 45–60 years: Results from a survey in France, Germany, the UK and the USA.**. *BJU Int* (2015.0) **115** 143-52. DOI: 10.1111/bju.12852
11. Mostafaei H, Sadeghi-Bazargani H, Hajebrahimi S, Salehi-Pourmehr H, Ghojazadeh M, Onur R. **Prevalence of female urinary incontinence in the developing world: A systematic review and meta-analysis—A Report from the Developing World Committee of the International Continence Society and Iranian Research Center for Evidence Based Medicine.**. *Neurourol Urodyn* (2020.0) **39** 1063-86. DOI: 10.1002/nau.24342
12. John G, Bardini C, Combescure C, Dällenbach P. **Urinary incontinence as a predictor of death: A systematic review and meta-analysis.**. *PLoS One* (2016.0) **11** 1-19. DOI: 10.1371/journal.pone.0158992
13. Holroyd-Leduc JM, Mehta KM, Covinsky KE. **Urinary Incontinence and Its Association with Death, Nursing Home Admission, and Functional Decline.**. *J Am Geriatr Soc* (2004.0) **52** 712-8. DOI: 10.1111/j.1532-5415.2004.52207.x
14. Berhe A, Alamer A, Negash K, Assefa B. **Urinary incontinence and associated factors among pregnant women attending antenatal care in public health facilities of Mekelle city, Tigray, Ethiopia**. *Women’s Heal* (2020.0) **16** 1-8. DOI: 10.1177/1745506520952009
15. Demissie E, Id B, Tesfaye W, Id T. **Symptomatic pelvic floor disorders and its associated factors in South-Central Ethiopia**. *PLoS One [Internet]* (2021.0) 1-15. DOI: 10.1371/journal.pone.0254050
16. Amanfo A, Ofori Osarfo, Joseph EK, Agbeno K, Azanu HS. **Prevalence and determinants of non-fistulous urinary incontinence among Ghanaian women seeking gynaecologic care at a teaching hospital**. *PLoS One [Internet]* (2020.0) 1-14. DOI: 10.1371/journal.pone.0237518
17. Akinlusi FM, Ottun TA, Oshodi YA, Oluwatoyin B. **Female Urinary Incontinence : Prevalence, Risk Factors and Impact on the Quality of Life of Gynecological Clinic Attendees in Lagos, Nigeria.**. *Nep J Obs Gynecol* (2020.0) **15** 31-8
18. Bodner-Adler B, Kimberger O, Laml T, Halpern K, Beitl C, Umek W. **Prevalence and risk factors for pelvic floor disorders during early and late pregnancy in a cohort of Austrian women**. *Arch Gynecol Obstet [Internet]* (2019.0) **300** 1325-30. DOI: 10.1007/s00404-019-05311-9
19. MacLennan AH, Taylor AW, Wilson DH, Wilson D. **The prevalence of pelvic floor disorders and their relationship to gender, age, parity and mode of delivery**. *British Journal of Obstetrics and Gynaecology* (2000.0) **107** 1460-70. DOI: 10.1111/j.1471-0528.2000.tb11669.x
20. Yohay D, Weintraub AY, Mauer-Perry N, Peri C, Kafri R, Yohay Z. **Prevalence and trends of pelvic floor disorders in late pregnancy and after delivery in a cohort of Israeli women using the PFDI-20**. *Eur J Obstet Gynecol Reprod Biol [Internet]* (2016.0) **200** 35-9. DOI: 10.1016/j.ejogrb.2016.02.037
21. Bowling CB, Munoz O, Gerten K, Mann M, Taryor R, Norman A. **Characterization of Pelvic Floor Symptoms in Women of Northeastern Liberia.**. *Neurourol Urodyn* (2009.0) **109** 251-3
22. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD. **The PRISMA 2020 statement: An updated guideline for reporting systematic reviews**. *BMJ* (2021.0) 372
23. Luchini Claudio, Stubbs Brendon, Marco Solmi NV. **Assessing the quality of studies in meta-analyses: Advantages and limitations of the Newcastle Ottawa Scale Luchini C, Stubbs B, Solmi M, Veronese N. Assessing the quality of studies in meta-analyses: Advantages and limitations of the Newcastle Ottawa Sca.**. *World J Psychiatry* (2012.0) **2** 26-32. PMID: 24175165
24. Higgins JPT, Thompson SG. **Quantifying heterogeneity in a meta-analysis.**. *Stat Med* (2002.0) **21** 1539-58. DOI: 10.1002/sim.1186
25. Bowling CB, Munoz O, Gerten KA, Mann M. **NIH Public Access**. *Int J Gynaecol Obs* (2010.0) **109** 251-3
26. Gashugi P LQ. **Prevalence and Risk Factors of Urinary Incontinence among Adult Rwandan Women**. *SA J Physiother* (2005.0) **61** 6-14
27. Patel Hiten D., Kamara Thaim B., Kushner Adam L.. **Estimating the prevalence of urinary and fecal incontinence in a nationally representative survey in Sierra Leone Hiten.**. *Int J Gynaecol Obs* (2014.0) **126** 175-6
28. Irshad A, Irshad S, Adesiyun AG, Sulayman HU. **The burden of urinary incontinence in late pregnancy : Antenatal clinic experience in a tertiary hospital in Northern Nigeria.**. *Arch Int Surg* (2021.0) 47-51
29. J, Rabiu A AIG. **PREVALENCE Of POSTPARTUM URI NARY INC ONTINE NC E AMONG WOME N ATTE NDI NG POSTNATAL C LI NIC AT Ai\ II NU KANOTEACH ING HOSPITAL.**. *Trop J Obstet Gynaecol* (2015.0) **32**
30. J P Madombwe SK. **High prevalence of urinary incontinence and poor knowledge of pelvic floor exercises among women in Ladysmith**. *SAJOG* (2005.0) **16** 18-21
31. Masenga GG, Benjamin C Shayo, Msuya S, Rasch V. **Urinary incontinence and its relation to delivery circumstances : A population-based study from rural Kilimanjaro, Tanzania**. *PLoS One [Internet]* (2019.0) 1-12. DOI: 10.1371/journal.pone.0208733
32. Njoku CO, Njoku AN, Emechebe CI, Okpe AE, Iklaki CI. **Pattern and risk factors of non-fistulous urinary incontinence among gynaecological clinic attendees in a Nigeria tertiary health institution.**. *Int J Reprod Contracept Obs Gynecol* (2020.0) **9** 2323-7
33. Obioha KC, Ugwu EO, Obi SN, Dim CC, Oguanuo TC. **Prevalence and predictors of urinary / anal incontinence after vaginal delivery : prospective study of Nigerian women.**. *Int Urogynecol J* (2015.0) **26** 1347-1354. DOI: 10.1007/s00192-015-2690-0
34. 34Oladosu A. Ojengbede, Imran O. Morhason-Bello, B. O. A., & Kolade, N. S. O. and C. O. (2010). urinary incontinence among 5000 black women community survey. BJU International, 1793–1800. 10.1111/j.1464-410X.2010.09758.x. DOI: 10.1111/j.1464-410X.2010.09758.x
35. Okunola TO, Rosiji B. **Prevalence and risk factors for urinary incontinence in pregnancy in Ikere-Ekiti, Nigeria**. *Neurourol Urodyn* (2018.0) 1-7. PMID: 29975425
36. Bello OO. **Prevalence of Non-Fistulous Urinary Incontinence among Nonparturient Women in A Tertiary Hospital**. *J woman’s Reprod Heal* (2018.0) **2** 35
37. Skaal L MM. **The Prevalence of Urinary Incontinence and its Impact on Quality of Life among the University Female Staff in South Africa**. *SA J Physiother* (2011.0) 45-9
38. Usifoh SF, Odilim VU, Udezi WA. **Prevalence of urinary incontinence in a sample of women living in Benin city, Prevalence of urinary incontinence in a sample of women living in Benin city, Nigeria**. *J Med Res Pr* (2012.0) **1** 29-32
39. Y Yağmur SG. **Urinary Incontinence in Women aged 40 and Older: Its Prevalence, Risk Factors, and Effect on Quality of Life**. *Niger J Clin Pract* (2021.0)
40. Abiola OO, Idowu A, Ogunlaja OA, Williams-abiola OT, Ayeni SC. **Prevalence, quality of life assessment of urinary incontinence using a validated tool (ICIQ-UI SF) and bothersomeness of symptoms among rural community : dwelling women in Southwest, Nigeria**. *Int J Community Med Public Heal* (2016.0) **3** 989-97
41. Adaji SE, Shittu OS, Bature SB, Nasir S, Olatunji O. **European Journal of Obstetrics & Gynecology and Reproductive Biology Suffering in silence : pregnant women ‘ s experience of urinary incontinence in**. *Eur J Obstet Gynecol* (2010.0) **150** 19-23. DOI: 10.1016/j.ejogrb.2010.02.008
42. Adanu RMK, Delancey JOL. **The physical finding of stress urinary incontinence among African women in Ghana.**. *Int Urogynecol J.* (2005.0)
43. Badejoko OO, Bola-oyebamiji S, Awowole IO. **Urinary incontinence : prevalence, pattern, and opportunistic screening in Ile-Ife, Nigeria.**. *Int Urogynecol J* (2015.0)
44. Balde FB, Diallo AB, Toure A, Kante D. **Risk Factors for Urinary Incontinence after Obstetric Vesicovaginal Fistula Closure in.**. *Surg Sci* (2021.0) **12** 1-8
45. Bekele A, Adefris M, Demeke S. **Urinary incontinence among pregnant women, following antenatal care at University of Gondar Hospital, North West Ethiopia**. *BMC Pregnancy Childbirth [Internet]* (2016.0) **16** 2-7. DOI: 10.1186/s12884-016-1126-2
46. Xue K, Palmer MH, Zhou F. **Prevalence and associated factors of urinary incontinence in women living in China : a literature review.**. *BMC Urol [Internet]* (2020.0). DOI: 10.1186/s12894-020-00735-x
47. Batmani S, Jalali R, Mohammadi M. **Prevalence and factors related to urinary incontinence in older adults women worldwide : a comprehensive systematic review and meta-analysis of observational studies**. *BMC Geriatr* (2021.0) **21** 1-17. PMID: 33388045
48. Melville JL, Katon W, Delaney K, Newton K. **Urinary Incontinence in US Women**. *Arch Intern Med* (2005.0) **165** 537-42. DOI: 10.1001/archinte.165.5.537
49. Budu E, Abdul-aziz Seidu, Armah-ansah EK. **Women ‘ s autonomy in healthcare decision- making and healthcare seeking behaviour for childhood illness in Ghana : Analysis of data from the 2014 Ghana Demographic and Health Survey**. *PLoS One [Internet]* (2020.0) 1-14. DOI: 10.1371/journal.pone.0241488
50. Zhou H, Shu B, Liu T, Wang X, Yang Z, Guo Y. **Association between parity and the risk for urinary incontinence in women.**. *Medicine (Baltimore)* (2018.0) **97**. DOI: 10.1097/MD.0000000000011443
51. Wood LN, Anger JT. **Urinary incontinence in women**. *BMJ* (2014.0) **349**. DOI: 10.1136/bmj.g4531
52. Maeda T, Tomita M, Nakazawa A, Sakai G, Funakoshi S, Komatsuda A. **Female Functional Constipation Is Associated with Overactive Bladder Symptoms and Urinary Incontinence**. *Biomed Res Int* (2017.0) **2017**. DOI: 10.1155/2017/2138073
53. Chen S, Yang C, Chien C. **Neuroscience Letters Colorectal distension enforce acute urinary bladder distension-induced hepatic vasoconstriction in the rat**. *Neurosci Lett* (2008.0) **443** 257-60. PMID: 18602450
54. Bump RC, Sugerman HJ, Fantl JA, Mcclish DK. **Obesity and lower urinary tract function in women : Effect of surgically induced weight loss**. *Am J Obs Gynecol* (1992.0) 392-9. DOI: 10.1016/s0002-9378(11)91418-5
55. Subak Leslee L., Holly E. Richter SH. **Obesity and Urinary Incontinence: Epidemiology and Clinical Research Update**. *J Urol* (2009.0) **182** 3-5. DOI: 10.1016/j.juro.2009.08.071
|
---
title: Assessment of Mongolian dietary intake for planetary and human health
authors:
- Dashzeveg Delgermaa
- Miwa Yamaguchi
- Marika Nomura
- Nobuo Nishi
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021422
doi: 10.1371/journal.pgph.0001229
license: CC BY 4.0
---
# Assessment of Mongolian dietary intake for planetary and human health
## Abstract
### Background
Healthy diets that consider environmental pressures are required to meet sustainable development goals in Mongolia. This study aimed to clarify the extent of planetary and human health on Mongolian dietary intake.
### Methods
The intake of eight food groups (g/day) was investigated using the national database of the Household socio-economic survey (HSES) 2019 in Mongolia. The boundary intake of the Planetary health diet (PHD) proposed by the EAT-Lancet Commission was considered $100\%$ adequate. The adequacy (%) of food groups in the HSES were calculated in two areas (urban and rural), during the two seasons (cold and warm), and the total by each boundary of the PHD. The differences between the recommended dietary intake (RDI) in Mongolia and the PHD were also investigated in the same manner.
### Results
The adequacy of red meat (i.e., beef, mutton, and horsemeat) in whole areas of Mongolia indicated more than 17 times higher intake (1,$738\%$) than the PHD. The adequacy of vegetables ($20\%$) and fruits ($8\%$) in Mongolia indicated an intake shortage compared to the PHD. These discrepancies in dietary adequacy were greater in rural areas and during the cold seasons than in urban areas and during the warm seasons, respectively. The animal-based protein sources, especially red meat (1,$091\%$), in the RDI of Mongolia were higher than those in the PHD.
### Conclusion
This study found a highly excessive intake of red meat and a low intake of vegetables and fruits compared with the PHD among Mongolian people, especially in rural areas and during the cold seasons. The limited diversity of food in severe geographic conditions, poor accessibility of food retailers, and insufficient nutrition education may have led to these results. Therefore, improvements in the food environment and nutritional education are required.
## Introduction
Food production negatively influences global environmental change, which threatens food security at the same time [1]. Mongolia has experienced noticeable climate change, with an increase in the average temperature of ≥ 2°C and a decline in rainfall between 1940 to 2015 [2]. Extreme weather, such as drought in summer, results in the loss of livestock and threatens food security and the population’s health [2]. Food production generates greenhouse-gas (GHG) emissions, nitrogen and phosphorus pollution, biodiversity loss, and water and land use [3]. The global GHG emissions from food production include $57\%$ from animal-based food (including livestock feed), $29\%$ from plant-based foods, and $14\%$ from other uses [4].
In addition to securing natural resources, dietary habits are also important when considering population health. In Mongolia, the prevalence of obesity ($20.9\%$ in men and $26.5\%$ in women) among adults aged 18 years and older is higher than in the Asian region ($17\%$ in men and $20\%$ in women), in addition to the double burden of malnutrition among children and women [5]. Overconsumption of animal-based foods may threaten population health. A review indicated that total protein and animal protein were associated with the risk of cardiovascular diseases and diabetes [6]. Prospective cohort studies and meta-analyses reported that total protein intake was positively associated with all-cause mortality, and higher animal protein intake was associated with mortality from cardiovascular disease [7].
The EAT-Lancet Commission proposed the Planetary health diet (PHD) [3], a framework of planetary boundaries that indicates the intake ranges of food groups to ensure human health and environmental sustainability. The framework recommends the predominant consumption of plant-based foods (vegetables, greens, fruits, and whole grains) and small amounts of animal-based foods (meat, fish, and eggs). This framework has been used in several studies. Studies in Brazil developed the PHD index, confirmed its validity and reliability [8], and found that high adherence to the index was associated with a lower prevalence of obesity [9]. A study in India reported that people consumed cereals, fruits, and vegetables but not enough protein compared with those in the PHD [10]. A study in Denmark proposed the development of a Danish diet adapted to a healthy plant-based diet aligned with the PHD [11]. The benefits of the PHD have been reported not only for health but also for socio-economic reasons. For example, an Australian study reported that a PHD basket was less expensive and more affordable than a typical Australian diet basket [12].
Although some countries have assessed their diets based on the PHD, there is no evidence of the PHD in Mongolia. Traditional Mongolian food is based on the products of nomadic animal herders who raise Mongolian steppe meat and milk [13]. Nomadic culture continues to be practiced in rural areas. Dietary habits in urban areas changed with socio-economic growth, and dietary patterns were positively associated with body mass index [13]. Food consumption is seasonal, particularly in rural areas. Dairy products and meat are highly consumed during winter [14]. In this context, assessing Mongolian diets in urban and rural areas and during cold and warm seasons is needed from the perspective of both planetary and human health.
Therefore, this study aimed to clarify how Mongolian dietary intake was aligned with the PHD using a national survey and compare it between areas (i.e., urban and rural) and seasons (i.e., cold and warm). In addition, to understand the sustainability of national dietary recommendations, this study investigated the differences between national dietary targets and the PHD.
## Study setting
Mongolian dietary intake was investigated using open-source data from the Household socio-economic survey (HSES) performed in 2019 [15]. The health boundary of the PHD was set as the benchmark for dietary intake. The recommended dietary intake (RDI) in Mongolia was used to assess the sustainability of the national dietary recommendations [17].
## Target population
The HSES is a nationally representative survey that estimates and monitors a country’s level of poverty and people’s living standards. The HSES 2019 was conducted following the procedure of the HSES 2016 [15, 16].This survey used the sampling frame developed by the National Statistics Office, based on population figures obtained from administrative records.
## Data collection
The 936 households were randomly surveyed each month from January 1st 2019 to January 1st 2020, for the HSES 2019, and a total of 11,232 households were selected. Of the total households, 11,197 participated in the survey ($99.7\%$ participation rate).
Urban and rural areas were classified according to the following steps. First, geographic domains were classified into four residential zones: Ulaanbaatar as an urban area, and as rural areas, 20 aimag (province) centers, 306 soum centers (i.e., a secondary subdivision outside Ulaanbaatar), and 891 Bags. Second, a primary sampling unit was selected in each zone using the probability proportional estimated size. Finally, 3,600 households in urban areas and 7,597 households in rural areas were randomly chosen from primary sampling units.
The HSES investigated dietary intake during the cold (October to March) and warm (April to September) seasons. However, there was no information on whether dietary intake during the two seasons was investigated in all 11,197 households.
## Questionnaires
The core questionnaire of household socio-economic data and household food consumption was made according to the previous surveys [15, 16]. In the household socio-economic data, to indicate the socio-demographics, this study employed age (18 years and older), sex (men), number of household members (four [median] or more), type of dwelling (Ger), raised or owned livestock (herding, poultry, or any animal) (yes), owned agricultural land (yes), and household enterprise (yes). In the household food consumption, the field offices transmitted the data and provide additional clarification to a survey team in the National Statistics Office through field supervisors. The survey team in the National Statistics Office performed logical and consistent checks for all data. A representative household reported a dietary record and some households were asked to revise their answers whenever the field office found an error. A 30-day dietary record compiled by a researcher every 10 days, three times during a single month, was recorded to capture the household’s food consumption in urban areas. A 7-day dietary record was administered to all provinces in rural areas using the following question: “how much food item have you consumed in total during the past 7 days?” In both urban and rural areas, the household representative answered the question, “did you or any member of your household spend the following items during the past month/ or the past 7 days?” If the answer was “no,” the major place of “restaurants, cafes” or “canteens in schools, works canteens” was selected.
## EAT-Lancet reference diet
This study used the PHD as a reference diet to assess the dietary intake of Mongolian [3]. The PHD indicated the scientific targets and ranges of intake of 11 food groups (e.g., 300 g/day ranged from 200 to 600 g/day in all vegetables), added fats (unsaturated and saturated oils), and added sugars (all sugars) per 2,500 kcal/day of total energy. In addition, six food groups (beef, lamb, and pork; chicken and other poultry; eggs; fish; legumes; and nuts) were included as protein sources.
## Recommended dietary intake in Mongolia
The RDI in Mongolia indicates the daily target consumption of meat, meat products, flour, bakery products and various types of rice based on the average daily consumption in Mongolians [17]. The targets of total daily energy intake and energy from fat and nutrients (i.e., unsaturated and saturated fats) were referenced according to the human energy requirement in $\frac{2001}{2002}$ proposed by the Ministry of Health in Mongolia [18].
## Food groups
This study classified food items of the HSES 2019 based on the food-based dietary guidelines “Ger” into the PHD food groups (Table 1) [19]. The national food guide is designed to shape a Mongolian wooden tent “Ger.” The food guide is divided into three food group layers: cereals and cereal products are placed at the bottom of the tent; vegetables, meat, fish, and eggs are placed at the second level; and fruits and dairy products are placed at the final level [19]. Although the EAT-Lancet Commission named the food group “beef, lamb, and pork” red meat [20], beef, mutton, horse meat, and camel meat are among the major red meats in Mongolia [15]. Therefore, this study changed the name to “red meat” to be understandable. Confectionery products and sugar were included in the “added sugar” in the PHD food group since the HSES indicated the intake of the two foods as one food group. We did not classify whole grains because the information was not available in the HSES.
**Table 1**
| Planetary health diets | Household socio-economic survey |
| --- | --- |
| Food group | Food item |
| Whole grains | Not available |
| Tubers or starchy vegetables | Potatoes, chips, and potato products |
| All vegetables | Green, yellow, and other vegetables |
| All fruits | Fresh fruit (e.g., wild fruit, apple, orange, kiwi, banana, pineapple, watermelon, seabucktorn, blueberry) |
| Dairy foods | Milk, cheese, curd, cream, Mongolian cheese, yogurt, mare milk, other dairy products, and other milks |
| Protein sources | |
| Red meat | Beef, mutton, horsemeat, goat meat, camel meat, pork, ham and sausages, other meats |
| Chicken and other poultry | Chicken and duck |
| Egg | Egg, other poultry eggs, and dried eggs |
| Fish | Fresh fish, salmon, white fish, canned fish, and smoked fish |
| Legumes | Corn, tofu, bean, peas, and preserved pea |
| Nuts | Tree nuts, ground nuts, walnuts, and any seeds |
| Added fats | |
| Unsaturated oils | Margarine and vegetable oils |
| Saturated oils | Butter, animal fats, lard, and suet |
| Added sugars | Sugar and confectionery products |
## Socio-demographics
The proportion of people aged 18 years and older and men was calculated by individuals. The proportion of four or more household members, Ger of dwelling, raised or owned livestock, owned agricultural land, and household enterprise was calculated per household. The result was shown as a whole area because no data were available on the living areas (i.e., urban and rural) in which the houesholds lived.
## Adequacy of dietary intake against the PHD
The health boundary and range (g/day) of each food group against 2,500 (kcal/day) proposed by the PHD were converted by the total energy intake (3,085 kcal/day) among Mongolian people aged 18 years and older. The adequacy (%) of each food group was calculated by dividing dietary intake (g/day) by the converted health boundary (g/day). Similarly, we calculated the health boundary, range, and adequacy in two areas (urban and rural), during the two seasons (warm and cold), and the RDI in Mongolia using each dietary (g/day) and energy intake (2,863 kcal/day in urban areas, 3,529 kcal/day in rural areas, 3,057 kcal/day during the cold season, 3,111 kcal/day during the warm season, and 2,400 kcal/day in the RDI in Mongolia). The food groups of chicken and other poultry, egg, fish, legumes, and nuts were not used in these two areas because this information was not presented in the database. For the same reason, this study did not use nuts in either season or RDI in Mongolia.
## Ethics
This study used tabulated and published information on the HSES. The National Statistics Office obtained informed consent from the household representatives.
## Characteristics of the population
The $64\%$ of individuals in the households were over 18 years ($48\%$ men) (Table 2). Over half of households lived with four or more members, and $41\%$ of them lived in Ger. The $34\%$ and $5.5\%$ of households possessed the livestock and agricultural land, respectively. A household enterprise was present in $12\%$ of households.
**Table 2**
| Characteristics a | N (%) |
| --- | --- |
| Individual b | |
| Age, ≥ 18 years old | 26182 (64) |
| Sex, men | 19909 (48) |
| Household | |
| The number of household members, ≥ 4 c | 6015 (54) |
| Type of dwelling, Ger | 4636 (41) |
| Raised or owned livestock (i.e., herding, poultry, or any animal), yes | 3810 (34) |
| Owned agricultural land, yes | 620 (5.5) |
| Household enterprise, yes | 1318 (12) |
## Dietary intake in whole and two areas in comparison to the PHD
Table 3 compares Mongolian dietary intake with the PHD in whole areas, urban areas, and rural areas. In whole areas, one of the major differences between the two diets was red meat which was more than 17 times higher intake (300 g/day, 1,$738\%$ adequacy) in Mongolia than that recommended by the PHD (17 g/day ranged from 0 to 35 g/day). On the other hand, all vegetables and fruits in Mongolia were lower (73 g/day, $20\%$ adequacy in all vegetables and 20 g/day, $8.1\%$ adequacy in all fruits) than these in the PHD (370 g/day ranged from 247 to 740 g/day in all vegetables and 247 g/day ranged from 123 to 370 g/day in all fruits). Other food groups indicated $100\%$ and more adequacy (105–$146\%$ adequacy) but almost within the range of each PHD, except for unsaturated oils ($13\%$ adequacy). High intake of red meat and low intake of all vegetables and fruits were more evident in rural areas than urban areas.
**Table 3**
| Unnamed: 0 | Whole areas | Unnamed: 2 | Unnamed: 3 | Urban areas | Unnamed: 5 | Unnamed: 6 | Rural areas | Unnamed: 8 | Unnamed: 9 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Food groups | Dietary intake (g/day)a | PHD boundary (range) (g/day)b | Adequacy (%)c | Dietary intake (g/day)a | PHD boundary (range) (g/day)b | Adequacy (%)c | Dietary intake (g/day)a | PHD boundary (range) (g/day)b | Adequacy (%)c |
| Whole grains | N.A. | 286 | N.A. | N.A. | 266 | N.A. | N.A. | 328 | N.A. |
| Tubers or starchy vegetables | 90 | 62 (0, 123) | 146 | 97 | 57 (0, 115) | 169 | 83 | 71 (0, 141) | 118 |
| All vegetables | 73 | 370 (247, 740) | 20 | 87 | 344 (229, 687) | 25 | 57 | 424 (282, 847) | 14 |
| All fruits | 20 | 247 (123, 370) | 8.1 | 23 | 229 (115, 344) | 10 | 17 | 282 (141, 424) | 6.0 |
| Dairy foods | 366 | 309 (0, 617) | 119 | 273 | 286 (0, 573) | 95 | 480 | 353 (0, 706) | 136 |
| Protein sources | | | | | | | | | |
| Red meat | 300 | 17 (0, 35) | 1738 | 247 | 16 (0, 32) | 1540 | 367 | 20 (0, 40) | 1963 |
| Chicken and other poultry | N.A. | 36 (0, 72) | N.A. | N.A. | 33 (0, 66) | N.A. | N.A. | 41 (0, 82) | N.A. |
| Egg | N.A. | 16 (0, 31) | N.A. | N.A. | 15 (0, 29) | N.A. | N.A. | 18 (0, 35) | N.A. |
| Fish | N.A. | 35 (0, 123) | N.A. | N.A. | 32 (0, 115) | N.A. | N.A. | 40 (0, 141) | N.A. |
| Legumes | N.A. | 93 (0, 123) | N.A. | N.A. | 86 (0, 115) | N.A. | N.A. | 106 (0, 141) | N.A. |
| Nuts | N.A. | 62 (0, 93) | N.A. | N.A. | 57 (0, 86) | N.A. | N.A. | 71 (0, 106) | N.A. |
| Added fats | | | | | | | | | |
| Unsaturated oils | 6.6 | 49 (25, 99) | 13 | 10 | 46 (23, 92) | 22 | 6.7 | 57 (28, 113) | 12 |
| Saturated oils | 16 | 15 (0, 15) | 110 | 13 | 14 (0, 14) | 96 | 20 | 17 (0, 17) | 121 |
| Added sugars | 40 | 38 (0, 38) | 105 | 43 | 36 (0, 36) | 121 | 40 | 44 (0, 44) | 92 |
## Dietary intake during two seasons in comparison to the PHD
The differences in dietary intake during the cold and warm seasons compared to the PHD are shown in Table 4. The adequacy of red meat during the cold season (324 g/day, 1,$905\%$ adequacy) was higher than that during the warm season (313 g/day, 1,$812\%$ adequacy). On the other hand, other consumptions of protein sources, such as chicken and other poultry, fish, and legumes during the cold and warm seasons were lower (0.2–$15\%$ adequacy) than those in the PHD, except for eggs ($86\%$ adequacy during the cold season and $97\%$ adequacy during the warm season).
**Table 4**
| Unnamed: 0 | Cold seasons | Unnamed: 2 | Unnamed: 3 | Warm seasons | Unnamed: 5 | Unnamed: 6 |
| --- | --- | --- | --- | --- | --- | --- |
| Food groups | Dietary intake (g/day)a | PHD boundary (range) (g/day)b | Adequacy (%)c | Dietary intake (g/day)a | PHD boundary (range) (g/day)b | Adequacy (%)c |
| Whole grains | N.A. | 284 | N.A. | N.A. | 289 | N.A. |
| Tubers or starchy vegetables | 94 | 61 (0, 122) | 154 | 92 | 62 (0, 123) | 148 |
| All vegetables | 84 | 367 (244, 734) | 23 | 85 | 370 (247, 740) | 23 |
| All fruits | 31 | 244 (122, 367) | 13 | 36 | 247 (123, 370) | 15 |
| Dairy foods | 231 | 305 (0, 610) | 76 | 250 | 309 (0, 617) | 81 |
| Protein sources | | | | | | |
| Red meat | 324 | 17 (0, 34) | 1905 | 313 | 17 (0, 35) | 1812 |
| Chicken and other poultry | 4.2 | 35 (0, 71) | 12 | 5.4. | 36 (0, 72) | 15 |
| Egg | 14 | 16 (0, 31) | 86 | 16. | 16 (0, 31) | 97 |
| Fish | 0.6 | 34 (0, 122) | 1.8 | 0.7 | 35 (0, 123) | 2.0 |
| Legumes | 0.2 | 92 (0, 122) | 0.2 | 0.2 | 93 (0, 123) | 0.2 |
| Nuts | N.A. | 61 (0, 92) | N.A. | N.A. | 62 (0, 93) | N.A. |
| Added fats | | | | | | |
| Unsaturated oils | 19 | 49 (24, 98) | 38 | 18 | 50 (25, 100) | 36 |
| Saturated oils | 7.3 | 14 (0, 14) | 51 | 7.1 | 15 (0, 15) | 48 |
| Added sugars | 47 | 38 (0, 38) | 124 | 53 | 38 (0, 38) | 137 |
## National dietary recommendations in comparison to the PHD
The RDI in Mongolia, compared with the PHD, is presented in Table 5. The three most abundant protein sources were red meat (120 g/day, 1,$091\%$ adequacy), eggs (20 g/day, $200\%$ adequacy), and chicken and other poultry (40 g/day, $182\%$ adequacy). The adequacy of other protein sources ranged from $70\%$ to $143\%$. RDIs in all protein sources were within the range of each PHD, except for red meat (120 g/day in RDI in Mongolia and 0–21 g/day in the PHD). All vegetables and fruits in the RDI in Mongolia were 260 g/day ($113\%$ adequacy) and 200 g/day ($131\%$ adequacy), respectively, and were within the range of each PHD (153–459 g/day for all vegetables and 77–230 g/day for all fruits).
**Table 5**
| Food groups | RDI in Mongolia (g/day)a | PHD boundary (range) (g/day)b | Adequacy (%)c |
| --- | --- | --- | --- |
| Whole grains | N.A. | 178 | N.A. |
| Tubers or starchy vegetables | 120 | 38 (0, 77) | 315 |
| All vegetables | 260 | 230 (153, 459) | 113 |
| All fruits | 200 | 153 (77, 230) | 131 |
| Dairy foods | 340 | 191 (0, 383) | 178 |
| Protein sources | | | |
| Red meat | 120 | 11 (0, 21) | 1091 |
| Chicken and other poultry | 40 | 22 (0, 44) | 182 |
| Egg | 20 | 10 (0, 19) | 200 |
| Fish | 30 | 21 (0, 77) | 143 |
| Legumes | 40 | 57 (0, 77) | 70 |
| Nuts | N.A. | 38 (0, 57) | N.A. |
| Added fats | | | |
| Unsaturated oils | 23 | 31 (15, 61) | 74 |
| Saturated oils | 10 | 9.0 (0, 9.0) | 111 |
| Added sugars | 33 | 24 (0, 24) | 137 |
## Discussion
This study clarified that Mongolian people have an extremely high intake of red meat and a low intake of vegetables and fruits based on the PHD recommendations. These results were more evident in rural areas and during the cold season than in urban areas and during the warm season, respectively. This is the first study to indicate the extent of the discrepancy between the current Mongolian dietary intake and the PHD for planetary and human health.
The present results of high consumption of red meat and low consumption of vegetables and fruits were similar to the global trend [20], except for countries such as India where a vegetarian diet is practiced [10]. Nevertheless, the discrepancy observed in this study was much larger than the global trend [20], Brazil [9], and Denmark [11]. Furthermore, the population intake of red meat, vegetables, and fruits did not meet the RDI in Mongolia in the present study. Among multifactorial interactions, such as limited food availability and accessibility, a lack of nutrition and health knowledge is likely to be a key factor in these results. A review found success in increasing vegetables and fruits intake in many countries by improving nutrition education [21]. A basic and robust system to disseminate nutrition education is required to reach the target of RDI in Mongolia. A study reported that even medical professionals lacked accurate knowledge of the recommended daily salt intake (5 g/day) [22]. Therefore, accurate nutrition education considering sustainable healthy diets to health professionals and communities is required to nudge them to choose healthier and more sustainable food.
This study showed that a high intake of red meat and dairy products and a low intake of vegetables and fruits were more evident in rural areas than urban areas. A similar result was reported in the previous study that a “Nomadic” dietary pattern indicated a high consumption of dairy products, milk, red meat, and refined grains, and low juice and sugar-sweetened beverages, processed meat, and fruit [13]. In addition, the Nomadic pattern was associated with increased iron and zinc intake and decreased fiber intake [13]. Nomadic dietary patterns may result in obesity and a high risk of cardiovascular diseases [6, 7]. According to the geographic characteristics in this study, people living with four or more family members in Ger and owning livestock are likely to have traditional dietary habits. As rural populations practice, traditional Mongolian diets are characterized by a high intake of dairy products (i.e., milk and natural yogurt), fats and oil, sugar, confectionery, and horsemeat [15]. In addition, nomadic herders usually feed themselves, especially on meat and dairy products [13]. These traditional dietary cultures imply that the limited accessibility, availability, and affordability of food retail and variety of food, such as fresh vegetables and fruits, are present in rural areas compared with urban areas.
The environmental impact may differ between urban and rural areas. The GHG emissions from livestock, mainly consumed in urban areas, come from several sources, such as emissions related to feed (e.g., fertilizer and land use), processing, transport, on-farm energy use, and enteric fermentation [23]. The environmental impact generated by livestock and wild game consumed by self-sufficiency in rural areas may be lower than that generated by livestock consumed by people in urban areas, even if there is high consumption of red meat in rural areas.
A higher intake of red meat during the cold season than during the warm season would have a specific background in a severe climate. During the coldest season, the traditional dietary pattern of high red meat (i.e., horsemeat) and fats is usually consumed [24], especially in rural areas, to preserve sufficient energy reserves at severe temperatures [14]. Therefore, the difference in dietary intake between the two seasons should be considered a strategy for sustainable healthy diets.
In this study, animal-based protein sources in the RDI in Mongolia were higher than those in the PHD, particularly red meat. Studies in America and Italy have reported differences between national dietary guidelines and the PHD due to their dietary habits and traditional cultures [25, 26]. A review suggested that national food-based dietary guidelines could be sustainable and healthy to some extent, even if dietary goals are not completely aligned with global health and environmental targets [27]. A new RDI in Mongolia may be required to originally achieve planetary and human health with consideration of feasibility, such as food culture, geographic characteristics, and food variety.
This study did not indicate the intake of whole grains due to the limited information from the HSES. A study reported that the usual mean intake (g/2,500kcal/day) of whole grains (ranging from 2.2 to 20) was lower than that of refined grains (ranging from 361 to 461) in eight provinces in Mongolia [13]. The consumption of refined grains, such as bread, pasta, and rice, may be associated with westernization [15].
The present results are generalizable because this study used an open-source national survey [15]. However, this study has some limitations that warrant mention. First, this study did not include food groups because the lack of information. The intake of whole grains has not been investigated in the HSES and the RDI in Mongolia. Some protein sources in two areas (i.e., chicken and other poultry, eggs, fish, legumes, and nuts) and during the two seasons (i.e., nuts) were investigated but not published in detail. Furthermore, the results of added sugar did not reflect only sugar intake, as this study included confectionery products. The fifth National Nutrition Survey reported the nutritional status of the Mongolian population but did not publish data on food-based dietary consumption [28]. More data should be considered in future studies to clarify the overall status of food intake. Second, the different processes of the 30-day dietary record in urban areas and the 7-day dietary record in rural areas made precise comparison between the two areas difficult. The survey in rural areas was conducted using simple methods compared to urban areas because of the limited resources for the survey, such as manpower. Third, this study did not necessarily adapt Mongolian food culture to the PHD, such as the characteristics of dietary cultures and food availability. The EAT-Lancet Commission recommends the local interpretation and adaptation of the universally applicable PHD [20]. According to the RDI in Mongolia [17], the recommended intake of red meat in the PHD may not be feasible for Mongolian diets. Fourth, the PHD targeted adults aged 18 years and older. Given that the data were available, the interpretation classified by sex and age group (i.e., children and older adults) would differ from the present results. Fourth, the assessment of sustainable healthy diets in Mongolia was indecisive only from this study, using one measurement. The measurement of GHG emissions, water and land use, and nitrogen and phosphorus fertilizer application may help us deeply understand this comprehensive assessment.
## Conclusions
This study indicated an extremely high intake of red meat and a low intake of vegetables and fruits compared to the recommended intake of the PHD among Mongolian people. This discrepancy was larger in rural areas and during the cold season than in urban areas and during the warm season, respectively. To prevent health inequality due to the geographic and seasonal situation of planetary and human health, further policies for multi-sectoral interventions, such as fields of infrastructure and education systems, are required to improve the accessibility, availability, and affordability of healthy food, as well as nutrition education.
## References
1. Myers SS, Smith MR, Guth S, Golden CD, Vaitla B, Mueller ND. **Climate Change and Global Food Systems: Potential Impacts on Food Security and Undernutrition**. *Annu Rev Public Health* (2017.0) **38** 259-277. DOI: 10.1146/annurev-publhealth-031816-044356
2. 2The World Bank Group and the Asian Development Bank. Climate Risk Country Profile: Mongolia. The World Bank Group and the Asian Development Bank. 2021. Available from: https://www.adb.org/sites/default/files/publication/709901/climate-risk-country-profile-mongolia.pdf. (2021.0)
3. Willett W, Rockström J, Loken B, Springmann M, Lang T, Vermeulen S. **Food in the Anthropocene: the EAT-Lancet Commission on healthy diets from sustainable food systems**. *Lancet* (2019.0) **393** 447-492. DOI: 10.1016/S0140-6736(18)31788-4
4. Xu X, Sharma P, Shu S, Lin TS, Ciais P, Tubiello FN. **Global greenhouse gas emissions from animal-based foods are twice those of plant-based foods**. *Nature Food* (2021.0) **2** 724-732. DOI: 10.1038/s43016-021-00358-x
5. 5Minister of Health in Mongolia, The National Center for Public Health, and World Health Organization. Fourth national STEPS Survey on Prevalence of Noncommunicable Disease and injury Risk Factors-2019. World Health Organization. 2020. Available from: https://cdn.who.int/media/docs/default-source/ncds/ncd-surveillance/data-reporting/mongolia/mongolia-steps-survey—2019_brief-summary_english.pdf?sfvrsn=5ba7a1d3_1&download=true
6. Mariotti F. **Animal and Plant Protein Sources and Cardiometabolic Health**. *Adv Nutr* (2019.0) **10** S351-s66. DOI: 10.1093/advances/nmy110
7. Chen Z, Glisic M, Song M, Aliahmad HA, Zhang X, Moumdjian AC. **Dietary protein intake and all-cause and cause-specific mortality: results from the Rotterdam Study and a meta-analysis of prospective cohort studies**. *Eur J Epidemiol* (2020.0) **35** 411-429. DOI: 10.1007/s10654-020-00607-6
8. Cacau LT, De Carli E, de Carvalho AM, Lotufo PA, Moreno LA, Bensenor IM. **Development and Validation of an Index Based on EAT-Lancet Recommendations: The Planetary Health Diet Index.**. *Nutrients* (2021.0) 13. DOI: 10.3390/nu13051698
9. Cacau LT, Benseñor IM, Goulart AC, Cardoso LO, Lotufo PA, Moreno LA. **Adherence to the Planetary Health Diet Index and Obesity Indicators in the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil).**. (2021.0) 13. DOI: 10.3390/nu13113691
10. Sharma M, Kishore A, Roy D, Joshi K. **A comparison of the Indian diet with the EAT-Lancet reference diet**. *BMC Public Health* (2020.0) **20** 812. DOI: 10.1186/s12889-020-08951-8
11. Lassen AD, Christensen LM, Trolle E. **Development of a Danish Adapted Healthy Plant-Based Diet Based on the EAT-Lancet Reference Diet.**. *Nutrients* (2020.0) 12. DOI: 10.3390/nu12030738
12. Goulding T, Lindberg R, Russell CG. **The affordability of a healthy and sustainable diet: an Australian case study.**. *Nutr J* (2020.0) **19** 109. DOI: 10.1186/s12937-020-00606-z
13. Bromage S, Daria T, Lander RL, Tsolmon S, Houghton LA, Tserennadmid E. **Diet and Nutrition Status of Mongolian Adults.**. *Nutrients* (2020.0) 12. DOI: 10.3390/nu12051514
14. 14Food and Agriculture Organization of the United Nations. FAO/UNICEF/UNDP Report: Joint Food Security Assessment Mission to Mongolia. Food and Agriculture Organization of the United Nations. 2007. Available from: https://www.fao.org/3/j9883e/j9883e00.htm. (2007.0)
15. 15The National Statistical Office of Mongolia. Household Socio-economic survey 2019; 2022 [cited 2022 June 16]. Database: Statistical Microdata [Internet]. Available from: http://web.nso.mn/nada/index.php/catalog/HSES
16. 16The National Statistics Office of Mongolia. Household Socio Economic Survey 2016; 2019 [cited 2022 July 24]. Database: International Household Survey Network [Internet]. Available from: https://catalog.ihsn.org/catalog/8346/study-description
17. 17Nutrition department, National Center for Public Health, Ministry of Health in Mongolia. The recommended dietary intake in Mongolia. 2021 [cited 2022 June 16]. In: legalinfo.mn [Internet]. Ulaanbaatar: Interactive. Available from: https://legalinfo.mn/mn/detail?lawId=207686&showType=1 (in Mongolian)
18. 18Nutrition department, National Center for Public Health, Ministry of Health in Mongolia. Human energy requirement. 2021 [cited 2022 June 16]. In: legalinfo.mn [Internet]. Ulaanbaatar: Interactive. Available from: https://legalinfo.mn/mn/detail/12552. (in Mongolian)
19. 19Nutrition department, National Center for Public Health, Ministry of Health in Mongolia. Food-based dietary guidelines–Mongolia. 2022 [cited 2022 June 16]. In: Food-based dietary guidelines [Internet]. Rome: FAO. Available from: https://www.fao.org/nutrition/education/food-based-dietary-guidelines/regions/countries/mongolia/fr/.
20. 20EAT-Lancet Commission. Summary Report of the EAT-Lancet Commission, Healthy Diets From Sustainable Food Systems. EAT-Lancet Commission. 2019 [cited 2022 June 16]. In: The EAT-Lancet Commission on Food, Planet, Health [Internet]. Stockholm: EAT. Available from: https://eatforum.org/eat-lancet-commission/eat-lancet-commission-summary-report/.. (2019.0)
21. Pem D, Jeewon R. **Fruit and Vegetable Intake: Benefits and Progress of Nutrition Education Interventions- Narrative Review Article.**. *Iran J Public Health* (2015.0) **44** 1309-1321. PMID: 26576343
22. Hikita N, Batsaikhan E, Sasaki S, Haruna M, Yura A, Oidovsuren O. **Factors Related to Lacking Knowledge on the Recommended Daily Salt Intake among Medical Professionals in Mongolia: A Cross-Sectional Study.**. *Int J Environ Res Public Health* (2021.0) 18. DOI: 10.3390/ijerph18083850
23. 23Barthelmie. Impact of Dietary Meat and Animal Products on GHG Footprints: The UK and the US. Climate 2022; 10:43. 10.3390/cli10030043. DOI: 10.3390/cli10030043
24. Dugee O, Khor GL, Lye MS, Luvsannyam L, Janchiv O, Jamyan B. **Association of major dietary patterns with obesity risk among Mongolian men and women.**. *Asia Pac J Clin Nutr* (2009.0) **18** 433-40. PMID: 19786392
25. Blackstone NT, Conrad Z. **Comparing the Recommended Eating Patterns of the EAT-Lancet Commission and Dietary Guidelines for Americans: Implications for Sustainable Nutrition**. *Curr Dev Nutr.* (2020.0) **4** nzaa015. DOI: 10.1093/cdn/nzaa015
26. Tucci M, Martini D, Del Bo C, Marino M, Battezzati A, Bertoli S. **An Italian-Mediterranean Dietary Pattern Developed Based on the EAT-Lancet Reference Diet (EAT-IT): A Nutritional Evaluation**. *Foods* (2021.0) 10. DOI: 10.3390/foods10030558
27. Springmann M, Spajic L, Clark MA, Poore J, Herforth A, Webb P. **The healthiness and sustainability of national and global food based dietary guidelines: modelling study**. *BMJ* (2020.0) **370** m2322. DOI: 10.1136/bmj.m2322
28. 28Ministry of Health, National Center for Public Health, UNICEF. Nutrition Status of the Population of Mongolia, Fifth National Nutrition Survey Report. UNICEF; 2017. Available from: https://www.unicef.org/mongolia/media/1116/file/NNS_V_undsen_tailan_EN.pdf. (2017.0)
|
---
title: 'Patterns of non-communicable comorbidities at start of tuberculosis treatment
in three regions of the Philippines: The St-ATT cohort'
authors:
- Sharon E. Cox
- Tansy Edwards
- Benjamin N. Faguer
- Julius P. Ferrer
- Shuichi J. Suzuki
- Mitsuki Koh
- Farzana Ferdous
- Naomi R. Saludar
- Anna-Marie C. G. Garfin
- Mary C. Castro
- Juan A. Solon
journal: PLOS Global Public Health
year: 2021
pmcid: PMC10021424
doi: 10.1371/journal.pgph.0000011
license: CC BY 4.0
---
# Patterns of non-communicable comorbidities at start of tuberculosis treatment in three regions of the Philippines: The St-ATT cohort
## Abstract
Diabetes and undernutrition are common risk factors for tuberculosis (TB), associated with poor treatment outcomes and exacerbated by TB. Limited data exist describing patterns and risk factors of multiple comorbidities in persons with TB. Nine-hundred participants ($69.6\%$ male) were enrolled in the Starting Anti-TB Treatment (St-ATT) cohort, including 133 ($14.8\%$) initiating treatment for multi-drug resistant TB (MDR-TB). Comorbidities were defined as: diabetes, HbA1c ≥$6.5\%$ and/or on medication; hypertension, systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg and/or on medication; anaemia (moderate/severe), haemoglobin <11g/dL; and, undernutrition (moderate/severe) body-mass-index <17 kg/m2. The most common comorbidities were undernutrition $23.4\%$ ($\frac{210}{899}$), diabetes $22.5\%$ ($\frac{199}{881}$), hypertension $19.0\%$ ($\frac{164}{864}$) and anaemia $13.5\%$ ($\frac{121}{899}$). Fifty-eight percent had ≥1 comorbid condition ($\frac{496}{847}$), with $17.1\%$ having ≥2; most frequently diabetes and hypertension ($$n = 57$$, $6.7\%$). Just over half of diabetes ($54.8\%$) and hypertension ($54.9\%$) was previously undiagnosed. Poor glycemic control in those on medication (HbA1c≥$8.0\%$) was common ($$n = 50$$/73, $68.5\%$). MDR-TB treatment was associated with increased odds of diabetes (Adjusted odds ratio (AOR) = 2.48, $95\%$ CI: 1.55–3.95); but decreased odds of hypertension (AOR = 0.55, $95\%$ CI: 0.39–0.78). HIV infection was only associated with anaemia (AOR = 4.51, $95\%$ CI: 1.01–20.1). Previous TB treatment was associated with moderate/severe undernutrition (AOR = 1.98, $95\%$ CI: 1.40–2.80), as was duration of TB-symptoms before starting treatment and household food insecurity. No associations for sex, alcohol or tobacco use were observed. MDR-TB treatment was marginally associated with having ≥2 comorbidities (OR = 1.52, $95\%$ CI: 0.97–2.39). TB treatment programmes should plan for large proportions of persons requiring diagnosis and management of comorbidities with the potential to adversely affect TB treatment outcomes and quality of life. Dietary advice and nutritional management are components of comprehensive care for the above conditions as well as TB and should be included in planning of patient-centred services.
## Introduction
Tuberculosis (TB) remains the leading cause of death globally from an infectious disease with strong poverty-associated social determinants including malnutrition [1, 2]. Philippines is a high burden country for both drug-sensitive and drug-resistant TB, with one of the highest estimated incidence rates ($\frac{554}{100}$,000) globally in the context of low incidence of HIV [2]. Many low- and middle-income countries, similar to Philippines, are undergoing nutrition transition with rapidly increasing nutrition-related non-communicable diseases (NCDs) such as diabetes and hypertension, not limited to higher income groups and associated with poor-quality but energy-dense diets, sedentary lifestyles and other behavioural risk factors such as alcohol and tobacco use, also risk factors for TB [3, 4].
Undernutrition is an important risk factor for developing TB disease [5], can also result from its physiological and socio-economic consequences [6] and is associated with adverse TB treatment outcomes including death [7]. Diabetes is also a risk factor for developing active TB disease and adverse treatment outcomes including death and relapse or recurrent TB [8–10]. In the Philippines undernutrition and diabetes are estimated to be the leading population level risk factors for TB, estimated to affect $13.3\%$ and $7.1\%$ of adults [3, 11]. There is an increasing recognition of the burden and impact of multimorbidity (the co-existence of 2 or more medical conditions) [12] for TB and other health outcomes, but also of opportunities within TB programmes for increasing screening and uptake of NCD health services [13]. However, despite many NCDs and TB sharing nutrition-related risk factors and management strategies, consideration of nutrition screening and linkage with nutrition services for TB and NCDs is rarely considered.
Limited systematic data of multimorbidity in TB exists and often relies on retrospective data, which due to high levels of undiagnosed NCDs and lack of routine screening or data capture/reporting (e.g., body mass index (BMI)) are probable under-estimates. Furthermore, reports often focus on only one condition, most often diabetes [14] and more recently depression or common mental disorders [15]. A recent secondary analysis of WHO Global Health *Survey data* (collected in 2003) indicates that up to two-thirds of people with TB may have one or more NCD [16]. However, these data included NCDs based on self-report and TB disease based on self-reported TB symptoms of cough and or haemoptysis. Moreover, this survey did not collect information on hypertension or nutritional status (under- or over-weight) or anaemia. Previously, within a cross-sectional study of persons on TB treatment we reported that up to $40\%$ of Filipino persons registered for TB treatment had at least one comorbidity of diabetes, moderate or severe anaemia (haemoglobin <11g/dl) or moderate or severe undernutrition (Body Mass Index (BMI) <17 kg/m2) [17].
For implementation of the END TB strategy, it is recommended that “All persons with TB need to be assessed for nutritional status and receive nutritional counselling and care according to need” and additionally, “all persons with TB should be screened for diabetes” and, “in addition to HIV/AIDS, other co-morbidities and health risks associated with TB require integrated management” [18]. However, to plan the delivery of such services, integrated within predominantly vertical, donor-led TB programmes, better data is needed to quantify the extent of multi-morbidities experienced by persons with TB.
The aims of this cross-sectional study were to quantify the prevalence, patterns and identify predictors of non-communicable comorbidities of Filipinos newly starting TB treatment enrolled in the Starting Anti-TB Treatment (St-ATT) cohort in three regions of the Philippines.
## Study design
A cross-sectional analysis of data collected at enrolment into a facility-based prospective cohort study conducted in public (government) TB Directly Observed Treatment (DOT) clinics in the Philippines (ISRCTN16347615).
## Setting
Participants were enrolled from 13 TB-DOTS clinics in Metro Manila ($$n = 3$$), Cebu (Region VII, $$n = 5$$) and Negros Occidental (Region VI, $$n = 5$$), including 4 centers for the programmatic management of drug resistant TB (PMDT) centers which were also HIV referral centers (2 in Negros Occidental, 1 In Metro Manila and 1 in Cebu. Metro *Manila is* identified as a high HIV category area compared to other areas of the country. Clinics in Cebu and Negros included urban, peri-urban and rural catchment areas (S1 Fig).
## Participants
All non-pregnant adults (≥18 years) registered at participating clinics and within 5 days of starting their TB treatment for pulmonary TB (bacteriologically confirmed or clinically diagnosed) on either standard treatment or on the WHO shorter regimen for multi-drug resistant TB (MDR-TB) were potentially eligible to participate. Exclusion criteria included current imprisonment, non-pulmonary TB, resident outside of study-designated barangays (village level administrative areas; some were excluded due to safety/transport limitations for field staff) or a medical or psychiatric disorder that, in the opinion of the investigators, precluded informed consent or ability to adhere to the study protocol. Written informed consent was obtained before enrolment in the local language (Filipino, Cebuano, Hilagaynon) or English.
## Comorbidity outcome [19]
Comorbidities assessed included: acute undernutrition defined as moderate or severe (body mass index, BMI <17.0 kg/m2) [20]; diabetes, defined as HbA1c >$6.5\%$ [21] or currently on a recognised treatment for diabetes; moderate or severe anaemia (haemoglobin <11 g/dL) [22]; hypertension defined as systolic blood pressure (SBP) ≥140 mmHg or a diastolic blood pressure (DBP) of ≥90 mmHg, equivalent to stage 2 hypertension by the 2017 American College of Cardiology and American Heart Association guidelines [23], or currently taking anti-hypertensive medications hypertension; and, HIV (positive HIV screening test or reported existing diagnosis). Severity of raised BP and hypertension was further categorized into elevated BP (systolic BP 120–129 mmHg & diastolic BP <80), stage 1 (systolic BP 130–139 mmHg, or diastolic BP 80–89 mmHg) and stage 2 (as above).
## Data collection
Enrolment took place between 1st August 2018 and 21st February 2020. Trained research nurses interviewed participants, completed all study assessments using structured questionnaires and extracted information recorded on individuals’ National TB Program treatment cards, using direct electronic data capture with tablets using Open Data Kit software [24]. Data were uploaded to a secure server daily. Household food security was assessed and categorized as food secure or mild food insecurity (raw score 0–2), moderate food insecurity (raw score 3–5) and severe food insecurity (raw score 6–9) using the Adapted U.S. Adult Household Food Security Survey Module (US HFSSM) [19, 25].
Research nurses conducted anthropometry, including weight (to the nearest 0.1 kg; Seca 803 Clara Digital Personal Non-Medical Scale) on a flat surface with the patient standing upright and unassisted without shoes. Heights were taken (to nearest 0.5 cm; Seca 216 Mechanical Stadiometer) without shoes or socks, with the patient standing unsupported and positioned fully upright with the lower border of the left orbit and the upper margin of the external auditory meatus horizontal. BMI was categorised for analysis according to WHO criteria; overweight/obese = BMI>25.0 kg/m2; normal = BMI 18.5–25.0, BMI<18.5 = underweight (BMI<18.5 - ≥17.0 = mild underweight, BMI <17-≥16.0 = moderate underweight, BMI<16.0 kg/m2 = severe underweight.
Waist and hip circumferences were measured (to the nearest 0.5 cm; Seca 201 measuring tape), midway between the uppermost border of the iliac crest and lower border of costal margin with tape parallel to the floor. Hip circumference was measured at the widest portion of the buttocks with the tape parallel to the floor [26]. Waist-to-hip ratio was calculated by dividing waist circumference by hip circumference. A high waist-to-hip ratio of ≥0.85 for females and 0.9 for males was used an indicator of central obesity [26].
Blood pressure was measured twice 5 minutes apart with the participant seated and at rest using an automated blood pressure monitor (Omron HEM-907, Kyoto, Japan). If measurements were ≥ 5mmHg apart, a third measure was taken and the average of the 2 closest values used. Fingerprick blood samples were used to obtain haemoglobin (HemoCue 301, Ängelholm, Sweden), HbA1C, (Trinity Biotech) and conduct HIV screening (Standard Diagnostics Bioline HIV-$\frac{1}{2}$ Ag/Ab Combo Rapid Test kits) for those with unknown status and who provided additional consent.
## NCD health risk behaviour definitions
Participants were asked if they were current smokers or had ever smoked, their average number of cigarettes per day and the number of years smoked; used to calculate the total number of pack years. Participants were asked if they regularly drank alcohol before their current TB diagnosis and if yes, how frequently on average either, daily, weekly or monthly and this was used to number of drinks/yr. For assessing smoking and alcohol use on risk of diabetes or hypertension, these were categorised as 4-level variables; none/never or irregular then total pack-years or total drinks/year divided into tertiles. Participants were also asked if they had changed their smoking or alcohol use since the current diagnosis. Questions related to drug abuse were not asked for political and legal reasons.
## Sample size
The sample size was determined by the primary objective of the cohort, which was to investigate associations between each of diabetes and undernutrition and treatment outcome. A minimum sample size of 800 was estimated to provide $90\%$ power (1−ß) and $5\%$ (α = 0.05) significance to detect associations between moderate or severe undernutrition (BMI<17 kg/m2, odds ratio ≥2.1) or diabetes (odds ratio ≥2.8) at start of treatment with adverse TB treatment outcomes vs treatment completion or cure, as defined programmatically [27], assuming 10–$20\%$ experience an adverse treatment outcome, 25–$30\%$ are undernourished and 10–$12\%$ are diabetic. Post-hoc sample size calculations were not performed for the analyses reported here.
## Analysis
Characteristics of participants were tabulated overall and by TB treatment regimen (drug sensitive or multi-drug resistant) and by area (Negros Ocidental, Cebu and Manila). Differences in categorical characteristic variables between areas were tested using chi-squared/test. Logistic regression was used for univariable and multivariable analysis of factors associated with each co-morbidity. Multivariable models for each outcome comorbidity were developed using forward step-wise selection of variables in three blocks, starting with socio-demographic characteristics, then nutrition-related variables and finally TB-related variables. Variables in each block were tested for inclusion in order of the most significant in univariable analysis. Selection of a final model for each outcome was based on inclusion of factors associated with an outcome based on a likelihood ratio test (LRT) p-value of ≤0.1 and retained if meeting this criterion after adjustment for other covariates.
## Ethical approval
The study was approved by the Nationally accredited Ethical Review Committee of the Asian Eye Institute, Manila, Philippines (ref 2018–008) and the Ethical Review Committee of Nagasaki University School of Tropical Medicine and Global Health, Japan and the London School of Hygiene & Tropical Medicine, UK (ref 14894).
## Results
Of 6,981 potentially eligible individuals started on either the standard DS-ATT or the WHO shorter MDR regimen at the 13 participating clinics during the enrolment period, 903 persons ($12.9\%$) were enrolled (S1 Fig & S1 Table in S1 File). Fewer potentially eligible persons were approached for enrolment in Negros Occidental ($7.1\%$) compared to Cebu ($35.6\%$) and Metro Manila ($41.9\%$). This was due to one very large clinic site in the provincial capital of Negros Occidental, with a large catchment area in this mostly rural province, some of which was outside of our recruitment area. However, there was a similar age and sex distribution between those enrolled and registered and between the sites. The analysis population included 900 participants, (Fig 1) two-thirds male with a mean age of 44.7 years (standard deviation, sd 16.4), ranging from 18–87 years. Fifteen percent were initiated on the WHO shorter regimen for drug-resistant TB, with a significantly lower proportion in Metro Manila compared to Cebu and Negros Occidental sites ($p \leq 0.001$).
**Fig 1:** *Flow chart of enrolment.*
Participant characteristics vary across regions (S2 Table in S1 File). Participants in Negros Occidental were generally older, less educated, less likely to be employed, have lower incomes but more likely to have health insurance. Food insecurity was highest in Metro Manila ($36\%$), followed by Negros Occidental ($25\%$) and Cebu ($19\%$). Households were larger in Negros Occidental, followed by Cebu, but household density (persons/number of rooms) was higher in Metro Manila. TB-related characteristics also differed significantly by region; bacteriologically confirmed TB was $51\%$ in Cebu and Negros Occidental compared to $41\%$ in Metro Manila. A history of previous TB treatment was more common in Negros Occidental ($38\%$) compared to Cebu ($34\%$) and Metro Manila ($28\%$), with $20\%$ of participants in Negros Occidental reporting one or more household contacts, ever having been previously diagnosed with TB, compared to $8\%$ in Cebu. Participants in Negros Occidental also had a longer duration and number of symptoms at start of treatment, with much higher prevalence of participants reporting current appetite loss, weight loss, night sweats and haemoptysis, possibly indicating more severe disease in these participants.
Socio-demographic and TB-related characteristics did not appear to differ between participants initiated on drug-sensitive (DS-TB) and MDR-TB regimens, apart from a longer duration and number of TB-related symptoms and a higher proportion of participants with a history of previous TB treatment ($26\%$ vs $75\%$, $p \leq 0.001$) in those initiating MDR TB treatment (Table 1). All but 1 of the 133 participants on MDR-TB treatment were bacteriologically confirmed and $5.5\%$ of New TB cases were initiated on MDR treatment, ranging from $6.9\%$ in Cebu to $2.5\%$ in Metro Manila (S3 Table in S1 File).
**Table 1**
| Characteristic | Characteristic.1 | All (N = 900) | DS-TB (N = 767) | MDR-TB (N = 133) | p-value |
| --- | --- | --- | --- | --- | --- |
| Socio-demographic characteristics | Socio-demographic characteristics | | | | |
| Age group | 18–40 years old | 380 (42.2) | 325 (42.4) | 55 (41.4) | 0.081 |
| | 41–65 years old | 419 (46.6) | 349 (45.5) | 70 (52.6) | |
| | ≥ 65 years old | 101 (11.2) | 93 (12.1) | 8 (6.0) | |
| Marital status | Single | 374 (41.6) | 320 (41.7) | 54 (40.6) | 0.501 |
| | Married | 428 (47.6) | 359 (46.8) | 69 (51.9) | |
| | Divorced/separated/widowed | 98 (10.9) | 88 (11.5) | 10 (7.5) | |
| Education | Primary | 258 (28.7) | 229 (29.9) | 29 (21.8) | 0.333 |
| | Secondary | 420 (46.7) | 350 (45.6) | 70 (52.6) | |
| | Tertiary or vocational | 214 (23.8) | 181 (23.6) | 32 (24.1) | |
| Employed | No | 560 (62.2) | 461 (60.1) | 99 (74.4) | 0.002 |
| | Yes | 339 (37.7) | 305 (39.8) | 34 (25.6) | |
| Family income | Less than 5,000 PHP | 372 (41.3) | 312 (40.7) | 60 (45.1) | 0.934 |
| | 5000–9999 PHP | 243 (27.0) | 211 (27.5) | 32 (24.1) | |
| | 10,000–14,999 PHP | 171 (19.0) | 146 (19.0) | 25 (18.8) | |
| | 15,000–19,999 PHP | 41 (4.6) | 36 (4.7) | 5 (3.8) | |
| | 20,000 PHP or more | 44 (4.9) | 37 (4.8) | 7 (5.3) | |
| | Don’t know | 28 (3.1) | 24 (3.1) | 4 (3.0) | |
| Food Insecurity | Food Secure | 675 (75.0) | 581 (75.7) | 94 (70.7) | 0.392 |
| | Moderate Food Insecurity | 169 (18.8) | 141 (18.4) | 28 (21.1) | |
| | Severe Food Insecurity | 56 (6.2) | 45 (5.9) | 11 (8.3) | |
| Have Health Insurance | | 557 (61.9) | 484 (63.1) | 73 (54.9) | 0.072 |
| Median household size (IQR) | Adult (18+) | 2 (1–3) | 2 (1–3) | 2 (1–3) | 0.362 |
| | Young (5–18) | 1 (0–2) | 1 (0–2) | 1 (0–2) | 0.066 |
| | Children (0–5) | 0 (0–1) | 0 (0–1) | 0 (0–1) | 0.863 |
| | Whole household | 3 (2–5) | 3 (2–5) | 4 (2–5) | 0.113 |
| Median Household density (IQR) | Median Household density (IQR) | 2.0 (1.4) | 2.0 (1.4) | 2.3 (1.8) | 0.866 |
| Region | Metropolitan Manila | 372 (41.3) | 311 (40.5) | 61 (45.9) | 0.033 |
| | Negros | 344 (38.2) | 288 (37.5) | 56 (42.1) | |
| | Cebu | 184 (20.4) | 168 (21.9) | 16 (12.0) | |
| TB-related characteristics | TB-related characteristics | | | | |
| New Tx or Relapse | New | 598 (66.4) | 565 (73.7) | 33 (24.8) | <0.001 |
| | Relapse/Failure/TALF/PTOU | 302 (33.6) | 202 (26.3) | 100 (75.2) | |
| Basis of diagnosis | Clinical diagnosis | 457 (50.8) | 456 (59.5) | 1 (0.8) | <0.001 |
| | Bacteriologically confirmed | 443 (49.2) | 311 (40.5) | 132 (99.2) | |
| DSSM grade | Negative | 258 (28.7) | 253 (33.0) | 5 (3.8) | <0.001 |
| | 1+ | 65 (7.2) | 64 (8.3) | 1 (0.8) | |
| | 2+ | 30 (3.3) | 25 (3.3) | 5 (3.8) | |
| | ≥3+ | 60 (6.7) | 43 (5.6) | 3 (2.3) | |
| Household TB history | ≥1 HHC ever diagnosed TB | 131 (14.6) | 110 (14.3) | 21 (15.8) | 0.002 |
| Median duration (days) symptoms before start of Tx (IQR) | Median duration (days) symptoms before start of Tx (IQR) | 48.0 (30.0–77.0) | 47.0 (30.0–75.0) | 53.0 (28.0–89.0) | 0.094 |
| Current TB symptoms | Cough | 795 (88.3) | 667 (87.0) | 128 (96.2) | <0.001 |
| | Fatigue | 572 (63.6) | 487 (63.5) | 85 (63.9) | <0.001 |
| | Fever | 350 (38.9) | 292 (38.1) | 58 (43.6) | <0.001 |
| | Night sweats | 292 (32.4) | 246 (32.1) | 46 (34.6) | <0.001 |
| | Reduced appetite | 350 (38.9) | 300 (39.1) | 50 (37.6) | <0.001 |
| | Chills | 186 (20.7) | 157 (20.5) | 29 (21.8) | <0.001 |
| | Chest pain | 317 (35.2) | 265 (34.6) | 52 (39.1) | <0.001 |
| | Weight loss | 522 (58.0) | 434 (56.6) | 88 (66.2) | <0.001 |
| | Haemoptysis | 275 (30.6) | 218 (28.4) | 57 (42.9) | <0.001 |
| | Other | 95 (10.6) | 80 (10.4) | 15 (11.3) | <0.001 |
## Prevalence of NCD risk factors: Smoking and alcohol behaviours
By self-report, $22\%$ were current smokers and $36\%$ past smokers, with median total pack-years of 10 [interquartile range [IQR] 2.5–22]. More individuals in Negros Occidental were current or ex-smokers ($64\%$, $p \leq 0.004$ by region) and there were more current smokers in Cebu ($28\%$ $p \leq 0.001$ by region). There was no evidence of different smoking behavior between MDR and DS-TB treatment regimens (Fig 2). Amongst current smokers, $90\%$ reported decreasing their usual smoking since their TB diagnosis ($\frac{183}{202}$) and $83\%$ of ex-smokers may have quit as a result of their TB symptoms or diagnosis ($\frac{267}{303}$). Overall, one-third of participants reported usually drinking alcohol at least once per month, with more in Cebu and Negros compared to Metro Manila (chi-squared test $$p \leq 0.002$$) but did not differ by MDR vs DS TB treatment regimen ($$p \leq 0.555$$). Frequency of alcohol intake, varied by region and MDR vs DS TB treatment ($p \leq 0.001$, $$p \leq 0.036$$) (Fig 2). Of those that reported regular alcohol consumption, $94\%$ reported decreasing their alcohol intake since diagnosis ($\frac{293}{311}$).
**Fig 2:** *Prevalence of risk behaviours by region and by drug sensitive or multi-drug resistant TB treatment regimen.Footnotes: Smoking status: Prevalence of smoking differed significantly by region (chi2, p = 0.004) but not by MDR vs DS TB treatment regimen (chi2, p = 0.92). The relative proportion of current compared to ex-smoking behaviour differed significantly by region (chi2, p<0.001) but not by MDR vs DS TB treatment regimen (chi2, p = 0.31). Alcohol intake: Prevalence of reported regular intake of alcohol differed significantly by region (chi2, p = 0.002) but not by MDR vs DS TB treatment regimen (chi2, p = 0.55). The frequency of alcohol intake in those that reported regular drinking differed significantly by region (chi2, p<0.001) and by MDR vs DS TB treatment regimen (chi2, p = 0.036).*
## Diabetes
At the start of TB treatment $22.6\%$ ($\frac{199}{881}$; $95\%$ CI: 19.8–$25.4\%$) had study-defined diabetes (Fig 3A). There was no evidence of a difference in prevalence by region, but there was between those on DS-TB vs MDR-TB treatment (Fig 3A; $20.6\%$ vs $34.1\%$; $$p \leq 0.001$$). Of those with diabetes, $45\%$ ($\frac{90}{199}$) reported a previous clinical diagnosis which differed by region, being highest in Cebu and lowest in Negros Occidental. Of 90 participants with a previous diabetes diagnosis, 49 ($54\%$) reported regular diabetes follow-up visits at a health center or with a doctor, whilst 77 ($86\%$) reported currently taking diabetes medication of whom 68, were taking metformin, glycazide, or in combination with insulin. However, amongst those reporting current diabetic medication, only $\frac{23}{73}$ ($31.5\%$) were controlled (HbA1c ≤$8.0\%$; [28]) with HbA1c ranging to over $14\%$ in the remaining (Fig 4). HbA1c values were higher in previously diagnosed versus newly diagnosed diabetes (median HbA1c% = $9.6\%$ [IQR: $7.5\%$ - $11.6\%$] vs. $7.5\%$ [IQR: $6.7\%$ - $10.8\%$]); Wilcoxon rank sum test $$p \leq 0.007$$) (Fig 4). Although the majority of diabetes cases had normal BMI ($\frac{117}{199}$, $59\%$), or were overweight/obese ($\frac{28}{199}$, $14\%$), there were 54 ($27\%$) who were malnourished (14 with moderate (BMI 16 - <17 kg/m2) and 15 with severe undernutrition (BMI<16 kg/m2)) (Fig 4. When separated by TB treatment regimen, there were no significant differences in HbA1c levels in those who had diabetes or in the ratio of previous to new diagnoses (S2 Fig). Overall, there were small increases in HbA1c with increasing age (ß-coefficient = 0.02, $p \leq 0.001$), but this association was not consistent when assessed in those with diabetes (ß-coefficient = -0.03, $$p \leq 0.026$$) and cases of diabetes were not limited to those in the older age group with $\frac{36}{199}$ ($18.1\%$) occurring in participants aged 40 or less, giving rise to a diabetes prevalence of $9.5\%$ ($\frac{15}{380}$) in this age group. ( S3 Fig)
**Fig 3:** *A. Prevalence, severity and existing and new diagnoses of diabetes and hypertension comorbidities by region and by drug sensitive or multi-drug resistant TB treatment regimen. Diabetes: HbA1c 6.5% or currently taking recognised diabetes medication. Previous diagnosis = reported diagnosis prior to TB diagnosis. Prevalence of diabetes did not differ by region (chi2 p = 0.24) but was higher in MDR vs DS TB treatment (chi2, p = 0.001). The proportion of new vs. previous diabetes diagnoses differed significantly by region (chi2, p<0.001) but not by MDR vs DS TB treatment regimen (chi2, p-0.21). Hypertension: defined as stage 2 hypertension: Systolic BP>140 mmHg OR diastolic BP>90 mmHg or currently taking recognised anti-hypertensive medication. Previous diagnosis = reported diagnosis prior to TB diagnosis. The prevalence of hypertension did not differ significantly by region (chi2, p = 0.91) but was lower in MDR vs DS TB treatment (chi2, p = 0.005). The proportion of new vs previous hypertension diagnoses did not differ by region or MDR vs DS TB treatment regimen (chi2, p = 0.741; p = 0.280). Grades of Hypertension: Elevated BP = elevated systolic BP ≥ 120 mmHg < 130 mmHg & normal diastolic BP <80 mmHg; Stage 1 Hypertension = systolic BP ≥ 130 mmHg < 140 mmHg OR diastolic BP ≥ 80 mmHg < 90 mmHg; Stage 2 hypertension: Systolic BP>140 mmHg or diastolic BP>90 mmHg. The prevalence of elevated BP/hypertension (all grades combined) did not differ significantly by region (chi2, p = 0.170) but did by MDR vs DS TB treatment regimen (chi2, p<0.001). The relative proportion of grades of hypertension (elevated, stage 1, stage 2) did not differ by region or MDR vs DS TB treatment regimen (chi2, p = 0.19, p = 0.84). B. Prevalence and severity of malnutrition and anaemia comorbidities by region and by drug sensitive or multi-drug resistant TB treatment regimen. Footnotes: Undernutrition: Mild = Body Mass Index (BMI) <18.5–17 kg/m2; Moderate = BMI <17–16 kg/m2; Severe = BMI<16 kg/m2. The prevalence of all grades of undernutrition differed significantly by region (chi2, p = 0.004) and by MDR vs DS TB treatment regimen (chi2, p = 0.005). The relative proportions of grade of undernutrition did not differ by region (chi2, p = 0.085) or by MDR vs DS TB treatment regimen (chi2, p = 0.12). Anaemia: Mild = Haemoglobin (Hb) <13.0 [male] <12.0 [female] g/dl—>11.0 g/dl; Moderate Hb ≤11.0–8.0 g/dl; Severe Hb <8.0 g/dl. The prevalence of all grades of anaemia did not vary significantly by region or by MDR vs DS TB treatment regimen (chi2, p = 0.49 & p = 0.77).* **Fig 4:** *HbA1c by BMI for new and previous diabetes diagnoses at start of TB treatment.Footnotes: y-axis: 6.5% cut-off to determine diabetes status. x-axis: BMI cut offs indicate WHO cut-offs of nutritional status; <14 = very severe thinness, <16 severe thinness, <17 = moderate thinness, <18.5 mild thinness, <25 normal, <30 overweight, ≥30 obese.*
Hypertension. The prevalence of hypertension was $19.0\%$ ($\frac{164}{864}$; $95\%$ CI: 16.3–$21.5\%$) and did not differ by region (Fig 3A) but was lower in those on MDR-TB than DS-TB regimens (Fig 3A, $p \leq 0.001$). The relative proportion of new vs previous hypertension diagnoses did not differ by region or by MDR vs DS TB treatment regimen (Fig 3A). When all grades of abnormal BP were considered (elevated BP, stage 1 or stage 2 hypertension by AHA criteria [23]), the overall prevalence increased to $43.6\%$, ($\frac{377}{864}$) which did not vary significantly by region but was lower in those enrolled on MDR TB treatment (Fig 3A). The relative proportions of grade of abnormal BP did not differ by region or MDR vs DS TB treatment regimen (Fig 3A). Forty-six percent ($\frac{76}{164}$) of those with hypertension reported a previous diagnosis with 70 of these reporting taking a recognised anti-hypertensive medication (losartan and amlodipine being the most common). However, of these only $\frac{38}{70}$ ($54\%$) were controlled (BP < $\frac{140}{90}$ mmHg) with systolic BP ranging up to 198 mmHg in the remaining (Fig 5). There was strong evidence of higher systolic and diastolic BP values in persons with new versus previous hypertension diagnoses (mean SBP = 147.2 mmHg [sd: 15.1] vs. 136.9 mmHg [sd: 20.7]; $p \leq 0.001$; & mean DBP = 89.6 mmHg [sd: 9.7] vs. 81.9 mmHg [sd: 11.9]; $p \leq 0.001$) (Fig 5). Both systolic and diastolic BP were positively associated with age ($p \leq 0.001$), but cases of study defined hypertension were not limited to older adults, with $\frac{17}{164}$ ($10.4\%$) occurring in participants aged 40 years or less (Fig 5), giving rise to a hypertension prevalence of $4.5\%$ ($\frac{17}{380}$) in this age group.
**Fig 5:** *Systolic and diastolic blood pressure by age in those with and without hypertension and new vs pre-existing diagnoses.Footnotes: Hypertension = SPB ≥ 140 mmHg OR DBP ≥ 90 mmHg OR previous diagnosis and currently taking recognised anti-hypertensive medication. Systolic blood pressure (SBP): y-axis: ≥120 mmHg indicates cut-off for elevated SBP; ≥130 mmHg indicates cut off for stage 1 hypertension; ≥140 mmHg indicates cut-off for stage 2 hypertension. Diastolic BP (DBP): y-axis: ≥80 indicates cut-off for elevated DBP; ≥90 indicates cut off for stage 2 hypertension (AHA 2017 criteria [22]).*
Undernutrition. The prevalence of all grades of undernutrition defined as BMI < 18.5 kg/m2 was $43.7\%$ ($\frac{393}{899}$; $95\%$ CI: 40.-$47.0\%$) and significantly differed by region, being most prevalent in Negros Occidental ($49.9\%$, $$p \leq 0.004$$) and more common in participants enrolled on MDR compared to DS TB treatment regimens ($54.9\%$ vs $41.8\%$, $$p \leq 0.005$$) (Fig 3B). The relative proportion of those who had severe undernutrition (BMI < 16 kg/m2) was possibly higher in Metro Manila ($39.2\%$, $$p \leq 0.085$$) (Fig 3B) and there was no evidence of a difference by TB treatment regimen ($$p \leq 0.120$$) (Fig 3B). The overall prevalence of moderate or severe undernutrition (BMI < 17.0 kg/m2), selected to define clinically significant co-morbidity was $23.4\%$ ($\frac{210}{899}$; $95\%$ CI: 20.6–$26.1\%$), which also significantly differed by region ($$p \leq 0.004$$) and by MDR vs DS TB treatment regimen ($$p \leq 0.004$$).
The prevalence of all grades of anaemia (Mild = haemoglobin (Hb) <13.0 [male] <12.0 [female] g/dl—>11.0 g/dl; Moderate Hb ≤11.0–8.0 g/dl; Severe Hb <8.0 g/dl.) was $40.3\%$ ($\frac{362}{899}$; $95\%$ CI: 37.1–$43.5\%$) and did not differ significantly by region or by MDR vs DS TB treatment regimens, nor did the relative proportion of grades of anaemia (Fig 3B). The overall prevalence of moderate or severe anaemia (haemoglobin <11 g/dl), selected to define clinically significant co-morbidity, was $13.5\%$ ($\frac{121}{899}$; $95\%$ CI: 11-2-$15.7\%$), possibly higher in Negros Occidental ($16.5\%$, $$p \leq 0.085$$) compared to the other regions, but not different by treatment regimen (Fig 3B).
After adjustment, older age, being married versus single, weight increase or loss in the past 3–6 months, MDR-TB, high waist-to-hip ratio and MDR versus DS treatment regimen were associated with increased odds of diabetes (Table 2). Moderate and severe food insecurity compared to none or mild and underweight BMI classification compared to normal were associated with lower odds of diabetes. In the final multivariable model high waist-to-hip ratio, as an indicator of central obesity, accounted for the univariable association observed for overweight or obese by BMI.
**Table 2**
| Unnamed: 0 | Unnamed: 1 | Diabetes1 (N = 881) | Hypertension2 (N = 864) | Anaemia (Mod/Severe)3 (N = 899) | Undernutrition (Mod/Severe)4 (N = 899) |
| --- | --- | --- | --- | --- | --- |
| Socio-demographic characteristics | Socio-demographic characteristics | Socio-demographic characteristics | Socio-demographic characteristics | Socio-demographic characteristics | Socio-demographic characteristics |
| Age3 | 18–40 years old | 1 | 0.13 (0.08–0.23) | 1 | 2.19 (1.51–3.18) |
| Age3 | 41–64 years old | 2.58 (1.61–4.14) | 1 | 1.66 (1.06–2.59) | 1 |
| Age3 | 65+ years old | 2.51 (1.31–4.81) | 2.68 (1.63–4.43) | 1.91 (1.00–3.54) | 1.41 (0.82–2.43) |
| Marital status | Single | 1 | - | - | - |
| Marital status | Married | 2.09 (1.31–3.31) | - | - | - |
| Marital status | Divorced/separated | 0.75 (0.23–2.48) | - | - | - |
| Marital status | Widowed | 0.85 (0.40–1.84) | - | - | - |
| Education level achieved | Tertiary/vocational | - | - | - | 1 |
| Education level achieved | Secondary | - | - | - | 1.83 (1.14–2.94) |
| Education level achieved | Primary | - | - | - | 1.85 (1.09–3.13) |
| Food insecurity level | None/mild | 1 | - | - | 1 |
| Food insecurity level | Moderate | 0.49 (0.30–0.82) | - | - | 1.04 (0.68–1.61) |
| Food insecurity level | Severe | 0.83 (0.38–1.81) | - | - | 2.23 (1.21–4.10) |
| Nutrition-related risk factors | Nutrition-related risk factors | Nutrition-related risk factors | Nutrition-related risk factors | Nutrition-related risk factors | Nutrition-related risk factors |
| Weight change in last 3–6 months | No change | 1 | - | - | |
| Weight change in last 3–6 months | Weight increase | 0.92 (0.41–2.18) | - | - | |
| Weight change in last 3–6 months | Weight decrease | 1.97 (1.19–3.27) | - | - | |
| BMI classification5 | Normal | 1 | 1 | - | |
| BMI classification5 | Underweight | 0.50 (0.33–0.75) | 0.43 (0.27–0.66) | - | |
| BMI classification5 | Overweight/obese | 1.53 (0.83–2.82) | 3.11 (1.73–5.62) | - | |
| BMI classification with undernutrition classification6 | Normal | - | - | 1 | |
| BMI classification with undernutrition classification6 | Mild underweight | - | - | 2.11 (1.26–3.54) | |
| BMI classification with undernutrition classification6 | Moderate underweight | - | - | 3.34 (1.86–5.99) | |
| BMI classification with undernutrition classification6 | Severe underweight | - | - | 2.49 (1.39–4.47) | |
| BMI classification with undernutrition classification6 | Overweight/obese | - | - | 0.64 (0.22–1.88) | |
| Appetite related food intake in past month vs pre-TB symptoms | No change | - | 1 | 1 | 1 |
| Appetite related food intake in past month vs pre-TB symptoms | Moderate/severe decrease | - | 0.52 (0.33–0.81) | 1.65 (1.06–2.57) | 1.66 (1.14–2.42) |
| Appetite related food intake in past month vs pre-TB symptoms | Increase | - | 1.12 (0.68–1.87) | 1.28 (0.70–2.32) | 1.04 (0.61–1.77) |
| TB-related risk factors | TB-related risk factors | TB-related risk factors | TB-related risk factors | TB-related risk factors | TB-related risk factors |
| Waist-to-hip ratio7 | Normal | 1 | - | - | - |
| Waist-to-hip ratio7 | High | 2.41 (1.64–3.56) | - | - | - |
| TB treatment regimen | DS | 1 | 1 | - | - |
| TB treatment regimen | MDR | 2.48 (1.55–3.95) | 0.55 (0.39–0.78) | - | - |
| HIV status | Negative | - | - | 1 | - |
| HIV status | Unknown | - | - | 1.48 (0.93–2.35) | - |
| HIV status | Positive | - | - | 4.51 (1.01–20.10) | - |
| TB type | New diagnosis | - | | - | 1 |
| TB type | Relapse/Failure | - | | - | 1.98 (1.40–2.80) |
| Duration of TB symptoms | <1 month | - | - | - | 1 |
| Duration of TB symptoms | 1–2 months | - | - | - | 1.29 (0.83–2.00) |
| Duration of TB symptoms | 2–3 months | - | - | - | 1.88 (1.15–3.06) |
| Duration of TB symptoms | >3 months | - | - | - | 1.92 (1.17–3.13) |
## Risk factors for co-morbid conditions
Detailed univariable analysis results are presented for each outcome of diabetes, hypertension, anaemia and undernutrition in S4-S7 Tables in S1 File.
## Hypertension
After adjustment, those aged 65 years and above compared to those aged 41–64 years and those who were overweight or obese compared to normal had increased odds of hypertension. Persons aged 18–40 vs. 41–64 years, who were underweight versus normal, who experienced a moderate or severe decrease in appetite related food intake and on a MDR vs DS treatment regimen had reduced odds of hypertension (Table 2).
## Anaemia
After adjustment, older age groups, those with moderate and severe decreases in recent, appetite-related food intake and mild, moderate or severe undernutrition compared to normal had increased odds of anemia. HIV status was retained in the model with a non-significantly increased odds in those with unknown status and in the few individuals with known positive status (Table 2).
## Undernutrition
After adjustment, those aged 18–40 and more than 65 years compared to 41–64 years, those with lower levels of education and those with a history of previous TB treatment and those with longer duration of TB symptoms before start of current treatment had increased odds of undernutrition. MDR versus DS-TB treatment regimen was no longer associated after adjustment for a history of previous TB treatment. Severe food insecurity, recent weight loss and recent, appetite-related reduced food intake were also associated with increased odds of undernutrition (Table 2).
## Multi-morbidity
Out of 847 individuals with complete data for the 4 co-morbidities, 496 ($58.6\%$) had at least one co-morbidity and 127 ($15.0\%$) individuals had 2 and 18 ($2.1\%$) individuals had 3 or more (Fig 6). The most common combination was diabetes with hypertension (57, $6.7\%$), followed by moderate or severe undernutrition with moderate or severe anaemia in 45 individuals ($5.3\%$) (Fig 6A). The pattern of co-morbidities appeared to differ in those on MDR-TB treatment regimens with $22.4\%$ ($\frac{29}{129}$) of individuals with 2 or more co-morbidities and the most common combination of co-morbidities being undernutrition and anaemia ($$n = 10$$; $7.7\%$) compared to $16.2\%$ ($\frac{116}{718}$) in those on DS-TB treatment with the most common combination of co-morbidities being diabetes and hypertension in 45 individuals ($6.8\%$) (Fig 6B & 6C).
**Fig 6:** *Venn diagrams of comorbidities in those with complete data for all 4 comorbidities for [A] All participants, N = 847; [B] participants initiating drug sensitive TB treatment regimens, N = 718; [C] participants initiating multi drug resistant TB treatment regimens, N = 129. Footnotes: Diabetes: HbA1c≥6.5% OR on current recognised medication; Hypertension: Stage 2, SPB ≥ 140 mmHg OR DBP ≥ 90 mmHg OR previous diagnosis and currently taking recognised anti-hypertensive medication; Anaemia: Moderate or severe haemoglobin ≤11.0 g/dL; Malnutrition: moderate or severe, BMI <17 kg/m2.*
## Discussion
The results of this study summarise the burden and predictors of NCD co-morbidities and their inter-relationships in Filipino persons initiating TB treatment. To our knowledge the current data represents the most comprehensive data available for the Philippines, and more globally for undernutrition and anaemia amongst persons with TB. The WHO End TB strategy includes the need to meet the complete healthcare needs of persons with TB. This is important both for the potential to improve, directly or indirectly, TB treatment outcomes, e.g. from food supplementation for undernutrition or better management of diabetes), but also from a health equity perspective and to protect others in NCD programmes from being exposed to persons with potentially infectious TB.
Overall, nearly $60\%$ of the study population had 1 or more co-morbid conditions, but with relatively little overlap between them. Of the four co-morbidities, moderate or severe undernutrition was the most common, closely followed by diabetes, both affecting just under a quarter of participants, followed by hypertension and lastly anaemia. When mild grades of undernutrition or anaemia were considered the prevalence of these conditions roughly doubled to $44\%$ and $40\%$ of the study population. Just over half of those with diabetes and hypertension were previously undiagnosed, whilst of those that were already on treatment more than half had poor control of hyperglycemia or BP.
The prevalence of diabetes in persons just starting TB treatment was considerably higher than in our previous cross-sectional study in persons at different stages of TB treatment ($22.6\%$, vs. $9.1\%$) [17]. Some of this difference may be due to TB-induced hyperglycemia causing elevated Hba1c that resolves over the course of TB treatment [28]. However, if TB-induced hyperglycaemia were the explanation we would expect to observe: i) longer duration of TB symptoms before start of treatment associated with higher HbA1c levels; and/or, ii) a higher proportion of newly diagnosed vs pre-existing diabetes to be observed in this study, neither of which we observed [17]. Alternatively, a greater loss to follow-up during treatment of those with comorbid TB-DM might explain the higher proportion of DM in this study. Investigating the proportion of people with elevated HbA1c who self-resolve with TB treatment will be addressed as a secondary objective in the main longitudinal St-ATT cohort, and effects of diabetes and glycemic control on TB treatment outcomes is a primary objective.
Similar to our previous findings, the data suggest that only a minority of those on diabetes medication achieved glycemic control in the period preceding their TB diagnosis, ($31.5\%$) even using the pragmatic <$8\%$ HbA1c cut-off as recently recommended for TB-DM [28]. This is not dissimilar to that reported in Filipino persons without TB, receiving standard care at health facilities of a similar level ($37\%$ HbA1c < $7.0\%$; $$n = 164$$) [29]. The odds of diabetes were approximately 2.5 times greater in those initiating an MDR-TB treatment regimen, but this did not appear to be mediated through any association with previous history of TB history, similar to our previous finding but in contrast with the multi-country TANDEM study [30]. Furthermore, there was no indication in an increase in the relative proportion of new vs, pre-existing DM diagnoses between DS- and MDR-TB treatment regimens, which might be expected if MDR-TB was associated with increased TB-induced hyperglycaemia.
Central obesity as indicated by a raised waist-to-hip ratio was strongly associated with diabetes suggesting that this may be a better predictor and potentially useful screening tool for TB-DM in this population [17, 31]. Low BMI was protective, but cannot exclude DM, as just over a quarter of all DM cases were undernourished (BMI<18.5). Importantly, even after adjustment for sample weighting for location, age and sex the prevalence of diabetes in these TB patients is considerably greater at $16.6\%$ ($95\%$ CI:13.1–$20.1\%$) than the estimated national adult prevalence of $7.1\%$ [11].
The prevalence of hypertension was lower than that of diabetes, and in contrast to diabetes was significantly lower in those initiating MDR-TB treatment regimens and independent of the large effects of age and under- or overweight status. There is currently no evidence to suggest that hypertension is associated with increased risk of TB [32]. The proportion of participants with both diabetes and hypertension was surprisingly low, although this proportion was higher in those initiating MDR- than DS-TB treatment. It is possible that in this population the relatively younger age of those with diabetes compared to the general population may underlie this observation. Although, the degree of overlap may have been considerably greater if we had used a lower BP cut-off in our hypertension definition. Similar to diabetes, reported use of anti-hypertensive medication did not correlate well with lower BP measurements.
Anaemia was the only condition observed to be associated with HIV status, whilst undernutrition was the only comorbid condition to be independently associated with any TB-related exposures, including increased odds of a previous history of TB and longer duration of TB-related symptoms before starting TB treatment. The prevalence of both these conditions were slightly higher than our previous observations, which is to be expected as TB treatment should contribute to some resolution of these conditions; this is something we will explore within the longitudinal data analysis of this cohort.
Data from the three regions suggest undernutrition, diabetes, hypertension and anemia are important co-morbidities across all regions and in both MDR and DS-TB, but that diabetes and hypertension tend to cluster together in those who are better nourished, whilst undernutrition and anaemia cluster in those who are less educated and poorer–as indicated by household food security. There is some evidence to suggest that diabetes may increase the risk of MDR-TB, rather than more TB-induced hyperglycemia occurring in MDR-TB. Although undernutrition was more common in MDR-TB than DS-TB this was accounted for by the higher proportion of those with a previous history of TB in those on MDR-TB treatment. Unfortunately, without further data it is not possible to determine the relative causal direction of this association. Effective TB treatment alone may not be enough to support nutritional recovery in patient populations at high risk of TB associated catastrophic costs. Failure of nutritional recovery is also known to be an indicator of inadequate treatment, although the mechanisms are still to be elucidated.
Somewhat surprisingly, there was no evidence of association between smoking and alcohol use and any of the comorbidities assessed. However, just under a quarter of participants were current smokers, although most reported decreased smoking as a result of their TB symptoms or diagnosis and many others had recently quit. The proportion of those that remain as non-smokers when TB-symptoms improve is unknown. This demonstrates the need for smoking cessation support to be included within TB programmes in the Philippines.
The data from this study builds on our previously reported cross-sectional study, [17] providing more definitive data on TB-comorbidity at TB treatment initiation, including systematic assessment of hypertension.
## Study strengths, limitations and further research
A strength of this study is the systematic measurement of comorbidities and description of their overlap and management in a large, well-described cohort of persons with TB representative of those initiating new treatment regimens in public facilities across a range of urban, peri-urban and rural settings in a high TB burden country in Asia, in which there is a different pattern of NCDs and HIV compared to most African countries. Study limitations include the high proportion of participants with unknown HIV status due to refusal of testing, although HIV and HIV co-infection in TB in the *Philippines is* known to be low. Due to logistical considerations, it was not possible to randomly select TB-DOTs clinics or patients from the clinics and this could have led to selection bias. However, despite the smaller proportion of patients enrolled from one of the very large clinics, age and sex distributions of the enrolled vs registered participants were reassuringly similar. Another increasingly recognized comorbidity of depression and anxiety was not included here as this was assessed in a subset and will be reported separately. Finally, causality between risk factors is difficult to infer from the multivariable analyses conducted and requires further research. How these comorbid conditions change over the course of TB treatment and their effect on TB treatment outcomes will be reported within the main findings of the cohort.
## Conclusions
More than half of persons initiating TB treatment regimens have one or more comorbid conditions requiring management as part of patient centered care to improve TB-related treatment outcomes and quality of life. The planning of such services needs better data to design services to improve both the diagnosis and management of these. Nutritional advice and management should be a core component of diabetes and hypertension management [33] but is often missing from programmes due to human resource limitations. Health staff trained in nutritional counselling or management are usually limited to those working in maternal and child health programmes, focusing on undernutrition. We propose that there is a need for an increased cadre of health workers trained in nutrition who could work across integrated programmes of TB and NCDs which would require training in managing nutrition-related chronic diseases and infection-related undernutrition of adults (HIV or TB). Further research is needed on how to optimize management in persons with multimorbidities of which nutrition should be considered a core component.
## References
1. Duarte R, Lonnroth K, Carvalho C, Lima F, Carvalho ACC, Munoz-Torrico M. **Tuberculosis, social determinants and co-morbidities (including HIV)**. *Pulmonology* (2018.0) **24** 115-9. DOI: 10.1016/j.rppnen.2017.11.003
2. 2WHO. Global TB Report. 2020.
3. 3Food and Nutrition Research Institute DoSaT, the Philippines,. 8th National Nutrition Survey2015 12th Jan 2019. Available from: http://www.fnri.dost.gov.ph/index.php/nutrition-statistic/19-nutrition-statistic/118-8th-national-nutrition-survey.
4. Mendenhall E, Kohrt BA, Norris SA, Ndetei D, Prabhakaran D. **Non-communicable disease syndemics: poverty, depression, and diabetes among low-income populations**. *Lancet* (2017.0) **389** 951-63. DOI: 10.1016/S0140-6736(17)30402-6
5. Lonnroth K, Williams BG, Cegielski P, Dye C. **A consistent log-linear relationship between tuberculosis incidence and body mass index**. *Int J Epidemiol* (2010.0) **39** 149-55. DOI: 10.1093/ije/dyp308
6. Cegielski JP, McMurray DN. **The relationship between malnutrition and tuberculosis: evidence from studies in humans and experimental animals**. *Int J Tuberc Lung Dis* (2004.0) **8** 286-98. PMID: 15139466
7. Waitt CJ, Squire SB. **A systematic review of risk factors for death in adults during and after tuberculosis treatment**. *Int J Tuberc Lung Dis* (2011.0) **15** 871-85. DOI: 10.5588/ijtld.10.0352
8. Al-Rifai RH, Pearson F, Critchley JA, Abu-Raddad LJ. **Association between diabetes mellitus and active tuberculosis: A systematic review and meta-analysis**. *PLoS One* (2017.0) **12** e0187967. DOI: 10.1371/journal.pone.0187967
9. Dooley KE, Chaisson RE. **Tuberculosis and diabetes mellitus: convergence of two epidemics**. *Lancet Infect Dis* (2009.0) **9** 737-46. DOI: 10.1016/S1473-3099(09)70282-8
10. Golub JE, Mok Y, Hong S, Jung KJ, Jee SH, Samet JM. **Diabetes mellitus and tuberculosis in Korean adults: impact on tuberculosis incidence, recurrence and mortality**. *Int J Tuberc Lung Dis.* (2019.0) **23** 507-13. DOI: 10.5588/ijtld.18.0103
11. 11International Diabetes Foundation. IDF Diabete Atlas, 2019. 2019.
12. 12The Academy of Medical Sciences. Multimorbidity: a priority for global health research 2018.
13. Siddiqi K, Stubbs B, Lin Y, Elsey H, Siddiqi N. **TB multimorbidity: a global health challenge demanding urgent attention**. *International Journal of Tuberculosis and Lung Disease* (2021.0) **25** 87-90. DOI: 10.5588/ijtld.20.0751
14. Noubiap JJ, Nansseu JR, Nyaga UF, Nkeck JR, Endomba FT, Kaze AD. **Global prevalence of diabetes in active tuberculosis: a systematic review and meta-analysis of data from 2.3 million patients with tuberculosis**. *Lancet Glob Health.* (2019.0) **7** e448-e60. DOI: 10.1016/S2214-109X(18)30487-X
15. Koyanagi A, Vancampfort D, Carvalho AF, DeVylder JE, Haro JM, Pizzol D. **Depression comorbid with tuberculosis and its impact on health status: cross-sectional analysis of community-based data from 48 low- and middle-income countries.**. *BMC Med.* (2017.0) **15** 209. DOI: 10.1186/s12916-017-0975-5
16. Stubbs B, Siddiqi K, Elsey H, Siddiqi N, Ma R, Romano E. **Tuberculosis and non-communicable disease multimorbidity: An analysis of the World Health Survey in 48 low- and middle-income countries.**. *Lancet Public Health* (2020.0)
17. White LV, Edwards T, Lee N, Castro MC, Saludar NR, Calapis RW. **Patterns and predictors of co-morbidities in Tuberculosis: A cross-sectional study in the Philippines**. *Sci Rep. 2020* **10** 4100. DOI: 10.1038/s41598-020-60942-2
18. 18WHO. Implementing the END TB Strategy: The Essentials. Geneva: World Health Organisation, 2015.
19. 19Economic Research Service. U.S. Adult food security survey module: three-stage design with screeners 2012. Available from: https://www.ers.usda.gov/media/8279/ad2012.pdf
20. 20WHO. Physical status: The use and interpretation of anthropometry: Report of a WHO expert committee. World Health Organisation, 1995.
21. **2010 Diagnosis and Classification of diabetes mellitus**. *Diabetes Care* (2010.0) **33** S62-S9. DOI: 10.2337/dc10-S062
22. 22WHO. Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity.
Geneva:
World Health Organisation, 2011.. *Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity.* (2011.0)
23. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: Executive Summary: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Circulation* (2018.0) **138** e426-e83. DOI: 10.1161/CIR.0000000000000597
24. Hartung C, Anokwa Y, Brunette W, Lerer A, Tseng T, Borriello G. **Open Data Kit: Tools to build information services for developing regions.**. *ICTD* (2010.0)
25. Melgar-Quinonez HR, Zubieta AC, MkNelly B, Nteziyaremye A, Gerardo MF, Dunford C. **Household food insecurity and food expenditure in Bolivia, Burkina Faso, And the Philippines**. *The Journal of nutrition* (2006.0) **136** 1431S-7S. DOI: 10.1093/jn/136.5.1431S
26. 26WHO. Waist Circumference and Waist-Hip Ratio: Report of a WHO Expert Consultation. Geneva: World Health Organisation, 2008.. *Waist Circumference and Waist-Hip Ratio: Report of a WHO Expert Consultation* (2008.0)
27. 27WHO. Definitions and Reporting Framework for Tuberculosis.
Geneva:
World Health Organisation, 2013 Contract No.: WHO/HTM/TB/2013.2.. *Definitions and Reporting Framework for Tuberculosis.* (2013.0)
28. Lin Y, Harries AD, Kumar AMV, Critchley JA, van Crevel R, Owiti P. **Management of diabetes mellitus-tuberculosis: A guide to the essential practice.**. *Paris: International Union Against Tuberculosis and Lung Disease (The Union),* (2019.0)
29. Ku GM, Kegels G. **Effects of the First Line Diabetes Care (FiLDCare) self-management education and support project on knowledge, attitudes, perceptions, self-management practices and glycaemic control: a quasi-experimental study conducted in the Northern Philippines**. *BMJ Open.* (2014.0) **4** e005317. DOI: 10.1136/bmjopen-2014-005317
30. Ugarte-Gil C, Alisjahbana B, Ronacher K, Riza AL, Koesoemadinata RC, Malherbe ST. **Diabetes mellitus among pulmonary tuberculosis patients from four TB-endemic countries: the TANDEM study**. *Clin Infect Dis* (2019.0). DOI: 10.1093/cid/ciz284
31. Grint D, Alisjhabana B, Ugarte-Gil C, Riza AL, Walzl G, Pearson F. **Accuracy of diabetes screening methods used for people with tuberculosis, Indonesia, Peru, Romania, South Africa**. *Bull World Health Organ* (2018.0) **96** 738-49. DOI: 10.2471/BLT.17.206227
32. Seegert AB, Rudolf F, Wejse C, Neupane D. **Tuberculosis and hypertension-a systematic review of the literature**. *Int J Infect Dis* (2017.0) **56** 54-61. DOI: 10.1016/j.ijid.2016.12.016
33. 33World Health Organisation. WHO package of essential noncommunicable (PEN) disease interventions for primary health care Geneva: WHO; 2020. Available from: https://www.who.int/activities/integrated-management-of-ncds.
|
---
title: 'Prevalence and determinants of anaemia among men in rural India: Evidence
from a nationally representative survey'
authors:
- Aditya Singh
- Sumit Ram
- Shivani Singh
- Pooja Tripathi
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021440
doi: 10.1371/journal.pgph.0001159
license: CC BY 4.0
---
# Prevalence and determinants of anaemia among men in rural India: Evidence from a nationally representative survey
## Abstract
Anaemia among men is a significant health issue which has not been given due importance. Only a handful of studies have captured the prevalence of anaemia among men. There is dearth of evidence base on anaemia among men in India. Therefore, this study attempts to fill this research gap by examining the socioeconomic, geographic, health-related, and behavioural differentials of anaemia among rural men in India. We analysed a cross-sectional sample of 61,481 men aged between 15–54 and living in rural areas from the National Family Health Survey (NFHS-5), conducted in 2019–21. Bivariate statistics and multivariable logistic regression were employed to assess the factors associated with anaemia. In rural India, three out of ten men were found to be anaemic. Older men [49–54 years] (Odds Ratio: 1.10, $95\%$ CI, 1.00–1.21), men without a formal education (OR: 1.36, $95\%$ CI, 1.26–1.47), those from Scheduled Tribes (OR: 1.48, $95\%$ CI, 1.39–1.58) and men who belonged to the poorest wealth quintile (OR: 1.24, $95\%$ CI: 1.25–1.35) had a higher risk of anaemia. Men who were underweight were more likely to be anaemic (OR: 1.36, $95\%$ CI: 1.30–1.43). When compared to the central region, men from the eastern (OR: 1.47, $95\%$ CI: 1.39–1.55) parts of India had higher a risk of anaemia. The findings suggest the need to recognise anaemia among men as a public health issue. When developing policy, significant variation in socioeconomic, geographic, health-related, and behavioural factors must be taken into account. Men should also be screened on a regular basis in order to reduce the national burden of anaemia.
## Introduction
According to the World Health Organization (WHO), anaemia is a disorder in which the number or haemoglobin concentration of red blood cells is below normal which subsequently results in the decreased oxygen-carrying capacity of blood [1] *Haemoglobin is* an iron-containing protein in the red blood cells (RBC) that transports oxygen from the lungs to the tissues and carries carbon dioxide from tissues back to lungs [2]. Nutritional deficiencies, particularly iron deficiency is the main reason behind this disease although deficiencies in vitamins B9, B12 and A may also cause anaemia. Acute and chronic infections, and genetic haemoglobin disorders are also found to lead to anaemia [3]. Although the degree to which anaemia in a population can be attributed to these causes varies across populations [4].
Despite great breakthroughs in science and healthcare, anaemia continues to be a significant global public health issue. Every fourth person in the world ($27\%$) has anaemia, with the developing countries alone accounting for more than $89\%$ of the burden [4]. Although anaemia is a condition that affects all age groups, it is more common among pregnant women and children. Therefore, there have been global efforts to reduce anaemia among adolescent girls and boys, and women of reproductive age (WRA), especially pregnant and lactating women. Reducing anaemia among WRA by $50\%$ is one of the prime goals in six global nutrition targets for 2025 endorsed by WHO [5] and putting an end to all forms of malnutrition has been listed as one of the targets of the Sustainable Development Goal (SDGs) 2 [6].
Developing countries like India have also made continuous efforts to reduce the anaemia among children and women. However, as suggested by the Global Nutrition Report [2021], there has been little progress in combatting anaemia and malnutrition since 2016 [7]. It is also evident in India’s rank (94th among 107 countries) in the recently published Global Hunger Index. Moreover, none of the programs and interventions have addressed anaemia among men, despite the fact that one in every four men in India suffer from anaemia [8].
While anaemia during pregnancy and early childhood is linked to a variety of negative outcomes for the child, including low birth weight, delayed mental and cognitive development, mortality and has the potential to cause maternal mortality [9–12], anaemia in men is generally not considered a disease or a significant problem. Though anaemia among men is rarely fatal, still it can cause fatigue, difficulty concentrating, and lethargy, which not only reduces quality of life but is also thought to decrease economic productivity [13,14]. As per the India State-Level Disease Burden Initiative, iron-deficiency anaemia is one of the major causes resulting to burden of morbidity among men across all states and union territories (UTs) of India [15]. Therefore, anaemia among men should also be treated as a serious public health issue.
While there is an abundant literature from developing and underdeveloped countries including India that focuses on anaemia among WRA and their children, [16–18], there is dearth of evidence base on anaemia among men, especially in India [19]. The literature is much more limited particularly for rural men.
A study in 2019 focused on the differentials of anaemia prevalence among men in India [19]. A recent study by Kumar et al. [ 2021] attempted to decompose the factors that contribute to socioeconomic inequality in anaemia among men in India [20]. Also, Kumar et al. assessed the correlates and geographic distribution of anaemia in men in the empowered action group (EAG) states of India [21]. However, no study has ever focused on the prevalence of anaemia among men in rural India or the risk factors associated with it.
Moreover, anaemia among men is on the rise, according to the most recent National Family Health Survey, 2019–21 (NFHS-5), and is more prevalent in rural areas than in urban areas [8]. Because the characteristics and health behaviours of rural males differ significantly from those of urban men, it is crucial that they be researched separately from urban men. In order to be able to design and implement targeted interventions to reduce anaemia among rural men, it is crucial to identify not only the geographical regions where anaemia prevalence is high but also the vulnerable groups of rural men who are more likely to suffer from anaemia. Therefore, using data from the nationally representative survey, this study aims to examine socioeconomic, health-related and behavioural differentials in the prevalence of anaemia among rural men in India and the factors associated with it. The spatial distribution of anaemia among rural men across Indian states and districts is also discussed in this study.
## Data source
The data comes from the latest round of National Family Health Survey (NFHS-5) carried out by International Institute for Population Sciences during 2019–2021 under the supervision of Ministry of Health & Family Welfare, Government of India. NFHS, the Indian version of the Demographic and Health Surveys (DHS), is a nationally representative cross-sectional survey that collects data on a wide range of demographic, socioeconomic, and health-related issues.
Using a two-stage stratified random sampling method, NFHS-5 interviewed 724115 women aged 15 to 49 years and 101839 men aged 15–54 years from 636699 households. Response rate was $97\%$ and $92\%$ for women and men respectively. Through a series of biomarker tests and measurements, the clinical, anthropometric, and biochemical (CAB) component of the NFHS-5 provided critical estimates of the prevalence of malnutrition, anaemia, hypertension, high blood glucose levels, waist and hip circumference, Vitamin D3, HbA1c, and malaria parasites. The survey covered 707 districts from 28 states and 8 UTs. A uniform sample design, which is representative at the national, state/UT, and district level, was adopted in each round of the survey [8].
The Biomarker Schedule contained measurements of height, weight, and haemoglobin levels for children; measurements of height, weight, waist and hip circumference, and haemoglobin levels for women aged 15–49 years and men aged 15–54 years; and blood pressure and random blood glucose levels for women and men aged 15 years and over. Additionally, both men and women were requested to provide a few more drops of blood from a finger prick for laboratory testing for HbA1c, malaria parasites, and Vitamin D3 [8].
We made a request to the DHS Program to provide us with the NFHS data. Once we received the permission to use the data, we downloaded the Men’s data file (MR) and the Household *Member data* file (PR). MR and PR datasets were then merged to avail the information on anaemia among men in India.
## Ethics approval and consent to participate
The present study has used secondary data, which is available in the public domain. The dataset has no identifiable information of the survey participants. Therefore, no ethical approval is required for this study.
## Sample
Fig 1 depicts the process of sample selection for the present study. Out of the 111,179 eligible men aged 15–54 years selected for the state module, 101839 men who were normally inhabitants and spent the night before the survey in their homes were interviewed. 92820 men consented to have their haemoglobin levels checked. For this study, 31339 out of the 92820 men were excluded as they belonged to urban areas. Our study was limited to remaining 61481 men residing in rural areas.
**Fig 1:** *Sample selection for the present study.*
## Anaemia testing
The authors did not collect blood specimens for anaemia testing for this study. These were collected under NFHS-5 by health investigators from eligible men aged 15 to 54 with their consent. Blood samples were drawn from a drop of blood taken from a finger prick (or a heel prick for children age 6–11 months) [8] and collected in a microcuvette, a single-use pipette. Concentration of haemoglobin was analysed on-site with HemoCue Hb 201+ analyser. Introduced in 1990, the HemoCue Hb 201+ is a battery-operated portable device used for quantitative determination of haemoglobin level in undiluted, capillary or venous blood. It converts the haemoglobin into methemoglobin and combines it with azide to form azidemethemoglobin followed by measurement of transmittance and haemoglobin absorbance [22–24].
## Dependent variable
The dependent variable in this study was whether or not the respondents were anaemic. Men were considered to have anaemia in any form if their haemoglobin concentration was less than 13.0 g/dL, mildly anaemic if it was 12.0–12.9 g/dL, moderately anaemic if it was 9.0–11.9 g/dL, and severely anaemic if it was less than 8.9 g/dL, according to WHO criteria [25]. The study developed a dichotomous variable for prevalence of anaemia. Men with a haemoglobin level less than 13g/dl were considered ‘anaemic’ and coded ‘1’ while men having a haemoglobin level greater than 13g/dl were classified as ‘not anaemic’ and coded ‘0’. We did not take into account the three categories of anaemia: mild, moderate, and severe.
## Independent variables
Definitions, categories and codes of independent variables are enlisted in Table 1. A wide range of variables were found to predict anaemia among men [26–28]. To illustrate, as a proxy for household income, the wealth index was chosen as a gauge of economic inequality. This is a measure of household wealth that is determined to be reliable based on both expenditure and income metrics [8]. Wealth index was one of the key predictors of this study. BMI was also one of the significant determinants of anaemia among men which was classified into four categories i.e., underweight (<18.5 kg/m2), normal (18.5–24.99 kg/m2), overweight (25–32 kg/m2), and obese (>32kg/m2). Age, education, social group, religion, and alcohol consumption were other important factors of anaemia in men. These variables could be categorised into four domains namely socioeconomic factors, community variables, health-related variables and behavioural characteristics. All the variables in present study were selected only after extensive review of literature and according to data availability. Fig 2 depicts a conceptual framework that shows the factors affecting anaemia among men.
**Fig 2:** *Conceptual framework showing factors affecting anaemia.* TABLE_PLACEHOLDER:Table 1
## Statistical analysis
Bivariate statistics was used to analyse the prevalence of anaemia among rural men by their background characteristics. The analysis was weighted for two-stage sampling design. Thus, weighted estimates were presented. Sampling weights (importance weight: iw) were included in the study. The ‘svyset’ command was used in Stata to account for clustering at the PSU level.
Since our dependent variable was binary in nature, we employed binary logistic regression to assess the effects of the predictor variables on the dichotomous dependent variable of the study. Chi-square test was performed to check if the independent variables were associated with the dependent variable. Only those variables that were found statistically significant ($p \leq 0.05$) were included in the regression models. We applied four models, i.e., model 1 included socioeconomic variables; model 2 included statistically significant variables from model 1 variables and geographic variable. Model 3 contained significant variables from model 2 and health-related variables. The final model included significant variables from model 3 and behavioural variables. The equation of a single-level binary logistic regression model can be specified as: Log(P/1‐P)=β0+β1x1+…+β2x2 Where, P indicates the probability of an event (prevalence of anaemia in this study), β0 is the intercept on y axis, βi indicates the regression coefficients associated with the reference group, xi is the independent variable.
The results of logistic regression models are presented in the form of odds ratios with p-values and $95\%$ confidence intervals (CI). We calculated Variance Inflation Factors (VIFs) for the final model to check whether multicollinearity among the independent variables existed. The VIFs for all the independent variables were considerably small (below 2.5) indicating that multicollinearity was not a problem for the models (results not shown). Stata 16 was used for analysing the unit level data [29]. ArcMap (version 10.5) was used to create the choropleth maps [30].
## Profile of the respondents
Table 2 present the sociodemographic profile of the men in rural India. About $17\%$ men were aged between 15–19 years. One in every seven men had no formal education. About a quarter of all men belonged to SC group, while $12.4\%$ men belonged to ST group. The Hindu faith was practised by the vast majority of men ($81\%$). Nearly one-fourth men belonged to the poorest wealth quintile. Around $30\%$ men were from the eastern region of India. At the time of the survey, roughly two-thirds of men were married. About $18\%$ of men were underweight (BMI less than 18.5 kg/m2). Around $28\%$ of men used smokeless tobacco, while $45\%$ smoked cigarettes.
**Table 2**
| Background Characteristics | n | % |
| --- | --- | --- |
| Age (in years) | | |
| 15–19 | 10329.0 | 16.8 |
| 20–29 | 17061.0 | 27.8 |
| 30–39 | 15592.0 | 25.4 |
| 40–49 | 13157.0 | 21.4 |
| 50–54 | 5349.0 | 8.7 |
| Level of education | | |
| No education | 8853.0 | 14.4 |
| Primary | 8423.0 | 13.7 |
| Secondary | 35597.0 | 57.9 |
| Higher | 8607.0 | 14.0 |
| Social group | | |
| SC | 14202.0 | 23.1 |
| ST | 7624.0 | 12.4 |
| OBC | 27543.0 | 44.8 |
| Others | 12173.0 | 19.8 |
| Religion | | |
| Hindu | 8546.0 | 81.0 |
| Muslim | 8546.0 | 13.9 |
| Christian | 1721.0 | 2.8 |
| Others | 1414.0 | 2.3 |
| Household wealth | | |
| Poorest | 14878.0 | 24.2 |
| Poorer | 16047.0 | 26.1 |
| Middle | 14694.0 | 23.9 |
| Richer | 11067.0 | 18.0 |
| Richest | 4734.0 | 7.7 |
| Marital status | | |
| Never married | 22555.0 | 34.2 |
| Currently married | 42465.0 | 64.3 |
| Work status | | |
| No | 14817.0 | 24.1 |
| Yes | 46664.0 | 75.9 |
| Region | | |
| South | 14264.0 | 23.2 |
| North | 5164.0 | 8.4 |
| Central | 7808.0 | 12.7 |
| East | 17891.0 | 29.1 |
| West | 12235.0 | 19.9 |
| North-East | 4119.0 | 6.7 |
| Body Mass Index | | |
| Underweight | 11067.0 | 18.0 |
| Normal | 38057.0 | 61.9 |
| Overweight | 10390.0 | 16.9 |
| Obese | 1967.0 | 3.2 |
| Media exposure | | |
| No | 24285.0 | 39.5 |
| Yes | 37196.0 | 60.5 |
| Consumption non-vegetarian food | | |
| No | 19182.0 | 31.2 |
| Yes | 42299.0 | 68.8 |
| Alcohol consumption | | |
| Never | 47340.0 | 77.0 |
| Less than once a week | 5841.0 | 9.5 |
| Once a week | 6087.0 | 9.9 |
| Almost every day | 2213.0 | 3.6 |
| Currently smokes | | |
| No | 33876.0 | 55.1 |
| Yes | 27605.0 | 44.9 |
| Use of smokeless tobacco | | |
| No | 44020.0 | 71.6 |
| Yes | 17461.0 | 28.4 |
| Total | 61481.0 | 100.0 |
## Differentials in prevalence of anaemia by background characteristics
Table 3 depicts the prevalence of anaemia among rural men in the country by various background characteristics. Overall, about one-fourth men in India were found to have anaemia. One out of every five urban men while three out of every ten rural men were anaemic in India (Fig 3). The prevalence of anaemia among rural men was highest in the age group 50–54 ($34.1\%$) followed by the age group 15–19 years ($33.8\%$). Men aged 20–29 years ($22.9\%$) had the lowest prevalence of anaemia. Prevalence of anaemia decreased with increase in education. Men with no education had the highest prevalence of anaemia. ST men ($30.9\%$) showed the highest prevalence of anaemia among the social groups. Anaemia prevalence was significantly higher among Muslim men and lower among Christian men. A steady decline was observed in the prevalence of anaemia with increase in household wealth. About one-third rural men belonging to the poorest households had anaemia.
**Fig 3:** *Trend of anaemia prevalence among rural men and men overall in India, NFHS.* TABLE_PLACEHOLDER:Table 3 Anaemia prevalence was highest in the eastern region ($34.1\%$) while lowest in the southern region ($18.5\%$). The north, west, north-west and central regions reported $27.2\%$, $28.9\%$, $26.9\%$ and $25\%$ prevalence of anaemia, respectively. Anaemia was inversely related with BMI as prevalence of anaemia was $34.7\%$ among underweight men versus $19.3\%$ among men who were overweight. Men who drank alcohol daily and used smokeless tobacco had slightly higher occurrence of anaemia than who did not consume it.
## Estimates from logistic regression analysis for anaemia among men in rural India
The results from the logistic regression models are presented in Table 4. The men within the age bracket 20–29 years and 30–39 years were $33\%$ and $26\%$ less likely to be anaemic than men aged 15–19. Men aged 50–54 years were slightly more likely to be anaemic (OR: 1.10, $95\%$ CI, 1.00–1.21). Men with no formal education were $36\%$ more likely ($95\%$ CI, 1.26–1.47) to be anaemic than men who obtained higher education. Men with primary education were a quarter times more likely ($95\%$ CI, 1.15–1.34) to have anaemia as compared to men with higher education. Men from to ST category (OR: 1.48, $95\%$ CI, 1.39–1.58) had significantly higher likelihood of being anaemic as compared to men of ‘Others’ category. The odds of being anaemic were $36\%$ higher among Muslim men ($95\%$ CI, 1.27–1.45) and $48\%$ lower among Christian men ($95\%$ CI, 0.48–0.57) as compared to Hindu men. The more the wealth, the lesser the risk to suffer from anaemia. Men from the richest wealth quintile were $29\%$ less likely to suffer from anaemia ($95\%$ CI, 0.66–0.78) than those from the poorest wealth quintile.
**Table 4**
| Unnamed: 0 | Model 1 | Model 1.1 | Model 1.2 | Model 1.3 | Model 2 | Model 2.1 | Model 2.2 | Model 2.3 | Model 3 | Model 3.1 | Model 3.2 | Model 3.3 | Model 4 | Model 4.1 | Model 4.2 | Model 4.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | CI (95%) | CI (95%) | | | CI (95%) | CI (95%) | | | CI (95%) | CI (95%) | | | CI (95%) | CI (95%) | |
| Independent Variables | OR | Upper | Lower | P value | OR | Upper | Lower | P value | OR | Upper | Lower | P value | OR | Upper | Lower | P value |
| Socioeconomic variables Age | | | | | | | | | | | | | | | | |
| 15–19 | Ref. | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| 20–29 | 0.62 | 0.58 | 0.66 | 0.000 | 0.61 | 0.57 | 0.65 | 0.000 | 0.66 | 0.62 | 0.70 | 0.000 | 0.67 | 0.63 | 0.71 | 0.000 |
| 30–39 | 0.65 | 0.60 | 0.70 | 0.000 | 0.65 | 0.60 | 0.70 | 0.000 | 0.72 | 0.67 | 0.78 | 0.000 | 0.74 | 0.69 | 0.80 | 0.000 |
| 40–49 | 0.79 | 0.73 | 0.86 | 0.000 | 0.79 | 0.73 | 0.86 | 0.000 | 0.88 | 0.81 | 0.96 | 0.002 | 0.91 | 0.84 | 0.99 | 0.032 |
| 50–54 | 0.95 | 0.86 | 1.04 | 0.027 | 0.95 | 0.87 | 1.04 | 0.286 | 1.06 | 0.96 | 1.16 | 0.252 | 1.10 | 1.00 | 1.21 | 0.052 |
| Level of education | | | | | | | | | | | | | | | | |
| No education | 1.36 | 1.26 | 1.46 | 0.000 | 1.38 | 1.28 | 1.18 | 0.000 | 1.32 | 1.13 | 1.43 | 0.000 | 1.36 | 1.26 | 1.47 | 0.000 |
| Primary | 1.26 | 1.17 | 1.36 | 0.000 | 1.25 | 1.16 | 1.35 | 0.000 | 1.21 | 1.12 | 1.30 | 0.000 | 1.24 | 1.15 | 1.34 | 0.000 |
| Secondary | 1.18 | 1.11 | 1.25 | 0.000 | 1.17 | 1.10 | 1.24 | 0.000 | 1.14 | 1.08 | 1.21 | 0.000 | 1.16 | 1.09 | 1.23 | 0.000 |
| Higher | Ref. | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| Social group | | | | | | | | | | | | | | | | |
| Others | Ref. | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| SC | 1.03 | 0.97 | 1.09 | 0.345 | 1.08 | 1.01 | 1.15 | 0.016 | 1.06 | 0.99 | 1.12 | 0.072 | 1.07 | 1.00 | 1.13 | 0.044 |
| SC | 1.43 | 1.34 | 1.52 | 0.000 | 1.47 | 1.38 | 1.57 | 0.000 | 1.47 | 1.38 | 1.56 | 0.000 | 1.48 | 1.39 | 1.58 | 0.000 |
| OBC | 0.94 | 0.90 | 0.99 | 0.029 | 1.01 | 0.96 | 1.07 | 0.725 | 1.00 | 0.94 | 1.05 | 0.877 | 0.99 | 0.94 | 1.05 | 0.783 |
| Religion | | | | | | | | | | | | | | | | |
| Hindu | Ref. | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| Muslim | 1.40 | 1.31 | 1.49 | 0.000 | 1.35 | 1.27 | 1.44 | 0.000 | 1.38 | 1.29 | 1.47 | 0.000 | 1.36 | 1.27 | 1.45 | 0.000 |
| Christian | 0.51 | 0.48 | 0.56 | 0.000 | 0.52 | 0.47 | 0.56 | 0.000 | 0.52 | 0.48 | 0.57 | 0.000 | 0.52 | 0.48 | 0.57 | 0.000 |
| Others | 0.98 | 0.91 | 1.06 | 0.618 | 0.91 | 0.84 | 0.99 | 0.025 | 0.94 | 0.86 | 1.02 | 0.109 | 0.94 | 0.87 | 1.02 | 0.139 |
| Household wealth | | | | | | | | | | | | | | | | |
| Poorest | Ref. | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| Poorer | 0.85 | 0.81 | 0.89 | 0.000 | 0.88 | 0.84 | 0.92 | 0.000 | 0.90 | 0.86 | 0.94 | 0.000 | 0.89 | 0.85 | 0.94 | 0.000 |
| Middle | 0.76 | 0.72 | 0.80 | 0.000 | 0.81 | 0.77 | 0.86 | 0.000 | 0.85 | 0.80 | 0.90 | 0.000 | 0.84 | 0.80 | 0.89 | 0.000 |
| Richer | 0.68 | 0.64 | 0.72 | 0.000 | 0.73 | 0.69 | 0.78 | 0.000 | 0.78 | 0.73 | 0.83 | 0.000 | 0.77 | 0.72 | 0.82 | 0.000 |
| Richest | 0.63 | 0.58 | 0.68 | 0.000 | 0.66 | 0.61 | 0.72 | 0.000 | 0.72 | 0.66 | 0.79 | 0.000 | 0.71 | 0.66 | 0.78 | 0.000 |
| Work status | | | | | | | | | | | | | | | | |
| No | Ref. | | | | | | | | | | | | | | | |
| Yes | 0.94 | 0.90 | 0.99 | 0.013 | | | | | | | | | | | | |
| Marital status | | | | | | | | | | | | | | | | |
| Never in union | Ref. | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| Currently in union | 0.96 | 0.90 | 1.02 | 0.162 | 0.93 | 0.88 | 0.98 | 0.013 | 0.96 | 0.90 | 1.02 | 0.162 | 0.97 | 0.91 | 1.03 | 0.286 |
| Formerly in union | 1.20 | 1.04 | 1.38 | 0.012 | 1.18 | 1.02 | 1.36 | 0.023 | 1.20 | 1.04 | 1.38 | 0.014 | 1.22 | 1.05 | 1.40 | 0.007 |
| Geographic variable Region | | | | | | | | | | | | | | | | |
| Central | | | | | Ref. | | | | Ref. | | | | Ref. | | | |
| South | | | | | 0.78 | 0.73 | 0.84 | 0.000 | 0.80 | 0.75 | 0.86 | 0.000 | 0.81 | 0.76 | 0.87 | 0.000 |
| North | | | | | 1.25 | 1.18 | 1.32 | 0.000 | 1.27 | 1.20 | 1.35 | 0.000 | 1.27 | 1.20 | 1.35 | 0.000 |
| East | | | | | 1.44 | 1.36 | 1.52 | 0.000 | 1.45 | 1.37 | 1.53 | 0.000 | 1.47 | 1.39 | 1.55 | 0.000 |
| West | | | | | 1.32 | 1.24 | 1.40 | 0.000 | 1.30 | 1.22 | 1.39 | 0.000 | 1.28 | 1.20 | 1.36 | 0.000 |
| Northeast | | | | | 1.16 | 1.09 | 1.24 | 0.000 | 1.21 | 1.13 | 1.29 | 0.000 | 1.24 | 1.16 | 1.33 | 0.000 |
| Health-related variables Body Mass Index | | | | | | | | | | | | | | | | |
| Normal | | | | | | | | | Ref. | | | | Ref. | | | |
| Underweight | | | | | | | | | 1.36 | 1.30 | 1.43 | 0.000 | 1.36 | 1.30 | 1.43 | 0.000 |
| Overweight | | | | | | | | | 0.76 | 0.72 | 0.80 | 0.000 | 0.76 | 0.72 | 0.80 | 0.000 |
| Obese | | | | | | | | | 0.76 | 0.68 | 0.86 | 0.000 | 0.77 | 0.68 | 0.86 | 0.000 |
| Blood sugar | | | | | | | | | | | | | | | | |
| Normal | | | | | | | | | Ref. | | | | | | | |
| High | | | | | | | | | 1.07 | 0.99 | 1.15 | 0.086 | | | | |
| Very High | | | | | | | | | 1.07 | 0.98 | 1.17 | 0.141 | | | | |
| Behavioural variables Media exposure | | | | | | | | | | | | | | | | |
| No | | | | | | | | | | | | | Ref. | | | |
| Yes | | | | | | | | | | | | | 1.03 | 1.00 | 1.08 | 0.087 |
| Consumption of non-vegetarian food | | | | | | | | | | | | | | | | |
| No | | | | | | | | | | | | | Ref. | | | |
| Yes | | | | | | | | | | | | | 1.03 | 0.98 | 1.07 | 0.217 |
| Alcohol consumption | | | | | | | | | | | | | | | | |
| Never | | | | | | | | | | | | | Ref. | | | |
| Less than once a week | | | | | | | | | | | | | 0.85 | 0.80 | 0.90 | 0.000 |
| Once a week | | | | | | | | | | | | | 0.91 | 0.85 | 0.96 | 0.001 |
| Almost every day | | | | | | | | | | | | | 0.96 | 0.88 | 1.05 | 0.412 |
| Currently smokes | | | | | | | | | | | | | | | | |
| No | | | | | | | | | | | | | Ref. | | | |
| Yes | | | | | | | | | | | | | 1.15 | 1.08 | 1.22 | 0.022 |
| Use of smokeless tobacco | | | | | | | | | | | | | | | | |
| No | | | | | | | | | | | | | Ref. | | | |
| Yes | | | | | | | | | | | | | 1.07 | 1.01 | 1.13 | 0.016 |
The odds of anaemia among men from the east region of the country were $47\%$ ($95\%$ CI, 1.39–1.55) higher than those from the central region. Men belonging to the north, west and north-east regions were $27\%$, $28\%$, and $24\%$ more likely to suffer from anaemia. However, the odds of anaemia among men were lower by $19\%$ ($95\%$ CI, 0.76–0.87) in the south region. Men who were underweight had $36\%$ ($95\%$ CI, 1.30–1.43) more likelihood of being anaemic whereas obese men were $23\%$ ($95\%$ CI, 0.68–0.86) less likely to suffer from anaemia as compared to men with normal BMI. The risk of anaemia among men using smokeless tobacco was more (OR:1.15, $95\%$ CI, 1.08–1.22) than those not using the same.
## Spatial analysis
Figs 4 and 5 show the spatial distribution of prevalence of anaemia among men across the Indian states and districts respectively using choropleth map. It is a technique through which unequal distribution of an element within a geographic area is depicted through gradients of the same colour. The higher the value or prevalence, the darker the shade [31].
**Fig 4:** *Map showing state wise prevalence anaemia among rural men in India, NFHS-5, 2019–21.(All the maps used are the authors’ own creations. As far as the base layer is concerned, we used a free GIS file from https://spatialdata.dhsprogram.com/boundaries/#view=table&countryId=IA for national and sub-national boundaries).* **Fig 5:** *Map showing district wise prevalence of anaemia among rural men in India, NFHS-5, 2019–21.(All the maps used are the authors’ own creations. As far as the base layer is concerned, we used a free GIS file from https://spatialdata.dhsprogram.com/boundaries/#view=table&countryId=IA for national and sub-national boundaries).*
The states as well as districts of India were classified into six categories according to prevalence of anaemia (%). West Bengal, Tripura, Assam, UT Jammu and Kashmir (>$35\%$) were found to have highest anaemia prevalence among rural men followed by Bihar, Jharkhand, Chhattisgarh, Odisha, Gujarat ($28\%$-$35\%$). South Indian states i.e., Andhra Pradesh, Karnataka, Tamil Nadu, and Kerala,) as well as Manipur and Nagaland showed lowest prevalence.
Districts of West Bengal, Bihar, Odisha, Chhattisgarh, Jammu & Kashmir (>$50\%$) showed the highest prevalence of anaemia among rural men followed by some districts of Uttar Pradesh and Madhya Pradesh ($40\%$-$50\%$). The lowest prevalence was found in some districts of Manipur, Nagaland, Karnataka, and Tamil Nadu (<$10\%$).
## Discussion
An estimated one-fourth of Indian rural men aged 15 to 54 years were found to be anaemic. In the last four years, the prevalence of anaemia has only risen [8]. The findings suggested that the prevalence of anaemia varies by sociodemographic characteristics among rural men. Rural men who were 49–54 years old, had no formal education, belonged to ST group, were Muslim, or from the poorest wealth quintile, lived in the eastern region, were underweight, and consumed alcohol and smokeless tobacco on a daily basis were more likely to have anaemia. At both the state and district levels, there was significant geographical variation in the prevalence of anaemia.
Older men, aged 50–54 years, were more likely to be anaemic, followed by adolescents (15–19 years). However, the prevalence of anaemia was lowest in the 20–29 age group. Older men are more vulnerable to anaemia, possibly as a result of suffering from other chronic diseases such as diabetes, chronic kidney diseases, hypertension [32]. Previous research has yielded similar results [26,33], although none of these studies have been conducted specifically on rural men.
Education level emerged to be a significant determinant of anaemia. Rural men with no education were most vulnerable to anaemia. Men with a higher level of education were less likely to develop anaemia. Previous studies also corroborate the same [19,34]. Disease awareness and knowledge, as well as the necessity of sanitation and health care, are raised through education. It also encourages people to listen to and accept the advice of health professionals [35,36].
Men in the ST category were more likely to be anaemic than men from other social groups. STs have a long history of being marginalised, with the majority of them still living in remote areas of the country. As a result, poor diet and a lack of access to healthcare services could be linked to their increased risk of anaemia [37]. Various studies have found that SCs and STs generally have poor health outcomes, underscoring the importance of caste prejudices. Despite affirmative actions by the Indian government post-1947, people from SC/ST groups remain deprived in a variety of areas, including health [38–40].
Muslim men had greater risk of anaemia, which is similar to the findings from an earlier study [20]. This study also found that Christian men living in rural India were at significantly lower risk of anaemia, which was not highlighted previously by any scientific study and requires further investigation. However, these results need to be cautiously interpreted as the sample size for Muslim and Christian population was small and sample distribution was highly skewed.
Household wealth was strongly associated with anaemia. The wealthier the household, the lesser the risk of anaemia. Several researches have explored the linkage between poverty and malnutrition. Poverty makes it difficult for people to eat a healthy diet and get health care [41,42]. Low socioeconomic position can exacerbate the prevalence of anaemia in a variety of ways, including a poor living and working environment, unhealthy habits such as smoking and limited access to health care, and a lack of health literacy [40]. In developing countries, people from poorer households are more likely to suffer anaemia than those from wealthy homes [43] due to factors such as substandard housing, hunger, and increased disease exposure.
A significant geographical variation in the prevalence of anaemia among rural men was also noted in this study. The likelihood of being anaemic was maximum among men belonging to the eastern region. A recent study on anaemia among men also pointed out higher prevalence of anaemia among men in the eastern India [21]. A previous study on anaemia among children also highlighted that people from central and eastern region were associated with higher risk of being anaemic [44]. However, this study found that men from the north, west and north-east region were more likely to suffer from anaemia compared to men in the central region. The differentials in population composition from one region to another could play a vital role in spatial variation in anaemia.
Anaemia was significantly associated with BMI in this study. Rural men with lower BMI, who were underweight, had a higher risk of anaemia. It is a well-researched fact that underweight persons have higher likelihood to be anaemic, as low BMI is caused by a lack of a balanced and healthy diet [45]. Men who drank alcohol on a regular basis had a higher risk of anaemia than men who did not consume alcohol. A population based study on India offered a similar finding [19]. Frequent consumption of alcohol leads to deterioration of health and a number of other chronic illnesses which are linked with anaemia.
There was no significant relationship found between anaemia prevalence and media exposure, blood sugar, and non-vegetarian food consumption. Previous studies have found a link between anaemia and these variables. It has been noted that as blood sugar levels rise, the likelihood of being anaemic also rises [46]. Another study on women in Afghanistan found a strong negative correlation between anaemia risk and meat consumption frequency [47]. Since the above-mentioned study was conducted on women, we should exercise caution while comparing the results of the current study with the findings from other studies. It was also found that women who took iron tablets or syrup at regular intervals had a lower risk of anaemia [48]. Thus, further research is needed to investigate the effect of such interventions in men.
The Indian government has devised a number of programmes and policies aimed at reducing the prevalence of anaemia in the country. Almost all of them, however, primarily target women of reproductive age and children. The National Nutritional Anaemia Control Program (NNACP), for example, was established in 1970 with the goals of encouraging regular consumption of iron-rich foods, providing iron and folate supplements to susceptible groups, and identifying and treating severely anaemic patients [49]. The 12-by-12 Initiative [2007] was launched in collaboration with the Ministry of Health and Family Welfare, the World Health Organization (WHO), UNICEF, and the Food and Agriculture Organization of the United Nations, with the goal of every child having a haemoglobin level of 12 grams by the age of 12 years by 2012 [50]. The National Iron Plus Initiative (NIPI) was launched in 2013 with the goal of providing free iron and folic acid supplements to adolescent boys and girls (10–19 years) as well as women of reproductive age, such as pregnant and breastfeeding women. The Union Ministry of Health and Family Welfare started the Weekly Iron and Folic Acid Supplementation (WIFS) initiative in 2013 under the National Health Mission to prevent anaemia among teenagers (NHM). Another initiative, Anaemia Mukt Bharat [2018], aimed to cut anaemia in young children, teenage boys and girls, pregnant and breastfeeding women, and women of reproductive age by half [50]. It is therefore recommended that the government and policymakers expand the scope of programs such as Anaemia Mukt Bharat and Iron supplementation programs to include sub-strata of rural aged men who are more susceptible to anaemia. Additionally, more research is needed to design such health interventions for men, keeping the sociodemographic and cultural context in sight.
This study has a few limitations that should be mentioned. First, because the data for this study comes from a cross-sectional survey, the relationship between dependent and independent variables demonstrated in this paper should be interpreted as association, and not as causality. Second, the model in this study uses only those variables that were available in the dataset. Some predictor variables may have been left out which may have resulted in what is known as omitted variable bias. We were unable to include folate, vitamin B12, or vitamin A intake as predictor variables in the model due to a lack of data. Third, the nationwide representative survey measured haemoglobin concentrations with a battery-operated portable HemoCueHb 201+ analyser, which may have underestimated the results when compared to laboratory testing (Didzun et al., 2019). Future research should take these limitations into account to get a more accurate and comprehensive picture of the prevalence of anaemia among rural men in India.
## Conclusion
Anaemia among rural men, just like it is among women and children, is a serious public health concern in India. Anaemia was found in three out of ten rural men. High-risk groups were older men, men without education, Muslim and STs, men from the poorest households, and men who were underweight. The benefits of existing programs and policies related to anaemia eradiation should be extended to men as well. In addition, targeted interventions among susceptible groups of rural men are advised as a way to reduce the prevalence of anaemia. Men’s haemoglobin levels should be checked on a regular basis and for that purpose appropriate screening facilities should to be made available closer to their residences so that they can be screened easily. When developing policies, it is important to keep geographical regions with high anaemia prevalence in mind. A comprehensive strategy based on the aforementioned proposals could be beneficial to reduce burden of anaemia among men in rural India.
## References
1. 1WHO. Health Topics. Anaemia. World Health Organisation. 2022.. *Health Topics. Anaemia* (2022.0)
2. Marengo-rowe AJ. **Bumc0019-0239**. *Proc Baylor Univ Med Cent* (2006.0) **19** 239-245
3. 3WHO. Nutritional Anaemias: Tools for Effective Prevention. World Health Organization.
2017.. *Nutritional Anaemias: Tools for Effective Prevention* (2017.0)
4. Kassebaum NJ. **The Global Burden of Anemia GBD 2013 Anemia Collaborators and Nicholas J Kassebaum**. *Hematol Clin* (2016.0) **30** 247-308
5. 5WHO. Global nutrition targets 2025: Anaemia policy brief (WHO/NMH/NHD/14.4). Geneva: World Health Organization; 2014. Geneva; 2014.. *Global nutrition targets 2025: Anaemia policy brief (WHO/NMH/NHD/14.4)* (2014.0)
6. 6Goal 2 | Department of Economic and Social Affairs.
7. **Global Nutrition Report: 2021**. *Glob Nutr Rep* (2021.0)
8. 8International Institute for Population Sciences (IIPS) and ICF. 2021. National Family Health Survey (NFHS-5), 2019–21: India. Mumbai;
2021.. *National Family Health Survey (NFHS-5), 2019–21: India. Mumbai;* (2021.0) **2021**
9. Haider BA, Olofin I, Wang M, Spiegelman D, Ezzati M, Fawzi WW. **Anaemia, prenatal iron use, and risk of adverse pregnancy outcomes: Systematic review and meta-analysis**. *BMJ* (2013.0) **347** 1-19. DOI: 10.1136/bmj.f3443
10. Mireku MO, Davidson LL, Koura GK, Ouédraogo S, Boivin MJ, Xiong X. **Prenatal hemoglobin levels and early cognitive and motor functions of one-year-old children**. *Pediatrics* (2015.0) **136** e76-e83. DOI: 10.1542/peds.2015-0491
11. Scott SP, Chen-Edinboro LP, Caulfield LE, Murray-Kolb LE. **The impact of anemia on child mortality: An updated review**. *Nutrients* (2014.0) **6** 5915-5932. DOI: 10.3390/nu6125915
12. Daru J, Zamora J, Fernández-Félix BM, Vogel J, Oladapo OT, Morisaki N. **Risk of maternal mortality in women with severe anaemia during pregnancy and post partum: a multilevel analysis**. *Lancet Glob Heal* (2018.0) **6** e548-e554. DOI: 10.1016/S2214-109X(18)30078-0
13. Horton S, Ross J. *Corrigendum to: ‘“The Economics of iron deficiency”‘ [Food Policy 28 (2003) 51–75]* (2007.0) **32** 141-143. DOI: 10.1016/j.foodpol.2006.08.002
14. **Reversing productivity losses from iron deficiency: The economic case**. *J Nutr* (2002.0) **132** 794S-801S. DOI: 10.1093/jn/132.4.794S
15. **India: Health of the Nation ‘ s States- The India State-Level Disease Burden Initiative**. *New Delhi, India. New Delhi, India.;* (2017.0)
16. Islam GMR. **Inequality, chronic undernutrition, maternity, and diabetes mellitus as the determinant of anemia among ever-married women in Bangladesh**. *BMC Public Health* (2021.0) **21** 1-11. DOI: 10.1186/s12889-021-10362-2
17. Stevens GA, Finucane MM, De-Regil LM, Paciorek CJ, Flaxman SR, Branca F. **Global, regional, and national trends in haemoglobin concentration and prevalence of total and severe anaemia in children and pregnant and non-pregnant women for 1995–2011: A systematic analysis of population-representative data**. *Lancet Glob Heal* (2013.0) **1** 16-25. DOI: 10.1016/S2214-109X(13)70001-9
18. Nguyen PH, Scott S, Avula R, Tran LM, Menon P. **Trends and drivers of change in the prevalence of anaemia among 1 million women and children in India, 2006 to 2016**. *BMJ Glob Heal* (2018.0) **3** 1-12. DOI: 10.1136/bmjgh-2018-001010
19. Didzun O, De Neve JW, Awasthi A, Dubey M, Theilmann M, Bärnighausen T. **Anaemia among men in India: a nationally representative cross-sectional study**. *Lancet Glob Heal* (2019.0) **7** e1685-e1694. DOI: 10.1016/S2214-109X(19)30440-1
20. Kumar P, Sharma H, Sinha D. **Socio-economic inequality in anaemia among men in India: a study based on cross-sectional data**. *BMC Public Health* (2021.0) **21** 1-12. DOI: 10.1186/s12889-021-11393-5
21. Kumar P, Sharma H, Patel KK. **Prevalence and risk factors of anaemia among men: A study based on Empowered Action Group states, India**. *Nutr Health* (2021.0) **27** 191-198. DOI: 10.1177/0260106020982348
22. Cohen AR, Seidl-Friedman J. **HemoCue system for hemoglobin measurement. Evaluation in anemic and nonanemic children**. *Am J Clin Pathol* (1988.0) **90** 302-305. DOI: 10.1093/ajcp/90.3.302
23. Vanzetti G.. **An azide-methemoglobin method for hemoglobin determination in blood**. *J Lab Clin Med* (1966.0) **67** 116-126. PMID: 5900720
24. Sanchis-Gomar F, Cortell-Ballester J, Pareja-Galeano H, Banfi G, Lippi G. **Hemoglobin Point-of-Care Testing**. *The HemoCue System. J Lab Autom* (2013.0) **18** 198-205. DOI: 10.1177/2211068212457560
25. Khusun H, Ray Y, Schultink W, Dillon DHS. **World health organization hemoglobin cut-off points for the detection of anemia are valid for an Indonesian population**. *J Nutr* (1999.0) **129** 1669-1674. DOI: 10.1093/jn/129.9.1669
26. Awaluddin SM, Shahein NA, Rahim NCA, Zaki NAM, Nasaruddin NH, Saminathan TA. **Anemia among men in Malaysia: A population-based survey in 2019**. *Int J Environ Res Public Health* (2021.0) 18. DOI: 10.3390/ijerph182010922
27. Kant S, Kumar R, Malhotra S, Kaur R, Haldar P. **Prevalence and determinants of anemia among adult males in a rural area of Haryana, India**. *J Epidemiol Glob Health* (2019.0) **9** 128-134. DOI: 10.2991/jegh.k.190513.001
28. Singh RK, Patra S. **Extent of Anaemia among Preschool Children in EAG States, India: A Challenge to Policy Makers**. *Anemia* (2014.0) **2014**. DOI: 10.1155/2014/868752
29. 29StataCorp. Stata: Statistical Software. College Station, TX: College Station, TX: StataCorp LLC.;
2019.. *Stata: Statistical Software* (2019.0)
30. 30ESRI. ArcMap Software. Redlands, CA: ESRI INC, 2016.; 2016.. *ArcMap Software* (2016.0)
31. Chien TW, Wang HY, Hsu CF, Kuo SC, Liu M. **Choropleth map legend design for visualizing the most influential areas in article citation disparities: A bibliometric study**. *Med (United States)* (2019.0) **98**. DOI: 10.1097/MD.0000000000017527
32. Paul B, Wilfred NC, Woodman R, DePasquale C. **Prevalence and correlates of anaemia in essential hypertension**. *Clin Exp Pharmacol Physiol* (2008.0) **35** 1461-1464. DOI: 10.1111/j.1440-1681.2008.05031.x
33. Duman TT, Aktas G, Meryem Atak B, Kocak MZ, Kurtkulagi O, Bilgin S. **General characteristics of anemia in postmenopausal women and elderly men**. *Aging Male* (2021.0) **23** 780-784. DOI: 10.1080/13685538.2019.1595571
34. Adamu AL, Crampin A, Kayuni N, Amberbir A, Koole O, Phiri A. **Prevalence and risk factors for anemia severity and type in Malawian men and women: Urban and rural differences**. *Popul Health Metr* (2017.0) **15** 1-15. DOI: 10.1186/s12963-017-0128-2
35. Sunuwar DR, Sangroula RK, Shakya NS, Yadav R, Chaudhary NK, Pradhan PMS. **Effect of nutrition education on hemoglobin level in pregnant women: A quasi-experimental study**. *PLoS One* (2019.0) **14** 1-12. DOI: 10.1371/journal.pone.0213982
36. Raghupathi V, Raghupathi W. **The influence of education on health: An empirical assessment of OECD countries for the period 1995–201**. *Arch Public Heal* (2020.0) **78** 1-18. DOI: 10.1186/s13690-020-00402-5
37. Singh A, Kumar A, Kumar A. **Determinants of neonatal mortality in rural India, 2007–2008**. (2013.0) 2007-2008. DOI: 10.7717/peerj.75
38. Van De Poel E, Speybroeck N. **Decomposing malnutrition inequalities between Scheduled Castes and Tribes and the remaining Indian population**. *Ethn Heal* (2009.0) **14** 271-287. DOI: 10.1080/13557850802609931
39. Dommaraju P, Agadjanian V, Yabiku S. **The pervasive and persistent influence of caste on child mortality in India**. *Popul Res Policy Rev* (2008.0) **27** 477-495. DOI: 10.1007/s11113-008-9070-0
40. Vart P, Jaglan A, Shafique K. **Caste-based social inequalities and childhood anemia in India: Results from the National Family Health Survey (NFHS) 2005–2006 Chronic Disease epidemiology**. *BMC Public Health* (2015.0) **15** 1-8. DOI: 10.1186/s12889-015-1881-4
41. Kibret KT, Chojenta C, D’Arcy E, Loxton D. **Spatial distribution and determinant factors of anaemia among women of reproductive age in Ethiopia: A multilevel and spatial analysis**. *BMJ Open* (2019.0) **9**. DOI: 10.1136/bmjopen-2018-027276
42. Sunuwar DR, Singh DR, Adhikari B, Shrestha S, Pradhan PMS. **Factors affecting anaemia among women of reproductive age in Nepal: A multilevel and spatial analysis**. *BMJ Open* (2021.0) **11**. DOI: 10.1136/bmjopen-2020-041982
43. Balarajan Y, Ramakrishnan U, Özaltin E, Shankar AH, Subramanian S V. **Anaemia in low-income and middle-income countries**. *Lancet* (2011.0) **378** 2123-2135. DOI: 10.1016/S0140-6736(10)62304-5
44. Sharma H, Singh SK, Srivastava S. **Socio-economic inequality and spatial heterogeneity in anaemia among children in India: Evidence from NFHS-4 (2015–16)**. *Clin Epidemiol Glob Heal* (2020.0) **8** 1158-1171. DOI: 10.1016/j.cegh.2020.04.009
45. Pal A, De S, Sengupta P, Maity P, Dhara PC. **An investigation on prevalence of Anaemia in relation to BMI and nutrient intake among adult rural population of West Bengal, India**. *Epidemiol Biostat Public Heal* (2014.0) **11** 1-10. DOI: 10.2427/8915
46. Soliman AT, De Sanctis V, Yassin M, Soliman N. **Iron deficiency anemia and glucose metabolism**. *Acta Biomed* (2017.0) **88** 112-118. DOI: 10.23750/abm.v88i1.6049
47. Flores-Martinez A, Zanello G, Shankar B, Poole N. **Reducing anemia prevalence in Afghanistan: Socioeconomic correlates and the particular role of agricultural assets**. *PLoS One* (2016.0) **11** 1-23. DOI: 10.1371/journal.pone.0156878
48. Wendt A, Stephenson R, Young M, Webb-Girard A, Hogue C, Ramakrishnan U. **Individual and Facility-Level Determinants of Iron and Folic Acid Receipt and Adequate Consumption among Pregnant Women in Rural Bihar, India**. *PLoS One* (2015.0) **10** e0120404. DOI: 10.1371/journal.pone.0120404
49. Kumar A.. **National nutritional anaemia control programme in India**. *Indian J Public Health* (1999.0) **43** 3-5. PMID: 11243085
50. Bhatia PV, Sahoo DP, Parida SP. **India steps ahead to curb anemia: Anemia Mukt Bharat**. *Indian J Community Heal* (2018.0) **30** 312-316
|
---
title: 'Prevalence of SARS-CoV-2 infection in Baja California, Mexico: Findings from
a community-based survey in February 2021 in the Mexico-United States border'
authors:
- Oscar E. Zazueta
- Richard S. Garfein
- J. Oggun Cano-Torres
- César A. Méndez-Lizárraga
- Timothy C. Rodwell
- Raquel Muñiz-Salazar
- Diego F. Ovalle-Marroquín
- Neiba G. Yee
- Idanya Rubí Serafín-Higuera
- Susana González-Reyes
- Jesus Rene Machado-Contreras
- Lucy E. Horton
- Steffanie A. Strathdee
- Ruth Rodríguez
- Linda Hill
- Ietza Bojórquez-Chapela
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021449
doi: 10.1371/journal.pgph.0000820
license: CC BY 4.0
---
# Prevalence of SARS-CoV-2 infection in Baja California, Mexico: Findings from a community-based survey in February 2021 in the Mexico-United States border
## Abstract
Between March 2020 and February 2021, the state of Baja California, Mexico, which borders the United States, registered 46,118 confirmed cases of COVID-19 with a mortality rate of 238.2 deaths per 100,000 residents. Given limited access to testing, the population prevalence of SARS-CoV-2 infection is unknown. The objective of this study is to estimate the seroprevalence and real time polymerase chain reaction (RT-PCR) prevalence of SARS-CoV-2 infection in the three most populous cities of Baja California prior to scale-up of a national COVID-19 vaccination campaign. Probabilistic three-stage clustered sampling was used to conduct a population-based household survey of residents five years and older in the three cities. RT-PCR testing was performed on nasopharyngeal swabs and SARS-CoV-2 seropositivity was determined by IgG antibody testing using fingerstick blood samples. An interviewer-administered questionnaire assessed participants’ knowledge, attitudes, and preventive practices regarding COVID-19. In total, 1,126 individuals (unweighted sample) were surveyed across the three cities. Overall prevalence of SARS-CoV-2 infection by RT-PCR was $7.8\%$ ($95\%$ CI 5.5–11.0) and IgG seroprevalence was $21.1\%$ ($95\%$ CI 17.4–25.2). There was no association between border crossing in the past 6 months and SARS-CoV-2 prevalence (unadjusted OR 0.40, $95\%$CI 0.12–1.30). While face mask use and frequent hand washing were common among participants, quarantine or social isolation at home to prevent infection was not. Regarding vaccination willingness, $30.4\%$ ($95\%$ CI 24.4–3 7.1) of participants said they were very unlikely to get vaccinated. Given the high prevalence of active SARS-CoV-2 infection in Baja California at the end of the first year of the pandemic, combined with its low seroprevalence and the considerable proportion of vaccine hesitancy, this important area along the Mexico-United States border faces major challenges in terms of health literacy and vaccine uptake, which need to be further explored, along with its implications for border restrictions in future epidemics.
## Introduction
SARS-CoV-2 is a novel respiratory coronavirus that was first reported in Wuhan, China, in December 2019. It was declared a Public Health Emergency of International Concern by the World Health Organization (WHO) on January 31, 2020, and later characterized as a pandemic [1]. As of January 26, 2022, more than 360 million COVID-19 cases and 5.6 million deaths had been reported across the world [2]. However, case reporting depends on a myriad of factors, including testing capacities, type of tests used, surveillance system strategy and population health behaviors. Since many of SARS-CoV-2 infections are mild or asymptomatic, they are less likely to be detected by passive surveillance systems. Therefore, SARS-CoV-2 prevalence estimates might be more accurate using population-based studies [3].
Baja *California is* a state located in the northern part of Mexico that shares a border with California in the United States (U.S.). The border region is demographically and economically important to both countries, particularly at the Tijuana/San Diego and Mexicali/Calexico ports of entry. In 2019 alone, northbound border crossing estimates were 18.5 million pedestrians, 31.3 million personal vehicles and 1.4 million commercial vehicles annually [4]. These activities accounted for nearly 60 billion US dollars in bilateral trade [5]. As part of an international effort to reduce viral transmission in the border region, the U.S. reached agreements with Mexico to limit all non-essential travel across their borders starting March 20, 2020 [6].
The Mexican national surveillance system strategy employed during the COVID-19 pandemic was based on registering only symptomatic individuals who met the working definition of a suspected case and testing $10\%$ of mild cases and $100\%$ of severe acute respiratory illnesses [7]. Therefore, cases with mild or no symptoms went undetected, likely contributing to an increased spread of local SARS-CoV-2 transmission and underreporting of cases. As of January 26, 2022, through the Epidemiologic Surveillance System for Respiratory Diseases (SISVER, in Spanish), Baja California has registered 116,870 confirmed cases of COVID-19 [8]. Nationally, this state had the second-highest mortality rate after Mexico City, with 11,451 confirmed deaths (320 deaths per 100,000 population) [9] and the eleventh highest rate of excess mortality ($44.9\%$) due to COVID-19, out of 32 states [10]. Given the passive nature of the Mexican surveillance strategy and the limitations of a sentinel-based approach, population-based COVID-19 prevalence estimates in Baja California were needed to assess the state of the pandemic and inform future health policies going forward as COVID-19 vaccination efforts began. Based on the national policy on COVID-19 vaccine prioritization, during the month of February 2021 only health workers and adults aged 60 or more were to be vaccinated as vaccines arrived in the Baja California [11]. During the study period, only 11,476 residents of Baja California ($0.3\%$ of the total population) had received a vaccine for COVID-19 [12]. Our study aim was to estimate the prevalence of SARS-CoV-2 infection using a population-based survey in the three major cities of Baja California prior to implementation of a national vaccination campaign.
## Design
We conducted a population-based household survey to estimate the prevalence of SARS-CoV-2 infection by RT-PCR and the IgG seroprevalence in the general population of Baja California, Mexico.
## Participants and study settings
The survey was conducted in Baja California’s three most populous cities: Mexicali, Tijuana, and Ensenada from February 1st to February 19th of 2021, immediately following the state’s second wave of new cases of COVID-19 (Fig 1). Survey inclusion criteria were: Spanish speakers residing in Baja California for at least six months, age five years or older. Prior to participation, we obtained written informed consent (or assent and parental consent for minors) for biological sample collection and survey interview.
**Fig 1:** *Epidemic curve in Baja California during the COVID-19 pandemic in 2020–2021.*
The protocol was approved by the Institutional Review Board (IRB) of Tijuana General Hospital (No. CONBIOETICA-02-CEI-001-20170526). Since data shared with co-authors from San Diego were de-identified, the human subjects review was not required by the University of California, San Diego IRB.
## Sample size and design
A probability, stratified, three-stage clustered sample design was employed. The target sample size ($$n = 1$$,500) was calculated to estimate a prevalence of $4\%$ with a $30\%$ precision relative to the expected proportion ($1.2\%$ absolute precision), $95\%$ confidence limits and a design effect of 1.9. The sample was stratified by city, and within cities, using the primary sampling unit (PSU) termed “Basic Geo-Statistic Areas” (AGEB), which is the basic unit of Mexico’s National Institute of Geography and Statistics (INEGI) for the subdivision of municipal geostatistical areas. Within each stratum, 33 AGEBs were selected with probability proportional to the size (PPS) of the number of inhabitants in the AGEB. Within AGEBs, eight blocks were selected, also with PPS, and in each block four households were selected through systematic random sampling. In the final stage, one participant was selected at random from a list of eligible household members. If the selected household member was not home at the time, the survey teams returned at an appointed time to complete the survey procedures.
## Survey tool and materials
A questionnaire was developed to assess knowledge, attitudes, and practices related to the COVID-19 pandemic, along with sociodemographic data and other variables. Data were captured during a face-to-face interview with each participant using tablets and smartphones equipped with a digital platform for that purpose.
Three biological samples were obtained from each participant: Fingerstick whole blood that was placed onto Whatman 903 protein saver cards for IgG antibody testing of dried blood spots; a nasopharyngeal and oropharyngeal swab for RT-PCR, and a second nasopharyngeal swab for Panbio COVID-19 Ag rapid test device by Abbott (Lake Country, IL, U.S.A.).
The questionnaire and specimen collection were carried out by medical and health sciences students previously trained on the use of personal protection equipment, sampling, and interviewing techniques, with supervision by faculty and physicians from the research team. The questionnaire and the dataset for this survey is published and is open for consultation [13].
## Specimen testing
RT-PCR for all samples was conducted at Baja California Public Health State Laboratory. The extraction of total RNA from oropharyngeal and nasopharyngeal samples used a volume of 200 μL. A 200 μL sample of APEX BioResearch Products Water UltraPure Free of DNAse, RNAse, Proteases and Endonucleases molecular biology grade water was included as negative extraction control. Total RNA was extracted by the Bioneer ExiPrep 96 Viral DNA/RNA kit extraction with magnetic beads, using the Bioneer EP 96L-BXDOO7 automated extraction equipment with a 200 μL sample volume, eluted in 100 μL and stored at 4 °C.
RT-PCR was performed targeting the SARS-CoV-2-specific nucleocapsid (N1) gene and human RP gene. Real-time RT-PCR Primers and Probes (2021 Integrated DNA Technologies, Inc.). For N1 and RP, primers and probes came in a single reagent and were used per manufacturer’s instructions. Amplification for N1 and RP was carried out separately in a final volume of 20 μL with the following reagents: 10 μL of Master Mix (qPCR BIO Probe 1-Step Go Mix No-ROX), 1μL of 20X Rtase Go Probe qPCR BIO Probe 1-Step Go No-ROX, 1.5 μL of N1 and RP primers/probe premixed reagent, 2.5 μL of PCR grade H2O, and 5 μL of the extracted template. Thermal cycling conditions included 10 min at 50 °C for reverse transcription, 2 min at 95 °C for polymerase activation, followed by 45 cycles at 95 °C for 3 s and 55 °C for 30 s for denaturation and amplification/detection, respectively. The RT-PCR was performed on the ABI 7500 real-time PCR (Applied Biosystems, CA, USA). A sample was considered positive if the cycle threshold (Ct) for N1 was ≤ 40 and for RP amplification was Ct of <35 cycles. Samples with RP Ct values >35 were repeated from RNA extraction. If the result was the same, samples were reported as indeterminate. Each sample was evaluated once, positive, negative and non-template controls were included in each experiment.
For antibody detection, whole blood samples were obtained by fingerstick and collected in Whatman 903 protein saver cards. The samples were subsequently sent to the Broad Institute Serology Lab (BISL, Boston, U.S.), where anti-SARS-CoV-2 IgG presence and abundance were determined in dried blood spots by ELISA assay according to BILS protocol [14].
## Statistical analysis
For prevalence estimates, we employed weights reflecting the inverse of the selection probability, as well as an adjustment for non-response and calibration to match the 2020 Census population of each city. Means and standard deviations were obtained to describe quantitative variables, and absolute and relative frequency for categorical variables. Unadjusted and adjusted prevalence odds ratios (pOR) were calculated to identify sociodemographic, health, and behavioral factors associated with RT-PCR positivity and IgG seroprevalence for SARS-CoV-2 infection using logistic regression analysis. For all tests, p-values <0.05 were considered statistically significant. All statistical analyses were carried out using the complex sample module of Stata STATA 15 (StataCorp, College Station, TX, USA).
## Results
Out of 2,898 households visited, 1,283 ($44\%$) agreed to participate. At the individual level, 1,126 ($89\%$) of the 1,267 randomly selected persons agreed to participate (Fig 2).
**Fig 2:** *Flowchart of the study selection process.*
## General characteristics
The survey was applied to 1,126 consenting participants, $35\%$ of whom were from Mexicali, $35\%$ from Tijuana, and $30\%$ from Ensenada. After weighting, they represented a population of 2.8 million residents (Fig 2). The mean age was 37 years ($95\%$CI 35–40). Overall, the sample included $50\%$ women, of which $2\%$ reported being currently pregnant (Table 1). Half of the population ($53\%$) had less than a high school education, $26\%$ had completed high school, and $22\%$ had a college degree or higher. Regarding self-perceived health, only $1\%$ of the population reported their health as bad or very bad, with most participants ($76\%$) reporting good or very good health status. Since the population-wide vaccination campaign had not started yet in Baja California, less than $1\%$ reported being vaccinated against COVID-19 at the time of the survey. Crossing the border in the past six months was also uncommon ($5\%$). Health risks reported most often were obesity ($26\%$), hypertension ($21\%$), diabetes ($12\%$), and smoking ($16\%$). Seropositivity was marginally more frequent among females ($25\%$) as compared to males ($17\%$), and no other differences by sociodemographic characteristics were observed (Table 2).
## COVID-19 knowledge, attitudes and practices
Overall, $6\%$ of participants reported not knowing what COVID-19 was, $58\%$ correctly identified it as a disease caused by a virus, and a further $15\%$ knew it was an infectious disease.
In terms of attitudes on COVID-19, $62\%$ worried about family members or friends getting COVID-19, $49\%$ reported being worried about getting COVID-19 themselves, $12\%$ mentioned being worried about not having access to health services, $14\%$ worried about their job implications if they contracted the disease (such as losing their jobs and finding a new job), $7\%$ worried about not having a safe place to recover from the disease, and $3\%$ worried about being around other people at home.
Regarding prevention measures practiced in the past 6 months to prevent COVID-19 (Table 3), $74\%$ of participants reported hand washing or using hand-sanitizer very frequently, $86\%$ reported using face mask very frequently, $42\%$ reported adopting quarantine or social isolation at home or shelter very frequently, and $56\%$ avoided leaving their house very frequently.
**Table 3**
| Unnamed: 0 | Very frequently | Frequently | Occasionally | Rarely | Never |
| --- | --- | --- | --- | --- | --- |
| Washed hands regularly or used hand-sanitizer (%) | 74.0 | 18.3 | 6.3 | 1.2 | 0.2 |
| Quarantined or socially isolated at home or a shelter (stayed at home/shelter without any visitors from outside the household) | 42.3 | 20.8 | 13.0 | 9.0 | 14.9 |
| Avoided leaving the house | 56.3 | 21.0 | 13.9 | 5.7 | 3.1 |
| Used a facemask or face shield when in public spaces | 85.7 | 12.1 | 0.9 | 1.3 | 0.0 |
| Used public transportation | 9.9 | 2.8 | 9.5 | 19.9 | 57.9 |
| Visited a park, bar, or restaurant. Went to a concert or other crowded place | 5.7 | 2.3 | 8.1 | 24.8 | 59.1 |
| Visited family members in other households | 6.2 | 8.3 | 20.7 | 26.1 | 38.7 |
When asked which measures participants believed might help to prevent COVID-19, the most frequently mentioned responses were wearing a face mask ($89\%$), washing hands with soap and water ($66\%$), and keeping a distance of 1.5 meters from other people ($57\%$).
When asked if they would get vaccinated if they had free access to the vaccine against COVID-19, less than half ($45\%$) reported that it was very likely that they would and $30\%$ reported that it was very unlikely. When asked about their main concerns about the COVID-19 vaccine, $37\%$ had no concerns, whereas $41\%$ worried about side effects, $10\%$ did not think that the vaccine worked, $4\%$ worried that the development process of the vaccines was too fast, $4\%$ did not trust healthcare providers, $4\%$ worried that health authorities had economic rather than public health reasons to promote vaccines and $1\%$ had concerns about asking permission in their jobs.
## Prevalence and factors associated with SARS-CoV-2 infection
The overall prevalence of SARS-CoV-2 infection detected by RT-PCR was $7.8\%$ ($95\%$ CI 5.5–11.0) (Table 4), the highest documented in Ensenada ($22.2\%$, $95\%$ CI 15.0–31.6), followed by Tijuana ($6.4\%$, $95\%$ CI 3.4–11.6), and Mexicali ($5.5\%$, $95\%$ CI 2.9–10.3). The prevalence of IgG antibody seropositivity was $26\%$ ($95\%$ CI 20.6–32.3) in Mexicali, $18.7\%$ in Tijuana ($95\%$ CI 13.7–25.0), and $21.9\%$ in Ensenada ($95\%$ CI 16.2–28.9), with an overall prevalence of $21.1\%$ across the three cities ($95\%$ CI 17.4–25.2). We also examined the prevalence of SARS-CoV-2 infection by considering positivity for either RT-PCR or IgG antibodies and observed a weighted prevalence of $28.4\%$ ($95\%$ CI 22.5–35.2) in Mexicali, $23.3\%$ ($95\%$ CI 17.6–30.2) in Tijuana, $38.6\%$ ($95\%$ CI 31.0–46.8) in Ensenada, and $26.3\%$ ($95\%$ CI 22.2–30.9) overall.
**Table 4**
| LOCATION | Number of participants (n) | RT-PCR | RT-PCR.1 | IgG antibodies | IgG antibodies.1 | RT-PCR and IgG antibodies combined | RT-PCR and IgG antibodies combined.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| LOCATION | Number of participants (n) | Weighted prevalence (%) | 95% CI | Weighted prevalence (%) | 95% CI | Weighted prevalence (%) | 95% CI |
| Mexicali | 413 | 5.5 | 2.9, 10.3 | 26.0 | 20.6, 32.3 | 28.4 | 22.5, 35.2 |
| Tijuana | 373 | 6.4 | 3.4, 11.6 | 18.7 | 13.7, 25.0 | 23.3 | 17.6, 30.2 |
| Ensenada | 340 | 22.2 | 15.0, 31.6 | 21.9 | 16.2, 28.9 | 38.6 | 31.0, 46.8 |
| Overall | 1126 | 7.8 | 5.5, 11.0 | 21.1 | 17.4, 25.2 | 26.3 | 22.2, 30.9 |
After adjusting for potential confounders, higher odds of being positive for IgG was associated with being a current smoker (pOR = 3.0, $95\%$ CI 1.5–6.1), to be part of the older age group (pOR = 1.8, $95\%$ CI 1.0–3.2), and to have high school education as compared to elementary or less (pOR = 2.2, $95\%$ CI 1.2–4.1) (Table 5).
**Table 5**
| Exposure | pOR | 95% CI | Adjusted pORa | 95% CI.1 |
| --- | --- | --- | --- | --- |
| Gender (Ref.: male) | 1.7 | 1.1,2.6 | 1.4 | 0.9,2.2 |
| >60 years old | 1.7 | 1.1,2.7 | 1.8 | 1.0,3.2 |
| Educational level | | | | |
| Elementary or less (ref.) | | | | |
| Secondary | 1.8 | 0.9,3.8 | 1.7 | 0.9,3.4 |
| High school | 1.8 | 1.0,3.3 | 2.2 | 1.2,4.1 |
| College or more | 1.3 | 0.8,2.2 | 1.6 | 0.9,2.8 |
| Worked during past week | 1.2 | 0.7,1.9 | 1.0 | 0.6,1.6 |
| Crossed the border in the past 6 months | 0.4 | 0.1,1.3 | 0.4 | 0.1,1.4 |
| History of diabetes | 0.5 | 0.3,0.8 | 0.6 | 0.3,1.1 |
| History of obesity | 0.4 | 0.3,0.8 | 0.6 | 0.1,1.0 |
| History of hypertension | 0.5 | 0.3,0.9 | 0.8 | 0.4,1.5 |
| History of cardiovascular disease | 0.9 | 0.5,1.7 | 1.3 | 0.6,2.7 |
| History of chronic pulmonary disease | 0.9 | 0.4,1.9 | 1.1 | 0.5,2.4 |
| History of cáncer | 1.4 | 0.5,3.7 | 2.3 | 0.9,5.8 |
| Smoking | 3.8 | 1.9,7.5 | 3.0 | 1.5,6.1 |
## Discussion
This study shows that in February of 2021 the population-based weighted estimate of SARS-CoV-2 infection by RT-PCR in the three largest cities of Baja California was high ($7.8\%$), with a higher prevalence in Ensenada as compared with Mexicali and Tijuana. Overall, SARS-CoV-2 IgG weighted seroprevalence was $21.1\%$ and was similar across the three cities. Active smoking was associated with higher odds of infection based on serum IgG antibodies, while border crossing in the past 6 months, working during the last week, and history of diabetes, obesity, hypertension, cardiovascular disease, chronic pulmonary disease, or cancer were not associated with COVID-19 seroprevalence. COVID-19 knowledge in the general population was limited overall; however, risk reduction practices like wearing face masks, washing hands regularly, and social distancing were common, while few participants reported home isolation or quarantine and avoiding indoor locations to prevent infection. Hesitancy surrounding vaccine uptake before the Mexico national vaccination campaign was reported by more than half the participants, with approximately one-third reporting that they were unsure or unlikely to get vaccinated even if it was available free through public health services.
In this study, Ensenada had a considerable higher prevalence of SARS-CoV-2 infection by RT-PCR at the time of the survey as compared to Tijuana and Mexicali. This finding is consistent with official data from the National Epidemiological Surveillance System [8], and it is a reflection of the dynamics of the epidemic over time within the same territory. This differences could be attributed to the fact that Mexicali and Tijuana were experiencing the decrease phase of the second wave of COVID-19, while Ensenada had a slower decrease rate when the survey took place. This statement is supported by official data for the period of January 1st–February 19th of 2021, when the incidence rate in Ensenada (539 cases per 100,000 habitants) was 2.8 times higher than in Tijuana (189.3 cases per 100,000 habitants) and 2.1 times higher than in Mexicali (251.7 cases per 100,000 habitants) [8, 15]. Also, it is important to note that these incidence rates have to be interpreted carefully, since they come from a sentinel surveillance system. Furthermore, mortality data due COVID-19 also seems to support these conclusions. For this time period (January 1st–February 19th, 2021), Ensenada had a much higher mortality rate (107.9 deaths per 100,000 habitants) compared to Tijuana and Mexicali (39.7 and 50.2 deaths per 100,000 habitants, respectively) [8, 15].
Seroprevalence results obtained in this study were slightly lower than those reported by the National Survey on Health and Nutrition (ENSANUT in Spanish), conducted in Mexico in August-November 2020 [16]. However, it should be noted that while the nationwide seroprevalence of anti-SARS-CoV-2 IgG antibodies was $24.9\%$ ($95\%$CI 22.2–$26.7\%$), the Pacific-Northern region, which includes Baja California and four other border states, had a seroprevalence of $31\%$ ($95\%$CI 25–$36.6\%$) from a smaller sample ($$n = 851$$).
An additional cross-sectional study conducted in 34 clinical laboratories and 34 blood banks from the Mexican Institute of Social Security (IMSS, in Spanish) found that the overall seroprevalence based on IgG antibodies in Mexico was $3.5\%$ in February of 2020 but increased to $33.5\%$ by December 2020 [17]. They estimated a seroprevalence of $40.7\%$ ($95\%$ CI 36.9–$44.5\%$) in the Northwest region, which is significantly higher than what our study documented. The study analyzed 24,273 serum samples from across all 32 States; however, the observed differences may be explained by the study design and potential selection bias associated to including only samples from blood banks and clinical laboratories.
Conversely, the seroprevalence in Baja California was much higher than in other regions of Latin America, including Ecuador ($$n = 2$$,457, seroprevalence: $13.2\%$) [18], Brazil ($$n = 31$$,128, seroprevalence: $2.8\%$) [19], Chile ($$n = 1$$,367, seroprevalence: $13.3\%$) [20], and Argentina ($$n = 2$$,024, seroprevalence: $10.1\%$) [21]. Nonetheless, one study in Colombia conducted from September-October 2020 showed a higher prevalence in the three samples cities; Medellín ($$n = 1$$,832, seroprevalence: $27\%$), Barranquilla ($$n = 1$$,487, Seroprevalence: $55\%$), and Leticia ($$n = 1$$,417, seroprevalence: $59\%$) [22].
In this study, active smoking was significantly associated with the presence of anti-COVID-19 IgG antibodies in participants, suggesting it was a risk factor for natural infection. This is consistent with studies showing a higher prevalence of neutralizing SARS-CoV-2 antibodies reported among current smokers [23, 24]. However, lower blood levels of anti-SARS-CoV-2 IgG have also been documented in active smokers compared to non-smokers [25], and there is evidence that smoking can decrease serum levels of IgG [26]; indicating that the relationship between SARS-CoV-2 infection and cigarette smoking is complex and likely highly dependent on time since infection.
This study had several limitations. First, a high refusal rate may have impacted the representativeness of the sample. The possibility of selection bias as a result of this was partially addressed by weighting of the results, but still we cannot discard this possibility. Second, there were concerns about the sensitivity of the IgG assay based on dried blood spots instead of venous blood which could have resulted in an underestimate of the seroprevalence in these populations. Third, our study was limited to large urban areas, which may not adequately reflect the prevalence in semi-urban and rural areas within the state. However, we are confident that this effect was minimal given that the population living outside of the three sampled cities represents less than $5\%$ of Baja California’s population.
The results of this study have significant policy implications, given that border crossing restrictions were implemented in March of 2020, effectively limiting the entrance of non-citizens to the U.S. [6]. These restrictions were suspended on November 8, 2021, allowing fully vaccinated non-citizen travelers to enter the United States [27]. This study showed that regardless of being a border state, Baja California had a COVID-19 prevalence that was comparable to the rest of Mexico, and active border crossing in the past 6 months was not associated with a higher prevalence of SARS-CoV-2 infection. Population-based prevalence studies on the other side of the border, particularly in the counties of San Diego and Imperial in California, U.S., would be very valuable to compare the impact of the border restrictions during the pandemic, and would provide more information to assess the effectiveness of this strategy to limit the dynamics of infectious diseases with pandemic potential.
In terms of practice, this study showed that COVID-19 knowledge in the general population was low after the second peak of cases in Baja California, despite the vast amount of information available on the disease. While the use of facemasks and hand washing was frequent, adopting quarantine or home isolation as a risk reduction measure was less common, potentially due to lack of information or the need to conduct essential activities outside the home. Most importantly, $35\%$ of participants reported that they were “unlikely” or “very unlikely” to get the COVID-19 vaccine for a variety of reasons. Policymakers should be aware that over $50\%$ of the participants interviewed expressed concerns about the potential side effects and effectiveness of the vaccine, indicating more work to be done in this population for optimal vaccine uptake in the future.
In conclusion, this study showed that the seroprevalence of SARS-CoV-2 infection in Baja California in February 2021 was comparable to other states in Mexico and that being a border state with the U.S. did not seem to be associated with higher infection rates. The observed higher prevalence of SARS-CoV-2 infection by RT-PCR in Ensenada as compared to Mexicali and Tijuana emphasizes the dynamic nature of the epidemic within the same territory, and highlights the importance of repeated cross-sectional studies over time to capture these differences as the pandemic continues. Finally, health authorities face important challenges in terms of health literacy concerning prevention measures and future vaccination campaigns, as shown by the high level of vaccine hesitancy in this study. Additionally, decision-makers should address possible barriers regarding isolation/quarantine as part of a test and isolate strategy, as this remains an essential public health measure for control of communicable diseases. Additional incentives such as financial support during home isolation may be necessary to encourage compliance with isolation recommendations.
## References
1. Zhu N, Zhang D, Wang W, Li X, Yang B, Song J. **A Novel Coronavirus from Patients with Pneumonia in China, 2019**. *NEJM* (2020.0) **382** 727-33. DOI: 10.1056/NEJMoa2001017
2. 2World Health Organization (WHO). WHO Coronavirus disease (COVID-19) Pandemic 2020 [Internet]. 2020 [cited 2022 Jan 27]. https://covid19.who.int/
3. Munster VJ, Koopmans M, Van Doremalen N, Van Riel D, De Wit E. **A Novel Coronavirus Emerging in China—Key Questions for Impact Assessment**. *NEJM* (2020.0) **382** 692-4. PMID: 31978293
4. 4U.S. Department of Transportation. Border crossing Entry Data—Annual Data [Internet]. 2021 [cited 2021 Jun 5]. https://explore.dot.gov/views/BorderCrossingData/Annual
5. 5Committee on Binational Regional Opportunities (COBRO). California—Baja California 2021 Border Master Plan Joint Meeting [Internet]. CALTRANS. [cited 2021 Jun 5]. https://www.sandag.org/uploads/meetingid/meetingid_5660_28889.pdf
6. 6U.S. Department of Homeland Security. Joint Statement of US-Mexico Joint Initiative to Combat the COVID-19 Pandemic. [Internet, cited 2021 Jun 6] https://www.dhs.gov/news/2020/03/20/joint-statement-us-mexico-joint-initiative-combat-covid-19-pandemic
7. 7Dirección General de Epidemiología (DGE). Lineamiento estandarizado para la vigilancia epidemiológica y por laboratorio de la Enfermedad respiratoria Viral [Internet]. Mexico City. [cited 2021 Jun 6]. https://coronavirus.gob.mx/wpcontent/uploads/2020/04/Lineamiento_de_vigilancia_epidemiologica_de_enfermedad_respiratoria-_viral.pdf
8. 8Dirección General de Epidemiología (DGE). Sistema de Vigilancia Epidemiológica de Enfermedad Respiratoria Viral (SISVER) [Internet]. Mexico City [cited 2021 Jun 1]. https://sisver.sinave.gob.mx/influenza/
9. 9Instituto Nacional de Salud Pública (INSP). Tablero analítico de seguimiento de casos documentados de SARS-CoV-2 en México [cited 2022 Jan 27]. https://www.insp.mx/nuevo-coronavirus-2019.html
10. 10Gobierno de México. Exceso de mortalidad en México [Internet, cited 2022 Jan 27]. https://coronavirus.gob.mx/exceso-de-mortalidad-en-mexico/
11. 11Secretaría de Salud. Política Nacional de Vacunación Contra el Virus SARS-CoV-2 para la prevención de la COVID-19 en México. [Internet, cited 2022 May 24]. https://coronavirus.gob.mx/wp-content/uploads/2021/04/28Abr2021_13h00_PNVx_COVID_19.pdf
12. 12Secretaría de Salud de Baja California. Información oficial del nuevo Coronavirus (COVID-19), Actualización 20/02/2021. [Internet cited 2022 May 24]. http://www.bajacalifornia.gob.mx/coronavirus?id=1
13. 13Zazueta, O, Garfein R, Cano-Torres J, Méndez-Lizárraga C, Rodwell T, Muñiz-Salazar R, et al. (2022). COVID-19 survey in Baja California, Mexico, in February of 2021, Dryad, Dataset.
14. Taubner B, Peredo-Wende R, Ramani A, Singh G, Strle K, Cady NC. **Rapid and Quantitative Detection of Human Antibodies Against the 2019 Novel Coronavirus SARS CoV2 and its Variants as a Result of Vaccination and Infection**. *Microbiol Spectr* (2021.0) **9** e0089021. DOI: 10.1128/Spectrum.00890-21
15. 15Instituto Nacional de Estadística y Geografía. Censo de Población y Vivienda 2020. [Internet cited 2022 May 25]. https://www.inegi.org.mx/app/cpv/2020/resultadosrapidos/default.html
16. 16Shamah-Levy T, Romero-Martínez M, Barrientos-Gutiérrez T, Cuevas-Nasu L, Bautista-Arredondo S, Colchero MA, et al. Encuesta Nacional de Salud y Nutrición 2020 sobre Covid-19. Resultados nacionales. Cuernavaca, México: Instituto Nacional de Salud Pública, 2021. https://ensanut.insp.mx/encuestas/ensanutcontinua2020/doctos/informes/ensanutCovid19ResultadosNacionales.pdf
17. Muñoz-Medina JE, Grajales-Muñiz C, Salas-Lais AG, Fernandes-Matano L, López-Macías C, Monroy-Muñoz IE. **SARS-CoV-2 IgG Antibodies Seroprevalence and Sera Neutralizing Activity in MEXICO: A National Cross-Sectional Study during 2020**. *Microorganisms* (2021.0) **9** 850. DOI: 10.3390/microorganisms9040850
18. Acurio-Páez D, Vega B, Orellana D, Charry R, Gómez A, Obimpeh M. **Seroprevalence of SARS-CoV-2 Infection and Adherence to Preventive Measures in Cuenca, Ecuador, October 2020, a Cross-Sectional Study**. *Int J Environ Res Public Health* (2021.0) **18** 4657. DOI: 10.3390/ijerph18094657
19. Hallal PC, Hartwig FP, Horta BL, Silveira MF, Struchiner CJ, Vidaletti LP. **SARS-CoV-2 antibody prevalence in Brazil: results from two successive nationwide serological household surveys**. *Lancet Glob Health* (2020.0) **8** e1390-e1398. DOI: 10.1016/S2214-109X(20)30387-9
20. 20Universidad del Desarrollo. Tercer informe de estudio UDD sobre prevalencia de COVID-19 en la Región Metropolitana [Internet]. 2020 Dec 2 [Cited 2021 June 17]. https://www.udd.cl/noticias/2020/12/02/tercer-informe-de-estudio-udd-sobre-prevalencia-de-covid-19-en-la-region-metropolitana/
21. 21Dirección General de Estadística y Censos. Encuesta de Seroprevalencia de COVID-19 Ciudad de Buenos Aires [Internet]. 2020 Nov [Cited 2021 June 18]. https://www.estadisticaciudad.gob.ar/eyc/wp-content/uploads/2020/11/ir_2020_1501.pdf
22. 22Instituto Nacional de Salud y Grupo Colaborativo Estudio País. Seroprevalencia de SARS-CoV-2 Durante la epidemia en Colombia: Estudio país [internet]. 2020 Nov 26 [Cited 2021 June 18]. https://www.ins.gov.co/BibliotecaDigital/seroprevalencia-primer-reporte.pdf
23. Michos A, Tatsi EB, Filippatos F, Dellis C, Koukou D, Efthymiou V. **Association of total and neutralizing SARS-CoV-2 spike -receptor binding domain antibodies with epidemiological and clinical characteristics after immunization with the 1 st and 2 nd doses of the BNT162b2 vaccine**. *Vaccine* (2021.0) **39** 5963-5967. DOI: 10.1016/j.vaccine.2021.07.067
24. Watanabe M, Balena A, Tuccinardi D, Tozzi R, Risi R, Masi D. **Central obesity, smoking habit, and hypertension are associated with lower antibody titres in response to COVID-19 mRNA vaccine**. *Diabetes Metab Res Rev* (2022.0) **38** e3465. DOI: 10.1002/dmrr.3465
25. Carrat F, De Lamballerie X, Rahib D, Blanché H, Lapidus N, Artaud F. **Antibody status and cumulative incidence of SARS-CoV-2 infection among adults in three regions of France following the first lockdown and associated risk factors: a multicohort study**. *Int J Epidemiol* (2021.0) **50** 1458-1472. DOI: 10.1093/ije/dyab110
26. Gulsvik A, Fagerhoi MK. **Smoking and immunoglobulin levels**. *Lancet* (1979.0) **1** 449. DOI: 10.1016/s0140-6736(79)90934-6
27. 27U.S. Customs and Border Protection. Temporary Travel Restrictions to Land Borders and Ferry Service between the United States, Canada, and Mexico [Internet, cited 2022 January 27]. https://help.cbp.gov/s/article/Article-1694?language=en_US
|
---
title: Accuracy comparison of ARIMA and XGBoost forecasting models in predicting the
incidence of COVID-19 in Bangladesh
authors:
- Md. Siddikur Rahman
- Arman Hossain Chowdhury
- Miftahuzzannat Amrin
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021465
doi: 10.1371/journal.pgph.0000495
license: CC BY 4.0
---
# Accuracy comparison of ARIMA and XGBoost forecasting models in predicting the incidence of COVID-19 in Bangladesh
## Abstract
Accurate predictive time series modelling is important in public health planning and response during the emergence of a novel pandemic. Therefore, the aims of the study are three-fold: (a) to model the overall trend of COVID-19 confirmed cases and deaths in Bangladesh; (b) to generate a short-term forecast of 8 weeks of COVID-19 cases and deaths; (c) to compare the predictive accuracy of the Autoregressive Integrated Moving Average (ARIMA) and eXtreme Gradient Boosting (XGBoost) for precise modelling of non-linear features and seasonal trends of the time series. The data were collected from the onset of the epidemic in Bangladesh from the Directorate General of Health Service (DGHS) and Institute of Epidemiology, Disease Control and Research (IEDCR). The daily confirmed cases and deaths of COVID-19 of 633 days in Bangladesh were divided into several training and test sets. The ARIMA and XGBoost models were established using those training data, and the test sets were used to evaluate each model’s ability to forecast and finally averaged all the predictive performances to choose the best model. The predictive accuracy of the models was assessed using the mean absolute error (MAE), mean percentage error (MPE), root mean square error (RMSE) and mean absolute percentage error (MAPE). The findings reveal the existence of a nonlinear trend and weekly seasonality in the dataset. The average error measures of the ARIMA model for both COVID-19 confirmed cases and deaths were lower than XGBoost model. Hence, in our study, the ARIMA model performed better than the XGBoost model in predicting COVID-19 confirmed cases and deaths in Bangladesh. The suggested prediction model might play a critical role in estimating the spread of a novel pandemic in Bangladesh and similar countries.
## Introduction
The coronavirus disease 2019 (COVID-19) is a major global public health threat. A group of pneumonia infections caused by a newly found β-coronavirus occurred in Wuhan, China in December 2019 [1]. On January 12, 2020, the World Health Organization (WHO) labelled this coronavirus the 2019-novel coronavirus (2019-nCoV) [2, 3]. More than 222 nations, including Bangladesh, have reported more than 263.1 million confirmed COVID-19 cases as of November 30, 2021, resulting in 5.2 million fatalities worldwide [4]. On March 8, 2020, IEDCR detected the first COVID-19 case in Bangladesh. On March 9, 2020, the number of infected cases began to rise, and as of December 31, 2021, Bangladesh had 1.6 million infected cases and 28,072 fatalities [5].
In South Asia, especially Bangladesh, COVID-19 has portrayed a significant gap in public health preparedness and response to contagious disease risks and outbreaks [6]. The lack of a dependable public health surveillance system is noticeable [7]. Of the 222 countries, Bangladesh globally ranks 4th on the daily increase of COVID-19 deaths [8] and 3rd in fatality rate in South Asia [9]. Bangladesh is a densely populated country, with almost 161.4 million people living in overcrowded cities and villages, with a population density of over 1115 persons per square kilometre [10]. The healthcare system of *Bangladesh is* falling short of international standards due to a scarcity of competent workers and inadequate healthcare services, despite the Bangladesh government’s efforts to address these challenges in the health service [11]. Furthermore, there are insufficient Intensive Care Unit (ICU) beds for the population. The government faces an uphill battle to control the COVID-19 spread. The Impact of COVID-19 in Bangladesh on education is also noticeable. Due to the lengthy university shutdown and home confinement caused by COVID-19, students’ learning was severely disrupted [12]. Students had a higher psychological effect due to COVID-19 [13]. The spread of COVID-19 poses a tremendous challenge for any administration in terms of public health system capacity and management in the event of a catastrophic emergency [14]. As a result, knowing the exact prediction and usual pattern of this virus is crucial for Bangladesh. The prediction model can assist hospitals, healthcare administration and related stakeholders in public health planning and response during the emergence of the COVID-19 pandemic.
The autoregressive integrated moving average (ARIMA) model is commonly used in the modelling of contagious diseases [15], such as influenza viruses [16], malaria [17], and hemorrhagic fever [18]. Several studies regarding COVID-19 forecasting used ARIMA model for predicting the confirmed cases and examined it as the best model [19–22]. On the other hand, the eXtreme Gradient Boosting, a new approach, is an uptrend machine learning technique in time series modelling [23, 24]. The XGBoost model has performed admirably in many medical research sectors [25–28], but the application of XGBoost model in predicting COVID-19 incidence is scanty [29–32]. Time series forecasting methods play a critical role in estimating the spread of an epidemic. Therefore, this study aimed to (a) model the overall trend of COVID-19 confirmed cases and deaths in Bangladesh; (b) generate a short-term forecast of 8 weeks of confirmed COVID-19 cases and deaths; (c) compare the predictive accuracy of the Autoregressive Integrated Moving Average (ARIMA) and eXtreme Gradient Boosting (XGBoost) for precise modelling of non-linear features and seasonal trends of the time series (Fig 1). The findings of this study will help policymakers and government officials with effective public health interventions to control the spread of an epidemic.
**Fig 1:** *Proposed methodology.*
## Data source
Daily confirmed cases and deaths of COVID-19 in Bangladesh from March 08, 2020, to November 30, 2021 were collected from the Directorate General of Health Service (DGHS) and Institute of Epidemiology, Disease Control and Research (IEDCR) [33, 34]. The daily confirmed cases and deaths of COVID-19 of 633 days in Bangladesh were divided into several training and test sets. The ARIMA and XGBoost models were established using those training data, and the test sets were used to evaluate each model’s ability to forecast and finally averaged all the predictive performances to choose the best model.
## ARIMA model
The ARIMA model is frequently used for time series modelling of contagious diseases [35]. It is one of the most often used time-series models in a variety of sectors of data analysis because it accounts for changing trends, periodic variations, and random disturbances in the data. It’s utilized for forecasting and better interpreting the data [36]. ARIMA(p, d, q) is a combination of the Autoregressive (AR) and Moving Average (MA) models, with the ’I’ standing for integration; where p denotes the autoregressive order, d for differencing order, and q for moving average order [37]. Stationary is a discardable property for a time series analysis. The difference order d is used to make a nonstationary time series to stationary. It is estimated by the Augmented Dickey-Fuller (ADF) test. An ARMA (p, q) model combines AR(p) and MA(q) models, which is best suited to univariate time series analysis. The AR(p) model assumes that a variable’s future value is determined by a linear combination of p previous observations plus a random error term. The AR(p) model is represented mathematically as follows: Yt=C+∅1Yt−1+∅2Yt−2+∅3Yt−3+∅4Yt−4…..∅pYt−p+εt [1] Yt and εt denote the actual value and error terms at time t, ∅i ($i = 1$,2,3,4….) denotes model parameters, and c denotes a constant. The order of the model is a positive integer p. Unlike the AR(p) model, the MA(q) model includes a dependent variable for previous errors. Following is the MA(q) model: Yt=μ+θ1εt−1+θ2εt−2+θ3εt−3+θ4εt−4+⋯+θqεt−q+εt [2] Here, μ denotes the series’ mean, θj ($j = 1$, 2, 3… q) denotes model parameters, and q is the model’s order. A mathematical representation of an ARMA (p, q) model is as follows: Yt=C+μ+∅1Yt−1+∅2Yt−2+∅3Yt−3+∅4Yt−4…..+∅pYt−p+θ1εt−1+θ2εt−2+θ3εt−3+θ4εt−4+⋯+θqεt−q+εt [3]
## Seasonal ARIMA model
A seasonal ARIMA model collects information from seasonal components that the conventional ARIMA model cannot comprehend. The seasonal model may be split into two types based on its complexity: an additive model (simple seasonal model) and a product seasonal model. The mathematical expression of the simple seasonal model’s is: Xt=St+Tt+It [4] Where, St, Tt, and It denote seasonal information, trend information, and random fluctuation information in the data, respectively. To build a seasonal ARIMA model, the components of the nonseasonal part is identified first. After that, the seasonal part is identified. For the seasonal information, the time series data were plotted to see the seasonality pattern. Then the Box-Cox transformation was performed to reduce the variance of the original COVID-19 time series. At the same time, the long term trend and seasonal variations were fixed by performing first-order differencing and seasonal differencing. An Augmented Dickey-Fuller (ADF) test can be used to determine if the time series is stable. The potential values of the autoregressive order p, moving average order q, seasonal autoregressive order P, and the seasonal moving average order Q may be calculated using the graphs of the autocorrelation function (ACF) and partial ACF (PACF) determined by the Box- Jenkins order determination method [38]. The corrected Akaike information criterion (AICc) value was used to evaluate the benefits and drawbacks of the model fit, and the model with its least AICc value was deemed the best. The Ljung-Box test is thus used to determine the white noise of the residuals [18, 38].
## XGBoost model
Extreme Gradient boosting (XGBoost) technique is an optimized distributed Gradient boosting library that can rapidly assess the importance of all input features and is a scalable machine learning system for tree boosting. It has proven to be a qualified and competent problem solver for machine learning [39, 40]. Gradient boosting is a popular method for building a forecasting model and a quantifiable boosting algorithm [38]. It was initially developed by Chen Tianqi and Carlos Gestrin in 2011 and has since been improved and polished by numerous scientists in the follow-up study [41]. The core concept of boosting (enhancing machine learning models) is to merge hundreds of low-accuracy prediction models into a single high-accuracy model. Several models must frequently be integrated to obtain good prediction accuracy under tolerable parameter values. The model may need to be iterated or repeated multiple times or more to attain sufficient accuracy if the data collection is vast or complicated; the XGBoost model could better handle this problem [18]. XGBoost is a robust and effective gradient boosting machine algorithm [42, 43]. The objective function can be written as follows: Obj(t)=∑$i = 1$nl(yi,y^i(t−1)+ft(xi))+Ω(ft)+constant [5] Where yi is the observed values, y^i(t−1) is the predicted value of the last iteration, xi is the feature vector, n is the sample size, ft is a new function which model learns, Ω(ft) is the regularization term which saves the model from complexity. l denotes the loss function, which calculates the difference between the label and the prediction in the previous phase, the new tree’s output [38, 44].
## Evaluation parameter of models
A model’s real accuracy can be measured by comparing predicted and actual values. A variety of performance metrics can be performed to calculate accuracy [45]. We used four prominent forecasting parameters to assess the predictive efficacy of our model: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE), Mean Percentage Error (MPE), as follows: MAE=1n∑$i = 1$n|yi^−yi| [6] RMSE=1n∑$i = 1$n(yi^−yi)2 [7] MAPE=1n∑$i = 1$n|yi^−yiyi|×$100\%$ [8] MPE=1n∑$i = 1$n(y^i−yiyi)×$100\%$ [9] Where n denotes the number of observations, yi^−yi denotes the error between the forecasted and actual value. The mean of the actual forecasting error is calculated by taking the arithmetic average of the absolute errors between the prediction and the actual value. The root mean square error (RMSE) is a commonly used metric for comparing the values forecasted by a model or estimate to the values observed, and it’s the average squared error squared. The MAPE measure calculates accuracy as a percentage, computed as the actual values minus the forecasted values divided by the actual values for each time period [46].
## Data analysis
Statistical analyses were performed using RStudio (Version 4.1.0) [47]. The ’tseries’, ’TSstudio’ and stats packages were used to process the time series. ARIMA models were built with the ’forecast’ package using auto.arima function for choosing the best model based on the AICc values [48]. The ’forecastxgb’ package was used for fitting XGBoost model. The necessary codes are available at https://github.com/ [49].
## Results
In Bangladesh, 1.6 million cases and 27,983 deaths of COVID-19 of 633 days (91 weeks) were recorded from March 08, 2020 to November 30, 2021. The highest COVID-19 confirmed cases were recorded 16,230 and deaths 264 in Bangladesh (Table 1). The data vary considerably and show weekly seasonality and nonlinearity pattern in both cases and deaths. Although the number of confirmed cases and deaths fluctuated in different weeks, there was a highly upward trend between 70 and 80 weeks. After that, it began to alleviate (Fig 2). The ADF test confirms that the data are not smooth. The entire data set (COVID-19 confirmed cases and deaths) was split into several training and test sets (S1 Text).
**Fig 2:** *A 633 day (91 weeks) time series plot for confirmed COVID-19 cases and deaths in Bangladesh from March 08, 2020 to November 30, 2021.* TABLE_PLACEHOLDER:Table 1 To decrease anomalies such as non-normality and heteroscedasticity that the variances are not constant, Box & Cox [1964] presented a parametric power transformation technique [50]. The Box-Cox transformation was applied to each training data set to remove the non-normality and exhibit less variation [51]. The decomposed data shows a weekly seasonal pattern in both cases and deaths [52]. Table 2 illustrates the predictive performance of different ARIMA models built from seven different training sets and their average values for COVID-19 confirmed cases.
**Table 2**
| ARIMA model | Train | Train.1 | Train.2 | Train.3 | Test | Test.1 | Test.2 | Test.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| ARIMA model | RMSE | MAE | MPE | MAPE | RMSE | MAE | MPE | MAPE |
| Sample 1 | 258.50 | 166.16 | -2.53 | 15.00 | 421.92 | 359.07 | -2.63 | 22.46 |
| Sample 2 | 241.82 | 154.73 | -2.72 | 13.39 | 244.33 | 216.94 | -47.26 | 48.85 |
| Sample 3 | 224.11 | 139.44 | -2.67 | 12.80 | 3844.89 | 2988.57 | 75.95 | 75.95 |
| Sample 4 | 280.81 | 165.97 | -1.72 | 12.50 | 2160.95 | 2031.37 | -163.43 | 163.44 |
| Sample 5 | 275.28 | 173.14 | -1.96 | 12.82 | 6325.54 | 5481.82 | 49.68 | 50.74 |
| Sample 6 | 560.19 | 276.18 | -2.15 | 13.11 | 8984.24 | 8217.15 | -549.35 | 549.35 |
| Sample 7 | 558.91 | 280.01 | -2.47 | 12.82 | 67.31 | 55.24 | 1.02 | 21.36 |
| Average error measures | 342.80 | 193.66 | -2.32 | 13.21 | 3149.88 | 2764.31 | -90.86 | 133.16 |
The XGBoost model for COVID-19 confirmed cases were built by adjusting different parameters like seas_method= ‘dummies’, trend_method= ‘none’, power transformation parameter ‘lambda’ for each training set. Table 3 illustrates the predictive performances of different training and test sets of the XGBoost model and their average values for confirmed cases.
**Table 3**
| XGBoost model | Train | Train.1 | Train.2 | Train.3 | Test | Test.1 | Test.2 | Test.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| XGBoost model | RMSE | MAE | MPE | MAPE | RMSE | MAE | MPE | MAPE |
| Sample 1 | 47.19 | 31.66 | -0.11 | 2.30 | 520.76 | 436.20 | 11.44 | 23.97 |
| Sample 2 | 31.39 | 22.05 | -0.16 | 1.72 | 925.60 | 865.12 | -190.79 | 190.81 |
| Sample 3 | 64.91 | 42.74 | 0.36 | 3.10 | 3727.73 | 2874.31 | 71.05 | 71.12 |
| Sample 4 | 76.87 | 51.28 | -0.11 | 3.64 | 1989.87 | 1814.20 | -156.36 | 156.56 |
| Sample 5 | 53.71 | 35.71 | -0.24 | 2.53 | 7374.24 | 6413.23 | 59.46 | 59.46 |
| Sample 6 | 130.08 | 80.66 | -0.17 | 4.06 | 9683.20 | 9183.94 | -561.51 | 561.55 |
| Sample 7 | 168.82 | 105.89 | -0.14 | 4.64 | 196.60 | 185.76 | -76.20 | 76.20 |
| Average error measures | 81.85 | 52.86 | -0.08 | 3.14 | 3488.29 | 3110.39 | -120.42 | 162.81 |
For COVID-19 death data, we built five different ARIMA models for five different training and test sets. The appropriate model for each training data set was selected based on the AICc value. The predictive performances of ARIMA models of five different training and test data sets and their average values were shown in Table 4.
**Table 4**
| ARIMA model | Train | Train.1 | Train.2 | Train.3 | Test | Test.1 | Test.2 | Test.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| ARIMA model | RMSE | MAE | MPE | MAPE | RMSE | MAE | MPE | MAPE |
| Sample 1 | 5.84 | 4.35 | -6.69 | 23.46 | 46.32 | 33.48 | 49.72 | 69.15 |
| Sample 2 | 6.14 | 4.59 | -4.71 | 23.09 | 110.31 | 101.42 | -287.79 | 286.79 |
| Sample 3 | 7.23 | 5.37 | -4.64 | 24.34 | 100.32 | 90.70 | 47.81 | 47.81 |
| Sample 4 | 9.98 | 6.67 | -4.35 | 22.51 | 246.09 | 228.36 | -620.60 | 620.60 |
| Sample 5 | 9.64 | 6.60 | -4.98 | 21.22 | 6.00 | 5.40 | -152.45 | 154.40 |
| Average error measures | 7.77 | 5.52 | -5.07 | 22.92 | 101.81 | 91.87 | -192.66 | 235.75 |
We also built the XGBoost model for COVID-19 deaths by adjusting the parameters seas_method= ‘dummies’, trend_method= ‘none’, power transformation parameter ‘lambda’ for each training set. The predictive measures of different training and test set for XGBoost model and their average values for COVID-19 deaths were shown in Table 5.
**Table 5**
| XGBoost model | Train | Train.1 | Train.2 | Train.3 | Test | Test.1 | Test.2 | Test.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| XGBoost model | RMSE | MAE | MPE | MAPE | RMSE | MAE | MPE | MAPE |
| Sample 1 | 2.19 | 1.45 | 0.32 | 6.34 | 40.32 | 28.11 | 20.68 | 63.18 |
| Sample 2 | 1.95 | 1.39 | -1.03 | 6.65 | 49.17 | 45.01 | -131.30 | 132.18 |
| Sample 3 | 3.80 | 2.66 | 1.69 | 10.05 | 150.98 | 136.59 | 71.82 | 72.82 |
| Sample 4 | 2.70 | 1.92 | -1.49 | 7.37 | 179.88 | 169.85 | -444.68 | 445.68 |
| Sample 5 | 3.20 | 2.27 | -1.18 | 7.50 | 16.27 | 15.47 | -470.79 | 471.27 |
| Average error measures | 2.77 | 1.94 | -0.34 | 7.58 | 87.32 | 79.01 | -190.85 | 237.03 |
The average MAPE values of the ARIMA model for COVID-19 confirmed cases is comparatively lower than the XGBoost model indicating that ARIMA performs better than XGBoost in predicting COVID-19 confirmed cases in Bangladesh. On the other hand, the average MAPE value of the ARIMA model for COVID-19 deaths is smaller than XGBoost which also indicates that ARIMA performs better than XGBoost in predicting COVID-19 deaths in Bangladesh.
In our study, it was found that ARIMA model performs better than XGBoost in predicting COVID-19 confirmed cases and deaths in Bangladesh. The detailed procedure of ARIMA and XGBoost model fitting for COVID-19 confirmed cases and deaths were shown in S1 Text.
## Discussion
In our study, we found a weekly seasonality for daily COVID-19 confirmed cases and deaths in Bangladesh. Because of the weekend, fewer health care staffs were available to report new cases or fewer people are tested, which causes weekly seasonality [11, 53]. It was simpler to assess the seasonality and pattern of this disease using seasonal decomposition, which offered a reference for us to analyze, process, and stabilize data, laying the groundwork for building a mathematical model. The ARIMA models for cases and deaths were created using a linear regression model to expose the data’s dynamic rules and forecast future data values. The ARIMA model combines the trend components, cyclical factors, and random errors originally included in the time series. This model combines the benefits of autoregressive and moving average models, is unconstrained by data sources, has high adaptability, and has good short-term predictions [18]. Instead of requiring particular influencing elements, the ARIMA model uses merely historical data to comprehend the illness pattern and achieve a more accurate forecast impact. As a result, the ARIMA approach is simple to learn and frequently employed [38]. In this study, the ARIMA method is compared to the XGBoost model for its fairly mature time series prediction approach and widespread application. The ARIMA model performs well on the nonstationary time series after applying Box-Cox transformation and differencing adjustments, demonstrating the model’s capacity to forecast diseases. *In* general, the greater the number of differences utilized, the more data is lost. We built different ARIMA models for different training sets and selected the best for each training set based on the AICc value for both confirmed cases and deaths [53, 54]. Finally, we averaged all the error measures from all models. The average MAPE value of the training data sets for confirmed cases was $13.21\%$, whereas it was $133.16\%$ for the test data sets. On the other hand, the average MAPE value of the training sets for death data was $22.92\%$, whereas the test sets was $235.95\%$. On the other hand, we used the most popular machine learning model to fit the nonlinear data [55]. The XGBoost model, a relatively new approach, is a gradient boosting-based ensemble machine learning technique that utilizes decision trees. The XGBoost technique offers several benefits in terms of model prediction, including the lack of data preprocessing, a quick operation speed, complete feature extraction, a strong fitting effect, and high prediction accuracy. This study applied this new technique to predict COVID-19 confirmed cases and deaths in Bangladesh. We selected the most often used ARIMA time series model as the baseline of this study. But the XGBoost model did not perform well on the nonlinear data. The XGBoost model has a considerably worse influence on forecasting than the ARIMA model in this COVID-19 research in Bangladesh because the number of confirmed cases and deaths has increased significantly between 70 and 80 weeks. The number of confirmed cases in the country has also altered dramatically due to changes in government policies. In addition, there might have other climatic and environmental factors that impact the COVID-19 incidence observed from some previous studies which didn’t incorporate in our study [46, 56–58]. As a result, the proposed model was no longer produced accurate predictions for this change. In this study, we compared the models’ predictive performances to provide a reference for the country’s policymakers to take effective steps and strategies to control the outbreak of the deadly disease. The study findings are useful to all other endemic countries similar to Bangladesh.
## Conclusion
For controlling the spread of the COVID-19 pandemic in Bangladesh and similar settings elsewhere, we developed a seasonal ARIMA model and XGBoost model. These models were used to create short-term forecasts in this study. The ARIMA model performed better than the XGBoost model in predicting COVID-19 confirmed cases and deaths in Bangladesh.
## Limitations
We compared the predictive performance of XGBoost and ARIMA models in this study, and the results help choose the best model for COVID-19 prediction in Bangladesh. There are many different prediction models, and we need to keep experimenting with them to find the best one for predicting confirmed COVID-19 cases and deaths. We focused on the impact of time on both cases and deaths in our research, which allows our model easier to build and forecast. Therefore, a limitation of our study is that, for example, meteorological data such as temperature, humidity, and wind speed variables were not incorporated but which are known to impact COVID-19. As mentioned above, this will be explored progressively with increasing data.
## References
1. Peeri NC, Shrestha N, Siddikur Rahman M, Zaki R, Tan Z, Bibi S. **The SARS, MERS and novel coronavirus (COVID-19) epidemics, the newest and biggest global health threats: what lessons have we learned?**. *Int J Epidemiol* (2020.0) **49** 717-726. DOI: 10.1093/ije/dyaa033
2. Lu R, Zhao X, Li J, Niu P, Yang B, Wu H. **Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding**. *Lancet* (2020.0) **395** 565-574. DOI: 10.1016/S0140-6736(20)30251-8
3. Ruan S.. **Likelihood of survival of coronavirus disease 2019**. *Lancet Infect Dis* (2020.0) **20** 630-631. DOI: 10.1016/S1473-3099(20)30257-7
4. 4Coronavirus Disease (COVID-19) Situation Reports. [cited 30 Nov 2021]. Available: https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports
5. 5COVID-19 Situation Updates | IEDCR. [cited 30 Nov 2021]. Available: https://iedcr.gov.bd/covid-19/covid-19-situation-updates
6. Rahman MS, Karamehic-Muratovic A, Amrin M, Chowdhury AH, Selim Mondol M, Haque U. *COVID-19 Epidemic in Bangladesh among Rural and Urban Residents: An Online Cross-Sectional Survey of Knowledge, Attitudes, and Practices* (2020.0). DOI: 10.3390/epidemiologia201
7. Bhutta ZA, Basnyat B, Saha S, Laxminarayan R. **Covid-19 risks and response in South Asia**. *BMJ* (2020.0) **368** 1-2. DOI: 10.1136/bmj.m1190
8. 8Fatalities hit yet another high in Bangladesh as 258 die of Covid in a day | Dhaka Tribune. [cited 28 Aug 2021]. Available: https://archive.dhakatribune.com/bangladesh/2021/07/27/fatalities-hit-yet-another-high-in-bangladesh-as-258-die-of-covid-in-a-day
9. 9Bangladesh Covid case fatality rate third in South Asia. [cited 28 Aug 2021]. Available: https://www.newagebd.net/article/144760/bangladesh-covid-case-fatality-rate-third-in-south-asia
10. Satu MS, Howlader KC, Mahmud M, Shamim Kaiser M, Islam SMS, Quinn JMW. **Short-term prediction of covid-19 cases using machine learning models**. *Appl Sci.* (2021.0) 11. DOI: 10.3390/app11094266
11. Ahmed SM, Hossain MA, RajaChowdhury AM, Bhuiya AU. **The health workforce crisis in Bangladesh: Shortage, inappropriate skill-mix and inequitable distribution**. *Hum Resour Health* (2011.0) **9** 1-7. DOI: 10.1186/1478-4491-9-1
12. Bari R, Sultana F. **Second Wave of COVID-19 in Bangladesh: An Integrated and Coordinated Set of Actions Is Crucial to Tackle Current Upsurge of Cases and Deaths**. *Front Public Heal.* (2021.0) **9** 1275. DOI: 10.3389/FPUBH.2021.699918/BIBTEX
13. Faisal RA, Jobe MC, Ahmed O, Sharker T. **Mental Health Status, Anxiety, and Depression Levels of Bangladeshi University Students During the COVID-19 Pandemic**. *Int J Ment Health Addict.* (2021.0). DOI: 10.1007/s11469-020-00458-y
14. Li J, Guo K, Viedma EH, Lee H, Liu J, Zhong N. **Culture versus Policy: More Global Collaboration to Effectively Combat COVID-19**. *Innovation(China)* (2020.0) **1** 100023. DOI: 10.1016/j.xinn.2020.100023
15. Zhang X, Hou F, Qiao Z, Li X, Zhou L, Liu Y. **Temporal and long-term trend analysis of class C notifiable diseases in China from 2009 to 2014**. *BMJ Open* (2016.0) **6** 11038. DOI: 10.1136/bmjopen-2016-011038
16. He Z, Tao H. **Epidemiology and ARIMA model of positive-rate of influenza viruses among children in Wuhan, China: A nine-year retrospective study**. *Int J Infect Dis* (2018.0) **74** 61-70. DOI: 10.1016/j.ijid.2018.07.003
17. Anwar MY, Lewnard JA, Parikh S, Pitzer VE. **Time series analysis of malaria in Afghanistan: using ARIMA models to predict future trends in incidence**. *Malar J.* (2016.0) **15** 1-10. DOI: 10.1186/s12936-015-1044-1
18. Wang T, Liu J, Zhou Y, Cui F, Huang Z, Wang L. **Prevalence of hemorrhagic fever with renal syndrome in Yiyuan County, China, 2005-2014**. *BMC Infect Dis* (2016.0) **16** 69. DOI: 10.1186/s12879-016-1404-7
19. Alzahrani SI, Aljamaan IA, Al-Fakih EA. **Forecasting the spread of the COVID-19 pandemic in Saudi Arabia using ARIMA prediction model under current public health interventions**. *J Infect Public Health* (2020.0) **13** 914-919. DOI: 10.1016/j.jiph.2020.06.001
20. Khan FM, Gupta R. **ARIMA and NAR based prediction model for time series analysis of COVID-19 cases in India**. *J Saf Sci Resil.* (2020.0) **1** 12-18. DOI: 10.1016/j.jnlssr.2020.06.007
21. Singh S, Parmar KS, Makkhan SJS, Kaur J, Peshoria S, Kumar J. **Study of ARIMA and least square support vector machine (LS-SVM) models for the prediction of SARS-CoV-2 confirmed cases in the most affected countries**. *Chaos, Solitons & Fractals* (2020.0) **139** 110086. DOI: 10.1016/j.chaos.2020.110086
22. Hernandez-Matamoros A, Fujita H, Hayashi T, Perez-Meana H. **Forecasting of COVID19 per regions using ARIMA models and polynomial functions**. *Appl Soft Comput.* (2020.0) **96** 106610. DOI: 10.1016/j.asoc.2020.106610
23. Rahman MS, Pientong C, Zafar S, Ekalaksananan T, Paul RE, Haque U. **Mapping the spatial distribution of the dengue vector Aedes aegypti and predicting its abundance in northeastern Thailand using machine-learning approach**. *One Heal.* (2021.0) **13** 100358. DOI: 10.1016/j.onehlt.2021.100358
24. Li Z, Wang Z, Song H, Liu Q, He B, Shi P. **Application of a hybrid model in predicting the incidence of tuberculosis in a Chinese population**. *Infect Drug Resist* (2019.0) **12** 1011-1020. DOI: 10.2147/IDR.S190418
25. Dinh A, Miertschin S, Young A, Mohanty SD. **A data-driven approach to predicting diabetes and cardiovascular disease with machine learning**. *BMC Med Informatics Decis Mak 2019 191* (2019.0) **19** 1-15. DOI: 10.1186/s12911-019-0918-5
26. Liu L, Yu Y, Fei Z, Li M, Wu F-X, Li H-D. **An interpretable boosting model to predict side effects of analgesics for osteoarthritis**. *BMC Syst Biol 2018 126* (2018.0) **12** 29-38. DOI: 10.1186/s12918-018-0624-4
27. Liu Z, Zhou T, Han X, Lang T, Liu S, Zhang P. **Mathematical models of amino acid panel for assisting diagnosis of children acute leukemia**. *J Transl Med 2019 171* (2019.0) **17** 1-11. DOI: 10.1186/s12967-019-1783-9
28. Zou LS, Erdos MR, Taylor DL, Chines PS, Varshney A, Parker SCJ. **BoostMe accurately predicts DNA methylation values in whole-genome bisulfite sequencing of multiple human tissues**. *BMC Genomics 2018 191* (2018.0) **19** 1-15. DOI: 10.1186/s12864-018-4766-y
29. Yan L, Zhang H-T, Goncalves J, Xiao Y, Wang M, Guo Y. **An interpretable mortality prediction model for COVID-19 patients**. *Nat Mach Intell* (2020.0) **2** 283-288. DOI: 10.1038/s42256-020-0180-7
30. Li WT, Ma J, Shende N, Castaneda G, Chakladar J, Tsai J. **Using Machine Learning of Clinical Data to Diagnose COVID-19**. *BMC Med Informatics Decis Mak* (2020.0) **20** 247. DOI: 10.1101/2020.06.24.20138859
31. Chowdhury MEH, Rahman T, Khandakar A, Al-Madeed S, Zughaier SM, Doi SAR. **An Early Warning Tool for Predicting Mortality Risk of COVID-19 Patients Using Machine Learning**. *Cognit Comput.* (2021.0). DOI: 10.1007/s12559-020-09812-7
32. Romeo L, Frontoni E. **A Unified Hierarchical XGBoost Model for Classifying Priorities for COVID-19 Vaccination Campaign**. *Pattern Recognit.* (2021.0) **121** 108197. DOI: 10.1016/j.patcog.2021.108197
33. 33COVID-19. [cited 30 Nov 2021]. Available: http://dashboard.dghs.gov.bd/webportal/pages/covid19.php
34. 34IEDCR. [cited 30 Nov 2021]. Available: http://old.iedcr.gov.bd/
35. Sahai AK, Rath N, Sood V, Singh MP. **ARIMA modelling & forecasting of COVID-19 in top five affected countries**. *Diabetes Metab Syndr Clin Res Rev* (2020.0) **14** 1419-1427. DOI: 10.1016/J.DSX.2020.07.042
36. Brockwell PJ, Davis RA. *Introduction to Time Series and Forecasting* (2016.0). DOI: 10.1007/978-3-319-29854-2
37. Kumar N, Susan S. **COVID-19 Pandemic Prediction using Time Series Forecasting Models**. *2020 11th Int Conf Comput Commun Netw Technol ICCCNT 2020* (2020.0). DOI: 10.1109/ICCCNT49239.2020.9225319
38. Lv CX, An SY, Qiao BJ, Wu W. **Time series analysis of hemorrhagic fever with renal syndrome in mainland China by using an XGBoost forecasting model.**. *BMC Infect Dis* (2021.0) **21** 1-13. DOI: 10.1186/s12879-020-05706-z
39. Zheng Y, Zhu Y, Ji M, Wang R, Liu X, Zhang M. **A Learning-Based Model to Evaluate Hospitalization Priority in COVID-19 Pandemics**. *Patterns* (2020.0) **1** 100092. DOI: 10.1016/j.patter.2020.100092
40. Hu CA, Chen CM, Fang YC, Liang SJ, Wang HC, Fang WF. **Using a machine learning approach to predict mortality in critically ill influenza patients: a cross-sectional retrospective multicentre study in Taiwan**. *BMJ Open* (2020.0) **10** e033898. DOI: 10.1136/bmjopen-2019-033898
41. Li W, Yin Y, Quan X, Zhang H. **Gene Expression Value Prediction Based on XGBoost Algorithm**. *Front Genet* (2019.0) **10** 1-7. DOI: 10.3389/fgene.2019.00001
42. Shrivastav LK, Jha SK. **A gradient boosting machine learning approach in modeling the impact of temperature and humidity on the transmission rate of COVID-19 in India**. *Appl Intell* (2021.0) **51** 2727-2739. DOI: 10.1007/s10489-020-01997-6
43. Babajide Mustapha I, Saeed F. **Bioactive Molecule Prediction Using Extreme Gradient Boosting**. *Molecules* (2016.0) **21** 1-11. DOI: 10.3390/molecules21080983
44. Nishio M, Nishizawa M, Sugiyama O, Kojima R, Yakami M, Kuroda T. **Computer-aided diagnosis of lung nodule using gradient tree boosting and Bayesian optimization**. *PLoS One* (2018.0) **13** 1-13. DOI: 10.1371/journal.pone.0195875
45. Prajapati S, Swaraj A, Lalwani R, Narwal A, Verma K, Singh G. *Comparison of Traditional and Hybrid Time Series Models for Forecasting COVID-19 Cases* (2019.0) 8
46. Luo J, Zhang Z, Fu Y, Rao F. **Time series prediction of COVID-19 transmission in America using LSTM and XGBoost algorithms**. *Results Phys* (2021.0) **27** 104462. DOI: 10.1016/j.rinp.2021.104462
47. 47RStudio: Integrated Development Environment for R RStudio Team. In: RStudio, PBC, Boston, MA (2021).. *Integrated Development Environment for R RStudio Team* (2021.0)
48. Hyndman RJ, Khandakar Y. **Automatic Time Series Forecasting: The forecast Package for R**. *J Stat Softw.* (2008.0) **27** 1-22. DOI: 10.18637/JSS.V027.I03
49. 49Arman-Hossain-Chowdhury/Time-series. [cited 29 Dec 2021]. Available: https://github.com/Arman-Hossain-Chowdhury/Time-series
50. Sakia RM. **The Box-Cox Transformation Technique: A Review**. *Stat* (1992.0) **41** 169. DOI: 10.2307/2348250
51. Curran-Everett D.. **Explorations in statistics: The log transformation**. *Adv Physiol Educ* (2018.0) **42** 343-347. DOI: 10.1152/advan.00018.2018
52. Rosselló J, Sansó A. **Yearly, monthly and weekly seasonality of tourism demand: A decomposition analysis**. *Tour Manag* (2017.0) **60** 379-389. DOI: 10.1016/j.tourman.2016.12.019
53. Dehning J, Zierenberg J, Spitzner FP, Wibral M, Neto JP, Wilczek M. **Inferring change points in the COVID-19 spreading reveals the effectiveness of interventions**. *Science (80-)* (2020.0) 369. DOI: 10.1126/science.abb9789
54. Zeng Q, Li D, Huang G, Xia J, Wang X, Zhang Y. **Time series analysis of temporal trends in the pertussis incidence in Mainland China from 2005 to 2016**. *Sci Rep* (2016.0) **6** 1-8. DOI: 10.1038/s41598-016-0001-8
55. Wu W, Guo J, An S, Guan P, Ren Y, Xia L. **Comparison of two hybrid models for forecasting the incidence of hemorrhagic fever with renal syndrome in Jiangsu Province, China**. *PLoS One* (2015.0) **10** 1-13. DOI: 10.1371/journal.pone.0135492
56. Pal SK, Masum MH. **Effects of meteorological parameters on COVID-19 transmission trends in Bangladesh**. *Environ Sustain 2021 43* (2021.0) **4** 559-568. DOI: 10.1007/S42398-021-00195-5
57. Menebo MM. **Temperature and precipitation associate with Covid-19 new daily cases: A correlation study between weather and Covid-19 pandemic in Oslo, Norway**. *Sci Total Environ* (2020.0) **737** 139659. DOI: 10.1016/j.scitotenv.2020.139659
58. Hossain MS, Ahmed S, Uddin MJ. **Impact of weather on COVID-19 transmission in south Asian countries: An application of the ARIMAX model**. *Sci Total Environ* (2021.0) **761** 143315. DOI: 10.1016/j.scitotenv.2020.143315
|
---
title: Risk factors for melioidosis in Udupi District, Karnataka, India, January 2017-July
2018
authors:
- Akhileshwar Singh
- Ashok Talyan
- Ramesh Chandra
- Anubhav Srivastav
- Vasudeva Upadhya
- Chiranjay Mukhopadhyay
- Shyamsundar Shreedhar
- Deepak Sudhakaran
- Suma Nair
- Mohan Papanna
- Rajesh Yadav
- Sujeet Kumar Singh
- Tanzin Dikid
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021467
doi: 10.1371/journal.pgph.0000865
license: CC0 1.0
---
# Risk factors for melioidosis in Udupi District, Karnataka, India, January 2017-July 2018
## Abstract
We initiated an epidemiological investigation following the death of a previously healthy 17 year-old boy with neuro-melioidosis. A case was defined as a culture-confirmed melioidosis patient from Udupi district admitted to hospital A from January 2013—July 2018. For the case control study, we enrolled a subset of cases admitted to hospital A from January 2017- July 2018. A control was resident of Udupi district admitted to hospital A in July 2018 with a non-infectious condition. Using a matched case-control design, we compared each case to 3 controls using age and sex groups. We assessed for risk factors related to water storage, activities of daily living, injuries and environmental exposures (three months prior to hospitalization), using conditional regression analysis. We identified 50 cases with case fatality rate $16\%$. Uncontrolled diabetes mellitus was present in $84\%$ cases and $66\%$ of cases occurred between May and October (rainy season). Percutaneous inoculation through exposure to stagnant water and injury leading to breakage in the skin were identified as an important mode of transmission. We used these findings to develop a surveillance case definition and initiated training of the district laboratory for melioidosis diagnosis.
## Introduction
Melioidosis is caused by gram-negative intracellular bacteria Burkholderia pseudomallei, which can infect both humans and animals. This environmental saprophyte is widely distributed in soil and fresh surface water in endemic regions of South East Asia, Northern Australia, the Indian subcontinent and areas of South America [1–4]. Up to $20\%$ of community-acquired sepsis in the tropics is due to melioidosis and the overall mortality varies from 20–$50\%$ depending on the availability of healthcare services [5–8]. In 2015, the global burden of melioidosis was 4.6 million DALYs, which was higher than other common neglected tropical diseases [9]. Estimates suggest that the extent of melioidosis global distribution is widespread, and the cases are grossly under-reported in 45 countries currently reporting [10]. This disparity is partly due to under-recognition due to its diverse clinical manifestations and the inadequacy of conventional bacterial identification methods [5,6,8,9].
Studies from Australia and South East Asia indicate that environmental and host factors determine disease acquisition. The disease incidence increases during the rainy season and adverse weather conditions like tsunamis and cyclones; agricultural workers are commonly affected [6,8,11,12]. Host factors such as the presence of one or more preexisting conditions that alter immune response (such as long standing uncontrolled diabetes mellitus or chronic renal failure), severe or penetrating injury or near-drowning are favorable for the occurrence of melioidosis [8,12,13]. In India, cases of melioidosis have been recognized from different regions, however case identification is confined to few tertiary centres due to limited diagnostic facilities [14–16]. The incidence of melioidosis in *India is* unknown but could be substantial due to the high burden of diabetes mellitus [17] and long coastline prone to extreme weather conditions.
On July 23, 2018, the death of a boy aged 17 years was reported to Moodabettu Primary Health Centre (PHC), Udupi district of Karnataka. On July 24, a team from the district disease surveillance office visited hospital A, where the deceased received treatment and had been diagnosed with melioidosis. The surveillance team also searched for similar cases in the village and conducted an awareness program for reporting sudden deaths. National Centre for Disease Control (NCDC) was notified, and Epidemic Intelligence Service Officers (EISOs) from NCDC joined the district investigation on August 1, 2018. We investigated to describe the epidemiology and identify risk factors to inform the initial public health responses.
## I. Ethics statement
This investigation was undertaken as part of an emergency public health response to identify the cause of an outbreak for early intervention. All statutory permissions were obtained from NCDC and Integrated Disease Surveillance Programme. Ethical approval was exempted as the investigation was conducted consistent with applicable state and central government law (Epidemic Diseases Act no.3, 1897). Strict data protection protocols reviewed by NCDC were followed while collecting information from cases and controls.
## II. Case investigation
To ascertain the cause of death, we interviewed the treating physicians and family members of the deceased. Information was collected on clinical presentation, the timeline of events before the death, treatment and laboratory results. The patient’s CSF specimen was plated on routine media as well as in BacT/ALERT automated culture system (bioMérieux, Marcy-L’Etoile, France). The patient’s blood was also collected to rule out septicemia and cultured in BacT/ALERT automated culture system. The preliminary Gram stain of the CSF sample revealed bipolar gram negative staining leading to a presumptive diagnosis, which was conveyed to the treating doctors to initiate immediate empirical treatment. The culture isolate from CSF after 48 hours of incubation at 37°C with $5\%$ CO2 were examined by latex agglutination, matrix assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF), and type three secretion system (TTSS) polymerase chain reaction (PCR).
## Study site
Udupi district has a population of 1,177,908 and a literacy rate of $83\%$ [18]. The weather is hot and humid during summers and receives heavy rainfall from May-August. The population’s comprises of agricultural communities, and rice is the main crop grown in the region. Udupi district has 76 primary health centres, six community health centres and a district hospital in the urban area. Patients who are critically ill are referred from most of Udupi district, other parts of Karnataka state and nearby districts of Kerala state to hospital A, which is a tertiary care teaching hospital.
## Case definition for descriptive analysis
A case was defined as a diagnosis of culture-confirmed melioidosis (isolation of B. pseudomallei from any clinical sample and suggestive clinical features) in a patient from the Udupi district who was admitted to the ward/intensive care unit of hospital A from January 2013—July 2018.
## Case search and data collection
We reviewed inpatient records and laboratory registers from 2013–18 from hospital A. For patients meeting the case definition, we collected clinical and laboratory data and created a line list without patient identifiers. Clinical data was available for patients attending the hospital after 2016.
## IV. Case-control study
We conducted a 1:3 matched case-control study to identify risk factors on hospitalized patients meeting the above case definition from January 2017 to July 2018 to limit recall bias. A control was defined as a resident of Udupi district admitted to hospital A in July (rainy season) 2018 with a non-infectious condition. We matched each case with three hospital controls by age and sex group (males age 15–43 years, males >50 years and females >49 years). All controls were selected from the hospital admission records (eligible controls were listed and selected by random number generation using MS Excel using RANDBETWEEN function). We excluded patients receiving antibiotic treatment for pneumonia or sepsis.
A semi-structured questionnaire was used to collect data on socio-demographics, housing conditions, water storage practices, daily living activities, injuries, and environmental exposures (three months before hospitalization) from cases and controls during interviews. For the deceased cases, we conducted proxy interviews with family members.
## Data analysis
The Epi Info software 7.2 was used to analyze frequencies and proportions. For the identified risk factors, crude odds ratio (OR) with a $95\%$ confidence interval (CI) was calculated. The exact conditional logistic regression analysis was run to obtain matched odds ratios (mOR) and $95\%$ CI considering the small sample size using SAS version 9.4
## I. Case investigation
A 17-year-old boy from the Udupi district developed complaints of fever and vomiting on July 7, 2018. On July 8, his illness progressed by evening, and he developed high-grade fever, severe headache and dizziness. On July 9, the family noticed deviation at the angle of the mouth and consulted a local doctor. The boy’s condition deteriorated and he was taken to hospital A, where he was admitted on July 11 with difficulty in swallowing, change in voice and ataxia. He developed seizures and coma on July 14 and was transferred to intensive care. His condition deteriorated with neurological involvement, and he died on July 21 in hospital A. The patient had a history of fall on June 25 during a Kabaddi game (*Kabaddi is* a local team game of chasing the opponent team played barefoot on a muddy playground). No external injuries were apparent, but minor abrasions on limbs and hands were noted. The case had no history of diabetes, renal dysfunction, or other comorbidities. He was diagnosed with neuro-melioidosis by cerebrospinal fluid culture and PCR. His blood culture for bacteria including mycobacterium tuberculosis was negative. Contrast magnetic resonance imaging of the brain and spinal cord showed patchy T2 and FLAIR hyperintense lesions in the brainstem, middle cerebellar peduncles, and bilateral posterior limbs of internal capsule, in addition to few subcortical lesions. Post-contrast enhancement was seen in the brainstem lesions, along the trigeminal nerves, facial and abducens nuclei. This case was later reported as part of case series showing spectrum of nervous involvement in melioidosis with detailed clinical description [19].
## II. Descriptive epidemiology of cases
A total of 50 cases of melioidosis were identified at hospital A from 2013–18. The median age of cases was 52 (range 17–83) years; $80\%$ were males, and the case fatality rate was $16\%$. The cases were distributed (by residence) in all the three taluks (administrative subdivisions) of the Udupi district (Fig 1); $76\%$ of cases occurred in the villages closer to the coastline, and $66\%$ of cases occurred between May and October (Fig 2).
**Fig 1:** *Geographical distribution of melioidosis cases from 2013–2018 Udupi District, Karnataka, India (n = 50).Base map republished from [20] under a CC BY license, with permission from Karnataka State Remote Sending Application Centre (KSRC), original copyright KSRC, 2022.* **Fig 2:** *Monthly distribution of average melioidosis cases and rainfall from 2013–2018, Udupi district, Karnataka, India (n = 50).*
We analyzed the clinical presentation of all 19 cases reported from January 2017- July 2018. The most common presenting symptoms were fever ($89\%$), cough ($42\%$), joint pain ($37\%$) and abscess ($16\%$). Uncontrolled diabetes (HbA1c >$7\%$) was documented in $84\%$ [16] cases with overall median HbA1c of $9.5\%$ (range $5.8\%$ -$13\%$). The other chronic comorbidities included chronic kidney disease with diabetes ($10\%$), COPD with diabetes ($10\%$), liver disease with diabetes ($10\%$), tuberculosis ($5\%$), and cancer ($5\%$). Among 19 cases, the most common presentation was bacteremic melioidosis ($58\%$), followed by skin and soft tissue ($11\%$), septic arthritis ($11\%$), pneumonia ($5\%$), splenic abscess ($5\%$), neuromelioidosis ($5\%$) and no focus ($5\%$). Out of 19 cases, five had lung involvement; of these, one had focal lung involvement and four others had lung involvement with bacteremia. All five cases with pneumonia were exposed either to soil or stagnant water. Among the 14 non pneumonia cases, 13 had history of exposure to soil or stagnant water.
## III. Case-control study
We enrolled all 19 cases from 2017–18 and 57 hospital controls in the matched cases control analysis. The enrollment flowchart for study participants is in appendix page 1. Univariate analysis showed that melioidosis cases were more likely than controls to have an injury with breach of skin, contact with stagnant water, wet soil, and both. In matched analysis, injury with a breach in skin and contact with stagnant water had a significant association with illness. In addition, we also looked for activity-specific odds ratios for contact with stagnant water. The odds of exposure to swimming in stagnant water, working in paddy field, and walk in waterlogged areas were higher among cases compared to controls (Table 1).
**Table 1**
| Risk Factors | Case (n = 19) | Case (n = 19).1 | Control (n = 57) | Control (n = 57).1 | OR | (95% CI) | (95% CI).1 | mOR (95% CI) | mOR (95% CI).1 | p-value |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | n | (%) | n | (%) | | | | | | |
| Injury leading to breach of skin | 10 | (53) | 9 | (16) | 5.9 | (1.9–18.7) | (1.9–18.7) | 6.2 | (1.9–20.1) | 0.0014 |
| Multiple chronic comorbidities†† | 4 | (21) | 6 | (10) | 2.3 | (0.6–9.1) | (0.6–9.1) | 2.2 | (0.6–8.9) | 0.2464 |
| Contact with stagnant water‡ | 7 | (37) | 4 | (7) | 63.0 | (6.1–651) | (6.1–651) | 58.2 | (5.2->999) | <0.001 |
| Contact with wet soil† | 3 | (16) | 8 | (14) | 13.5 | (1.2–147) | (1.2–147) | 10.8 | (0.7–630) | 0.0887 |
| Contact with stagnant water and wet soil | 8 | (42) | 9 | (16) | 32.0 | (3.5–290) | (3.5–290) | 28.2 | (3.2->999) | 0.0004 |
| No contact with wet soil or stagnant water | 1 | (5) | 36 | (63) | Ref | | | Ref | | |
| Contact with stagnant water | | | | | | | | | | |
| Swimming in pond, work in paddy field and walk in waterlogged area | 6 | (32) | 0 | (0) | 93.0 | (8.4–4030) | (8.4–4030) | >999 | (<0->999) | 0.9941 |
| Swimming in pond and work in paddy field | 3 | (16) | 2 | (3) | 29.0 | (2.9–287) | (2.9–287) | 48.6 | (4.3–556) | 0.0018 |
| Swimming in pond and walk in waterlogged area | 2 | (10) | 2 | (3) | 19.5 | (1.7–219) | (1.7–219) | 16.9 | (1–198) | 0.0247 |
| Work in paddy field | 2 | (10) | 5 | (9) | 7.8 | (0.9–68) | (0.9–68) | 12.2 | (1.2–126) | 0.0362 |
| Swimming in pond | 2 | (10) | 5 | (9) | 7.8 | (0.9–68) | (0.9–68) | 11.2 | (1.1–110) | 0.0379 |
| Walk in waterlogged area | 2 | (10) | 4 | (8) | 9.7 | (1.1–89) | (1.1–89) | 10.1 | (1.1–96) | 0.0438 |
| No contact with stagnant water | 2 | (10) | 39 | (68) | Ref | Ref | | Ref | | |
## Discussion
The death of an adolescent due to neuromelioidosis led to a broader, multi-year retrospective investigation of admitted cases of melioidosis in one district. These cases were mostly males with uncontrolled diabetes. The results from the case-control study suggest that outdoor exposure to stagnant water and wet soil in rainy season are a risk factor for melioidosis in the Udupi district.
We observed that the majority ($66\%$) of the cases occurred during the rainy session (May-October). This is similar to the seasonality reported from the west coastal region of India, Singapore and Australia [11,16,21]. The descriptive analysis showed most ($76\%$) cases were from villages closer to the coast with paddy fields and low-lying areas that frequently flood during the rainy session. These environmental conditions could increase the chances of contact with contaminated wet soil and stagnant water during agricultural and non-agricultural activities associated with infection. In our case-control study, exposure to wet soil and stagnant water were significant risk factors for melioidosis ($p \leq 0.001$). Additionally, we also found that sustaining cut injuries was an independent predictor in multivariate analysis. Melioidosis cases were six times more likely to be exposed to cut injuries compared to controls. These findings, combined with environmental conditions, indicate that percutaneous inoculation is an important transmission mode for melioidosis in rural India. Similar epidemiological risk factors were identified among melioidosis cases in rural Thailand and northern Australia [11,12]. In contrast, recent findings have also recognized inhalation of B. pseudomallei and eating food contaminated with soil or dust as other important transmission modes [21,22].
People with long-standing diabetes with poor glycemic control are known to be at increased risk of acquiring melioidosis [6,12], and $84\%$ of the cases in this study detected between 2013–2018 had uncontrolled diabetes. A systematic review from India for the period 1991–2018 found diabetes mellitus to be a major predisposing condition in $70\%$ of reported cases [16]. There are multiple reports of melioidosis from various parts of India [15,16,23–25], but it is not a notifiable disease. Estimates suggest that melioidosis is endemic to India with an annual burden of ~52,006 cases and death count of 31,425 (13,405–75,601) cases, with further escalation of the mortality rate to $90\%$ if the disease remains undiagnosed and untreated [10]. Given the high burden of diabetes in India, inadequate diagnostic facilities in the microbiology laboratories, especially in rural parts [26] and low awareness among physicians and the public, the actual burden of melioidosis in *India is* expected to be high. Therefore, priorities include establishing state and national referral laboratory networks, training for diagnosis and surveillance for melioidosis, and increasing awareness in the community and among physicians.
Though neuromelioidosis is a rare (4–$5\%$) presentation of melioidosis [19,27,28], children and adolescents have poor outcomes, frequently resulting in either death or neurological impairment ($37\%$) attributed to severe sepsis and its complications, resulting from delay in treatment [5,28]. Studies from Australia [29] and Cambodia [30] have documented a case fatality rate of 7–$17\%$; notably children in the Australian study did not have any underlying comorbidities, similar to the young index patient in our study.
Our study has limitations, including recall bias for exposure due to the retrospective nature of the study design. We attempted to limit this by enrolling the most recent cases and asking for exposures from cases and controls within a three-month reference period. The enrollment period for cases and controls was different due to logistical issues. However, we ensured comparability of risks by enrolling controls during the rainy season, as more than half of the cases were reported during this season. Finally, the findings are from admitted cases from a single hospital were enrolled, potentially limiting the generalizability of our findings.
We recommended enhancing the education of medical doctors for early diagnosis and treatment of melioidosis among fever cases; and strengthening district public health labs in Karnataka state for melioidosis diagnosis. We also recommended communication campaigns targeting high-risk groups (diabetics in coastal areas, patients with chronic renal failures, agricultural workers and people with frequent soil exposure like daily labourers). These communication campaigns should focus on minimizing unnecessary exposures to soil and water through avoidance or protective clothing including use of footwear.
Following this investigation, the National Centre for Disease Control published an information bulletin to increase awareness among clinicians of melioidosis as a differential diagnosis in fever of unknown origin and community acquired pneumonia. To enhance surveillance, an expert group meeting was convened to develop surveillance case definitions, and district lab personnel training was initiated to strengthen melioidosis diagnostic capacity.
## References
1. Goodrick I, Todd G, Stewart J. **Soil characteristics influencing the spatial distribution of melioidosis in Far North Queensland**. *Australia. Epidemiol Infect* (2018.0) **146** 1602-1607. PMID: 29970213
2. Aldhous P.. **Tropical medicine: melioidosis? Never heard of it**. *Nature* (2005.0) **434** 692-693. DOI: 10.1038/434692a
3. Stone R.. **Infectious disease. Racing to defuse a bacterial time bomb**. *Science* (2007.0) **317** 1022-1024. DOI: 10.1126/science.317.5841.1022
4. Currie BJ, Dance DA, Cheng AC. **The global distribution of Burkholderia pseudomallei and melioidosis: an update**. *Trans R Soc Trop Med Hyg* (2008.0) **102** S1-4. DOI: 10.1016/S0035-9203(08)70002-6
5. White NJ. **Melioidosis**. *Lancet* (2003.0) **361** 1715-1722. DOI: 10.1016/s0140-6736(03)13374-0
6. Wiersinga WJ, Van der Poll T, White NJ, Day NP, Peacock SJ. **Melioidosis: insights into the pathogenicity of Burkholderia pseudomallei**. *Nature Rev* (2006.0) **4** 272-282. DOI: 10.1038/nrmicro1385
7. Currie BJ, Fisher DA, Howard DM, Burrow JN, Lo D, Selva-Nayagam S. **Endemic melioidosis in tropical northern Australia: a 10-year prospective study and review of the literature**. *Clin Infect Dis* (2000.0) **31** 981-986. DOI: 10.1086/318116
8. Hassan MR, Pani SP, Peng NP, Voralu K, Vijayalakshmi N, Mehanderkar R. **Incidence, risk factors and clinical epidemiology of melioidosis: a complex socio-ecological emerging infectious disease in the Alor Setar region of Kedah, Malaysia**. *BMC Infect Dis* (2010.0) **10** 302. DOI: 10.1186/1471-2334-10-302
9. Birnie E, Virk HS, Savelkoel J, Spijker R, Bertherat E, Dance DAB. **Global burden of melioidosis in 2015: a systematic review and data synthesis**. *Lancet Infect Dis* (2019.0) **19** 892-902. DOI: 10.1016/S1473-3099(19)30157-4
10. Limmathurotsakul D, Golding N, Dance DA, Messina JP, Pigott DM, Moyes CL. **Predicted global distribution of Burkholderia pseudomallei and burden of melioidosis**. *Nat Microbiol* (2016.0) **1** 15008
11. Currie B. J., Jacups S. P.. **Intensity of rainfall and severity of melioidosis, Australia**. *Emerg. Infect. Dis* (2003.0) **9** 1538-1542. DOI: 10.3201/eid0912.020750
12. Suputtamongkol Y, Chaowagul W, Chetchotisakd P, Lertpatanasuwun N, Intaranongpai S, Ruchutrakool T. **Risk factors for melioidosis and bacteremic melioidosis**. *Clin Infect Dis* (1999.0) **29** 408-13. DOI: 10.1086/520223
13. Chierakul W.. **Melioidosis in 6 tsunami survivors in Southern Thailand**. *Clin. Infect. Dis* (2005.0) **41** 982-990. DOI: 10.1086/432942
14. Saravu K, Mukhopadhyay C, Vishwanath S, Valsalan R, Docherla M, Vandana KE. **Melioidosis in southern India: epidemiological and clinical profile.**. *Southeast Asian J Trop Med Public Health* (2010.0) **41** 401-9. PMID: 20578524
15. Menon R, Baby P, Kumar V A, Surendran S, Pradeep M, Rajendran A. **Risk Factors for Mortality in Melioidosis: A Single-Centre, 10-Year Retrospective Cohort Study**. *Scientific World Journal* (2021.0) **2021** 8154810. DOI: 10.1155/2021/8154810
16. Mukhopadhyay C, Shaw T, Varghese GM, Dance DAB. **Melioidosis in South Asia (India, Nepal, Pakistan, Bhutan and Afghanistan)**. *Trop Med Infect Dis* (2018.0) **3** 51. DOI: 10.3390/tropicalmed3020051
17. **The increasing burden of diabetes and variations among the states of India: the Global Burden of Disease Study 1990–2016**. *Lancet Glob Health* (2018.0) **6** e1352-e1362. DOI: 10.1016/S2214-109X(18)30387-5
18. 18Government of Karnataka. About district. Uduppi district. Available at https://udupi.nic.in/en/about-district/ (accessed on 04 July, 2022).
19. Chatterjee A, Saravu K, Mukhopadhyay C, Chandran V. **Neurological Melioidosis Presenting as Rhombencephalitis, Optic Neuritis, and Scalp Abscess with Meningitis: A Case Series from Southern India**. *Neurol India* (2021.0) **69** 480-482. DOI: 10.4103/0028-3886.314590
20. 20Karnataka State Remote Sensing Application Centre. Avaliable at https://ksrsac.karnataka.gov.in/ (accessed on July 20, 2022).
21. Liu X, Pang L, Sim SH, Goh KT, Ravikumar S, Win MS. **Association of melioidosis incidence with rainfall and humidity, Singapore, 2003–2012**. *Emerg. Infect. Dis* (2015.0) **21** 159-162. DOI: 10.3201/eid2101.140042
22. Limmathurotsakul D, Kanoksil M, Wuthiekanun V, Kitphati R, deStavola B, Day NP. **Activities of daily living associated with acquisition of melioidosis in northeast Thailand: a matched case-control study**. *PLoS Negl Trop Dis* (2013.0) **7** e2072. DOI: 10.1371/journal.pntd.0002072
23. Ganesan V, Sundaramoorthy R, Subramanian S. **Melioidosis-Series of Seven Cases from Madurai, Tamil Nadu, India**. *Indian J Crit Care Med* (2019.0) **23** 149-151. DOI: 10.5005/jp-journals-10071-23139
24. Annamalai AK, Padmini K. **Melioidosis**. *Indian J Med Res* (2019.0) **149** 561-562. DOI: 10.4103/ijmr.IJMR_2018_17
25. Subramony H, Gunasekaran S, Paul Pandi VK. **Disseminated melioidosis with native valve endocarditis: a case report**. *Eur Heart J Case Rep* (2019.0) **3** ytz097. DOI: 10.1093/ehjcr/ytz097
26. 26Central Bureau of Health Intelligence. National health profile 2019. Available at https://www.cbhidghs.nic.in/showfile.php?lid=1147 (accessed on December 30, 2020).
27. Saravu K, Kadavigere R, Shastry AB, Pai R, Mukhopadhyay C. **Neurologic melioidosis presented as encephalomyelitis and subdural collection in two male labourers in India**. *J Infect Dev Ctries* (2015.0) **9** 1289-93. DOI: 10.3855/jidc.6586
28. Wiersinga WJ, Currie BJ, Peacock SJ. **Melioidosis**. *N Engl J Med* (2012.0) **367** 1035-44. DOI: 10.1056/NEJMra1204699
29. Kandasamy Y, Norton R. **Paediatric melioidosis in North Queensland, Australia**. *J Paediatr Child Health* (2008.0) **44** 706-8. DOI: 10.1111/j.1440-1754.2008.01410.x
30. Turner P, Kloprogge S, Miliya T, Soeng S, Tan P, Sar P. **A retrospective analysis of melioidosis in Cambodian children, 2009–2013**. *BMC Infect Dis.* (2016.0) **16** 688. DOI: 10.1186/s12879-016-2034-9
|
---
title: Missed opportunities for earlier diagnosis of HIV infection in people living
with HIV in Thailand
authors:
- Angsana Phuphuakrat
- Kanitin Khamnurak
- Sirawat Srichatrapimuk
- Wittaya Wangsomboonsiri
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021504
doi: 10.1371/journal.pgph.0000842
license: CC BY 4.0
---
# Missed opportunities for earlier diagnosis of HIV infection in people living with HIV in Thailand
## Abstract
HIV testing is the first step to making people living with HIV (PLHIV) aware of their status. Thailand is among the countries where antiretroviral therapy is initiated in PLHIV at the lowest CD4 cell counts. We aimed to quantify and characterize missed opportunity (MO) for earlier diagnosis of HIV infection in PLHIV in Thailand. The medical records of adults who were newly diagnosed with HIV between 2019 and 2020 at the two tertiary hospitals in Thailand were reviewed. A hospital visit due to an HIV clinical indicator disease but an HIV test was not performed was considered an MO for HIV testing. Of 422 newly diagnosed PLHIV, 60 persons ($14.2\%$) presented with at least one MO, and 20 persons ($33.3\%$) had more than one MO. In PLHIV with MO, the median (interquartile range) time between the first MO event and HIV diagnosis was 33.5 (7–166) days. The three most common clinical manifestations that were missed were skin manifestations ($25.0\%$), unexplained weight loss ($15.7\%$), and unexplained lymphadenopathy ($14.3\%$). Anemia was a factor associated with MO for HIV diagnosis [odds ratio (OR) 2.24, $95\%$ confidence interval (CI) 1.25–4.35; $$p \leq 0.018$$]. HIV screening reduced the risk of MO for HIV diagnosis (OR 0.53 $95\%$ CI 0.29–0.95; $$p \leq 0.032$$). In conclusion, MOs for earlier diagnosis of HIV infection occurred in both participating hospitals in Thailand. Skin manifestations were the most common clinical indicator diseases that were missed. HIV testing should be offered for patients with unexplained anemia. Campaigns for HIV screening tests should be promoted.
## Introduction
Treatment of human immunodeficiency virus (HIV) infection by antiretroviral therapy (ART) has reduced morbidity and mortality in people living with HIV (PLHIV) and prevented the risk of onward transmission. HIV testing and diagnosis is the first step to making PLHIV aware of their status and entering the cascade of care. Despite the free HIV testing programs supported by the government and non-governmental organizations in Thailand, it was estimated that $50\%$ of Thai PLHIV did not aware of their HIV status and $50\%$ of Thai PLHIV started ART at CD4 counts below 100 cells/mm3 [1]. Late detection of HIV infection leads to delayed treatment and results in increased morbidity and mortality from acquired immunodeficiency syndrome (AIDS), healthcare costs [2], and treatment complexity.
Thailand has been scaling up ART. The national guidelines on HIV/AIDS treatment and prevention have recommended treating all PLHIV since 2014, irrespective of CD4 cell count. Access to HIV care and treatment is also free of charge in government hospitals. However, the proportion of PLHIV with late ART initiation was still significant. A previous study reported a global trend of increased median CD4 cell counts at the start of ART in PLHIV who started ART between 2002 and 2015 [3]. Notwithstanding this, Thailand was among the countries in which ART was commenced at the lowest CD4 spectrum [4], and approximately $75\%$ of PLHIV started ART with CD4 cell count <200 cells/mm3 [3]. At Ramathibodi Hospital, the medians CD4 cell count at HIV diagnosis during 2011–2013, 2015–2017, and 2018 were 146 (45–298) [5], 159 (53–308) [6], and 116 (34–332) [7] cells/mm3, respectively.
There has been no study on the risk factors of late HIV diagnosis in Thailand. In this study, we quantified and characterized missed opportunity (MO) for earlier diagnosis of HIV infection in PLHIV in Thailand. We also studied the risk factors of the MO for earlier diagnosis of HIV infection.
## Study setting
This study was performed at Ramathibodi and Sawanpracharak hospitals in Thailand. Ramathibodi *Hospital is* a 1300-bed university hospital in Bangkok that serves a diverse population in central Thailand and is capable of super tertiary care. Sawanpracharak *Hospital is* a 700-bed tertiary hospital, which serves as a primary-level community hospital for Nakhon Sawan province and as a secondary and tertiary referral hospital for upper central Thailand. We retrospectively reviewed medical records of persons who were newly diagnosed with HIV infection between January 2019 and December 2020 by retrieving from the hospital databases. We included PLHIV whose age was at least 15 years old. PLHIV with previously known HIV-positive before 2019 were excluded. Sociodemographic data (e.g., gender, age at HIV diagnosis, HIV transmission route, and race) and clinical as well as laboratory data (e.g., CD4 cell counts, HIV viral load, clinical manifestations, and AIDS-defining illness at diagnosis) were included in the analyses.
The study protocol was reviewed and approved by the Ethical Clearance Committee on Human Rights Related to Research Involving Human Subjects of the Faculty of Medicine Ramathibodi Hospital, Mahidol University (MURA$\frac{2021}{64}$) and the Research and Journal Division of Sawanpracharak Hospital ($\frac{23}{2564}$). As the study was retrospective, the ethics committees waived the requirement for informed consent.
## Definitions
An MO was defined as a healthcare encounter due to a clinical manifestation that may be caused by HIV infection or as a healthcare encounter due to an AIDS-defining illness (clinical indicator disease) but HIV was not tested [8]. AIDS-defining illnesses were defined following Thailand National Guidelines on HIV/AIDS treatment and prevention 2017 [1]. Clinical indicator diseases for adult HIV infection were recorded following the UK National Guidelines for HIV testing 2008 [9] and Australian National HIV Testing Policy 2011 [10]. Late presenters were defined as PLHIV whose CD4 cell count at HIV diagnosis was less than 350 cells/mm3 [11]. HIV screening included an HIV test performed in a check-up program, preoperative screening, during antenatal care, before entry to inpatient care, before entry to prison, following the positive result of the partner, and according to the patient’s request. Anemia was defined following World Health Organization (WHO) criteria for anemia (hemoglobin <13 g/dL in men and <12 g/dL in women) [12].
## Statistical methods
All data were retrieved from patients’ medical records and were compiled in a tabular manner, using an Excel datasheet (see S1 Data). Statistical analyses were performed using Stata statistical software version 17.0 (StataCorp, College Station, TX, USA). Mean±standard deviation or median (interquartile range; IQR) were used to describe continuous variables. Student’s t-test, Mann-Whitney U-test, chi-square test, and Fisher’s exact test were used for the comparisons, as appropriate, with the level of significance set at a p-value of <0.05. A logistic regression model was performed to determine factors associated with MO of earlier diagnosis of HIV infection. Variables that presented a p-value of <0.2 from univariable logistic regression, or those that may affect missing opportunities despite their p-values of more than 0.2 were considered in a multivariable logistic regression model. Odds ratio (OR) and its $95\%$ confidence interval (CI) were estimated.
## Patient characteristics
During the study period, we identified 422 persons newly diagnosed with HIV infection (231 and 191 persons at Ramathibodi and Sawanpracharak hospitals, respectively). The mean age at the diagnosis was 39.3±13.4 years, and 290 persons ($68.7\%$) were men. Nearly all ($98.6\%$) were Thai. Of these, 340 persons ($80.6\%$) had available CD4 cell count data. The median CD4 at HIV diagnosis was 160 (44–313) cells/mm3. At the diagnosis, 275 persons ($80.9\%$) were late presenters and 196 persons ($57.7\%$) had CD4 cell counts less than 200 cells/mm3. The most common HIV transmission routes were heterosexual ($61.1\%$) followed by homosexual ($23.3\%$). The most common reason for testing was physician request (181 persons; $42.9\%$) and patients were mostly diagnosed in an outpatient care setting (253 persons; $60.0\%$). Nearly half of the PLHIV (205 persons, $48.6\%$) had clinical indicator diseases at an HIV diagnosis.
Sixty patients ($14.2\%$) had presented at least one MO. Table 1 shows characteristics of PLHIV at HIV diagnosis stratified by history of MO. There were no significant differences in age, gender, or body mass index (BMI) among PLHIV with or without MOs. There was a higher proportion of PLHIV with MO in persons whose risk was heterosexual, but the difference was not statistical significance ($$p \leq 0.424$$).
**Table 1**
| Unnamed: 0 | PLHIV without MO (n = 362) | PLHIV with ≥1 MO (n = 60) | p-value |
| --- | --- | --- | --- |
| Age, (years), mean±SD | 38.9±13.1 | 41.5±15.2 | 0.118 |
| Male, n (%) | 253 (69.9) | 37 (61.7) | 0.203 |
| Body mass index, (kg/m2) mean±SD | 21.9±4.3 | 21.2±3.4 | 0.253 |
| Hospital, n (%) | | | 0.426 |
| Ramathibodi | 201 (55.5) | 30 (50.0) | |
| Sawanpracharak | 161 (44.5) | 30 (50.0) | |
| Race, n (%) | | | 0.315 |
| Thai | 356 (98.3) | 60 (100.0) | |
| Foreigners | 6 (1.7) | 0 (0.0) | |
| HIV risk | | | 0.424 |
| Heterosexual | 215 (59.4) | 43 (71.7) | |
| Homosexual | 98 (27.1) | 13 (21.7) | |
| Bisexual | 1 (0.3) | 0 (0.0) | |
| Persons who inject drugs | 1 (0.3) | 0 (0.0) | |
| Unknown | 47 (13.0) | 4 (6.7) | |
| Clinical indicator diseases | | | 0.003 |
| No | 198 (54.7) | 19 (31.7) | |
| Disease may be caused by HIV infection | 97 (26.8) | 27 (45.0) | |
| AIDS-defining illness | 67 (18.5) | 14 (23.3) | |
| Reason for testing | | | 0.065 |
| Preoperative laboratory test | 63 (17.4) | 9 (15.0) | |
| Check-up program | 4 (1.1) | 0 (0.0) | |
| Antenatal care | 13 (3.6) | 0 (0.0) | |
| Screening on entry to inpatient care | 77 (21.3) | 14 (23.3) | |
| Screening on entry to prison | 2 (0.6) | 1 (1.7) | |
| Partner positive result | 19 (5.3) | 0 (0.0) | |
| Patient request | 37 (10.2) | 2 (3.3) | |
| Physician request | 147 (40.6) | 34 (56.7) | |
| Prior HIV test | 18 (5.0) | 1 (1.7) | 0.496 |
| Site of testing | | | 0.308 |
| Emergency department | 50 (13.8) | 5 (8.3) | |
| Inpatient care | 100 (27.6) | 14 (23.3) | |
| Outpatient care | 212 (58.6) | 41 (68.3) | |
| Ordering physician at HIV diagnosis | | | 0.153 |
| General practitioner | 72 (19.9) | 7 (11.7) | |
| Resident | 108 (29.8) | 14 (23.3) | |
| Fellow | 13 (3.6) | 4 (6.7) | |
| Specialist | 169 (46.7) | 35 (58.3) | |
| Hemoglobin, (g/dL), mean±SD | 11.9±2.5 | 11.1±2.3 | 0.011 |
| Anemia | 197 (54.4) | 46 (76.7) | 0.001 |
| WBC, (cells/mm3), median (IQR) | 6,800 (5,310–8,520) | 5,800 (4,750–7,410) | 0.013 |
| CD4, (cells/mm3), median (IQR) | 187 (45–332) | 84 (40–161) | 0.003 |
| CD4 cell counts <350 cells/mm3 | 225 (62.2) | 50 (83.3) | <0.001 |
| Platelets, (/mm3), mean±SD | 261,939±113,501 | 262,644±87,842 | 0.964 |
| Aspartate aminotransferase | 33 (23–54) | 37 (23–53) | 0.781 |
| Alanine aminotransferase | 25 (15–41) | 25 (16–37) | 0.445 |
| Estimated glomerular filtration rate | 101.5±26.5 | 102.5±22.8 | 0.807 |
The presence of clinical indicator diseases was different in PLHIV who had or did not have MO (Table 1). A higher proportion of PLHIV with at least one MO had clinical indicator diseases at HIV diagnosis ($$p \leq 0.003$$). In PLHIV with clinical indicator diseases, HIV testing was requested by a physician in 167 persons ($81.5\%$). In total, HIV testing was requested by physicians in 181 persons ($42.9\%$). Other reasons for testing were due to screening purposes. Of note, Sawanpracharak Hospital, but not Ramathibodi Hospital, has a mandatory HIV screening test on entry to inpatient care. This was able to detect 91 ($21.6\%$) new cases during the study period. In PLHIV without clinical indicator diseases, HIV screening tests detected a significant proportion of this population, compared to those detected by physician request ($93.6\%$ vs. $0.5\%$; $p \leq 0.001$).
Only 19 PLHIV ($4.5\%$) in this study had prior HIV testing, and there was no significant difference in the proportion between the two groups. Sites of testing and physicians who requested anti-HIV tests at HIV diagnosis were not significantly different between the two groups.
The proportion of late presenters was significantly higher in PLHIV with a history of MO ($83.3\%$ vs. $62.2\%$; $p \leq 0.001$). In PLHIV with and without MO, the median CD4 cell counts were 84 (40–161), and 187 (45–332) cells/mm3, respectively. CD4 cell counts were significantly higher in PLHIV without history of MO ($$p \leq 0.003$$). White blood cells were also significantly higher in PLHIV without history of MO (6,800 vs. 5,800 cells/mm3; $$p \leq 0.013$$). Hemoglobin levels were higher in PLHIV without history of MO (11.9 vs. 11.1; $$p \leq 0.011$$). There were no statistically significant differences in the numbers of platelets, levels of aspartate aminotransferase, alanine aminotransferase, and estimated glomerular filtration rate (Table 1).
## Missed opportunity events and clinical indicator diseases
Of the 60 PLHIV who had at least one MO, 20 persons ($33.3\%$) had more than one MO (Fig 1). The median duration between the first MO event and HIV diagnosis was 33.5 (IQR 7–166, range 1–1,472) days. The median number of MO event was one event per person (IQR 1–2, range 1–22).
**Fig 1:** *Numbers of missed opportunity events for earlier diagnosis of HIV infection in PLHIV at Ramathibodi and Sawanpracharak hospitals between 2019 and 2020.*
There were 117 MO events and 140 clinical indicator diseases in PLHIV with a history of MO. Of the clinical indicator diseases, four diseases ($2.9\%$) were AIDS-defining illnesses. Table 2 shows clinical indicator diseases that were missed in the study. The most common clinical indicator diseases that were missed included skin manifestations ($25.0\%$), unexplained weight loss ($15.7\%$), and unexplained lymphadenopathy ($14.3\%$) respectively (Table 2). For the skin manifestations, dermatitis/seborrheic dermatitis was the most common skin manifestation that HIV testing should be offered but missed. In PLHIV who had unexplained weight loss, pulmonary tuberculosis was the most common final diagnosis. In those who had unexplained lymphadenopathy, pathological diagnosis revealed reactive lymphadenopathy and lymphoma as the most common final diagnosis.
**Table 2**
| Clinical manifestations | Number (%) |
| --- | --- |
| Skin manifestations | 35 (25.0) |
| Dermatitis / seborrheic dermatitis | 28 (20.0) |
| Folliculitis | 2 (1.4) |
| Eczema | 5 (3.6) |
| Unexplained weight loss | 22 (15.7) |
| Unexplained lymphadenopathy | 20 (14.3) |
| Cervical dysplasia | 11 (7.9) |
| Herpes zoster in an individual <50 years old | 9 (6.4) |
| Sexually transmitted infection | 7 (5.0) |
| Unexplained leukocytopenia/thrombocytopenia persisting for >4 weeks | 7 (5.0) |
| Chronic cough | 6 (4.3) |
| Anal wart | 5 (3.6) |
| Unexplained oral candidiasis | 3 (2.1) |
| Subacute fever | 3 (2.1) |
| Prolonged fever | 3 (2.1) |
| Pulmonary tuberculosis | 2 (1.4) |
| Disseminated or extrapulmonary tuberculosis | 1 (0.7) |
| Hepatitis B | 1 (0.7) |
| Lymphoma | 1 (0.7) |
| Peripheral neuropathy of unknown origin | 1 (0.7) |
| Unexplained chronic diarrhea | 1 (0.7) |
| Tonsillitis (recurrent) | 1 (0.7) |
| Infectious mononucleosis | 1 (0.7) |
## Risk factors of missed opportunities for earlier HIV diagnosis
Univariable and multivariable analyses of factors associated with an MO for earlier diagnosis of HIV infection were performed (Table 3). In univariable analysis, HIV screening and anemia were associated with MO. In multivariable analysis, anemia showed an increased risk of MOs (OR 2.24; $95\%$ CI 1.15–4.35; $$p \leq 0.018$$). HIV screening reduced the risk of MO for HIV diagnosis (OR 0.53; $95\%$ CI 0.29–0.95; $$p \leq 0.032$$).
**Table 3**
| Factors | Univariable analysis | Univariable analysis.1 | Univariable analysis.2 | Multivariable analysis | Multivariable analysis.1 | Multivariable analysis.2 |
| --- | --- | --- | --- | --- | --- | --- |
| Factors | OR | 95% CI | p-value | OR | 95% CI | p-value |
| Age | 1.02 | 1.00–1.04 | 0.119 | 1.01 | 0.98–1.03 | 0.582 |
| Female | 1.44 | 0.82–2.54 | 0.205 | 1.07 | 0.54–2.12 | 0.843 |
| Heterosexual | 1.73 | 0.95–3.15 | 0.073 | 1.64 | 0.77–3.48 | 0.197 |
| HIV screening test | 0.52 | 0.30–0.91 | 0.021 | 0.53 | 0.29–0.95 | 0.032 |
| Specialist request | 1.60 | 0.92–2.78 | 0.096 | 1.70 | 0.96–3.01 | 0.068 |
| Anemia | 2.75 | 1.46–5.18 | 0.002 | 2.24 | 1.15–4.35 | 0.018 |
| Leukopenia | 2.03 | 0.91–4.53 | 0.085 | 1.61 | 0.70–3.71 | 0.263 |
## Discussion
This study demonstrated $14.2\%$ of MOs for earlier diagnosis of HIV infection occurred in our hospitals. In PLHIV who had at least one MO had lower CD4 count, and higher proportions of PLHIV with clinical indicator diseases at HIV diagnosis compared to those who did not have an MO. In this study, skin manifestations were the most common clinical indicator diseases that were missed. Anemia was significantly associated with an increased risk of MO, whereas HIV screening reduced the risk of MO.
A previous study showed that median CD4 cell counts at ART initiation increased over time in Asian PLHIV from 2007 to 2011 [13]. In our study, however, we revealed approximately $80\%$ of newly diagnosed PLHIV in Thailand were presented late. The proportion of late presenters in Thailand was higher than in other countries in Asia and the Pacific [14]. Previous studies showed $43\%$ of PLHIV in China [15] and $63\%$ of those in Malaysia [16] had a late presentation. Older age, male gender, and persons who inject drugs were factors associated with a presentation into care at CD4 cell count <200 cells/mm3 in Asian countries [17].
Factors associated with MO were different in various settings. In Israel, a low prevalence setting, there were no national guidelines concerning HIV testing except for the guidelines regarding pregnant women. Approximately one-third of PLHIV in the Israel study were diagnosed late; old age and heterosexual were the risk factors. Hematological diseases were the most common clinical indicator diseases that were missed in the study [18]. A study from Europe in the setting of an HIV outpatient clinic in Switzerland showed that $59\%$ of newly diagnosed PLHIV were late presenters, and $47\%$ had presented with at least one MO [11]. In this setting, MO was associated with individuals from sub-Saharan Africa, men who have sex with men, and patients under follow-up for chronic disease. In the United Kingdom, the study of MO in newly diagnosed HIV infection in Africans revealed that nearly half of the participants were diagnosed with HIV infection when CD4 cell count was <200 cells/mm3 [19]. Approximately $75\%$ of the participants attended general practitioner in the two years before HIV diagnosis. However, HIV testing was not offered for $82.4\%$ of those who accessed the services in the year before HIV diagnosis. A study in Malaysia reported $57\%$ of MOs for earlier HIV diagnosis [16]. Unexplained fever and/or fever lasting more than one month were the most common presenting symptom that was missed. Our study showed skin manifestations were the most common clinical indicator disease that were missed.
Multivariable analysis in our study demonstrated unexplained and/or unaware anemia was associated with MOs for earlier HIV diagnosis. Anemia was identified as a factor related to MOs for earlier HIV diagnosis in Canada [20]. In the study, $21\%$, $19\%$, and $18\%$ of anemia was related to iron deficiency anemia, vitamin B12 deficiencies, and unspecified causes, respectively. We identified an HIV screening test as a factor that reduced the risk of MOs, this factor agreed with the earlier study [21]. Of note, HIV screening on entry to inpatient care could detect many new cases in this study.
Thailand National Guidelines recommend HIV testing in persons who had signs and symptoms compatible with HIV infection or AIDS [1], but the clinical indicators of HIV infection were not described in detail. Adding the description of clinical indicator diseases to the guidelines might help remind clinicians when HIV testing should be offered. Unexplained anemia should be added to the conditions that anti-HIV should be tested. Thailand National Guidelines recommend HIV testing following the WHO guidelines on HIV testing services [22], in which counseling and consent are needed before HIV testing. In 2006, the United States Centers for Disease Control and Prevention (CDC) revised the HIV testing guidelines. The current CDC guidelines recommend an opt-out HIV screening in health-care settings [23]. This strategy does not require neither specific signed consent for HIV testing nor prevention counseling. HIV screening is included in the standard screening tests in all health-care settings unless the patient declines. The randomized clinical trial performed at the emergency department of a teaching hospital and regional trauma center showed that this opt-out approach significantly increased HIV testing acceptance [24]. In Thailand, the opt-out screening was previously studied in the setting of women undergoing treatment for cervical neoplasia with $100\%$ patient acceptance [25]. This suggested the feasibility of the opt-out approach implementation in health-care settings, e.g., screening before entry to inpatient care, in Thailand.
The strength of this study is that we compiled the newly diagnosed PLHIV in two hospitals located in different regions and had different screening programs. This study had some limitations. We were able to review the medical records only in these hospitals, and the patients might visit outside hospitals or services before presenting to the participating hospitals. Moreover, some data were incomplete, as with any retrospective study.
In conclusion, MOs for earlier diagnosis of HIV infection occur in both hospitals in Thailand and a large proportion of PLHIV presented late. Skin manifestations were the most common clinical indicator diseases that were missed. Unexplained and/or unaware anemia was associated with MOs for earlier HIV diagnosis, whereas HIV screening tests decreased the risk. HIV screening programs should be promoted, and the national guidelines should provide the details on unexplained anemia and clinical indicator diseases that HIV testing should be offered.
## References
1. 1Department of Disease Control, Mininstry of Public Heatlh. Thailand National Guidelines on HIV/AIDS Treatment and Prevention
2017. [cited 24 March 2022]. Available from: http://www.thaiaidssociety.org/images/PDF/hiv_thai_guideline_2560.pdf.. *Thailand National Guidelines on HIV/AIDS Treatment and Prevention* (2017.0)
2. Fleishman JA, Yehia BR, Moore RD, Gebo KA. **The economic burden of late entry into medical care for patients with HIV infection.**. *Med Care.* (2010.0) **48** 1071-9. DOI: 10.1097/MLR.0b013e3181f81c4a
3. **Global trends in CD4 cell count at the start of antiretroviral therapy: Collaborative study of treatment programs.**. *Clin Infect Dis* (2018.0) **66** 893-903. DOI: 10.1093/cid/cix915
4. Phuphuakrat A, Kiertiburanakul S, Sungkanuparph S. **Current status of HIV treatment in Asia and the Pacific region.**. *Sex Health.* (2014.0) **11** 119-25. DOI: 10.1071/SH13045
5. Pradubthai Y, Muennuch S, Aumkhyan S, Setthaudom C, Phuphuakrat A, Kiertiburanakul S. **Clinical characteristics and trends of CD4 cell counts of newly diagnosed HIV-infected patients at a university hospital in Thailand**. *J Infect Dis Antimicrob Agents* (2019.0) **36** 13-22
6. Eamsakulrat P, Kiertiburanakul S. **The impact of timing of antiretroviral therapy initiation on retention in care, viral load suppression and mortality in people living with HIV: A study in a university hospital in Thailand.**. *J Int Assoc Provid AIDS Care* (2022.0) **21**. DOI: 10.1177/23259582221082607
7. 7Boonyawairote R, Setthaudom C, Kiertiburanakul S. Clinical characteristics of newly diagnosed individuals with HIV infection in a university hospital. Abstract 3.13. 45th annual meeting of the Infectious Disease Association of Thailand; October 11–14, 2019; Chonburi, Thailand.
8. Hopkins C, Reid M, Gilmour J, Werder S, Briggs S. **Missed opportunities for earlier diagnosis of human immunodeficiency virus infection among adults presenting to Auckland District Health Board hospital services**. *Intern Med J* (2019.0) **49** 495-501. DOI: 10.1111/imj.14073
9. 9British HIV Association, British Association of Sexual Health and HIV, and British Infection Society. UK National Guidelines for HIV Testing 2008. [cited 24 March 2022]. Available from: https://www.bhiva.org/file/RHNUJgIseDaML/GlinesHIVTest08.pdf.
10. 10Australian Government, Department of Health. National HIV Testing Policy 2011. [cited 24 March 2022]. Available from: https://www1.health.gov.au/internet/main/publishing.nsf/Content/ohp-bbvs-hiv-testing-policy.
11. Lhopitallier L, Moulin E, Hugli O, Cavassini M, Darling KEA. **Missed opportunities for HIV testing among patients newly presenting for HIV care at a Swiss university hospital: A retrospective analysis**. *BMJ Open* (2018.0) **8** e019806. DOI: 10.1136/bmjopen-2017-019806
12. **Nutritional anaemias: Report of a WHO scientific group.**. *World Health Organ Tech Rep Ser* (1968.0) **405** 5-37. PMID: 4975372
13. Kiertiburanakul S, Boettiger D, Lee MP, Omar SF, Tanuma J, Ng OT. **Trends of CD4 cell count levels at the initiation of antiretroviral therapy over time and factors associated with late initiation of antiretroviral therapy among Asian HIV-positive patients**. *J Int AIDS Soc* (2014.0) **17** 18804. DOI: 10.7448/IAS.17.1.18804
14. 14UNAIDS. AIDSinfo 2020. [cited 30 March 2022]. Available from: https://aidsinfo.unaids.org/.
15. Sun C, Li J, Liu X, Zhang Z, Qiu T, Hu H. **HIV/AIDS late presentation and its associated factors in China from 2010 to 2020: A systematic review and meta-analysis.**. *AIDS Res Ther.* (2021.0) **18** 96. DOI: 10.1186/s12981-021-00415-2
16. Koh KC, Islam M, Chan WK, Lee WY, Ho YW, Alsagoff SAH. **Missed opportunities for earlier HIV-testing in patients with HIV infection referred to a tertiary hospital, a cross-sectional study.**. *Med J Malaysia* (2017.0) **72** 209-14. PMID: 28889131
17. Jeong SJ, Italiano C, Chaiwarith R, Ng OT, Vanar S, Jiamsakul A. **Late presentation into care of HIV disease and its associated factors in Asia: Results of TAHOD**. *AIDS Res Hum Retroviruses* (2016.0) **32** 255-61. DOI: 10.1089/AID.2015.0058
18. Levy I, Maor Y, Mahroum N, Olmer L, Wieder A, Litchevski V. **Missed opportunities for earlier diagnosis of HIV in patients who presented with advanced HIV disease: A retrospective cohort study**. *BMJ Open* (2016.0) **6** e012721. DOI: 10.1136/bmjopen-2016-012721
19. Burns FM, Johnson AM, Nazroo J, Ainsworth J, Anderson J, Fakoya A. **Missed opportunities for earlier HIV diagnosis within primary and secondary healthcare settings in the UK**. *AIDS* (2008.0) **22** 115-22. DOI: 10.1097/QAD.0b013e3282f1d4b6
20. Nanditha NGA, St-Jean M, Tafessu H, Guillemi SA, Hull MW, Lu M. **Missed opportunities for earlier diagnosis of HIV in British Columbia, Canada: A retrospective cohort study.**. *PLoS One.* (2019.0) **14** e0214012. DOI: 10.1371/journal.pone.0214012
21. Hu X, Liang B, Zhou C, Jiang J, Huang J, Ning C. **HIV late presentation and advanced HIV disease among patients with newly diagnosed HIV/AIDS in Southwestern China: A large-scale cross-sectional study**. *AIDS Res Ther* (2019.0) **16** 6. DOI: 10.1186/s12981-019-0221-7
22. 22World Health Organization. Consolidated guidelines on HIV testing services, 2019. [cited 21 June 2022]. Available from: https://www.who.int/publications/i/item/978-92-4-155058-1.. *Consolidated guidelines on HIV testing services* (2019.0)
23. Branson BM, Handsfield HH, Lampe MA, Janssen RS, Taylor AW, Lyss SB. **Revised recommendations for HIV testing of adults, adolescents, and pregnant women in health-care settings.**. *MMWR Recomm Rep.* (2006.0) **55** 1-17. PMID: 16988643
24. Montoy JC, Dow WH, Kaplan BC. **Patient choice in opt-in, active choice, and opt-out HIV screening: randomized clinical trial**. *BMJ* (2016.0) **532** h6895. DOI: 10.1136/bmj.h6895
25. Kietpeerakool C.. **Human immunodeficiency virus infection in women undergoing treatment for cervical neoplasia: prevalence and the feasibility of routine screening**. *Asian Pac J Cancer Prev* (2008.0) **9** 36-8. PMID: 18439069
|
---
title: Effects of sociodemographic and health factors on the self-management of non-communicable
diseases among Chilean adults during the Covid-19 pandemic
authors:
- Daniela Nicoletti-Rojas
- Rodrigo Retamal
- Ricardo Cerda-Rioseco
- Lorena Rodríguez-Osiac
- Mauricio Fuentes-Alburquenque
- Marcela Araya-Bannout
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021505
doi: 10.1371/journal.pgph.0000763
license: CC BY 4.0
---
# Effects of sociodemographic and health factors on the self-management of non-communicable diseases among Chilean adults during the Covid-19 pandemic
## Abstract
Individuals with non-communicable diseases (NCDs) are potentially at increased vulnerability during the Covid-19 pandemic and require additional help to reduce risk. Self-management is one effective strategy and this study investigated the effect of sociodemographic and health factors on the self-management of some non-communicable diseases, namely hypertension, type 2 diabetes mellitus and dyslipidemia, among Chilean adults during the Covid-19 pandemic. A cross-sectional telephone survey was carried out on 910 participants with NCDs, from Santiago, Chile. An adapted and validated version of the “Partners in Health” scale was used to measure self-management. Exploratory *Factor analysis* yielded five dimensions of this scale: Disease Knowledge, Healthcare Team Relationship, General Self-Management and Daily Routines, Drug Access and Intake, and Monitoring and Decision-Making. The average of these dimensions was calculated to create a new variable Self-Management Mean, which was used as a dependent variable together with the five separate dimensions. Independent variables included age, gender, years of schooling, number of diseases, the percentage of Multidimensional Poverty Index in the commune of residence, and self-rated health status. Beta regressions and ANOVA for the Beta regression residuals were utilized for analyses. Beta regression model explained $8.1\%$ of the variance in Self-Management Mean. Age, years of schooling, number of diseases and self-rated health status were statistically associated with Self-Management Mean and dimensions related to daily routines and health decision making, such as Disease Knowledge, General Self-Management and Daily Routines, and Monitoring and Decision-Making. Gender and the percentage of Multidimensional Poverty Index in the commune of residence were insignificant. Strategies for self-management of NCDs during a crisis should consider age, years of schooling, number of diseases, and self-rated health status in their design.
## Introduction
Currently global health is characterized by a high prevalence of non-communicable diseases [1] and the emergence of new infectious diseases such as Covid-19 with the resulting global pandemic [2]. Preventing and controlling infectious diseases, nutritional problems and associated chronic diseases require pertinent planning and strategies [3]. Individuals living with non-communicable diseases (NCDs) [4] are particularly vulnerable as a result of the pandemic. Covid-19 mortality rates reported worldwide have shown higher rates of morbidity, hospitalization and mortality in particularly for Type 2 diabetes and arterial hypertension, as well as cancer [5–7]. In addition, measures used to reduce the spread of Covid-19 including social distancing, mobility restrictions and lockdowns may impact on people with NCDs by making it more difficult to access health services and to follow prescribed diets or perform physical activity, as well as increasing smoking and excessive alcohol consumption [8]. In Chile, the 10th Report of Covid-19 Symptoms and Practices National Monitoring (in Spanish, Monitoreo Nacional de Síntomas y Prácticas Covid-19, or MOVID-19) [9], showed that only $26.1\%$ those with NCDs had access to healthcare services due to the lockdown and the reduction in routine health services. The lack of programmatic alignments for managing the NCD population during the Covid-19 pandemic may obstruct the fight against both pandemics, especially in low and middle income countries, and therefore, the challenge to keep healthcare continuity for chronic disease is relevant [8].
In Chile the primary healthcare system aims at promoting healthy lifestyle habits and capacities for individuals and communities. All risk factors related to lifestyle habits, such as diet, physical activity, and smoking, among others, constitute an important part of the burden of disease in Chile. The challenge of the primary healthcare system is to achieve changes in the behavior of individuals to improve lifestyle habits, to develop protecting factors, and to decrease the prevalence of risk factors. The lockdown has impacted on various primary healthcare programs [10], such as the Cardiovascular Health Program (in Spanish, Programa de Salud Cardiovascular, PSCV onwards), which aims to prevent and reduce morbidity, disability and premature mortality caused by cardiovascular diseases (arterial hypertension, type 2 diabetes mellitus and dyslipidemia) [11]. Implementation of strategies to transfer knowledge and skills to those with NCDs so that they can manage their diseases, improve health and quality of life, and prevent saturation of the healthcare system is important [12, 13].
One strategy that has shown some effectiveness in controlling NCDs is self-management [14] which is defined as a set of knowledge, attitudes, skills and behaviors that individuals develop to cope with their experience including identification and management of symptoms, treatment adherence, prevention of physical, social and emotional consequences, as well as adapting to these consequences [15–17]. Self-management is developed with the support of healthcare professionals, the healthcare system and other sectors [18, 19].
Self-management is a complex phenomenon determined by multiple factors (medical, biological, sociodemographic, and psychological, among others) [20–22], which may undermine or enhance self-management. To develop and promote self-management, it is necessary to know what factors influence self-management in the NCD population in the context of the Covid-19 pandemic, as well as a reformulation of the tasks and roles of healthcare teams. The Chilean primary healthcare system is a well-developed public system to support healthcare in an individual and community level [23]. It is possible that the knowledge of the effect of selected sociodemographic and health factors on self-management may help to Chilean and other countries to adjust and re-design programs aimed to provide support for self-management to the NCDs population during sanitary crisis such as the Covid-19 pandemic. Additionally, it may help to other countries that have primary healthcare systems to develop more resilient programs to the NCDs population during sanitary crisis. The objective of the present study is to investigate the effect of selected sociodemographic and health factors on NCD self-management.
## Study design
The sample size calculation for proportions indicated a minimum sample size of 666 participants from the users of the PSCV in the Metropolitan Region ($$n = 711$$,620), based on alpha of $5\%$, beta of $80\%$ and $99\%$ confidence interval (Fig 1) and oversampling by $30\%$ to yield a minimum sample of 900 individuals. Due to the Covid-19 lockdown restrictions data collection was performed via a telephone survey.
**Fig 1:** *Sample flow chart.*
Participants in the study were selected using a three-staged probabilistic design. In the first stage, the 32 communes that comprise the city of Santiago were divided in three tertiles utilizing the percentage of the Multidimensional Poverty Index (%MPI) of each commune. Three communes were randomly selected from each tertile, giving a total of 9 communes. In the second stage, two Family Health Care Centers (Centro de Salud Familiar or CESFAM) were randomly selected from each of these communes. In the third stage, individuals were selected from each CESFAM through a stratified random sampling by age and gender.
The inclusion criteria were people enrolled in the PSCV, to be older than 20 years of age, with access to a telephone or to be able to connect via video call. Exclusion criteria considered people that had difficulties to answer a phone call or to respond to the survey appropriately, such as participants with hearing, cognitive or mental health disabilities.
## Data collection
Due to the lockdown the survey was carried out by telephone. The decision of carrying out the survey by telephone was based on the fact that this was the unique source of contact to the participants. Surveys carried out by telephone have proven better response rates compared to internet surveys [24, 25]. In order to allow for non-response a total of 7,282 telephone numbers (landline and mobile) were obtained from the selected CESFAM. The survey was carried out between October 2020 and January 2021 (14 weeks). The phone numbers were divided between nine experienced and previously trained telephone operators, who made an average of 719 calls per week, averaging 222 effective calls per week. Informed consent was obtained before the start of the survey. The average call duration was 37 minutes (SD = 14 minutes) with a minimum duration of 10 minutes and a maximum duration of 112 minutes. Each call was recorded. Ineffective calls corresponded to either wrong or invalid numbers. The total number of effective calls gave a weekly average of 66 responses, the rest corresponded to participants that fell under exclusion criteria or declined to participate. Participants who refused to participate were not different in terms of gender, age, and commune of residence.
The survey obtained data from 922 individuals (477 women and 433 men). A total of twelve participants were excluded (Fig 1) due to non-compliance with the inclusion and exclusion criteria, or for having incomplete responses. Therefore, the final analysis was performed on a total sample of 910 participants.
## Variables
The study utilized the Self-management mean (SMM) and its dimensions as outcomes, while sociodemographic and health self-perception variables were used as independent variables. SMM was built from the “Chronic disease self-management in Covid-19 pandemic scale for people enrolled in the Cardiovascular Health Program”, which was designed based on the “Partners in Health (PIH)” scale developed in 2003 at the Flinders Human Behavior & Health Research Unit at the University of Flinders [26]. The original instrument is an 11-item scale comprising three dimensions: core self-management, condition knowledge and symptom monitoring, and it was designed to be self-administered. All the questions were adapted for telephone usage and placed in the context of the pandemic and the PSCV. The original score for each question was from 0 to 8 and it was modified from 0 to $100\%$ in $10\%$ intervals. The score was inverted, where 0 indicates absence of self-management and 100 indicates maximum possible presence of self-management. These modifications were made following the recommendations of Bandura [27] to measure self-efficacy, taking into consideration that self-efficacy and self-management are strongly related [28] and that in many cases the scales of self-efficacy are used to evaluate self-management [29].
New questions were included aiming to obtain information concerning support networks, including family, community and health teams making a 17 item scale. Finally, the scale contained 17 items (S1 Text).
The scale was submitted to a process of content validation. Six cognitive interviews [30] were performed, with the objective of assessing comprehension and sociocultural pertinence. The interviews were performed on three men and three women, of ages and diagnosis equivalent to the population to which the survey was applied. All participants showed understanding of the statements and the response scale.
The scale was also submitted to a psychometric properties analysis, utilizing an exploratory factor analysis. A total of two items were removed from the original scale, as they showed low commonality (items 8 and 16, S1 Text), which resulted in the final 15-items instrument (Cronbach’s α = 0.836). The exploratory factor analysis provided five dimensions: Disease Knowledge (DK), that refers to the self-reported knowledge beliefs that participants have regarding their condition in general, including causes, effects and relation to Covid-19; Healthcare Team Relationship (HTR), that considers the participant’s engagement in the decision-making process pertaining to treatment and perceived support from the healthcare team; General Self-Management and Daily Routines (SMDR), which refers to everyday management of actions related to treatment, such as diet, physical activity and emotional management; Drug Access and Intake (DAI), related to the possibility of accessing required medication for treatment and correct intake; and Monitoring and Decision-Making (MDM); related to the capacity of participants to self-assess disease signs and symptoms, and to make decisions based on these assessments. Finally, SMM was created using the mean of the five dimensions.
The sociodemographic factors used as independent variables were gender (man and woman), age (subtracting the date of birth to the date of the interview), the %MPI of the commune of residence, and the self-reported years of schooling categorized into five groups (incomplete primary education [< 8 years], incomplete secondary education [< 12 years], complete secondary education [≥ 12 years], complete technical education or incomplete higher education, and complete higher education). The health factors were the self-rated health status (divided in four categories: poor, fair, good and excellent) and the number of diseases that was calculated by adding the number of self-declared diseases attended by the PSCV: diabetes, arterial hypertension, and dyslipidemia. Other diseases such as neoplasia, blood disorders, endocrine diseases, mental or neurologic diseases, digestive diseases, respiratory diseases (not related to Covid-19), musculoskeletal disorders, genitourinary diseases, hearing disorders, infectious diseases (not related to Covid-19), immune diseases, ophthalmologic diseases and disorders, cardiovascular diseases, and metabolic diseases were also included.
## Data analysis
Beta regressions were used to determine the effect of each independent variable on the variability of the dependent variables. Beta regressions have been shown to give better data adjustment when dependent variables present asymmetrical, outlier values, given their higher flexibility when compared to standard linear regressions [31]. In addition, evidence indicates that beta distribution-based analysis is recommended in continuous answer formats, where responses must be marked on a continuous range, with no semantic references, as is the case of this survey [32–34]. Shapiro-Wilk tests performed on residuals from linear and beta regressions using data from this study rejected normal distribution. However, the visualization of residual values (q-q plots, not presented in this paper) and Akaike Information Criteria values (AIC beta regression model = -736.265; AIC linear regression model = -700.099) showed better adjustments in beta regressions compared to linear regressions. Analysis of variance (ANOVA) was performed using the residual values obtained from beta regressions with the aim of controlling for the effect of the other independent variables. Finally, Bonferroni correction was applied to correct for the number of statistical tests undertaken, since six beta regression models were performed using the same dependent variables (p-value<0.007).
## Ethics statement
The Project was approved by the ethical committee of the Faculty of Medicine, University of Chile (Project: Nº 100–2020). Verbal informed consent was obtained from all participants.
## Results
Table 1 provides the descriptive statistics and Fig 2 shows histograms and box-and-whiskers charts for the SMM and the dimensions. As can be seen from Table 1 and Fig 2, the general pattern of SMM and the dimensions showed negative skewness, as well as means and medians over 0.6. As can be seen from the interquartile ranges, MDM, HTR and DK dimensions showed high dispersion, while DAI, SMM, and SMDR dimensions showed low dispersion. The DAI dimension was worth highlighting since it presented a median close to the upper limit (>0.999), pronounced asymmetry and low variability in comparison to other dimensions.
**Fig 2:** *Histogram and box-plot of self-management mean and its dimensions.* TABLE_PLACEHOLDER:Table 1 Table 2 presents descriptive statistics for the independent variables. It was found that women were more likely to participate. The average age was 57.86 years, with a standard deviation of 18.17 years. Minimum and maximum ages were 21 years and 101 years respectively. The majority of participants declared had incomplete secondary education, two diseases, and good health. The %MPI in the commune of residence showed an average of $20.87\%$, with a standard deviation of $0.06\%$. The commune with the lowest %MPI was Santiago and the commune with the highest participation was Recoleta.
**Table 2**
| Variables | Categories | Frequency | Percent |
| --- | --- | --- | --- |
| Gender | Female | 477 | 52.42 |
| Gender | Male | 433 | 47.58 |
| Age (years) | 20–39 | 223 | 24.51 |
| Age (years) | 40–69 | 420 | 46.15 |
| Age (years) | 70+ | 267 | 29.34 |
| Years of schooling | Primary incomplete | 124 | 13.63 |
| Years of schooling | High school incomplete | 221 | 24.29 |
| Years of schooling | High school complete | 249 | 27.36 |
| Years of schooling | Bachelor incomplete / technical complete | 213 | 23.41 |
| Years of schooling | Bachelor complete | 103 | 11.32 |
| Number of diseases | 1 | 200 | 21.98 |
| Number of diseases | 2 | 281 | 30.88 |
| Number of diseases | 3 | 212 | 23.3 |
| Number of diseases | 4 | 118 | 12.97 |
| Number of diseases | 5 or more | 99 | 10.88 |
| Health status | Poor | 72 | 7.91 |
| Health status | Fair | 344 | 37.8 |
| Health status | Good | 412 | 45.27 |
| Health status | Very good | 82 | 9.01 |
| % Multidimensional Poverty Index by commune | Santiago (9.63%) | 117 | 12.86 |
| % Multidimensional Poverty Index by commune | Macul (13.47%) | 84 | 9.23 |
| % Multidimensional Poverty Index by commune | La Cisterna (17.82%) | 100 | 10.99 |
| % Multidimensional Poverty Index by commune | Recoleta (22.5%) | 125 | 13.74 |
| % Multidimensional Poverty Index by commune | Pudahuel (22.51%) | 124 | 13.63 |
| % Multidimensional Poverty Index by commune | Puente Alto (23.31%) | 107 | 11.76 |
| % Multidimensional Poverty Index by commune | Peñalolén (26.28%) | 104 | 11.43 |
| % Multidimensional Poverty Index by commune | Pedro Aguirre Cerda (26.76%) | 95 | 10.44 |
| % Multidimensional Poverty Index by commune | Conchalí (29.37%) | 54 | 5.93 |
Table 3 presents the results of the beta regressions of SMM and the dimensions as dependent variables and their association with the independent variables. Pseudo R-squared showed that the independent variables explained approximately $8.10\%$ of the SMM variation, while the SMDR dimension showed the highest pseudo R-squared value of $13.0\%$. The variables that were significantly associated with SMM were age, years of schooling, number of diseases, and health status. The DK dimension was significantly associated with gender, years of schooling, number of diseases, and very good health status. The SMDR dimension was significantly associated with age, higher complete education, and health status. The HTR dimension was only significantly associated with very good health status, while the MDM dimension was significantly associated with age and number of diseases. The DAI dimension was not significantly associated with any of the independent variables.
**Table 3**
| Variable | Self-Management Mean | Self-Management Mean.1 | Self-Management Mean.2 | Self-Management Mean.3 | Disease Knowledge | Disease Knowledge.1 | Disease Knowledge.2 | Disease Knowledge.3 | Healthcare Team Relationship | Healthcare Team Relationship.1 | Healthcare Team Relationship.2 | Healthcare Team Relationship.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Variable | Estimate | S.E. | Z | P-value | Estimate | S.E. | Z | P-value | Estimate | S.E. | Z | P-value |
| Intercept | -0.112 | 0.174 | -0.644 | n.s. | 0.620 | 0.237 | 2.616 | 0.009 | 0.197 | 0.274 | 0.72 | n.s. |
| Gender | -0.012 | 0.052 | -0.234 | n.s. | -0.191 | 0.071 | -2.707 | 0.007 | -0.166 | 0.082 | -2.03 | n.s. |
| Age | 0.008 | 0.002 | 4.947 | <0.001 | -0.004 | 0.002 | -1.700 | n.s. | 0.005 | 0.002 | 2.117 | n.s. |
| Years of schooling (<8 years) | -0.005 | 0.085 | -0.055 | n.s. | -0.243 | 0.115 | -2.105 | n.s. | 0.023 | 0.134 | 0.171 | n.s. |
| Years of schooling (12 years) | 0.221 | 0.072 | 3.075 | 0.002 | 0.345 | 0.098 | 3.532 | <0.001 | 0.033 | 0.113 | 0.289 | n.s. |
| Years of schooling (technical complete / higher incomplete) | 0.166 | 0.076 | 2.188 | n.s. | 0.446 | 0.104 | 4.301 | <0.001 | -0.047 | 0.12 | -0.396 | n.s. |
| Years of schooling (higher complete) | 0.390 | 0.096 | 4.059 | <0.001 | 0.346 | 0.128 | 2.697 | 0.007 | 0.316 | 0.148 | 2.137 | n.s. |
| Number of diseases | 0.067 | 0.022 | 3.046 | 0.002 | 0.110 | 0.030 | 3.706 | <0.001 | 0.044 | 0.034 | 1.304 | n.s. |
| Health status (bad) | -0.205 | 0.097 | -2.112 | n.s. | -0.033 | 0.134 | -0.248 | n.s. | -0.166 | 0.155 | -1.069 | n.s. |
| Health status (good) | 0.038 | 0.056 | 0.670 | n.s. | 0.033 | 0.077 | 0.435 | n.s. | -0.063 | 0.089 | -0.708 | n.s. |
| Health status (very good) | 0.510 | 0.099 | 5.129 | <0.001 | 0.428 | 0.130 | 3.284 | 0.001 | 0.569 | 0.148 | 3.85 | <0.001 |
| % Multidimensional Poverty Index by commune | 0.374 | 0.439 | 0.851 | n.s. | -0.789 | 0.595 | -1.326 | n.s. | 0.712 | 0.687 | 1.037 | n.s. |
| Phi | 6.693 | 0.297 | 22.51 | <0.001 | 2.671 | 0.111 | 24.05 | <0.001 | 1.438 | 0.057 | 25.15 | <0.001 |
| Pseudo R-squared | 0.081 | 0.081 | 0.081 | 0.081 | 0.069 | 0.069 | 0.069 | 0.069 | 0.035 | 0.035 | 0.035 | 0.035 |
| | General Self-Management and Daily Routines | General Self-Management and Daily Routines | General Self-Management and Daily Routines | General Self-Management and Daily Routines | Drug Access and Intake | Drug Access and Intake | Drug Access and Intake | Drug Access and Intake | Monitoring and Decision-Making | Monitoring and Decision-Making | Monitoring and Decision-Making | Monitoring and Decision-Making |
| | Estimate | S.E. | Z | P-value | Estimate | S.E. | Z | P-value | Estimate | S.E. | Z | P-value |
| Intercept | -0.702 | 0.21 | -3.338 | 0.001 | 1.133 | 0.263 | 4.315 | <0.001 | -1.070 | 0.293 | -3.646 | <0.001 |
| Gender | 0.007 | 0.063 | 0.114 | n.s. | 0.006 | 0.077 | 0.074 | n.s. | 0.047 | 0.087 | 0.538 | n.s. |
| Age | 0.019 | 0.002 | 9.918 | <0.001 | 0.005 | 0.002 | 2.039 | n.s. | 0.010 | 0.003 | 3.768 | <0.001 |
| Years of schooling (<8 years) | 0.058 | 0.103 | 0.557 | n.s. | 0.021 | 0.127 | 0.164 | n.s. | -0.030 | 0.144 | -0.208 | n.s. |
| Years of schooling (12 years) | 0.220 | 0.087 | 2.535 | n.s. | 0.152 | 0.107 | 1.417 | n.s. | 0.251 | 0.121 | 2.072 | n.s. |
| Years of schooling (technical complete / higher incomplete) | 0.095 | 0.092 | 1.039 | n.s. | 0.058 | 0.114 | 0.505 | n.s. | 0.184 | 0.128 | 1.436 | n.s. |
| Years of schooling (higher complete) | 0.315 | 0.115 | 2.749 | 0.006 | 0.191 | 0.140 | 1.361 | n.s. | 0.417 | 0.158 | 2.633 | n.s. |
| Number of diseases | -0.030 | 0.026 | -1.142 | n.s. | 0.064 | 0.032 | 1.973 | n.s. | 0.100 | 0.037 | 2.730 | 0.006 |
| Health status (bad) | -0.348 | 0.116 | -2.991 | 0.003 | -0.221 | 0.148 | -1.492 | n.s. | -0.126 | 0.166 | -0.755 | n.s. |
| Health status (good) | 0.255 | 0.068 | 3.749 | <0.001 | 0.064 | 0.084 | 0.762 | n.s. | 0.042 | 0.095 | 0.440 | n.s. |
| Health status (very good) | 0.756 | 0.119 | 6.368 | <0.001 | 0.224 | 0.141 | 1.586 | n.s. | 0.417 | 0.159 | 2.630 | n.s. |
| % Multidimensional Poverty Index by commune | 0.425 | 0.528 | 0.806 | n.s. | -0.020 | 0.653 | -0.031 | n.s. | 1.520 | 0.736 | 2.063 | n.s. |
| Phi | 4.026 | 0.175 | 23.010 | <0.001 | 1.428 | 0.076 | 18.76 | <0.001 | 0.955 | 0.036 | 26.620 | <0.001 |
| Pseudo R-squared | 0.130 | 0.130 | 0.130 | 0.130 | 0.035 | 0.035 | 0.035 | 0.035 | 0.042 | 0.042 | 0.042 | 0.042 |
Table 4 presents the results of sequential ANOVA using the residual values obtained from the beta regressions, with the aim of controlling for the other independent variables. The association between SMM and the independent variables remained similar, while DK was significantly associated with years of schooling and number of diseases, SMDR was significantly associated with age and health status, DAI was significantly associated with health status, and MDM was significantly associated with age. HTR showed no association with the independent variables. Eta squared showed small effect sizes of each independent variable on SMM and its dimensions.
**Table 4**
| Variable | Self-Management Mean | Self-Management Mean.1 | Self-Management Mean.2 | Disease Knowledge | Disease Knowledge.1 | Disease Knowledge.2 | Healthcare Team Relationship | Healthcare Team Relationship.1 | Healthcare Team Relationship.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Variable | F value | P-value | Eta-squared | F value | P-value | Eta-squared | F value | P-value | Eta-squared |
| Gender | 0.036 | n.s. | <0.001 | 5.099 | n.s. | 0.006 | 3.148 | n.s. | 0.003 |
| Age | 15.790 | <0.001 | 0.017 | 2.055 | n.s. | 0.002 | 2.689 | n.s. | 0.003 |
| Years of schooling | 11.040 | 0.001 | 0.012 | 17.070 | <0.001 | 0.018 | 0.905 | n.s. | 0.001 |
| Number of diseases | 7.607 | 0.006 | 0.008 | 8.672 | 0.003 | 0.009 | 1.169 | n.s. | 0.001 |
| Health status | 19.960 | <0.001 | 0.022 | 4.222 | n.s. | 0.005 | 5.900 | n.s. | 0.006 |
| % Multidimensional Poverty Index by commune | 0.154 | n.s. | <0.001 | 1.765 | n.s. | 0.002 | 0.371 | n.s. | <0.001 |
| | General Self-Management and Daily Routines | General Self-Management and Daily Routines | General Self-Management and Daily Routines | Drug Access and Intake | Drug Access and Intake | Drug Access and Intake | Monitoring and Decision-Making | Monitoring and Decision-Making | Monitoring and Decision-Making |
| | F value | P-value | Eta-squared | F value | P-value | Eta-squared | F value | P-value | Eta-squared |
| Gender | 0.011 | n.s. | <0.001 | 0.010 | n.s. | <0.001 | 0.247 | n.s. | <0.001 |
| Age | 52.930 | <0.001 | 0.055 | 5.511 | n.s. | 0.006 | 9.934 | 0.002 | 0.011 |
| Years of schooling | 2.588 | n.s. | 0.003 | 1.805 | n.s. | 0.002 | 4.877 | n.s. | 0.005 |
| Number of diseases | 0.662 | n.s. | 0.001 | 6.240 | n.s. | 0.007 | 5.848 | n.s. | 0.006 |
| Health status | 40.580 | <0.001 | 0.043 | 8.693 | 0.003 | 0.010 | 4.583 | n.s. | 0.005 |
| % Multidimensional Poverty Index by commune | 0.041 | n.s. | <0.001 | 0.002 | n.s. | <0.001 | 2.925 | n.s. | 0.003 |
## Discussion
Self-management is determined by multiple variables at various levels [20–22] and this study examined how of variation in sociodemographic and health factors impacted on self-management.
This study used a modified PIH scale to monitor how participants, during the Covid-19 pandemic, expressed their abilities to perform tasks by themselves and what they received from their support networks and healthcare systems. Although the SMM was observed to have a high score, its dimensions showed heterogeneity. For example, the DAI dimension showed that actions conducted with institutional support mechanisms established prior to the Covid-19 pandemic, presented higher scores and low variability. On the contrary, SMDR which includes physical activity and diet behaviors that depend on people and their environment presented lower scores and were more variable. These results can be explained by the difficulty in performing physical activity and diet-related behaviors compared to medication intake.
The SMM model accounted for $8.1\%$ of the variation of the self-management construct, while the independent variables explained between $3.5\%$ to $13\%$ of the studied dimensions indicating that there is an important amount of unexplained variation. Other studies [35–37] found that the nature of social support, self-efficacy, cultural values, self-awareness, empowerment, perceived barriers, health literacy, among others, improved self-management, thereby accounting for the higher percentage of variation explained in these studies.
In the present study, age, years of schooling level, number of diseases and self-rated health status were significantly associated with SMM, DK, SMDR, and MDM. These variables have been used in other research [38–45] aiming to understand the short-term modifiable effects on self-management.
Self-rated health status was statistically associated with SMM and the SMDR dimension. Additionally, after removing the effect of the other independent variables, self-rated health status presented a significant association with the DAI dimension. These results are consistent with previous studies, which found that self-rated health status is a variable significantly associated with other health indicators. Smith et al. [ 43] demonstrated that self-rated health was significantly associated with self-management measured with the PIH scale. The authors concluded that the probability of having high self-management is greater if the participant has a good or excellent self-rated health status. The significant association that exists between these two variables can be explained as a positive reciprocal influence, where better disease self-management generates improved self-rated health status, reinforcing self-management behaviors.
Years of schooling showed a significant association with SMM, as well as with the DK, SMDR, HTR, and MDM dimensions with the higher education category showing the largest contribution to higher self-management. This could be explained by a greater capacity of the population to understand messages, information, and in-depth reflection regarding prescriptions provided by healthcare teams. During the Covid-19 pandemic, this skill appears as an accumulated history that should be taken into account by healthcare teams when providing care to users. Heijmans et al. [ 38] found in a Dutch study that education level significantly influences healthcare literacy and consequently, self-management. Heijmans et al. [ 38] argued that communication, functional and critical thinking skills in the population that access healthcare services are a relevant resource for the structural foundations of a healthcare system. Kim et al. [ 46] proposed that it is important to consider the broad scope and skill-depth that participants can develop for self-management, the paths to search and obtain information, and how health information can be applied in their daily lives, which requires knowledge, motivational skills, access to official information and the capacity to develop strategies to implement sustainable changes.
Age showed significant and positive associations with SMM, SMDR and MDM. This suggests that the accumulation of advice and experience, routines and resilience acquired from the process of living with a NCD leads to improvements in disease management, which are expressed in concrete actions taken during Covid-19 lockdowns, including diet, physical activity and emotional management. The significant association between age and the MDM dimension reveals that age is positively correlated with greater performance in monitoring signs and symptoms and making decisions based on these assessments. These results are similar to those reported by Heijmans et al. [ 38], who also using the PIH scale, observed a positive association between age and self-management.
The number of diseases presented a positive association with SMM and the DK dimensions. Participants with a higher number of diseases have greater self-management and disease knowledge. Other studies [13, 38, 47] have observed that the number of diseases has a significant inverse effect on self-management. Fix et al. [ 47] observed that patients with comorbidities perceive an interdependence between their conditions and subsequently had problems separating information between their illnesses. However, the majority of conditions shared by participants in this study have a common set of characteristics (origins, indications, consequences and manner to be addressed by the healthcare system). Consequently, information and indications received for one disease can function as reinforcement for others leading to greater levels of knowledge and self-management. The results obtained from this study and previous literature [13, 38, 47] suggest that future research should consider the total number of diseases, as well as groups of diseases, in order to know the differential effect on self-management and its dimensions.
In this study self-reported gender was not significantly associated with SMM or its dimensions, which suggests that self-perception of practices, knowledge and experiences learned from living with a condition are not mediated by gender perception [48]. The effect of gender on NCD self-management is not consistent and some studies have observed significant associations [39, 41, 49, 50] while others have not [43, 51, 52], or an association that is exclusive to specific dimensions [38]. An explanation for these disparities in findings could be attributed to different cultural expectations associated with gender roles in the countries where studies were conducted. For instance, Alrahbi [49] observed that Omani men performed physical activity more regularly than women, due to social restrictions imposed on women. Hara et al. [ 50] observed a worsening of psychological symptoms associated with diabetes in women, which suggests that gender has a major impact on patient empowerment and that female patients may face more psychological or family communication issues. Both studies aim at social and cultural differences that are involved in NCD self-management. The different ways that men and women address their own health has been acknowledged and studied [43] and WHO [53] acknowledges the effect of gender on general health. Health practices and behaviors can be understood as activities built in association with gender roles [54, 55]. One possible explanation of the present result is that the PSCV program provides homogenous healthcare to men and women alike, without differentiating care in accordance to gender characteristics, and as a possible result, the response that is received is also homogenous.
There was no association between self-management and the extent of commune poverty in the present study. This lack of association could be due to the low variation in % poverty between communes and that household income might have been a better predictor. For example, the National Socioeconomic Characterization Survey (Encuesta de Caracterización Socioeconómica Nacional) 2017 [56] found a near two-fold increase in poverty and extreme poverty levels by household income, in the Región Metropolitana [57].
Finally, the HTR dimension did not show significant association with any independent variables. There is evidence that indicates that the relationship between healthcare teams and participants are determined by variables such as personality traits, communication skills and certain common characteristics [58, 59], which were not assessed in our study.
Our study has several limitations. Data acquired through a telephone survey, may affect the results compared to face-to-face survey. However, the limitations imposed by the lock-down due to the Covid-19 pandemic, prevented any type of in-person meetings. Another limitation is the cross-sectional nature of this research. The research was carried out between the first and the second waves of Covid-19 in Chile and it is not known whether self-management changed during this time due to the changing access to healthcare or support strategies. Additionally, this research did not consider other types of independent variables, such as other social and individual variables, cultural variables, or psychological conditions, among others. However, the purpose of this research was to focus understanding on the effects of selected sociodemographic and health factors in order to know what kind of social and individual barriers or enhancers have to be considered to implement strategies to improve the self-management. Finally, it is possible that some bias may have occurred due to refusal of some participants to take part in the research. While gender, age, and commune of residence were controlled by the sample design, other variables such as SMM and its dimensions, as well as years of schooling, number of diseases, and health status were not controlled, showing eventual differences between the participants and those who refused to participate.
## Conclusions
Self-management is a complex phenomenon determined by a large number of variables of diverse nature, which interact with each other.
The Covid-19 pandemic interrupted the regular health care system and this situation provided an opportunity to study the role that the health systems has in relationship to self-management in the context of NCDs.
This study showed that when the health system develops and consolidate practices focused on self-management, such as drug access, there is stability. On the contrary, other individual dependent dimensions that have not been reinforced before the pandemic, showed more differential responses and the PSCV needs to recognize its strengths and weaknesses in order to improve self-management as a crucial factor in the Chronic Care Model [57] adopted by the PSCV.
At the global level, this study shows some directions to improve the quality of healthcare systems [58] in countries experiencing crises. Strategies linked to self-management support should focus on providing structural and community mechanisms that allow strengthening self-management in the population before, during and after a pandemic or in future health, social or natural disaster crises, taking into consideration sociodemographic and health factors that modulate the expression of self-management.
## References
1. 1PAHO Noncommunicable diseases. Regional office for the Americas WHO [Internet]. https://www.paho.org/es/temas/enfermedades-no-transmisibles.
2. Yang C, Jin Z. **An Acute Respiratory Infection runs into the most common noncommunicable Epidemic- COVID-19 and Cardiovascular diseases**. *JAMA Cardiol* (2020.0) **5** 743-744. DOI: 10.1001/jamacardio.2020.0934
3. 3World Health Organization. Innovative Care for Chronic Conditions: Building Blocks for Action [Internet]. 2002. https://apps.who.int/iris/handle/10665/42500.
4. 4World Health Organization Noncommunicable diseases [Internet]. 2021. https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases#:~:text=Noncommunicable%20diseases%20(NCDs)%2C%20also,physiological%2C%20environmental%20and%20behavioural%20factors.
5. Clark A, Jit M, Warren- Gash C, Guthrie B, Wang HH, Mercer SW. **Global, regional and national estimates of the population at increased risk of severe COVID- 19 due to underlying conditions in 2020: a modelling study**. *Lancet Glob Health* (2020.0) **8** e1003-e1017. DOI: 10.1016/S2214-109X(20)30264-3
6. Clerkin K, Fried J, Raikhelkar J, Sayer G, Griffin J, Masoumi A. **COVID- 19 and Cardiovascular Disease**. *Circulation* (2020.0) **141** 1648-1655. DOI: 10.1161/CIRCULATIONAHA.120.046941
7. Bakouny Z, Hawley JE, Choueiri TK, Peters S, Rini BI, Warner JL. **COVID-19 and Cancer: Current Challenges and Perspectives**. *Cancer Cell* (2020.0) **38** 629-646. DOI: 10.1016/j.ccell.2020.09.018
8. Kluge HHP, Wickramasinghe K, Rippin HL, Mendes R, Peters DH, Kontsevaya A. **Prevention and control of non-communicable diseases in the COVID- 19 response**. *The Lancet* (2020.0) **395** 1678-1680. DOI: 10.1016/S0140-6736(20)31067-9
9. 9Medical College of Chile, Universidad Diego Portales, Pontificia Universidad Católica de Chile, Universidad San Sebastián, Universidad Central, Universidad de la Frontera. MCOVID-19 ¿Cuál ha sido el impacto de la pandemia en el acceso a atenciones de salud? Un análisis para la adaptación de nuestro sistema de salud. Santiago, Chile. [Internet] 2020 Oct 5. https://www.movid19.cl/publicaciones/decimo-informe/.
10. 10Latorre R. Healthcare resorts to Primary Care after hospital collapse and enables “prolonged observation” beds in clinics [Internet] La Tercera. 2021 Jun 9. https://www.latercera.com/earlyaccess/noticia/salud-recurre-a-la-atencion-primaria-tras-colapso-hospitalario-y-habilita-camas-de-observacion-prolongada-en-consultorios/KQ2FSNM2P5CIBJH25MKE4RLEFE/.
11. 11Ministerio de Salud, Gobierno de Chile. Orientación Técnica Programa de Salud Cardiovascular. [Internet] 2017. http://familiarycomunitaria.cl/FyC/wp-content/uploads/2018/05/Programa-de-salud-cardiovascular.-MINSAL-Chile-2017.pdf.
12. Kastner M, Hayden L, Wong G, Lai Y, Makarski J, Treister V. **Underlying mechanisms of complex interventions addressing the care of older adults with multimorbidity: a realist review**. *BMJ Open* (2019.0) **9** e025009. DOI: 10.1136/bmjopen-2018-025009
13. May CR, Eton DT, Boehmer K, Gallacher K, Hunt K, MacDonald S. **Rethinking the patient: using Burden of Treatment Theory to understand the changing dynamics of illness**. *Bmc Health Serv Res* (2014.0) **14** 281. DOI: 10.1186/1472-6963-14-281
14. Hardman R, Begg S, Spelten E. **What impact do chronic disease self-management support interventions have on health inequity gaps related to socioeconomic status: a systematic review**. *BMC Health Serv Res* (2020.0) **20** 150. DOI: 10.1186/s12913-020-5010-4
15. Corbin J, Strauss A. **Managing Chronic Illness at home: three lines of work**. *Qual Sociol* (1985.0) **8** 224-247
16. Coleman MT, Newton KS. **Supporting self-management in patients with chronic illness**. *Am Fam Physician* (2005.0) **72** 1503-10. PMID: 16273817
17. Pearce G, Parke HL, Pinnock H, Epiphaniou E, Bourne CL, Sheikh A. **The PRISMS taxonomy of self-management support: derivation of a novel taxonomy and initial testing of its utility**. *J Health Serv Res Policy* (2016.0) **21** 73-82. DOI: 10.1177/1355819615602725
18. 18World Health Organization. Preparation of Health Care Professionals for the 21st Century. The Challenge of Chronic Diseases, World Health Organization. 2005. https://apps.who.int/iris/handle/10665/43044
19. Bonal RR, López VN, Vargas P, Meoño MT, Brañas CRW. **Support to Chronic Conditions Self-management: a Challenge of Health Systems in Latin America**. *Finlay* (2017.0) **7** 268-277
20. Ochieng JM, Crist JD. **Social Determinants of Health and Health Care Delivery: African American Women’s T2DM Self-Management**. *Clin Nurs Res* (2021.0) **30** 263-272. DOI: 10.1177/1054773820916981
21. Hyman I, Shakya Y, Jembere N, Gucciardi E, Vissandjée B. **Provider- and patient-related determinants of diabetes self-management among recent immigrants: Implications for systemic change**. *Can Fam Physician* (2017.0) **63** e137-e144. PMID: 28209706
22. Khalesi S, Irwin C, Sun J. **Lifestyle and self-management determinants of hypertension control in a sample of Australian adults**. *Expert Rev Cardiovasc Ther* (2018.0) **16** 229-236. DOI: 10.1080/14779072.2018.1435272
23. 23OECD. OECD Reviews of Public Health: Chile: A Healthier Tomorrow. 2019. OECD Publishing, Paris.
24. Potoglou D, Kanaroglou PS, Robinson N. **Evidence on the Comparison of Telephone and Internet Surveys for Respondent Recruitment**. *The Open Transportation Journal* (2012.0) **6** 11-22
25. Manfreda KL, Bosnjak M, Berzelak J, Haas I, Vehovar V. **Web Surveys versus other Survey Modes: A Meta-Analysis Comparing Response Rates**. *International Journal of Market Research* (2008.0) **50** 79-104. DOI: 10.1177/147078530805000107
26. Battersby W, Ask A, Reece M, Markwick J, Collins P. **The Partners in Health scale: The development and psychometric properties of a generic assessment scale for chronic condition self-management**. *Australian Journal of Primary Health* (2003.0) **9** 41-52
27. Bandura A., Pajares F, Urdan T. *Self-efficacy beliefs of adolescents* (2006.0) 307-37
28. Breland JY, Wong JJ, McAndrew LM. **Are Common Sense Model constructs and self-efficacy simultaneously correlated with self-management behaviors and health outcomes: A systematic review**. *Health Psychol Open* (2020.0) **7**. DOI: 10.1177/2055102919898846
29. Banerjee A, Hendrick P, Bhattacharjee P, Blake H. **A systematic review of outcome measures utilised to assess self-management in clinical trials in patients with chronic pain**. *Patient Educ Couns* (2018.0) **101** 767-778. DOI: 10.1016/j.pec.2017.12.002
30. Willis GB. *Analysis of the cognitive interview in questionnaire design* (2015.0)
31. Cribari-Neto F, Zeileis A. **Beta Regression in R**. *Journal of Statistical Software* (2010.0) **34** 1-24
32. Noel Y, Dauvier B. **A Beta Item Response Model for Continuous Bounded Responses**. *Applied Psychological Measurement* (2007.0) **31** 47-73
33. Verkuilen J, Smithson M. **Mixed and Mixture Regression Models for Continuous Bounded Responses Using the Beta Distribution**. *Journal of Educational and Behavioral Statistics* (2012.0) **37** 82-113
34. Niekerk JV, Bekker A, Arashi M. **Beta regression in the presence of outliers—A wieldy Bayesian solution**. *Stat Methods Med Res* (2019.0) **28** 3729-3740. DOI: 10.1177/0962280218814574
35. Duman JGY. **Self-Management of Chronic Diseases: A Descriptive Phenomenological Study**. *Soc Work Public Health* (2021.0) **36** 300-310. DOI: 10.1080/19371918.2020.1859034
36. Thojampa S, Mawn B. **The moderating effect of social cognitive factors on self-management activities and HbA1c in Thai adults with type-2 diabetes**. *Int J Nurs Sci* (2016.0) **4** 34-37. DOI: 10.1016/j.ijnss.2016.12.006
37. Yang L, Li K, Liang Y, Zhao Q, Cui D, Zhu X. **Mediating role diet self-efficacy plays in the relationship between social support and diet self-management for patients with type 2 diabetes**. *Arch Public Health* (2021.0) **79** 14. DOI: 10.1186/s13690-021-00533-3
38. Heijmans M, Waverijn G, Rademakers J, van der Vaart R, Rijken M. **Functional, communicative and critical health literacy of chronic disease patients and their importance for self-management**. *Patient Educ Couns* (2015.0) **98** 41-8. DOI: 10.1016/j.pec.2014.10.006
39. Albright TL, Parchman M, Burge SK. **Predictors of self-care behavior in adults with type 2 diabetes: an RRNeST study**. *Fam Med* (2001.0) **33** 354-60. PMID: 11355645
40. Battersby M, Harris M, Smith D, Reed R, Woodman R. **A pragmatic randomized controlled trial of the Flinders Program of chronic condition management in community health care services**. *Patient Educ Couns* (2015.0) **98** 1367-75. DOI: 10.1016/j.pec.2015.06.003
41. Lee LY, Tung HH, Tsay SL, Chen YC, Lee HH, Zeng YX. **Predictors for self-management in older adults with type 2 diabetic nephropathy**. *J Clin Nurs* (2020.0) **29** 922-931. DOI: 10.1111/jocn.15154
42. Smith D, Lawn S, Harvey P, Battersby M. **Concurrent validity of the Partners in Health scale against general self-rated health in chronic conditions: A short report**. *Chronic Illn* (2019.0) **15** 74-77. DOI: 10.1177/1742395317743559
43. Smith D, Fairweather-Schmidt AK, Harvey P, Bowden J, Lawn S, Battersby M. **Does the Partners in Health scale allow meaningful comparisons of chronic condition self-management between men and women? Testing measurement invariance**. *J Adv Nurs* (2019.0) **75** 3126-3137. DOI: 10.1111/jan.14124
44. Schüz N, Walters JA, Cameron-Tucker H, Scott J, Wood-Baker R, Walters EH. **Patient Anxiety and Depression Moderate the Effects of Increased Self-management Knowledge on Physical Activity: A Secondary Analysis of a Randomised Controlled Trial on Health-Mentoring in COPD**. *COPD* (2015.0) **12** 502-9. DOI: 10.3109/15412555.2014.995289
45. Vaughan B, Grant M, Moroz J, Ngawaka C, Mulcahy J. **Self-management behaviour and knowledge of patients with musculoskeletal complaints attending an Australian osteopathy clinic: A consecutive sampling design**. *International Journal of Osteopathic Medicine* (2020.0) **37** 3-9. DOI: 10.1016/j.ijosm.2020.05.004
46. Kim S, Song Y, Park J, Utz S. **Patients’ Experiences of Diabetes Self-Management Education According to Health-Literacy Levels**. *Clin Nurs Res* (2020.0) **29** 285-292. DOI: 10.1177/1054773819865879
47. Fix GM, Cohn ES, Solomon JL, Cortés DE, Mueller N, Kressin NR. **The role of comorbidities in patients’ hypertension self-management**. *Chronic Illn* (2014.0) **10** 81-92. DOI: 10.1177/1742395313496591
48. Robertson S.. *Understanding men and health: masculinities, identity, and well-being* (2007.0)
49. Alrahbi H.. **Diabetes self-management (DSM) in Omani with type-2 diabetes**. *International Journal of Nursing Sciences* (2014.0) **1** 352-9
50. Hara Y, Iwashita S, Okada A, Tajiri Y, Nakayama H, Kato T. **Development of a novel, short, self-completed questionnaire on empowerment for patients with type 2 diabetes mellitus and an analysis of factors affecting patient empowerment**. *Biopsychosoc Med* (2014.0) **8** 19. DOI: 10.1186/1751-0759-8-19
51. Kurnia AD, Amatayakul A, Karuncharernpanit S. **Predictors of diabetes self-management among type 2 diabetics in Indonesia: Application theory of the health promotion model**. *Int J Nurs Sci* (2017.0) **4** 260-265. DOI: 10.1016/j.ijnss.2017.06.010
52. D’Souza MS, Karkada SN, Parahoo K, Venkatesaperumal R, Achora S, Cayaban ARR. **Self-efficacy and self-care behaviours among adults with type 2 diabetes**. *Appl Nurs Res* (2017.0) **36** 25-32. DOI: 10.1016/j.apnr.2017.05.004
53. 53World Health Organization. Gender and Health. [Consulted 2021 jul 15]. https://www.who.int/health-topics/gender#tab=tab_1.
54. Saltonstall R.. **Healthy bodies, social bodies: men’s and women’s concepts and practices of health in everyday life**. *Soc Sci Med* (1993.0) **36** 7-14. DOI: 10.1016/0277-9536(93)90300-s
55. 55Robertson S. Understanding men and health: Masculinities, identity and well-being: Masculinity, identity and well-being. McGraw-Hill Education. UK. 2007.
56. 56Ministry of Social Development, Government of Chile. Sample design methodology. 2018. http://observatorio.ministeriodesarrollosocial.gob.cl/casen-multidimensional/casen/docs/Diseno_Muestral_Casen_2017_MDS.pdf
57. 57Ministry of Social Development 2020. Summary of results: Poverty by income and income distribution. CASEN 2020. Social Observatory. Santiago, Chile. http://observatorio.ministeriodesarrollosocial.gob.cl.
58. Chandra S, Mohammadnezhad M, Ward P. **Trust and Communication in a Doctor- Patient Relationship: A Literature Review**. *Journal of Healthcare Communications* (2018.0) **3** 36. DOI: 10.4172/2472-1654.100146
59. Chipidza FE, Wallwork RS, Stern TA. **Impact of the Doctor-Patient Relationship**. *Prim Care Companion CNS Disord* (2015.0) **17**. DOI: 10.4088/PCC.15f01840
|
---
title: Do community-based active case-finding interventions have indirect impacts
on wider TB case detection and determinants of subsequent TB testing behaviour?
A systematic review
authors:
- Helena R. A. Feasey
- Rachael M. Burke
- Marriott Nliwasa
- Lelia H. Chaisson
- Jonathan E. Golub
- Fahd Naufal
- Adrienne E. Shapiro
- Maria Ruperez
- Lily Telisinghe
- Helen Ayles
- Cecily Miller
- Helen E. D. Burchett
- Peter MacPherson
- Elizabeth L. Corbett
journal: PLOS Global Public Health
year: 2021
pmcid: PMC10021508
doi: 10.1371/journal.pgph.0000088
license: CC BY 4.0
---
# Do community-based active case-finding interventions have indirect impacts on wider TB case detection and determinants of subsequent TB testing behaviour? A systematic review
## Abstract
Community-based active case-finding (ACF) may have important impacts on routine TB case-detection and subsequent patient-initiated diagnosis pathways, contributing “indirectly” to infectious diseases prevention and care. We investigated the impact of ACF beyond directly diagnosed patients for TB, using routine case-notification rate (CNR) ratios as a measure of indirect effect. We systematically searched for publications 01-Jan-1980 to 13-Apr-2020 reporting on community-based ACF interventions compared to a comparison group, together with review of linked manuscripts reporting knowledge, attitudes, and practices (KAP) outcomes or qualitative data on TB testing behaviour. We calculated CNR ratios of routine case-notifications (i.e. excluding cases identified directly through ACF) and compared proxy behavioural outcomes for both ACF and comparator communities. Full text manuscripts from 988 of 23,883 abstracts were screened for inclusion; 36 were eligible. Of these, 12 reported routine notification rates separately from ACF intervention-attributed rates, and one reported any proxy behavioural outcomes. Two further studies were identified from screening 1121 abstracts for linked KAP/qualitative manuscripts. $\frac{8}{12}$ case-notification studies were considered at critical or serious risk of bias. $\frac{8}{11}$ non-randomised studies reported bacteriologically-confirmed CNR ratios between 0.47 ($95\%$ CI:0.41–0.53) and 0.96 ($95\%$ CI:0.94–0.97), with $\frac{7}{11}$ reporting all-form CNR ratios between 0.96 ($95\%$ CI:0.88–1.05) and 1.09 ($95\%$ CI:1.02–1.16). One high-quality randomised-controlled trial reported a ratio of 1.14 ($95\%$ CI 0.91–1.43). KAP/qualitative manuscripts provided insufficient evidence to establish the impact of ACF on subsequent TB testing behaviour. ACF interventions with routine CNR ratios >1 suggest an indirect effect on wider TB case-detection, potentially due to impact on subsequent TB testing behaviour through follow-up after a negative ACF test or increased TB knowledge. However, data on this type of impact are rarely collected. Evaluation of routine case-notification, testing and proxy behavioural outcomes in intervention and comparator communities should be included as standard methodology in future ACF campaign study designs.
## Introduction
With over 1.4 million deaths per year [1], tuberculosis (TB) was second only to SARS-CoV-2 as an infectious cause of death globally in 2020. As many as three million people are living with undiagnosed TB disease [1]. Early diagnosis and treatment are fundamental to TB control efforts: the WHO End TB strategy includes targets of at least $90\%$ of people who develop TB being notified and treated within one year by 2025 [2]. Innovative approaches are needed to accelerate progress towards this target from the current estimate of $71\%$ [1].
WHO defines both patient-initiated care-seeking and provider-initiated systematic screening approaches to identify people living with undiagnosed TB [3, 4]. Screening pathways can be facility-based systematic screening or community-based “active case-finding” (ACF). Patient-initiated care-seeking can arise through people recognizing TB symptoms and presenting to a health facility (passive case-finding or PCF), or result from advocacy, communication and social mobilization activities (ACSM) that can prompt earlier care seeking for facility-based TB screening (enhanced case-finding or ECF).The key difference between ACF and ECF is that ACF implies individual interaction between a participant and healthcare worker in the community (e.g. where the participant completes a symptom screen, submits sputum for TB testing or undergoes a chest X-ray).
ACF interventions are designed to directly identify people living with undiagnosed TB in the community but may also have an indirect impact on wider TB case detection as seen in an 2011–14 ACF intervention in Blantyre, Malawi where routine facility-based case-notifications increased substantially over the intervention period [5]. Routine case notification rate (CNR) ratios with a comparison group (excluding those directly identified through the ACF) >1 would be an indication of indirect impact. This indirect impact could be due to enhanced diagnostics introduced through the intervention or an impact on subsequent community TB testing rates and behaviour. Enhanced diagnostics could increase routine case-notifications through improved test sensitivity, although this is likely to be limited to bacteriologically-confirmed TB and there may be a concurrent drop in clinically-diagnosed TB. Health workers could also offer more TB tests if aware of the enhanced diagnostic capacity, leading to higher testing rates. Higher TB testing rates could be due to changes in health worker or community behaviour.
ACF interventions can cover a wide range of activities including door-to-door visits or mobile clinics. They are almost invariably accompanied by ACSM activities, even if only to promote ACF participation and explain to the community the purpose of the intervention and the need for repeat testing if symptoms persist. As such, ACF could influence subsequent TB testing behaviour through the three elements of the COM-B behavioural theory (capacity, opportunity and motivation) and potentially increase TB case-notifications in health facilities through indirect effects (Fig 1). COM-B is a comprehensive model developed from a review of 19 existing behaviourial theories [6] that has been widely applied in assessing and developing public health interventions [6–9] including those for *Tuberculosis diagnosis* and prevention [10–12].
**Fig 1:** *Conceptual framework for how tuberculosis active case finding may affect subsequent healthcare-seeking behaviour.Footnote: (1) Capacity, Opportunity and Motivation are the three domains of the COM-B behavioural theory [6].*
The behavioural mechanisms by which ACSM delivered through ACF interventions may lead to increased knowledge about TB disease and services, or act as a prompt for symptomatic people to present to a health centre for TB testing, are not well understood. ACF interventions could affect knowledge, attitudes and practice (KAP), prompting more timely care-seeking and increasing levels of TB testing and case-notifications through health facilities. ACF interventions may also reduce TB stigma or change social norms and community perceptions around TB. These factors could influence the capacity, motivation and opportunity [6] for subsequent TB testing behaviour (Fig 1). The duration of any behaviour change from ACF is likely to be modified by characteristics of the target population, such as level of education, and ease of access to routine healthcare.
Previous systematic reviews by Kranzer et al [2013] [13], Mhimbira et al [2017] [14] and Burke et al [2021] [15] have shown that ACF interventions can initially increase TB case-notifications, but not invariably. The indirect effects of ACF on routine case-notifications however, has not previously been reviewed. We therefore aimed to systematically review the evidence of indirect effects of ACF on routine facility-based TB case-notifications and also accompanying quantitative proxy behavioural outcomes, such as KAP, that could inform the mechanisms underlying any effect on subsequent TB testing behaviour.
## Methods
We conducted a systematic review of studies reporting the indirect effect of community ACF for TB on routinely-diagnosed TB case-notifications and quantitative proxy behavioural outcomes, such as self-reported TB testing behaviour and KAP of TB.
## Definitions
Active case finding (ACF) was defined as systematic TB screening activities implemented in a specific population. The screening could take any form (e.g. symptom interview, radiology, microbiological testing, referral for specialist medical assessment, in any order) but required a personal interaction between a screener and the person being screened. Health promotion communication activities alone (e.g. leaflet delivery) were considered to be ECF and not ACF. Interventions based solely at a routine healthcare facility were considered systematic TB screening interventions, not ACF.
Routinely-diagnosed TB case-notifications were those identified through ongoing standard healthcare facility-based case-finding activities and excluding TB case-notifications identified through ACF activities (whether tested in the field or referred for testing after screening in the community).
Additionality represents the total increase above expected numbers in TB case-notifications following an active case-finding intervention. This captures all patients who would not have been identified during that time period in the absence of the intervention [16], and can be estimated from comparison of changes in case-notifications in the intervention population during the project compared to the control population or period [17].
Substitution represents the phenomenon of TB patients diagnosed by an active case-finding intervention who, in the absence of the intervention, would still have been identified through routine case-finding activities within the same time period. The extent to which substitution has occurred can be estimated from the number of patients directly diagnosed by ACF minus those identified as additional cases (additionality).
The quantitative proxy behavioural outcomes we examined were: TB knowledge, attitudes and practices (KAP) were what is known, believed and done in relation to TB [18], typically assessed through pre- and post-intervention surveys.
Testing for TB was when a person who has TB symptoms or signs suggestive of TB has a diagnostic test (through submitting sputum for microbiological testing, radiology or specialist medical assessment).
TB stigma was defined as a dynamic process of devaluation that significantly discredits an individual in the eyes of others due to their known or suspected TB status. Within particular cultures or settings, certain attributes are defined by others as discreditable or unworthy [19]. TB stigma could be assessed through a validated scale or through qualitative data.
TB social norms were rules and standards that are understood by members of a group, and that guide or constrain social behaviours around TB, without the force of law [20]. Social norms could be assessed through quantitative data using validated domains or vignettes, or qualitative approaches.
## Inclusion and exclusion criteria
We included studies evaluating an ACF intervention that compared epidemiological TB outcomes (TB case-notifications or TB prevalence) between populations exposed and not exposed to ACF and reported either routinely-diagnosed TB case-notifications or identified proxy behavioural outcomes. Routinely-diagnosed TB notification outcomes could either be directly reported or calculated if both direct ACF yield and overall case notifications were reported for the same period and relevant population. Applicable study designs included randomised controlled trials, studies with a parallel comparison group (controlled before-after studies) and studies with a time-based comparison (before-after studies). We included studies with adults aged 15 years or older that screened at least 1000 people (since the prevalence of active TB in a community will rarely exceed $1\%$). Interventions conducted in closed communities (e.g. prisons) and specific occupational groups (e.g. miners) were included but screening interventions for contacts of people with TB (contact tracing) were not. Studies published before 1 January 1980 and those not in English were excluded.
## Search strategies
The literature search included all studies identified in a previous review by Kranzer et al in 2013 [13], covering the period 1 Jan 1980 to Oct 13 2010, and an additional search of PubMed, EMBASE, Scopus and the Cochrane Library for papers published between 1 Nov 2010 and 4 Feb 2020 (subsequently updated to 13 April 2020) (search strategy in S1 Text). Studies identified through the updated search were title and abstract double screened for initial eligibility (original research, where ACF had taken place, written in English, French or Spanish) by FN, AES and LHC. The full text of eligible studies and all studies from the Kranzer and colleagues review were reviewed by two of HRAF, RMB and MN. Inclusion decisions were resolved by consensus and discussion with ELC and PM. Reference lists from eligible manuscripts were examined and expert opinion on other available papers was sought from members of the WHO TB Screening Guideline Development Group for this and the accompanying review on TB ACF effectiveness [15]. Data was extracted from studies independently by two of HRAF, RMB and MN and entered into a spreadsheet.
## Accompanying qualitative and KAP studies literature search
To increase the number of studies reporting proxy behavioural outcomes relevant to subsequent health seeking behaviour, a further search was conducted for additional secondary manuscripts on qualitative or KAP studies related to the ACF studies identified through the initial literature search (search strategy S2 Text). To be included, the study had to be part of the ACF intervention study identified through the main literature search and include qualitative or quantitative data on the impact of the ACF itself on community TB health seeking behaviour (KAP, TB testing behaviours, pathways to care, TB stigma or social norms). Studies not specifically demonstrating the impact of the ACF on these factors in the ACF target population were excluded, e.g. if the KAP measures were for a different population.
## Access to healthcare
We classified studies according to level of healthcare access within the target population based on distance to and cost of care, as indicated by the reported context or assumed from knowledge of the local health system (S3 Table), on a scale of ‘Standard’ (routine free healthcare available within catchment area), ‘Restricted’ (access restricted due to distance and/or cost) or ‘Hard to reach’ (populations specifically selected as hard to reach).
## Outcomes and risk of bias assessment
Outcomes were a comparison of routine case notification rates (excluding those identified through ACF) and a comparison of reported TB KAP scores (proxy behavioural measure) between groups exposed to and not exposed to the community-based ACF.
To establish routinely diagnosed case notification rates, person-years of follow-up and notified TB cases diagnosed only through routine screening activities were extracted or calculated from available data using simple arithmetic (see S3 Table for extracted data). Person-years were calculated for the target populations for which case-notifications were reported. For before-after studies if the size of the population was not reported separately for the pre- and post-intervention periods it was assumed the size of the population did not change. None of the studies presented case-notification ratios for routine diagnosis; we calculated these through subtracting the available ACF-specific case-notifications from the overall notification data. For randomised and before-after studies we calculated the CNR ratio (intervention vs control groups or baseline vs post-intervention populations) and for controlled before-after studies with a non-randomised comparison group the outcome measure was a comparison of the before to after TB CNR ratio in the two comparison groups: the ratio of the CNR ratios.
Where data was available confidence intervals were calculated using Stata. For studies affected by clustering, three possible values (0.01, 0.05 and 0.1) of the intra-cluster correlation coefficient (ICC) were used to calculate three possible $95\%$ confidence intervals using the Cochrane recommended method [21]. Only the narrowest intervals (ICC = 0.01) are presented in this text, with the others presented in Table 2. Confidence intervals for KAP scores are presented as reported by the authors.
For randomised studies, the Cochrane Risk of Bias (ROB) tool [22] was used to assess risk of bias. Non-randomised studies were assessed for risk of bias using ROBINS-I [23] and qualitative studies were assessed through the Critical Appraisal Skills Programme (CASP) checklist [24].
## Ethical approval and data availability
Ethical approval was not required for this study. All data is available within the results and supplementary materials tables.
## Results
From a total of 23,883 studies identified, full texts of 988 were assessed for inclusion (S1 Table), and 36 with a suitable community-based ACF study design for this review were identified, including 12 that reported case-notification data from both routine facilities and from ACF-identified notifications (Fig 2). Only one out of the 36 manuscripts reported any proxy behavioural outcomes [25], but the additional search identified 1121 manuscripts, of which four articles were eligible for inclusion as KAP/qualitative manuscripts after full text review, but two of these were excluded from full analysis following identification of additional documentation (S2 Table).
**Fig 2:** *Modified PRISMA diagram showing articles reviewed and main reasons for exclusion.*
## Routine TB case-notifications
Of the 12 studies identified for the review of ACF impact on routinely-identified case-notifications, one was a randomised controlled trial [26], six were controlled before-and-after studies (with a parallel comparison group) and five were before-after studies with no comparison group (Table 1). One of the controlled before-and-after studies (Cegielski 2013 [27]) was excluded from further analysis since no cases of TB were identified after the intervention period so meaningful case notification ratios could not be calculated. For all studies (except Miller 2010) the “after” or outcome notifications period was the period during the intervention and did not extend beyond.
**Table 1**
| Study | Design | Country | Population | Healthcare access | ACF | Qualitative / KAP studies |
| --- | --- | --- | --- | --- | --- | --- |
| Case-notifications outcomes | Case-notifications outcomes | Case-notifications outcomes | Case-notifications outcomes | Case-notifications outcomes | Case-notifications outcomes | |
| Miller 2010 | Cluster-randomised trial | Brazil | Urban slums | Standard | ACF (door to door) vs. usual case finding plus leafleting | |
| Aye 2018 | Controlled before-after | Myanmar | Urban slums (& “neighbourhood contacts”) | Standard | Door to door symptom screening and sputum collection for “neighbourhood contacts”, community mobilisation and sputum collection for others | |
| Cegielski 2013 | Controlled before-after | USA | General population—urban | Standard | Community mobilisation, TST screening, mobile clinic. | |
| Datiko 2017 (& Yassin 2013) | Controlled before-after | Ethiopia | Remote rural | Restricted | Community mobilisation, door to door symptom screening, sputum transport | Tulloch 2015 |
| Kan 2012 | Controlled before-after | China | General population—rural | Restricted | Schoolchildren reported symptoms in family members, home visits to symptomatic people, sputum transport. | |
| Parija 2014 | Controlled before-after | India | General population—rural | Restricted | Community mobilisation, mobile clinic, community health workers | |
| Vyas 2019 | Controlled before-after | India | Indigenous groups | Restricted | Door to door symptom screening, sputum collection | |
| Corbett 2010 | Before-after | Zimbabwe | General population—urban | Standard | Community mobilisation, door to door symptom screening or mobile clinics | |
| Fatima 2016 | Before-after | Pakistan | Urban slums “neighbourhood contacts” | Standard | Door to door, sputum collection. | |
| Fatima 2014 | Before-after | Pakistan | Urban slums perceived high risk or hard to reach | Hard to reach | Community mobilisation, mobile clinics (microscopy) | |
| Ford 2019 | Before-after | India | Remote rural | Restricted | Community mobilisation, mobile clinics (CxR). | |
| Lorent 2014 | Before-after | Cambodia | Urban slums—perceived high risk or hard to reach | Hard to reach | Community health workers, door to door symptom screening, sputum collection | Lorent 2015 |
| Behavioural outcomes (KAP) | Behavioural outcomes (KAP) | Behavioural outcomes (KAP) | Behavioural outcomes (KAP) | Behavioural outcomes (KAP) | Behavioural outcomes (KAP) | |
| Adane 2019 | RCT | Ethiopia | Prison | | Peer educators in prisons. People in prison with identified TB symptoms in control and intervention transferred to clinic for physician review | |
Populations varied from urban high-density neighbourhoods to rural communities with long distances to healthcare. From the limited information available, three studies were classified as having been conducted in a setting with “standard” access to routine healthcare, two were classified as specifically “hard-to-reach” and the rest were classified as having restricted access to routine care due to remoteness and/or cost (see S3 Table for extracted data).
ACF interventions combined different strategies including door-to-door screening (eight studies), sputum collection by volunteers or community health workers (seven studies) and community mobilisation for mobile screening clinics (four studies) (Table 1). Of the 11 studies analysed, four reported only bacteriologically-confirmed TB, two reported data for all forms TB (including clinically-diagnosed TB) and five reported both. Only Datiko 2017 and Lorent 2014 reported improving routinely available TB diagnostics as part of the intervention.
The included RCT was conducted by Miller et al comparing door-to-door ACF with leaflet delivered ECF in a Brazilian favela, with a staggered intervention delivered serially in pairs of clusters [26]. The total trial period was 283 days, including the complete intervention time through 60 days after ending ACF in the final clusters. Using calendar time-period, the CNR ratio was 1.14 ($95\%$ CI: 0.94–1.40) implying a $14\%$ relative increase in non-ACF-diagnosed case notification rate for ACF compared to ECF (Table 2). A before-during-after analysis, however, accounting for the staggered cluster pair-by-pair initiation design, showed data consistent with a degree of “substitution” whereby patients who would otherwise have been diagnosed routinely during the intervention period and immediately afterwards were found though ACF. The CNR ratio for ACF compared to ECF clusters was, 0.65 ($95\%$ CI: 0.36–1.19) during the intervention and 0.80 ($95\%$ CI: 0.51–1.27) for the 60 days immediately after the intervention, but 1.42 ($95\%$ CI: 1.12–1.82) for days outside this period (both pre intervention and >60days to end of follow-up) which accounted for $68.5\%$ of the 283-day total trial period (Fig 3). There were some concerns of bias due to missing data in this study.
**Fig 3:** *Case notification rates from Miller cluster-randomised trial in Brazil.Notes: ACF = Active case-finding; ECF = Enhanced case-finding. Relative CNR in days before intervention and >60 days after intervention unknown so presented as consistent.* TABLE_PLACEHOLDER:Table 2 Of the other included studies, the outcome measure of routinely-diagnosed CNR ratio or ratio of CNR ratios (depending on study design) ranged from 0.96 to 1.09 for all forms of TB and 0.47 to 0.96 for bacteriologically-confirmed TB (Table 2). These differences were only significant at the $p \leq 0.05$ level for three of the seven studies reporting all forms of TB: Aye 2018 1.09 ($95\%$ CI: 1.02–1.16) [28], Fatima 2016 1.04 ($95\%$ CI: 1.03–1.05) [29] and Fatima 2014 1.06 ($95\%$ CI: 1.03–1.09) [30]. Confidence intervals were not calculated for Ford 2019 [31] due to unavailability of data. Outcome measures did not appear to be associated with reported healthcare accessibility.
For all five before-after studies, the during intervention overall (both ACF and routine) case notification rates increased but the routine CNR change for bacteriologically-confirmed TB ranged from a $25\%$ reduction (Corbett 2010 [32]) to a $4\%$ reduction (Fatima 2016) (Fig 4), consistent with a degree of substitution or accelerated diagnosis of patients who would otherwise been diagnosed routinely. For all forms of TB, however, the change ranged from a $7\%$ reduction (Lorent 2014 [33]) to a $6\%$ increase (Fatima 2016 [29]). Lorent 2014 was the only before-after study reporting a decrease in all form routine TB CNR during intervention implementation.
**Fig 4:** *Routinely diagnosed case notification rates in non-randomised before-after studies.*
For the six controlled before-after studies increases or decreases in the routine TB CNR in the intervention group reflected the directional change in routine case notification rate in the control group for all studies except two bacteriologically-confirmed reports: Parija 2014 [34] ($1\%$ increase in control group and $14\%$ reduction in intervention group) and Datiko 2017 [35] ($8\%$ increase in control group and $49\%$ reduction in intervention group) (Fig 5). Both studies were conducted with remote rural communities and in Datiko 2017 participants with smear-negative ACF results were offered follow-up radiological TB diagnosis.
**Fig 5:** *Routinely-diagnosed case notification rates in controlled before-after studies.*
The majority of non-randomised studies were considered to be at critical (two studies) or serious risk of bias (6 studies) with three studies at moderate risk of bias (Corbett 2010 [32], Parija 2014 [34] and Vyas 2019 [36]) (Fig 6).
**Fig 6:** *Risk of bias and quality assessments for included studies.*
## Proxy behavioural outcomes
The included study from the search on proxy behavioural outcomes was a cluster-randomised trial of ACF provided through peer inmate educators in 16 selected prisons in Ethiopia that was classified to be at low risk of bias [25] (Fig 6). KAP scores were collected through a semi-structured post-intervention survey conducted with a randomly selected (process not reported) sample of 1218 inmates, using a pre-tested questionnaire detailed in a separate manuscript describing questionnaire development and baseline KAP survey results [37].
This study reported that the intervention group had higher levels of good TB knowledge and practice than the control group. Composite scores of overall knowledge ($p \leq 0.0001$) and good practice ($p \leq 0.0001$) were significantly higher for ACF compared to control prison respondents, even after adjustment for education, geographical location and cluster size in a generalised estimating equation (GEE) model (adjusted OR 2·54, $95\%$ CI 1·93–3·94 for good knowledge, and adjusted OR 1·84, 1·17–2·96 for good practice). There was no significant difference in the composite favourable attitude domain between the two groups (adjusted OR 0.80, $95\%$ CI 0.52–1.25).
## Linked KAP and qualitative studies
Of the four publications [38–41] initially identified, two were excluded from further analysis [38, 39] as additional documentation [42] demonstrated that KAP surveys were not aligned to the populations or timing of the ACF interventions. The two included qualitative studies provided insight into how ACF impacts subsequent TB testing and healthcare-seeking behaviours, although neither directly compared healthcare-seeking behaviours between ACF and routine diagnosis populations.
Tulloch et al conducted in-depth-interviews from May 2011 to February 2012 with participants in a door-to-door symptom screening ACF intervention in 19 districts of Sidama zone conducted in rural Ethiopia from Oct 2010 to 2015 [40, 43, 44]. From these data, researchers describe different healthcare-seeking pathways including those who have heard about TB services through the intervention activities, and then self-referred to a facility for testing. Some participants also acted as ongoing advocates: “There are some who have not heard, if so I always tell them at any opportunity” [40]. The study thus defines mechanisms through which an indirect effect of the ACF intervention could affect subsequent healthcare-seeking behaviour. In addition, the majority of undiagnosed participants were disappointed to have a negative result with an unresolved health problem: “I feel much sorrow. I gave them my sputum and they said I was negative but still I feel pain inside… I am not happy about the result.” [ 40].
Lorent et al. 2015 conducted a survey and interviews with patients diagnosed with TB through door-to-door ACF among high-risk urban populations in Cambodia [33, 41]. Approximately $20\%$ of TB patients diagnosed through the ACF intervention delayed treatment initiation so the main study focus was on exploring reasons for delayed or failed linkage to care, with a comparison of perspectives between those who delayed treatment initiation and those who started treatment without delay. Participants reported that ACF had removed barriers of access and cost and emphasised the need for health education on TB, including stronger peer-support networks.
## Discussion
To our knowledge, the potential indirect impact of TB active case finding interventions on routine TB case-notifications and subsequent TB testing behaviour has not previously been reviewed. In this systematic review, which has direct relevance to ACF campaigns for other respiratory pathogens such as SARS-CoV-2, we aimed to synthesise evidence from evaluations of TB ACF interventions relating to this indirect, but potentially important, impact. Our main finding was the need for more evidence: we found mixed weak evidence that TB ACF may be effective at indirectly increasing routine TB case notification rates for non-bacteriologically confirmed TB, and insufficient evidence to conclude whether or not ACF impacts subsequent TB testing behaviour. The small number of published studies that specifically address this important issue were at risk of bias introduced by the design or completeness of evaluation, and critical differences in study design precluded meta-analysis as well as firm conclusions. Reaching consensus on how to approach and address this question, including published draft protocols, questionnaires, analysis plans, and key-word suggestions would facilitate the rapid accumulation of high-quality harmonised publications able to support meta-analysis in subsequent systematic reviews. ACF implementers should aim to routinely include prospective qualitative and quantitative assessment of indirect effects, given the critical importance of behavioural change as a key driver of respiratory disease care and prevention [45].
In this review a routine CNR ratio >1 gives an indication of an indirect effect of ACF on routine case-notifications. This was seen in the Miller 2010 RCT (1.14, CI:0.94–1.40) and four of the other studies for all form TB notifications: Aye 2018 (1.09, CI:1.02–1.16), Fatima 2016 (1.04, CI:1.03–1.05), Fatima 2014 (1.06, CI:1.03–1.09), and Ford 2019 (1.02, no CI) but not in any of the bacteriologically-confirmed TB reports. This suggests any indirect impact was unlikely to be due to improved diagnostics implemented through the ACF since this would be expected to be seen primarily in bacteriologically-confirmed rates, but instead may be due to increased TB testing rates and changes in TB testing behaviour. In addition, an indirect effect was not observed in the only two studies which did report improved diagnostics (Datiko 2017 & Lorent 2014). case-notifications. The limited evidence available suggests that there may be a difference in impact between the two forms of TB (Table 2, Figs 4 and 5).
Routine bacteriologically-confirmed TB notifications mostly decreased during the ACF, consistent with a degree of “substitution” (see Methods) whereby ACF identifies some patients who would otherwise have been identified by routine services–although they may have benefited through earlier diagnosis and treatment. Consequently, overall bacteriologically-confirmed CNR increased with ACF but the CNR for routinely diagnosed bacteriologically-confirmed cases decreased (CNR ratio range 0.47–0.96). However, for all forms of TB, routine TB CNRs tended to remain at a similar or slightly higher-level during the community ACF interventions (CNR ratio range 0.93–1.09), which could be explained either by ACF promoting early presentation for clinical diagnosis (when patients are not readily confirmed) or by false positive diagnoses, or a combination of the two.
This difference between bacteriologically-confirmed and all forms TB could be due to the desire identified in Tulloch et al. [ 40] for participants with negative bacteriological TB results from the ACF to have some resolution for their health problem. These participants could subsequently attend a facility looking for a diagnosis and then be clinically diagnosed with either extra-pulmonary or pulmonary TB. Datiko et al. [ 35] and Lorent et al. [ 33] showed a decrease in routine all forms TB CNR but in the Datiko study, researchers actively followed up ACF participants with negative results by offering them further radiological examination and clinical diagnosis, whilst participants in the Lorent study were selected as the ‘most hard-to-reach’, suggesting they may have found it difficult to visit a facility for a later clinical diagnosis.
It should be noted that a CNR ratio of ≤1 in this review does not preclude an indirect impact of the ACF on case-notifications as this could still occur but be masked by the “substitution” effect, especially when the CNR ratio is 1 or only slightly below (as in Vyas 2019 (1.00, CI:0.95–1.05), Datiko 2017 (0.96, CI:0.88–1.05) and Lorent 2014 (0.93, CI:0.89–0.97) for all forms TB, and Fatima 2016 (0.96, CI:0.94–0.97) and Fatima 2014 (0.93, CI:0.90–0.95) for bacteriologically-confirmed TB). When the CNR ratio is substantially smaller (e.g. Datiko 2017 (0.47, CI:0.41–0.53) and Corbett 2010 (0.75, CI:0.63–0.89) for bacteriologically-confirmed TB) this suggests there is no indirect impact.
Where it occurs, the indirect impact of ACF on routine TB case-notifications could extend beyond the period of the ACF intervention itself. However, the Miller et al RCT [26] was the only study to specifically assess impact after the end of ACF in a study that reported bacteriologically-confirmed cases only and compared ACF with an ECF intervention. As expected, during the intervention period (mean 27 days) and the 60 days directly afterwards, ECF (leaflets) was associated with increased numbers of TB patients diagnosed through the routine health services. However, the ACF arm had total routine case-notifications beyond those seen with ECF. This could reflect a longer-lasting indirect ACF impact or could just reflect ongoing higher CNRs in the ACF arm since the relative contributions of the pre-intervention and >60 days post-intervention periods are unknown. Personal interaction has been shown to be more effective than purely written information in multiple disciplines [46–48] so temporary in-person community TB diagnosis services could potentially create a longer-term impression than providing literature alone.
We found no evidence that the nature of target populations and levels of healthcare access were important effect modifiers, but cannot conclude that these do not influence the indirect effectiveness of ACF due to the limited number of studies, lack of consistent reporting, and heterogeneity of both populations and interventions.
Disappointingly, we found no studies reporting TB testing rates which would have allowed us to distinguish whether increases in routine TB case-notifications were likely due to an increase in testing or enhanced sensitivity of improved diagnostics with a constant testing rate. In addition, only one study included proxy behavioural outcomes as an integral part of the study design. This Ethiopian cluster randomised trial set in prisons used KAP outcomes as a proxy for subsequent healthcare-seeking behaviour [25] and was assessed as being at low risk-of-bias. TB knowledge and intended care seeking for TB symptoms was improved among inmates provided with the peer-educator intervention, and the study protocol and outcome measures provide a template for subsequent similar interventions and evaluations. Two additional reports provided some qualitative insights supportive of possible impact of ACF on subsequent health seeking behaviour, but conclusions were limited by lack of non-ACF or before-after comparators.
There were several limitations to this review. Despite a literature search covering 40 years and >25,000 titles and abstracts, we found only 12 studies with suitable routine TB case notification data, all of which had very heterogenous interventions and study designs. Just one study specifically addressed outcomes related to subsequent TB testing behaviour following an ACF intervention. As such, we could not conduct meta-analysis, assess generalisability, or quantify the likely impact of behaviour change from ACF on key variables that define the reproduction number for TB and drive epidemiology [49]. Due to resource and time constraints, we only included manuscripts published in English, and did not include unpublished data or grey literature. Notably, TB REACH (http://www.stoptb.org/global/awards/tbreach/) has funded numerous ACF projects since 2010 with reporting that meets many of our criteria, but we were unable to access unpublished data within the short time available for this review. In addition, the Kranzer et al review used for articles published between 1980 to 2010 did not focus on proxy behavioural outcomes so studies reporting on these could have been missed for this period, but as these outcomes are likely to always be secondary to core outcomes of TB notifications and epidemiology (which were included) the likelihood is low. Statistical limitations include limited availability to adjust for confounders as these data were not consistently reported. We also assumed that ACF diagnoses are a subset of the total notifications but an ACF diagnosis could then become a notification in another population for example through population movement, although this is not reported by any of the studies.
Our main recommendations are to strengthen the evidence regarding ACF and indirect effects on subsequent TB notifications and testing behaviour. Qualitative and quantitative assessment of the indirect effects of ACF should be conducted prospectively. Testing rates would be a better outcome measure than case-notifications to establish indirect impact on TB testing behaviour but these are not routinely collected. Case-notifications, and TB testing where available, from both ACF and routine diagnostic services should be reported separately, ideally including pre-ACF, during-ACF and post-ACF periods, evaluated against a comparator population. The inclusion of a comparator is critical, as this is what allows attribution of impact to the ACF intervention itself. To better understand the mechanisms through which ACF potentially impacts TB testing behaviour, relevant outcomes including TB KAP, test initiator (patient or health worker), stigma and norms should be investigated and reported, ideally through repeated cross-sectional sampling before and after implementation. Accompanying qualitative research would provide the rich detail needed to understand how the ACF intervention creates these indirect impacts on subsequent TB case detection.
## Conclusions
In conclusion, the available literature is insufficient, providing only weak evidence for an indirect effect of ACF on clinically diagnosed routine TB case-notifications and insufficient quantitative evidence to assess whether or not ACF impacts subsequent TB testing behaviour. The few available data suggest that ACF can increase TB knowledge and intention to seek early TB diagnosis, together with a desire for diagnosis in those with negative bacteriological ACF results, with potential to impact on future TB testing and case detection rates. Future ACF intervention studies should incorporate assessment of any indirect impact of ACF on facility-based testing and notifications, and other factors with potential to influence TB testing behaviour including KAP, stigma and social norms.
## References
1. 1World Health Organisation. Global Tuberculosis Report 2020. 2020.. *Global Tuberculosis Report 2020* (2020)
2. 2World Health Organisation. The End TB Strategy. 2015.. *The End TB Strategy* (2015)
3. **Early detection of Tuberculosis. An overview of approaches, guidelines and tools**. (2011)
4. 4World Health Organisation. Systematic screening for active tuberculosis. Principles and recommendations. 2013.. *Systematic screening for active tuberculosis. Principles and recommendations* (2013)
5. MacPherson PW E., Choko A., Nliwasa M., Mdolo A., Chiume L., Corbett E.. **Pre and post-intervention prevalence surveys will be conducted to establish whether the ACF leads to a reduction in TB prevalence World Conference on Lung Health of the International Union Against Tuberculosis and Lung Disease (The Union); Cape Town**. (2015)
6. Michie S, van Stralen MM, West R. **The behaviour change wheel: a new method for characterising and designing behaviour change interventions**. *Implement Sci* (2011) **6** 42. DOI: 10.1186/1748-5908-6-42
7. Kredo T, Cooper S, Abrams A, Muller J, Volmink J, Atkins S. **Using the behavior change wheel to identify barriers to and potential solutions for primary care clinical guideline use in four provinces in South Africa.**. *BMC Health Serv Res* (2018) **18** 965. DOI: 10.1186/s12913-018-3778-2
8. Leon N, Namadingo H, Bobrow K, Cooper S, Crampin A, Pauly B. **Intervention development of a brief messaging intervention for a randomised controlled trial to improve diabetes treatment adherence in sub-Saharan Africa**. *BMC Public Health* (2021) **21** 147. DOI: 10.1186/s12889-020-10089-6
9. Williams KN, Thompson LM, Sakas Z, Hengstermann M, Quinn A, Díaz-Artiga A. **Designing a comprehensive behaviour change intervention to promote and monitor exclusive use of liquefied petroleum gas stoves for the Household Air Pollution Intervention Network (HAPIN) trial**. *BMJ Open* (2020) **10** e037761. DOI: 10.1136/bmjopen-2020-037761
10. Ayakaka I, Ackerman S, Ggita JM, Kajubi P, Dowdy D, Haberer JE. **Identifying barriers to and facilitators of tuberculosis contact investigation in Kampala, Uganda: a behavioral approach**. *Implement Sci* (2017) **12** 33. DOI: 10.1186/s13012-017-0561-4
11. Van Ginderdeuren E, Bassett J, Hanrahan C, Mutunga L, Van Rie A. **Health system barriers to implementation of TB preventive strategies in South African primary care facilities**. *PLoS One* (2019) **14** e0212035. DOI: 10.1371/journal.pone.0212035
12. Oliwa JN, Nzinga J, Masini E, van Hensbroek MB, Jones C, English M. **Improving case detection of tuberculosis in hospitalised Kenyan children-employing the behaviour change wheel to aid intervention design and implementation**. *Implement Sci* (2020) **15** 102. DOI: 10.1186/s13012-020-01061-4
13. Kranzer K, Afnan-Holmes H, Tomlin K, Golub JE, Shapiro AE, Schaap A. **The benefits to communities and individuals of screening for active tuberculosis disease: a systematic review**. *Int J Tuberc Lung Dis* (2013) **17** 432-46. DOI: 10.5588/ijtld.12.0743
14. Mhimbira FA, Cuevas LE, Dacombe R, Mkopi A, Sinclair D. **Interventions to increase tuberculosis case detection at primary healthcare or community-level services**. *Cochrane Database Syst Rev* (2017) **11** Cd011432. DOI: 10.1002/14651858.CD011432.pub2
15. Burke RM, Nliwasa M, Feasey HRA, Chaisson LH, Golub JE, Naufal F. **Community-based active case-finding interventions for tuberculosis: a systematic review.**. *Lancet Public Health* (2021). DOI: 10.1016/S2468-2667(21)00033-5
16. Blok L, Creswell J, Stevens R, Brouwer M, Ramis O, Weil O. **A pragmatic approach to measuring, monitoring and evaluating interventions for improved tuberculosis case detection**. *International health* (2014) **6** 181-8. DOI: 10.1093/inthealth/ihu055
17. Creswell J, Sahu S, Blok L, Bakker MI, Stevens R, Ditiu L. **A multi-site evaluation of innovative approaches to increase tuberculosis case notification: summary results**. *PLoS One* (2014) **9** e94465. DOI: 10.1371/journal.pone.0094465
18. **Advocacy, communication and social mobilization for TB Control / A Guide to Developing Knowledge, Attitude and Practice Surveys**. (2008)
19. Challenge TB. *TB Stigma Measurement Guidance* (2018)
20. Dempsey RC, McAlaney J, Bewick BM. **A Critical Appraisal of the Social Norms Approach as an Interventional Strategy for Health-Related Behavior and Attitude Change**. *Frontiers in Psychology* (2018) **9**. DOI: 10.3389/fpsyg.2018.02180
21. Higgins JPT TJ, Chandler J, Cumpston M, Li T, Page MJ, Welch VA. **Cochrane Handbook for Systematic Reviews of Interventions version 6.1 (updated September 2020)**. (2020)
22. Sterne JAC, Savović J, Page MJ, Elbers RG, Blencowe NS, Boutron I. **RoB 2: a revised tool for assessing risk of bias in randomised trials**. *BMJ* (2019) **366** l4898. DOI: 10.1136/bmj.l4898
23. Sterne JA, Hernán MA, Reeves BC, Savović J, Berkman ND, Viswanathan M. **ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions**. *BMJ* (2016) **355** i4919. DOI: 10.1136/bmj.i4919
24. **CASP Qualitative Checklist**. (2018)
25. Adane K, Spigt M, Winkens B, Dinant G-J. **Tuberculosis case detection by trained inmate peer educators in a resource-limited prison setting in Ethiopia: a cluster-randomised trial**. *The Lancet Global Health* (2019) **7** e482-e91. DOI: 10.1016/S2214-109X(18)30477-7
26. Miller AC, Golub JE, Cavalcante SC, Durovni B, Moulton LH, Fonseca Z. **Controlled trial of active tuberculosis case finding in a Brazilian favela**. *The international journal of tuberculosis and lung disease: the official journal of the International Union against Tuberculosis and Lung Disease* (2010) **14** 720-6. PMID: 20487610
27. Cegielski JP, Griffith DE, McGaha PK, Wolfgang M, Robinson CB, Clark PA. **Eliminating tuberculosis one neighborhood at a time**. *American journal of public health* (2013) **103** 1292-300. DOI: 10.2105/AJPH.2012.300781
28. Aye S, Majumdar SS, Oo MM, Tripathy JP, Satyanarayana S, Kyaw NTT. **Evaluation of a tuberculosis active case finding project in peri-urban areas, Myanmar: 2014–2016**. *Int J Infect Dis* (2018) **70** 93-100. DOI: 10.1016/j.ijid.2018.02.012
29. Fatima R, Qadeer E, Yaqoob A, Haq MU, Majumdar SS, Shewade HD. **Extending ’Contact Tracing’ into the Community within a 50-Metre Radius of an Index Tuberculosis Patient Using Xpert MTB/RIF in Urban, Pakistan: Did It Increase Case Detection?**. *PLoS One* (2016) **11** e0165813. DOI: 10.1371/journal.pone.0165813
30. Fatima R, Qadeer E, Enarson DA, Creswell J, Stevens RH, Hinderaker SG. **Success of active tuberculosis case detection among high-risk groups in urban slums**. *Pakistan. Int J Tuberc Lung Dis* (2014) **18** 1099-104. DOI: 10.5588/ijtld.14.0001
31. Ford D, Datta B, Prakash AK, Tripathy JP, Goyal P, Singh S. **Fifth year of a public-private partnership to improve the case detection of tuberculosis in India: A role model for future action?**. *Indian J Tuberc* (2019) **66** 480-6. DOI: 10.1016/j.ijtb.2019.09.005
32. Corbett EL, Bandason T, Duong T, Dauya E, Makamure B, Churchyard GJ. **Comparison of two active case-finding strategies for community-based diagnosis of symptomatic smear-positive tuberculosis and control of infectious tuberculosis in Harare, Zimbabwe (DETECTB): A cluster-randomised trial**. *The Lancet* (2010) **376** 1244-53. DOI: 10.1016/S0140-6736(10)61425-0
33. Lorent N, Choun K, Thai S, Kim T, Huy S, Pe R. **Community-based active tuberculosis case finding in poor urban settlements of Phnom Penh, Cambodia: a feasible and effective strategy**. *PLoS One* (2014) **9** e92754. DOI: 10.1371/journal.pone.0092754
34. Parija D, Patra TK, Kumar AMV, Swain BK, Satyanarayana S, Sreenivas A. **Impact of awareness drives and community-based active tuberculosis case finding in Odisha, India**. *International Journal of Tuberculosis and Lung Disease* (2014) **18** 1105-7. DOI: 10.5588/ijtld.13.0918
35. Datiko DG, Yassin MA, Theobald SJ, Blok L, Suvanand S, Creswell J. **Health extension workers improve tuberculosis case finding and treatment outcome in Ethiopia: a large-scale implementation study**. *BMJ Glob Health* (2017) **2** e000390. DOI: 10.1136/bmjgh-2017-000390
36. Vyas A, Creswell J, Codlin AJ, Stevens R, Rao VG, Kumar B. **Community-based active case-finding to reach the most vulnerable: tuberculosis in tribal areas of India**. *Int J Tuberc Lung Dis* (2019) **23** 750-5. DOI: 10.5588/ijtld.18.0741
37. Adane K, Spigt M, Johanna L, Noortje D, Abera SF, Dinant GJ. **Tuberculosis knowledge, attitudes, and practices among northern Ethiopian prisoners: Implications for TB control efforts**. *PLoS One* (2017) **12** e0174692. DOI: 10.1371/journal.pone.0174692
38. Thapa B, Chadha SS, Das A, Mohanty S, Tonsing J. **High and equitable tuberculosis awareness coverage in the community-driven Axshya TB control project in India**. *Public health action* (2015) **5** 70-3. DOI: 10.5588/pha.14.0105
39. Thapa B, Prasad BM, Chadha SS, Tonsing J. **Serial survey shows community intervention may contribute to increase in knowledge of Tuberculosis in 30 districts of India**. *BMC public health* (2016) **16** 1155. DOI: 10.1186/s12889-016-3807-1
40. Tulloch O, Theobald S, Morishita F, Datiko DG, Asnake G, Tesema T. **Patient and community experiences of tuberculosis diagnosis and care within a community-based intervention in Ethiopia: a qualitative study**. *BMC Public Health* (2015) **15** 187. DOI: 10.1186/s12889-015-1523-x
41. Lorent N, Choun K, Malhotra S, Koeut P, Thai S, Khun KE. **Challenges from Tuberculosis Diagnosis to Care in Community-Based Active Case Finding among the Urban Poor in Cambodia: A Mixed-Methods Study.**. *PLOS ONE* (2015) **10** e0130179. DOI: 10.1371/journal.pone.0130179
42. Karuna DS. **International Union Against Tuberculosis and Lung Disease. Knowledge, Attitude and Practice about Tuberculosis in India**. *A Midline Survey* (2014)
43. Datiko DG, Yassin MA, Theobald SJ, Blok L, Suvanand S, Creswell J. **Health extension workers improve tuberculosis case finding and treatment outcome in Ethiopia: a large-scale implementation study.**. *BMJ global health* (2017) **2** e000390. DOI: 10.1136/bmjgh-2017-000390
44. Yassin MA, Datiko DG, Tulloch O, Markos P, Aschalew M, Shargie EB. **Innovative community-based approaches doubled tuberculosis case notification and improve treatment outcome in Southern Ethiopia.**. *PLoS One* (2013) **8** e63174. DOI: 10.1371/journal.pone.0063174
45. Flaxman S, Mishra S, Gandy A, Unwin HJT, Mellan TA, Coupland H. **Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe.**. *Nature* (2020) **584** 257-61. DOI: 10.1038/s41586-020-2405-7
46. Knopf-Amelung S, Gotham H, Kuofie A, Young P, Manney Stinson R, Lynn J. **Comparison of Instructional Methods for Screening, Brief Intervention, and Referral to Treatment for Substance Use in Nursing Education**. *Nurse Educ* (2018) **43** 123-7. DOI: 10.1097/NNE.0000000000000439
47. Ali N, Zahra T, Vosogh MN, Vosoghi N, Zare M, Mardi A. **Effectiveness Comparison of Mothers’ In-person Versus Written Nutritional Education Intervention on Infant Growth in Iran.**. *Int J MCH AIDS* (2015) **3** 74-80. PMID: 27621988
48. Ramsaroop SD, Reid MC, Adelman RD. **Completing an advance directive in the primary care setting: what do we need for success?**. *J Am Geriatr Soc* (2007) **55** 277-83. DOI: 10.1111/j.1532-5415.2007.01065.x
49. Houben RMGJ, Dowdy DW, Vassall A, Cohen T, Nicol MP, Granich RM. **How can mathematical models advance tuberculosis control in high HIV prevalence settings?**. *The international journal of tuberculosis and lung disease: the official journal of the International Union against Tuberculosis and Lung Disease* (2014) **18** 509-14. DOI: 10.5588/ijtld.13.0773
|
---
title: 'Factors driving underweight, wasting, and stunting among urban school aged
children: Evidence from Merawi town, Northwest Ethiopia'
authors:
- Tilahun Tewabe
- Md. Moustafa Kamal
- Khorshed Alam
- Ali Quazi
- Majharul Talukder
- Syeda Z. Hossain
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021509
doi: 10.1371/journal.pgph.0000586
license: CC BY 4.0
---
# Factors driving underweight, wasting, and stunting among urban school aged children: Evidence from Merawi town, Northwest Ethiopia
## Abstract
Prior research identified malnutrition as one of the most common causes of morbidity and mortality among children globally. Furthermore, research revealed that over two thirds of deaths associated with inappropriate feeding practices occurred during the early years of life. Improper feeding practices impact a child’s health in many different ways. However, research on the possible factors driving underweight, wasting, and stunting among school aged children in developing countries is limited, hence warrant further attention. Against this backdrop, this research strives to identify and assess the determinants of underweight, wasting and stunting among school aged children of a developing country-Ethiopia. A community based cross-sectional study was conducted from April 1, 2018 to June 15, 2018 in Merawi town, Ethiopia. An interviewer-administered questionnaire was used to collect data from a sample of 422 children. Binary logistic regression technique was performed to examine the effect of each selected variable on the outcome measure. The prevalence of being underweight, wasting and stunting was found to be $5.7\%$, $9.8\%$, $10.4\%$, respectively. The age of the child [adjusted odds ratio (AOR) = 12.930 (2.350, 71.157)] and the number of children [AOR = 8.155 (1.312, 50.677)] were emerged as the key determinants for underweight, and the gender of the child was significantly associated with wasting [AOR = 0.455 (0.224, 0.927)]. Finally, the age of the child [AOR = 12.369 (2.522, 60.656)] was found to predict the risk of stunting. This study revealed the age, number of children and gender of the child to have a significant association with malnutrition. The findings of this research suggest that in improving the feeding practices of young school-aged children, special attention should be paid to female children and those coming from relatively large families.
## Introduction
Childhood malnutrition including underweight, wasting, and stunting and their consequences are the major global health priority, particularly in low- and middle-income countries [1]. Malnutrition is the most important risk factor for childhood illnesses and deaths globally, with hundreds of millions of young children being affected worldwide [2]. More importantly, the majority of stunting, underweight, wasting and including micronutrient deficiencies in children are occurring in developed regions such as in Asia and sub-Saharan Africa, mainly due to inadequate feeding to meet their growth demand and high burden of infectious diseases in the regions [3, 4].
School age children ranging between 6–12 years has been found to be a critical period of physical, cognitive, and social development of children [5, 6]. However, in case of a problem such as nutrition [7], these vital parameters would not be achieved and may result in chronic malnutrition, intellectual development delay, school failures, and delayed transition to safe adolescent and adulthood [8, 9]. Although, there is lack of comprehensive evidence about the magnitude of malnutrition among school age children, small scale surveys in developing countries indicate that a high burden of chronic malnutrition such as stunting in school age children is observed to be prevalent amongst $40\%$ India [10], $39\%$ in Indonesia [11], and $26\%$ in Bangladesh [12, 13], and $22.5\%$ in Nigeria [14].
Studies in an Ethiopian context have reported a high prevalence of malnutrition among school going children. For example, an study conducted in Addis Ababa [15] and a systematic review [16] reported up to $18\%$ prevalence of underweight among school age children. Also, other small scale surveys reported up to a $57\%$ stunting rate in Humbo located at Southwestern Ethiopia [17], $46\%$ in Northwestern Ethiopia [17], and $42\%$ in southern Ethiopia [18]. Age of child, mother education, parent occupation, family size, food insecurity and poverty were the main predictors of higher prevalence of chronic malnutrition in the regions [17–19]. However, research findings pertaining to the above issues are still inconsistent and inconclusive. While overall, substantial progress has been made in child health programs including school based feeding, nutrition education, awareness creation, periodic deworming, vitamin, a supplementation and vaccinations, the rate of this reduction in the level of malnutrition like stunting has been insufficient to achieve child health-related Sustainable Development Goals (SDGs) of the United Nations [20]. To achieve child health related goals, up-to-date evidence on nutritional status and related factors are needed to evaluate progresses and identify gaps for future intervention.
Using these premises, we conducted this community-based survey to assess nutritional status of school age children by collecting information on socio-demographics, economic status, child diet habits, food insecurity, and others leading to under nutrition.
## Study settings and period
A community-based cross-sectional study was conducted between April 1, 2018 and June 15, 2018 in Merawi town, which is located in Amhara regional state in Ethiopia, 535 km away from Addis Ababa, and 30 km from Bahir Dar. Based on the latest projections from the Central Statistical Agency of Ethiopia, *Merawi is* estimated to have a total population of 35,541. Of these, 18,479 are males and 17,062 are females. Most of the inhabitants ($98.91\%$) practiced Ethiopian Orthodox Christianity. The town has several private and public health clinics, including a public health center, and a public hospital.
The sample size was calculated using a single population proportion formula by considering the following assumptions: $$P \leq 50$$% proportion of malnutrition for school age children, margin of error (d) $5\%$, and confidence level (CL) $95\%$, and after considering a $10\%$ non-response rate, the final sample size stood at 422. All three kebeles (small administrative units) of the town were included in the study. The number of households in the town was obtained from the administrative offices. We proportionated the sample size based on the number of households in each kebele. Households in each kebele were then selected using computer generated random numbers. From the selected household, we interviewed mothers with school age children. The youngest child was included in the sample for those with more than one child in a similar age group in the same household. However, we moved to the immediate next household for households without school-age children to conduct our interview.
## Data collection tools and procedure
Data were collected by trained nurses through a structured questionnaire using back translation mechanism. The instrument was translated into Amharic from English, before being translated back and pre-tested for consistency. Data collectors were properly trained in collection techniques and procedures. The data were collected through face-to-face interviews with mothers along with their child (6–12 years of age) to assess socio-demographic variables and environmental characteristics, including maternal/child characteristics, and finally anthropometric measurements.
## Data quality assurance
All research related questionnaires of this study was developed after comprehensive review of the literature and was subsequently pre-tested using $5\%$ of the calculated sample size. Training was given in relation to each module of the questionnaire for data collectors. To ensure data quality, completeness, accuracy, and consistency, all collected data were checked every day by the investigator during the entire data collection period. Any errors related to clarity, ambiguity, incompleteness, and misunderstanding, were resolved on a daily basis.
## Dependent variable
The main outcome variable of the study is the nutritional status (underweight, wasting and stunting) among school age children.
## Independent variable
A large number of variables were used as independent variables in this study. These include [1] social and economic variables such as low food availability, dietary diversity, media access, misconception about certain feedings, inadequate feeding practice during illness, inadequate breastfeeding and weaning practices, late initiation of complementary feeding, access to iodized salts, family planning and number of children, [2] environmental factors such as unhygienic living conditions, agricultural, and food shortage; [3] child and maternal related factors such as age, sex, birth interval, birth size, breastfeeding, educational status, place of delivery and immunizations. These independent variables are drawn from prior research in similar fields [15, 20–30].
Fig 1 depicts the nexus between dependent and independent variables examined in this research. The research model (Fig 1) represents the category of variables assessed in this study. The diagram on the right-hand side represents the dependent (outcome) variable which is school age children nutritional status. This variable is measured by the level of underweight, wasting and stunting, diet habits and household food security. Weight of the child was measured using a digital weight scale (7506 digital scale) and was recorded in the nearest one decimal point, and height of the child was measured using a fixed non-bending wooden meter. Children were instructed to take off their shoes to stand upright on their heels, buttocks and shoulders touching the wall. Height of the child was measured and recorded in centimeters. After the measurements were performed, child’s nutritional status was defined as being underweight, wasting and stunting, i.e., when the child had one of the three characteristics such that weight for height, height for age, and weight for age less than -2SD from the reference population based on the WHO multicenter growth reference chart 2007 [23].
**Fig 1:** *A research model showing the possible links between the explanatory and outcome variables.*
On the other hand, the three boxes on the left-hand side of the model represents explanatory (independent) variables including socio-economic variables, environmental and child and maternal factors. The social and economic variables which are measured by low foods availability, dietary diversity, media access, misconceptions about certain feedings, inadequate feeding practices during illness, inadequate breastfeeding and weaning practices, late initiation of complementary feeding, access to iodized salts, family planning and number of children. The environmental factors, include sanitation, agriculture (irrigation), and food shortages in the household. Finally, the maternal and child-related factors included age, sex, birth interval, birth size, breastfeeding, educational status, maternal, place of delivery and immunization of the child.
## Data analysis
Collected data were entered and cleaned using *Epi data* version 3.1 and exported to WHO Anthro Plus and SPSS software version 21 for analysis. Both descriptive and inferential statistics were used to present the data. Binary logistic regression was performed to examine the effect of each independent variable on the outcome variable, and this is presented with adjusted odds ratios (AOR) and $95\%$ confidence intervals (CI). We also examined for the presence of collinearity issue among variables and the results were found to be in acceptable range; variance inflation factor (VIF) was in acceptable rage (VIF< 3). A non-significant Hosmer-Lemeshow model goodness fit test was also achieved. Statistical significance was set at the universal cut off point of p = <0.05.
## Ethics approval
Ethical clearance was obtained from the Amhara Public Health Institute. The data collectors informed each parent/guardian and confirmed their willingness to participate by signing an informed consent sheet. Thus, consent was obtained from each study participant and confidentiality was assured for all the information provided. Moreover, personal identifiers were not included in the questionnaire.
## Socio-demographic characteristics
Of all eligible participants, 392 constituted a response rate of $92.8\%$. Just over half ($50.8\%$) of the respondents were males, $47\%$ were between 6–8 years of age, and $30.3\%$ were first in birth order. As far as the age of the mother is concerned, an overwhelming majority ($96.5\%$) were between 18–45 years of age, $59.6\%$ had family members of five and above, $46.5\%$ had 3 to 4 children, and $65.2\%$ were orthodox Christians. Concerning educational status of mothers, $55.8\%$ were high school and above educated, $42.2\%$ were employed, and $82.3\%$ were married. Convresely, $64.7\%$ of husbands were educated at hiigh school level and above. Of these $46.9\%$ were employed, $12.6\%$ had farming land, and $68.7\%$ had an average monthly income of 2001 Ethiopian birr and above. A vast majority of mothers ($93.7\%$) had access to electronic information from different sources (Table 1).
**Table 1**
| Variable | Response | Total n (%) | Underweight n (%) | Stunted n (%) | Wasted n (%) |
| --- | --- | --- | --- | --- | --- |
| Sex of the child | Male | 201 (50.8) | 12 (3.8) | 17(4.3) | 28 (7.1) |
| Sex of the child | Female | 195 (49.2) | 6 (1.9) | 22 (5.6) | 13 (3.3) |
| Age of the child | 6–8 years | 186 (47.0) | 4 (1.3) | 3(0.8) | 25 (6.3) |
| Age of the child | 9–10 years | 134 (33.8) | 13 (4.1) | 18 (4.5) | 14(3.5) |
| Age of the child | 11–12 years | 76 (19.2) | 1(0.3) | 18 (4.5) | 2 (0.5) |
| Number of children | 1–2 | 179 (45.2) | 10 (3.1) | 18 (4.5) | 20 (5.1) |
| Number of children | 3–4 | 184 (46.5) | 5 (1.6) | 17 (4.3) | 20 (5.1) |
| Number of children | Five and above | 33 (8.3) | 3(0.9) | 4(1.0) | 1(0.3) |
| Number of family | 1–2 | 10 (2.5) | 10 (3.1) | 1(0.3) | 1(0.3) |
| Number of family | 3–4 | 150 (37.9) | 5(1.6) | 16 (4.0) | 16(4.0) |
| Number of family | 5+ | 236 (59.6) | 3 (0.9) | 22(5.6) | 22 (5.6) |
| Birth order of the child | First | 120 (30.3) | 5 (1.6) | 11 (2.8) | 12 (3.0) |
| Birth order of the child | Second and above | 276 (69.7) | 13 (4.1) | 28 (7.1) | 29 (7.3) |
| Age of mother | 18–45 | 381 (96.2) | 18 (5.7) | 34 (8.6) | 41 (10.4) |
| Age of mother | 46–60 | 15 (3.8) | 0 (0.0) | 5 (1.3) | 0 (0.0) |
| Religion | Orthodox | 258 (65.2) | 8 (2.5) | 21 (5.3) | 23(5.8) |
| Religion | Muslim | 100 (25.3) | 9(2.8) | 16(4.0) | 13 (3.3) |
| Religion | Protestant | 38 (9.6) | 1 (0.3) | 2(0.5) | 5(1.3) |
| Mother education | Educated | 221 (55.8) | 10(3.1) | 23 (5.8) | 23(5.8) |
| Mother education | Uneducated | 175 (44.2) | 8 (2.5) | 16 (4.0) | 18 (4.5) |
| Occupation of mother | Employed | 167 (42.2) | 7 (2.2) | 19 (4.8) | 17 (4.3) |
| Occupation of mother | Farmer | 13 (3.3) | 0 (0.0) | 1(0.3) | 0 (0.0) |
| Occupation of mother | Unemployed | 216 (54.5) | 11 (3.5) | 19 (4.8) | 24 (6.1) |
| Marital status | Married | 326 (82.3) | 16 (5.0) | 30 (7.6) | 34 (8.6) |
| Marital status | Unmarried | 70 (17.7) | 2 (0.6) | 9 (2.3) | 7 (1.8) |
| Husband level education | Educated | 233 (64.7) | 9 (2.8) | 26(6.6) | 26 (6.6) |
| Husband level education | Uneducated | 127 (35.3) | 9 (2.8) | 12 (3.1) | 15 (3.8) |
| husband’s occupation | Employed | 169 (46.9) | 7(2.2) | 18 (4.6) | 20(5.1) |
| husband’s occupation | Unemployed | 191 (53.1) | 11 (3.6) | 20(4.6) | 20 (5.2) |
| Farming land | Yes | 50 (12.6) | 2 (0.6) | 3(0.8) | 2 (0.5) |
| Farming land | No | 346 (87.4) | 16 (5.0) | 36 (9.1) | 39 (9.8) |
| Irrigation user | Yes | 30 (7.6) | 1(0.3) | 2(0.5) | 1 (0.3) |
| Irrigation user | No | 366 (92.4) | 17 (5.3) | 37 (9.3) | 40 (10.1) |
| Average monthly income | < 1000 | 40 (10.1) | 1 (0.3) | 3 (0.8) | 2 (0.5) |
| Average monthly income | 1001–2000 | 84 (21.2) | 5 (1.6) | 7 (1.8) | 8 (2.0) |
| Average monthly income | > 2001 | 272 (68.7) | 12 (3.8) | 29 (7.3) | 31 (7.8) |
| A radio and or television | Yes | 370 (93.4) | 17 (5.3) | 39 (9.8) | 39 (9.8) |
| A radio and or television | No | 26 (6.6) | 1(0.3) | 0(0.0) | 2(0.5) |
## Maternal and child health related characteristics
Regarding child and maternal health utilization characteristics, 364 ($91.9\%$) had ante-natal follow up, 272 ($68.7\%$) had TT vaccinations, 325 ($82.1\%$) received additional feeding during pregnancy, 361 ($91.2\%$) delivered in health institutions, 393 ($99.2\%$) had experience of breastfeeding, 314 ($79.3\%$) practiced breastfeeding exclusively, and 333 ($84.1\%$) mothers used family planning to control birth.
About 279 ($70.5\%$) children were fully vaccinated, 255 ($64.4\%$) took vitamin A supplement up to five years old, most ($86.4\%$) children were in school with nearly half below grade two. About 55 ($13.9\%$) children engaged in work; amongst them, 20 ($36.4\%$) spent more than three hours per day in work. About 297 ($75.0\%$) had a history of illness: diarrhea ($54.5\%$), pneumonia ($31.3\%$), measles ($3\%$), malaria ($2.7\%$) and others ($8.4\%$). About 171 ($43.2\%$) children had diarrheal morbidity in the past one year, 149 ($37.6\%$) took additional feeding during illness, 382 ($96.5\%$) children had breakfast regularly, and of them 336 ($85.1\%$) had their meals four and more times per day.
A total of 360 ($90.9\%$) mothers had access to child nutrition education, 61 ($15.4\%$) experienced water shortage for cooking, 367 ($92.7\%$) used pipe water, 372 ($93.9\%$) regularly kept child hygiene, 323 ($81.6\%$) regularly washed their hands, 352 ($88.9\%$) cut their nails, 323 ($81.6\%$) used iodized salt, 288 ($72.7\%$) used wood for cooking, and 288 ($72.7\%$) had a modern latrine facility (see Table 2 for details).
**Table 2**
| Variables | Response | Frequency | Percent |
| --- | --- | --- | --- |
| Antenatal care follows up | Yes | 364 | 91.9 |
| Antenatal care follows up | No | 32 | 8.1 |
| TT Immunization status | Completed | 272 | 68.7 |
| TT Immunization status | Incomplete | 101 | 25.5 |
| TT Immunization status | Not vaccinated | 23 | 5.8 |
| Additional feeding during pregnancy | Yes | 325 | 82.1 |
| Additional feeding during pregnancy | No | 71 | 17.9 |
| Place of delivery | Health facility | 361 | 91.2 |
| Place of delivery | Home | 35 | 8.8 |
| History of breastfeeding | Yes | 393 | 99.2 |
| History of breastfeeding | No | 3 | 0.8 |
| EBF | Yes | 314 | 79.3 |
| EBF | No | 82 | 20.7 |
| Vaccination status | Completed | 279 | 70.5 |
| Vaccination status | Not completed | 109 | 27.5 |
| Vaccination status | Not vaccinated | 8 | 2.0 |
| Vitamin A supplementation | Yes | 255 | 64.4 |
| Vitamin A supplementation | Not completed | 128 | 32.3 |
| Vitamin A supplementation | No | 13 | 3.3 |
| Ever used of FP methods | Yes | 333 | 84.1 |
| Ever used of FP methods | No | 63 | 15.9 |
| Is your child on school | Yes | 342 | 86.4 |
| Is your child on school | No | 54 | 13.6 |
| Level/ grade of education | 1–2 grade | 172 | 49.7 |
| Level/ grade of education | Grade three above | 174 | 50.3 |
| Is your child engaged in work | Yes | 55 | 13.9 |
| Is your child engaged in work | No | 341 | 86.1 |
| If engaged in work for how many hours | 1–3 hours | 31 | 55.4 |
| If engaged in work for how many hours | Above three hours | 25 | 44.6 |
| How can the child get feeding if he engages in work for more than 3 hours | Yes | 20 | 36.4 |
| How can the child get feeding if he engages in work for more than 3 hours | No | 35 | 63.6 |
| Child history of illness | Yes | 297 | 75.0 |
| Child history of illness | No | 99 | 25.0 |
| Type of illness | Pneumonia | 93 | 31.3 |
| Type of illness | Diarrhea | 162 | 54.5 |
| Type of illness | Measles | 9 | 3.0 |
| Type of illness | Malaria | 8 | 2.7 |
| Type of illness | Others /specify | 25 | 8.4 |
| Child diarrheal morbidity in the past year one year | Yes | 171 | 43.2 |
| Child diarrheal morbidity in the past year one year | No | 225 | 56.8 |
| Child diarrheal morbidity in the last two weeks | Yes | 27 | 6.8 |
| Child diarrheal morbidity in the last two weeks | No | 369 | 93.2 |
| Type feeding do you give for your child during illnesses | Regular family dish | 247 | 62.4 |
| Type feeding do you give for your child during illnesses | Additional feeding | 149 | 37.6 |
| Do the child eat breakfast regularly | Yes | 382 | 96.5 |
| Do the child eat breakfast regularly | No | 14 | 3.5 |
| Frequency of eating per day | Four and above | 336 | 85.1 |
| Frequency of eating per day | 1–3 | 59 | 14.9 |
| Access to child nutrition education | Yes | 360 | 90.9 |
| Access to child nutrition education | No | 36 | 9.1 |
| Water shortage for cooking | Yes | 61 | 15.4 |
| Water shortage for cooking | No | 335 | 84.6 |
| Type of water used for cooking | Pipe | 367 | 92.7 |
| Type of water used for cooking | Dam water River /follow water | 29 | 7.3 |
| Boiling water for serving a child | Yes | 22 | 5.6 |
| Boiling water for serving a child | No | 374 | 94.4 |
| Do you regularly keep your child hygiene | Yes | 372 | 93.9 |
| Do you regularly keep your child hygiene | No /pipe | 24 | 6.1 |
| Regular hand washing | Yes | 323 | 81.6 |
| Regular hand washing | No | 73 | 18.4 |
| Nail cutting | Yes | 352 | 88.9 |
| Nail cutting | Only mine | 35 | 8.8 |
| Nail cutting | No | 9 | 2.3 |
| Type of salt used for cooking | Iodized | 323 | 81.6 |
| Type of salt used for cooking | Normal/dish salt | 73 | 18.4 |
| Type cooking fuel | Wood | 288 | 72.7 |
| Type cooking fuel | Petroleum gas | 13 | 3.3 |
| Type cooking fuel | electricity | 95 | 24.0 |
| Type of toilet | Modern latrine | 288 | 72.7 |
| Type of toilet | Temporary toilet/ open | 108 | 27.3 |
## Factors driving malnutrition
Initially variables were tested using bivariate analysis to see their association with the outcome variables (underweight, wasting and stunting). Gender, diarrhea in the past year, age of the child, number of children, husband’s education, year of education, and type of salt used were the independent predictors of underweight. Gender, age of the child, farming land, irrigation, diarrheal illness, breakfast, and mother education about child feeding were the independent predictors of wasting. Vaccination, Vitamin A supplement, child work engagement, age of the child, mother’s age, child’s level of education, and food served at work were the independent predictors of stunting. Age of the child and number of children were the final predictors for being underweight, while it was gender for wasting, and age of the child for stunting.
After identifying variables in the bivariate analysis, we performed multivariate logistic regression analysis to determine the significant variables that determined nutritional status of school ages in the study area. Thus, in the binary logistic regression model, the age of the child emerged as the key determinant for being underweight. A child between 6–8 years of age has an almost 13 times higher chance of having a normal weight than a child of 9–12 years of age [AOR = 12.930 (2.350, 71.157)]. On the other hand, the number of children emerged as the determinant for being underweight. Children living within families with 3–4 children were almost eight times more likely to have a normal weight than those from families with five or more children [AOR = 8.155 (1.312, 50.677)] (see Table 3). By contrast, wasting was affected by the gender of the child. A male child had a $55\%$ lower chance to have a normal body mass index for his age than those of female children [AOR = 0.455(0.224, 0.927)] (see Table 4). This study revealed age of the child as the driving factor of height for age (stunting). Children within the age group of 6–8 years were almost 13 times more likely to be affected by stunting than those of children within the age group of 9–12 years [AOR = 12.369 (2.522, 60.656)] (Table 5).
## Discussion
Prevalence of being underweight stood at $5.7\%$ in this research. This result is much lower than that of the prior studies conducted: $16\%$ in Adds Ababa [15], $28\%$ in rural Northwest Ethiopia [31], and from neighboring country Kenya ($14.9\%$) [28]. This could be attributed to the fact that in the study area, child eating habits were extended to almost three times per day, while fasting occurred at the early age group, and there was an absence of school-based feeding.
This research revealed the age of the child to be the determinant factor for being underweight. A child between 6–8 years of age is almost 13 times more likely to have a normal weight for their age than that of a 9–12-year-old child. On the other hand, the number of children in a family was the determinant for being underweight. Children living with 3–4 children in a family were almost eight times more likely to have normal weight than those from families with five or more children. This finding is consistent with the research findings in Addis Ababa [15], a systematic review in Ethiopia [16] and in Nairobi [28], and may be attributed to the fact that when the child is at a young school age there is a risk of poor appetite in relation to the school environment, and when the number of children is increased it may create acute food shortages in poor families resulting in negligent child-rearing habits.
The prevalence of wasting was $9.8\%$, which was similar with $9\%$ in Gonder [17] but lower than the that of the findings and other studies in Uganda [32], in Afghanistan [33] and in Kenya [28]. This may be due to accelerated growth in this age group, which may be related to good feeding, in terms of access to food and diversity of food. By contrast, wasting was affected by the sex of the child. A male child had a $55\%$ lower chance of having a normal body mass index for his age, which is comparable to the findings in Maputo, Mozambique [34], in Western Kenya [35], Nairobi [28], and India [25]. This finding points to a subconscious practice to pay preferential attention to male children and this practice is prevalent in many underdeveloped countries in Asia and Africa. In this study the prevalence of stunting was found to be at $10.4\%$ which is lower than those study findings in $16\%$ in Adds Ababa [15], $42\%$ Southern Ethiopia [18], $46\%$ in Gonder [17], and from Uganda of $23.8\%$ [32], in Kenya of $30.2\%$ [28], and in Abeokuta, Southwest Nigeria of $17.4\%$ [36]. This could be due to the socio-cultural and economic differences between countries from the referenced areas. In this study, the age of the child emerged as the determinant factor of height for age or stunting. Children within the age group of 6–8 years were almost 13 times more likely to be affected by stunting than those within the age group of 9–12 years. This finding corroborates the research finding in relation to Afghanistan [33, 37], India [25], in Gonder [33] and in Addis Ababa, Ethiopia [15]. This may be due to very young school age children being at risk of adopting school-based feeding, the distance of the school from the home, and a change of the environment at school from home.
The findings of the study make significant contributions to knowledge in relation to the critical roles played by age, number of children in a family, and the gender of the child in driving nutritional status among children in a developing country such as Ethiopia. In particular, this research has contributed to the realm of our knowledge gap in terms of feeding practices of young school-aged children especially female children often receiving comparatively less attention as well as children belonging to large families.
## Study limitations and potential for further study
Despite its unique contributions, this study has a number of limitations. First, this is a community-based study using information provided by mothers about their children to assess the nutritional status of their school age children, which might have information recall bias to some variables related to time. Second, estimates of a child’s nutrition were measured using the internationally used standard tools and definitions, which did not consider country specific parameters. This might underestimate or overestimate the local situation. Third, this research only used a structured interviewer-administered questionnaire and measurements that might not reflect culture and perception of food taboos in the community. Fourth, this research was confined to a particular region of Ethiopia with a relatively small sample size which might affect the generalizability of the findings across the borders in the country and beyond. However, future research could use the result to conduct more broad-based sample drawn from several regions in Ethiopia and beyond to come up with better findings for practice and research about child nutrition in developing regions.
## Concluding remarks
This research addressed some critical issues in the public health arena that were not well covered in the existing literature, especially in the context of a developing African country, Ethiopia where malnutrition is highly prevalent in school aged children. In particular, this study identified the possible determinants of underweight, wasting and stunting among school aged children (6–12 years). The outcomes of this study have an important bearing on designing appropriate policies to address the current problems pertaining to the above issues examined in this research. One particular issue emerging from this research that warrants the attention of policymakers is that female children receive lesser attention than their male counterparts in terms of feeding. Furthermore, attention can be directed to children belonging to large families as they are more likely to suffer from malnutrition and are underweight for an obvious reason. Age of children is another issue deserving attention because malnutrition leading to underweight is more common in older students than their younger counterparts. Although the data of this study was obtained from mothers and might involve recall bias to some variable which suggests the importance of prospective and rigorous to investigate causal associations, this study shows that school age children feeding practices need to be improved and adjusted to the children cohorts identified in this research. Finally, since, healthy work force is vital for accelerating the pace of economic growth of a transitional economy such as Ethiopia, policymakers should address the above-mentioned issues as an early intervention strategy towards developing healthy human resources for Ethiopia to reap the benefits of demographic dividend.
## References
1. de Onis M, Branca F. **Childhood stunting: a global perspective**. *Matern Child Nutr* (2016.0) **12** 12-26. DOI: 10.1111/mcn.12231
2. 2Children: improving survival and well-being [Internet]. [cited 2021 Jan 29]. https://www.who.int/news-room/fact-sheets/detail/children-reducing-mortality
3. Cusick SE, Kuch AE. **Determinants of Undernutrition and Overnutrition among Adolescents in Developing Countries**. *Adolesc Med State Art Rev* (2012.0) **23** 440-56. PMID: 23437681
4. 4WHO | Global strategy for infant and young child feeding [Internet]. WHO. World Health Organization; [cited 2021 Jan 29]. http://www.who.int/nutrition/publications/infantfeeding/9241562218/en/
5. Fink G, Rockers PC. **Childhood growth, schooling, and cognitive development: further evidence from the Young Lives study**. *Am J Clin Nutr* (2014.0) **100** 182-8. DOI: 10.3945/ajcn.113.080960
6. 6Onis M. Child Growth and Development. In 2017. p. 119–41.
7. Bryan J, Osendarp S, Hughes D, Calvaresi E, Baghurst K, van Klinken JW. **Nutrients for cognitive development in school-aged children**. *Nutr Rev* (2004.0) **62** 295-306. DOI: 10.1111/j.1753-4887.2004.tb00055.x
8. Asmare B, Taddele M, Berihun S, Wagnew F. **Nutritional status and correlation with academic performance among primary school children, northwest Ethiopia**. *BMC Research Notes* (2018.0) **11** 805. DOI: 10.1186/s13104-018-3909-1
9. Acharya Y, Luke N, Haro MF, Rose W, Russell PSS, Oommen AM. **Nutritional status, cognitive achievement, and educational attainment of children aged 8–11 in rural South India**. *PLOS ONE* (2019.0) **14** e0223001. DOI: 10.1371/journal.pone.0223001
10. Pal D, Kanungo S. **Malnutrition Scenario among School Children in Eastern-India-an Epidemiological Study**. *Epidemiology* (2016.0) **06**
11. Lestari S, Fujiati II, Keumalasari D, Daulay M. **The prevalence and risk factors of stunting among primary school children in North Sumatera, Indonesia**. *IOP Conference Series: Earth and Environmental Science* (2018.0) **125** 012219
12. Sanin KI, Haque A, Nahar B, Mahfuz M, Khanam M, Ahmed T. **Food Safety Practices and Stunting among School-Age Children—An Observational Study Finding from an Urban Slum of Bangladesh**. *Int J Environ Res Public Health* (2022.0) **19** 8044. DOI: 10.3390/ijerph19138044
13. Danjin M, Adewoye S, Sawyerr H. **Prevalence and Socio-demographic Determinants of Stunting among School Age Children (SAC) in Gombe State, Nigeria**. *Journal of Advances in Medicine and Medical Research* (2020.0) 22-34
14. Bogale TY, Bala ET, Tadesse M, Asamoah BO. **Prevalence and associated factors for stunting among 6–12 years old school age children from rural community of Humbo district, Southern Ethiopia**. *BMC Public Health* (2018.0) **18** 653. PMID: 29793479
15. Degarege D, Degarege A, Animut A. **Undernutrition and associated risk factors among school age children in Addis Ababa, Ethiopia**. *BMC Public Health* (2015.0) **15** 375. DOI: 10.1186/s12889-015-1714-5
16. Assemie MA, Alamneh AA, Ketema DB, Adem AM, Desta M, Petrucka P. **High burden of undernutrition among primary school-aged children and its determinant factors in Ethiopia; a systematic review and meta-analysis**. *Italian Journal of Pediatrics* (2020.0) **46** 118. DOI: 10.1186/s13052-020-00881-w
17. Getaneh Z, Melku M, Geta M, Melak T, Hunegnaw MT. **Prevalence and determinants of stunting and wasting among public primary school children in Gondar town, northwest, Ethiopia**. *BMC Pediatrics* (2019.0) **19** 207. DOI: 10.1186/s12887-019-1572-x
18. Tariku EZ, Abebe GA, Melketsedik ZA, Gutema BT. **Prevalence and factors associated with stunting and thinness among school-age children in Arba Minch Health and Demographic Surveillance Site, Southern Ethiopia**. *PLoS One* (2018.0) **13** e0206659. DOI: 10.1371/journal.pone.0206659
19. Lisanu Mazengia A, Andargie Biks G. **Predictors of Stunting among School-Age Children in Northwestern Ethiopia**. *Journal of Nutrition and Metabolism* (2018.0) **2018** 1-7. DOI: 10.1155/2018/7521751
20. 20WHO | Sustainable Development Goal 3: Health [Internet]. WHO. World Health Organization; [cited 2021 Jan 27]. http://www.who.int/topics/sustainable-development-goals/targets/en/
21. 21WHO | Indicators for assessing infant and young child feeding practices [Internet]. WHO. World Health Organization; [cited 2021 Jan 29]. https://www.who.int/maternal_child_adolescent/documents/9789241596664/en/
22. Müller O, Krawinkel M. **Malnutrition and health in developing countries**. *CMAJ* (2005.0) **173** 279-86. DOI: 10.1503/cmaj.050342
23. Turck D, Michaelsen KF, Shamir R, Braegger C, Campoy C, Colomb V. **World Health Organization 2006 Child Growth Standards and 2007 Growth Reference Charts: A Discussion Paper by the Committee on Nutrition of the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition**. *Journal of Pediatric Gastroenterology & Nutrition* (2013.0) **57** 258-64. PMID: 23880630
24. Wang Y, Monteiro C, Popkin BM. **Trends of obesity and underweight in older children and adolescents in the United States, Brazil, China, and Russia**. *Am J Clin Nutr* (2002.0) **75** 971-7. DOI: 10.1093/ajcn/75.6.971
25. Srivastava A, Mahmood SE, Srivastava PM, Shrotriya VP, Kumar B. **Nutritional status of school-age children—A scenario of urban slums in India**. *Arch Public Health* (2012.0) **70** 8. DOI: 10.1186/0778-7367-70-8
26. Rayhan MI, Khan MSH. **Factors causing malnutrition among under five children in Bangladesh**. *Pak J Nutr* (2006.0) **5** 558-62
27. Kassahun Alemu KM. **Prevalence of Malnutrition and Associated Factors Among Children Aged 6–59 Months at Hidabu Abote District, North Shewa, Oromia Regional State**. *J Nutr Disorder Ther* (2013.0) **03**
28. Chesire EJ, Orago ASS, Oteba LP, Echoka E. **Determinants of under nutrition among school age children in a Nairobi peri-urban slum**. *East Afr Med J* (2008.0) **85** 471-9. DOI: 10.4314/eamj.v85i10.9671
29. Shubair ME, Yassin MM, Al-Hindi AI, Al-Wahaidi AA, Jadallah SY. **Intestinal parasites in relation to haemoglobin level and nutritional status of school children in Gaza**. *Journal of the Egyptian Society of parasitology* (2000.0) **30** 365-75. PMID: 10946498
30. Medhi GK, Barua A, Mahanta J. **Growth and Nutritional Status of School Age Children (6–14 Years) of Tea Garden Worker of Assam**. *Journal of Human Ecology* (2006.0) **19** 83-5
31. Degarege A, Erko B. **Association between intestinal helminth infections and underweight among school children in Tikur Wuha Elementary School, Northwestern Ethiopia**. *J Infect Public Health* (2013.0) **6** 125-33. DOI: 10.1016/j.jiph.2012.11.008
32. Kikafunda JK, Walker AF, Collett D, Tumwine JK. **Risk factors for early childhood malnutrition in Uganda**. *Pediatrics* (1998.0) **102** E45. DOI: 10.1542/peds.102.4.e45
33. Frozanfar MK, Yoshida Y, Yamamoto E, Reyer JA, Dalil S, Rahimzad AD. **Acute malnutrition among under-five children in Faryab, Afghanistan: prevalence and causes**. *Nagoya J Med Sci* (2016.0) **78** 41-53. PMID: 27019527
34. Prista A, Maia JAR, Damasceno A, Beunen G. **Anthropometric indicators of nutritional status: implications for fitness, activity, and health in school-age children and adolescents from Maputo, Mozambique**. *Am J Clin Nutr* (2003.0) **77** 952-9. DOI: 10.1093/ajcn/77.4.952
35. Bloss E, Wainaina F, Bailey RC. **Prevalence and predictors of underweight, stunting, and wasting among children aged 5 and under in western Kenya**. *J Trop Pediatr* (2004.0) **50** 260-70. DOI: 10.1093/tropej/50.5.260
36. Senbanjo IO, Oshikoya KA, Odusanya OO, Njokanma OF. **Prevalence of and Risk factors for Stunting among School Children and Adolescents in Abeokuta, Southwest Nigeria**. *J Health Popul Nutr* (2011.0) **29** 364-70. DOI: 10.3329/jhpn.v29i4.8452
37. Burki T.. **Ethiopia shows it is single-minded in tackling disease**. *The Lancet Infectious Diseases* (2016.0) **16** 153-4. DOI: 10.1016/S1473-3099(16)00021-9
|
---
title: 'The availability of essential medicines for cardiovascular diseases at healthcare
facilities in low- and middle-income countries: The case of Bangladesh'
authors:
- Shariful Hakim
- Muhammad Abdul Baker Chowdhury
- Md. Ashiqul Haque
- Nasar U. Ahmed
- Gowranga Kumar Paul
- Md. Jamal Uddin
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021517
doi: 10.1371/journal.pgph.0001154
license: CC BY 4.0
---
# The availability of essential medicines for cardiovascular diseases at healthcare facilities in low- and middle-income countries: The case of Bangladesh
## Abstract
Long-term, often lifelong care for cardiovascular disease (CVD) patients requires consistent use of medicine; hence, the availability of essential medicine for CVD (EM-CVD) is vital for treatment, quality of life, and survival. We aimed to assess the availability of EM-CVD and explore healthcare facility (HCF) characteristics associated with the availability of those medicines in Bangladesh. This study utilized publicly available cross-sectional data from the 2014 and 2017 waves of the Bangladesh Health Facilities Survey (BHFS). The analysis included 204 facilities (84 from the 2014 BHFS and 120 from the 2017 BHFS) that provide CVD diagnosis and treatment services. The outcome variable "EM-CVD availability" was calculated as a counting score of the following tracer medicines: angiotensin-converting enzyme (ACE) inhibitors (enalapril), thiazide, beta-blockers (atenolol), calcium channel blockers (amlodipine and nifedipine), aspirin, and simvastatin/atorvastatin. A multivariable Poisson regression model was used to identify the HCF characteristics associated with EM-CVD availability. The number of Bangladeshi HCFs that provide CVD screening and treatment services increased just a little between 2014 and 2017 (from $5.4\%$ to $7.9\%$). Since 2014, there has been an increase in the availability of calcium channel blockers (from $37.5\%$ to $38.5\%$), aspirin (from $25.3\%$ to $27.9\%$), and simvastatin/atorvastatin (from $8.0\%$ to $30.7\%$), whereas there has been a decrease in the availability of ACE inhibitors (enalapril) (from $12.5\%$ to $6.5\%$), thiazide (from $15.7\%$ to $11.1\%$), and beta-blockers (from $42.5\%$ to $32.5\%$). The likelihood of EM-CVD being available was higher among private and urban facilities than among public and rural facilities. Furthermore, facilities that had 24-hour staff coverage and performed quality assurance activities had a higher chance of having EM-CVD available than those that did not have 24-hour staff coverage and did not undertake quality assurance activities. Government authorities should think about a wide range of policy implications, such as putting more emphasis on public and rural facilities, making sure staff is available 24 hours a day, and performing quality assurance activities at facilities to make EM-CVD more available.
## Introduction
The 2030 Agenda for Sustainable Development Goals (SDGs) of the United Nations (UN) intends to reduce premature deaths from non-communicable diseases (NCDs) by 2030 [1]. Cardiovascular diseases (CVDs) are the most frequent causes of NCDs deaths, accounting for approximately 17.9 million deaths per year, representing $32\%$ of all deaths. Heart attacks and strokes account for over 4 out of every 5 CVD deaths, and a third of such deaths happen prematurely in those younger than the age of 70. If the current trend continues, it is predicted that it will reach 23.3 million by 2030 [2]. CVDs also place a significant financial burden on the patient, the healthcare system, and the nation’s economy. By 2030, global spending on CVD health is forecasted to exceed $1,044 billion, up from $863 billion in 2010 [3].
While the growth in CVD prevalence is a global phenomenon, it has been quickest in low- and middle-income countries (LMICs) over the previous decade. The burden of CVDs continues excessively in LMICs at a greater rate than in high-income countries (HICs), as over $75\%$ of CVD deaths happen in LMICs [2]. Between 2011 and 2025, an estimated cumulative economic loss from all NCDs will be $7.28 trillion in LMICs, and CVDs will be responsible for about half of this estimated loss [4].
Bangladesh, an LMIC, continues to have a high disease burden even with the shift from infectious diseases to chronic diseases [5]. One of the most common diseases in *Bangladesh is* CVD [6]. CVDs alone account for $25.1\%$ of total deaths in Bangladesh, and this estimate will increase to $37.2\%$ by 2030 [7]. Recent research indicated that $15.5\%$ of Bangladeshi aged 40–69 years are susceptible to CVDs [6]. CVDs and their known associates account for $13.4\%$ of lost disability-adjusted life years (DALYs) [8].
CVD prevention necessitates significant governmental initiatives, infrastructure, and system changes, and the creation of a conducive environment and individual behavioral changes, which can be a long-term and challenging task, particularly in low-resource nations like Bangladesh. However, disease control and health maintenance remain the main feasible options for millions of people with CVD for the time being, and this choice requires treating CVD patients with essential medicines.
In Bangladesh, the healthcare system is shared by the governmental and private sectors. In the public sector, there is a lack of medicine stocks, and hence almost all Bangladeshis acquire their essential medicines through direct out-of-pocket (OOP) purchases from the private sector [9]. In 1995, private OOP health spending accounted for $60\%$ of total health spending; in 2014, the figure was $67\%$ [10]. To run health systems effectively, essential medicines should always be available in sufficient quantities to meet the basic healthcare needs of the population [11]. Access to essential medicines and vaccines that are safe, effective, of high quality, and affordable for all is one of the SDGs [12].
According to Penchansky and Thomas, the idea of access subsumes five dimensions: availability, accessibility, accommodation, acceptability, and affordability [13]. Availability is not only an essential first step, but it is also a challenging aspect, mainly in poorly funded public health systems [14]. A central component of universal health coverage is the continuous availability of essential medicines in the delivery of standard health services [15]. To figure out what to do next, like how to get access to essential medicines for CVD (EM-CVD), it is important to study the availability of the EM-CVD and the challenges they pose for health care systems. Long-term, often lifelong care for CVD patients requires consistent use of medication. Hence, the availability EM-CVD of is vital for treatment, quality of life, and survival.
In 2010, a study investigated the availability of five cardiovascular medicines in 36 countries and described a mean availability of $26.3\%$ in the public sector and $57.3\%$ in the private sector [16]. Prior research in 35 LMICs discovered the mean availability of anti-hypertensive generics at the lowest price of $34.7\%$ in the public sector and $57.1\%$ in the private sector [17]. According to Khatib et al. recent analysis of the availability and affordability of four CVD medicines in pharmacies collected from nearly 600 communities in 18 countries, availability of all four medicines in the urban community was 95 percent, 80 percent, 62 percent, and 25 percent in high-income, upper-middle-income, and low-income countries, respectively [18]. The percentages in the rural community were 90 percent, 73 percent, 37 percent, and 3 percent, respectively. In the West Region of Cameroon in 2014, a study by Jingi et al. found that the availability of drugs for CVD and diabetes ranged from 36.4 percent to 59.1 percent in urban settings and from 9.1 percent to 50 percent in rural settings [19]. In a 2019 study, Kaiser et al. investigated the availability of essential diabetes and CVD medicines, discovering that two anti-diabetics and nine anti-hypertensives had high availability ($80\%$), while the remaining anti-diabetics and anti-hypertensives had availability levels that were mostly under $50\%$ for the rest of the surveyed anti-diabetics and anti-hypertensives in three Zambian provinces [20].
Using data from the 2014–2015 Tanzania Service Provision Assessment (SPA) survey, Bintabara and Ngajilo looked at how ready healthcare facilities (HCFs) were to offer outpatient care for NCDs. They observed that the private higher-level facilities had significantly more basic diagnostic tools and medicines available than their public counterparts [21]. A study by Armstrong et al. revealed that facility region, facility type, managing authority, and range of HIV services are significant correlates of NCD’s medicine availability utilizing 196 Ugandan HCF data [22]. Another study that used the 2015 Nepal SPA showed that private and urban facilities were more prepared to offer services for CVD and diabetes than public facilities and facilities in rural areas [23].
A study by Biswas et al. investigated the preparedness of HCFs for diabetes and cardiovascular services in Bangladesh using the 2014 Bangladesh Health Facility Survey (BHFS) [24]. This study mentioned that among the facilities that offer CVD services, only $0.9\%$ had all four of the four service preparedness factors (guidelines, trained staff, equipment, and medicine) in Bangladesh. To our knowledge, no research has attempted to thoroughly investigate the availability of EM-CVD in Bangladesh to find the HCF characteristics linked with essential medicine availability. We need a deeper understanding of the factors impacting the availability of EM-CVD at HCFs in Bangladesh to accomplish the SDGs by assuring a consistent supply of those medicines. The present study sought to examine the availability of EM-CVD and its associated factors in Bangladesh.
## Study population and setting
This study included data from the latest survey rounds of the 2014 and 2017 Bangladesh Health Facilities Survey (BHFS). The BHFS was conducted in 2014 and 2017 by the U.S. Agency for International Development (USAID) and used standardized questionnaires from the service provision assessment (SPA) component of the DHS Program. The National Institute of Population Research and Training (NIPORT) and the Ministry of Health and Family Welfare (MOHFW) ran the survey with help from the Bangladesh government and the United States Agency for International Development. ICF International and the Associates for Community and Population Research (ACPR) collected the data. The Demographic and Health Survey (DHS) runs the survey in collaboration with Bangladesh by the National Institute of Population Research and Training (NIPORT) with technical assistance provided by ICF, USA. The surveys employed identical questionnaires for both the facility inventory and the interviews with the health providers.
The samples for the 2014 and 2017 BHFS were designed to include establishments from across the country’s administrative divisions. The sampling frames were a list of 19,184 and 19,811 registered HCFs from the 2014 and 2017 BHFS, respectively, furnished by NIPORT and MOHFW.
From the complete formal-sector health facilities, 1,596 health facilities for the 2014 BHFS and 1,600 health facilities for the 2017 BHFS were selected using stratified random sampling. Interviewers were unable to survey some of the facilities because they were not open or operating at the time of the survey. Finally, the 2014 and 2017 BHFS contain information on 1,548 and 1,524 facilities, respectively. The sampling strategy and study design are detailed elsewhere [25, 26].
## Selection of study samples
Of all the facilities, only those that offer NCD services-especially those where providers diagnose and treat CVD-were included. From the 2014 BHFS, 84 facilities were selected for the availability of EM-CVD, and 120 facilities were included as study samples for the availability of EM-CVD from the 2017 BHFS. The screening process used to select a study sample is illustrated in Fig 1.
**Fig 1:** *Flow chart of study samples selection.The star (stars) in the figure is a sign indicating the number of healthcare facilities in the 2014 and 2017 BHFS, respectively. One star indicates the facilities from the 2014 survey, and the sign with two stars indicates the facilities from the 2017 survey. Numbers in the parentheses represent weighted data.*
## Outcome variable
According to the WHO-Service Availability and Readiness Assessment (SARA) reference document [24], the availability of EM-CVD was assessed using a list of tracer drugs: ACE inhibitors (enalapril), thiazide, beta-blockers (atenolol), calcium channel blockers (amlodipine/nifedipine), aspirin, and simvastatin/atorvastatin. Availability was marked out as the presence of at least one legitimate tracer medicine in a facility on the day of the visit, which could be seen by the people collecting the data. Essential medicine availability measures the number of non-expired (valid) tracer drugs for CVD in a health facility. The outcome variable ‘EM-CVD availability’ was calculated as a counting score of the tracer medicines ranging from 0 to 6, where higher scores indicated the greater availability of essential medicines. The scores reflect how many of the six EM-CVD a facility had in stock on the survey day. Therefore, the outcome variable quantifies the number of valid essential drugs for CVD in a health facility.
## Potential associated factors
Potential associated factors of interest include managing authority (public, private), location (urban, rural), administrative division (Barisal, Chittagong, Dhaka, Khulna, Rajshahi, Rangpur, Sylhet, Mymensingh), external supervision (not received, received within the past 6 months, received more than 6 months ago), routine user fee or charges for client service (not available, available), 24-hour staff coverage (not available, available), system to elicit clients’ opinions about the health facility or its service (not available, available), and guidelines for the diagnosis and management of CVD (not available, available).
## Data analysis
With the help of the Chi-square test, we were able to assess the proportion of EM-CVD availability between the groups of several potential associated factors. The multivariable Poisson regression model was used to identify the health facility characteristics associated with EM-CVD availability. Multicollinearity was checked to see if there was any strong association between the potential factors, and it was determined to be absent. All data management and analyses were conducted using Stata 13 (StataCorp, College Station, TX, USA). To account for the complex survey design, we weighted all of our analyses using the weight option in Stata with the sampling weights provided in the dataset. For modeling exercises, we used the "svy" command of Stata to account for the survey design, primary sampling unit, and cluster.
## Ethics statement
Because we used publicly available, de-identified data from online data repositories, our study was exempt from the ethical review. ICF International’s Institutional Review Board (IRB) and the Bangladesh Medical Research Council of the Ministry of Health and Family Welfare (MOH&FW) have evaluated and approved the methodology and questionnaires for DHS surveys, ensuring that the survey complies with US Department of Health and Human Services regulations for the protection of human subjects (45 CFR 46).
## EM-CVD availability
The count of various EM-CVD existing at every facility in both the surveys was skewed; scores were gathered at the lowest possible score and with a lengthy tail towards the highest possible score (Fig 2). Among facilities offering CVD screening and treatment services, $41.1\%$ (BHFS 2014) and $48.8\%$ (BHFS 2017) of facilities had no EM-CVD on-site at all.
**Fig 2:** *Number of essential medicines for CVD available in sampled facilities, BHFS 2014 and 2017.*
Fig 3 summarizes the six EM-CVDs by category and the medicines available at a sample of sites in the 2014 and 2017 BHFS. Between 2014 and 2017, the availability of calcium channel blockers (from $37.5\%$ to $38.5\%$), aspirin (from $25.3\%$ to $27.8\%$), and simvastatin/atorvastain (from $8.0\%$ to $30.7\%$) increased while ACE inhibitors (enalapril) (from $15.7\%$ to $11.1\%$), thiazide (from $15.7\%$ to $11.1\%$), and beta-blockers (from $42.5\%$ to $32.1\%$) fell.
**Fig 3:** *Availability of essential medicines for CVD at health facilities in Bangladesh, BHFS 2014 and 2017.*
## Background characteristics of surveyed facilities by the number of EM-CVD available
Table 1 depicts the proportion of EM-CVD availability among the categories of various characteristics of facilities. In 2014, $40.0\%$ of public facilities, $52.3\%$ of rural facilities, and $55.2\%$ of facilities that had no routine user fees or charges for client service did not have any EM-CVD, compared to $44.3\%$, $56.3\%$, and $71.6\%$, respectively, in 2017. Between 2014 and 2017, the availability of four or more EM-CVD decreased among facilities that were situated in Dhaka division (from $30.0\%$ to $22.7\%$) and that had guidelines for the diagnosis and management of CVD (from $19.7\%$ to $11.1\%$).
**Table 1**
| Unnamed: 0 | BHFS 2014 (n = 253) | BHFS 2014 (n = 253).1 | BHFS 2014 (n = 253).2 | BHFS 2014 (n = 253).3 | BHFS 2017 (n = 335) | BHFS 2017 (n = 335).1 | BHFS 2017 (n = 335).2 | BHFS 2017 (n = 335).3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Factors | Number of available essential medicines for CVD, n (%) | Number of available essential medicines for CVD, n (%) | Number of available essential medicines for CVD, n (%) | P-value* | Number of available essential medicines for CVD, n (%) | Number of available essential medicines for CVD, n (%) | Number of available essential medicines for CVD, n (%) | P-value* |
| Factors | None available | 1–3 available | 4 or more available | P-value* | None available | 1–3 available | 4 or more available | P-value* |
| Managing authority | | | | <0.001 | | | | <0.001 |
| Public | 68 (40.0) | 88 (51.8) | 14 (8.2) | | 105 (44.3) | 111 (46.8) | 21 (8.9) | |
| Private | 20 (24.1) | 33 (39.8) | 30 (36.1) | | 12 (12.1) | 39 (39.4) | 48 (48.5) | |
| Location of facility | | | | <0.001 | | | | <0.001 |
| Urban | 42 (25.5) | 84 (50.9) | 39 (23.6) | | 46 (21.9) | 104 (49.5) | 60 (28.6) | |
| Rural | 46 (52.3) | 37 (42.0) | 5 (5.7) | | 71 (56.3) | 46 (36.5) | 9 (7.1) | |
| Administrative division | | | | 0.03 | | | | 0.12 |
| Barishal | 9 (22.0) | 27 (65.9) | 5 (12.2) | | 15 (30.0) | 29 (58) | 6 (12.0) | |
| Chattogram | 28 (40.0) | 31 (44.3) | 11 (15.7) | | 29 (33.3) | 30 (34.5) | 28 (32.2) | |
| Dhaka | 15 (25.0) | 27 (45.0) | 18 (30.0) | | 15 (34.1) | 19 (43.2) | 10 (22.7) | |
| Khulna | 7 (29.2) | 12 (50.0) | 5 (20.8) | | 15 (41.7) | 15 (41.7) | 6 (16.7) | |
| Rajshahi | 11 (55.0) | 7 (35.0) | 2 (10.0) | | 11 (34.4) | 13 (40.6) | 8 (25.0) | |
| Rangpur | 6 (37.5) | 8 (50.0) | 2 (12.5) | | 6 (20) | 17 (56.7) | 7 (23.3) | |
| Sylhet | 12 (54.5) | 9 (40.9) | 1 (4.5) | | 16 (44.4) | 18 (50) | 2 (5.6) | |
| Mymensingh | | | | | 10 (47.6) | 9 (42.9) | 2 (9.5) | |
| External supervisory visit to facility | | | | 0.31 | | | | 0.01 |
| Not received | 1 (12.5) | 2 (25.0) | 5 (62.5) | | 1 (20.0) | 0 (0.0) | 4 (80.0) | |
| Received, within the past 6 months | 78 (35.5) | 106 (48.2) | 36 (16.4) | | 106 (35.7) | 139 (46.8) | 52 (17.5) | |
| Received, more than 6 months ago | 9 (36.0) | 13 (52.0) | 3 (12.0) | | 10 (29.4) | 11 (32.4) | 13 (38.2) | |
| Routine user fee or charges for client service | | | | | | | | <0.001 |
| Not available | 53 (55.2) | 37 (38.5) | 6 (6.3) | | 63 (71.6) | 24 (27.3) | 1 (1.1) | |
| Available | 35 (22.3) | 84 (53.5) | 38 (24.2) | | 54 (21.8) | 126 (50.8) | 68 (27.4) | |
| 24-hours staff coverage | | | | | | | | <0.001 |
| Not available | 36 (70.6) | 14 (27.5) | 1 (2.0) | | 38 (66.7) | 19 (33.3) | 0 (0.0) | |
| Available | 52 (25.7) | 107 (53.0) | 43 (21.3) | | 79 (28.3) | 131 (47.0) | 69 (24.7) | |
| System to elicit clients’ opinions about the health facility or its services | | | | 0.01 | | | | <0.001 |
| Not available | 48 (47.1) | 38 (37.3) | 16 (15.7) | | 44 (55.7) | 23 (29.1) | 12 (15.2) | |
| Available | 40 (26.5) | 83 (55.0) | 28 (18.5) | | 73 (28.4) | 127 (49.4) | 57 (22.2) | |
| Routine quality assurance activities | | | | 0.01 | | | | <0.001 |
| Not performed | 40 (50.0) | 29 (36.3) | 11 (13.8) | | 53 (58.2) | 26 (28.6) | 12 (13.2) | |
| Performed | 48 (28.1) | 91 (53.2) | 32 (18.7) | | 62 (25.6) | 123 (50.8) | 57 (23.6) | |
| Guideline for the diagnosis and management of cardio-vascular disease | | | | 0.08 | | | | 0.09 |
| Available | 14 (23.0) | 35 (57.4) | 12 (19.7) | | 8 (22.2) | 24 (66.7) | 4 (11.1) | |
| Not available | 74 (38.5) | 86 (44.8) | 32 (16.7) | | 109 (36.3) | 126 (42.0) | 65 (21.7) | |
## Factors associated with EM-CVD availability
Table 2 indicates the factors associated with the EM-CVD availability using the Poisson regression model. The likelihood of EM-CVD availability scores was $81\%$ more among the private facilities compared to public facilities in 2014 (relative risk (RR): 1.81, confidence interval (CI): 1.25–2.62, p-value: < 0.001). Also, similar findings were found for private facilities compared to public facilities in 2017 (RR: 2.59, CI: 2.02–3.32, p-value: <0.001). The likelihood of EM-CVD availability was $33\%$ lower in rural facilities than in urban sections (RR: 0.67, CI: 0.45–0.98, p-value: 0.04). The same result was also obtained in the 2017 survey (RR: 0.64, CI: 0.46–0.89, p-value: 0.01). When compared to facilities without 24-hour staff coverage, the availability of EM-CVD was 4.11 higher in 2014 (RR: 4.11, CI: 2.11–8.02, p-value: 0.001) and 5.18 higher in 2017 (RR: 5.18, CI: 2.93–9.16, p-value: 0.001). In 2014, facilities from the Sylhet division (RR: 0.44, CI: 0.23–0.85, p-value: 0.02) had lower odds of EM-CVD availability than facilities from other divisions. Moreover, the availability of EM-CVD was higher in the facilities that had routine user fees or charges for client service (RR: 1.67, CI: 1.02, 2.73, p-value: 0.04) and performed routine quality assurance activities (RR: 1.52, CI: 1.11, 2.09, p-value: 0.01) compared to their counterparts.
**Table 2**
| Factors | BHFS 2014 | BHFS 2014.1 | BHFS 2017 | BHFS 2017.1 |
| --- | --- | --- | --- | --- |
| Factors | RR (95% CI) | P-value | RR (95% CI) | P-value |
| Managing authority | | | | |
| Public | Reference | | Reference | |
| Private | 1.81 (1.25, 2.62) | <0.001 | 2.59 (2.02, 3.32) | <0.001 |
| Location of facility | | | | |
| Urban | Reference | | Reference | |
| Rural | 0.67 (0.45, 0.98) | 0.04 | 0.64 (0.46, 0.89) | 0.01 |
| Administrative division | | | | |
| Barishal | Reference | | Reference | |
| Chattogram | 0.56 (0.35, 0.9) | 0.02 | 0.99 (0.72, 1.36) | 0.95 |
| Dhaka | 0.66 (0.41, 1.08) | 0.10 | 0.86 (0.57, 1.3) | 0.48 |
| Khulna | 0.64 (0.37, 1.12) | 0.12 | 0.71 (0.44, 1.15) | 0.17 |
| Rajshahi | 0.54 (0.22, 1.36) | 0.19 | 0.92 (0.62, 1.36) | 0.66 |
| Rangpur | 0.52 (0.28, 0.98) | 0.04 | 1 (0.70, 1.44) | 0.98 |
| Sylhet | 0.44 (0.23, 0.85) | 0.02 | 1.04 (0.68, 1.58) | 0.87 |
| Mymensingh | | | 0.75 (0.45, 1.24) | 0.25 |
| External supervisory visit to facility | | | | |
| Not received | Reference | | Reference | |
| Received, within the past 6 months | 0.88 (0.55, 1.4) | 0.59 | 0.74 (0.49, 1.13) | 0.16 |
| Received, more than 6 months ago | 0.91 (0.49, 1.67) | 0.75 | 0.82 (0.55, 1.21) | 0.31 |
| Routine user fee or charges for client service | | | | |
| Not available | Reference | | Reference | |
| Available | 1.17 (0.76, 1.82) | 0.47 | 1.67 (1.02, 2.73) | 0.04 |
| 24-hours staff coverage | | | | |
| Not available | Reference | | Reference | |
| Available | 4.11 (2.11, 8.02) | <0.001 | 5.18 (2.93, 9.16) | <0.001 |
| System to elicit clients’ opinions about the health facility or its services | | | | |
| Not available | Reference | | Reference | |
| Available | 0.96 (0.66, 1.41) | 0.85 | 1.03 (0.78, 1.37) | 0.82 |
| Routine quality assurance activities | | | | |
| Not performed | Reference | | Reference | |
| Performed | 1.03 (0.71, 1.5) | 0.87 | 1.52 (1.11, 2.09) | 0.01 |
| Guideline for the diagnosis and management of cardio-vascular disease | | | | |
| Available | Reference | | Reference | |
| Not available | 0.77 (0.54, 1.1) | 0.14 | 0.84 (0.56, 1.27) | 0.41 |
## Discussion
The current study examined the availability of EM-CVD and investigated facility characteristics connected with the availability of these medicines. Between 2014 and 2017, calcium channel blockers and aspirin became more available, while ACE inhibitors (enalapril), thiazides, and beta-blockers became less available. We identified a significant association between EM-CVD availability and managing authority of the facility, facility location, 24-hour staff coverage, and quality assurance activities.
The availability of EM-CVD was low in some countries (e.g., ACE inhibitors: $16.7\%$ in Uganda, $10.0\%$ in Tanzania; thiazide: $4.4\%$ in Nepal, $5.0\%$ in Tanzania; Beta blockers: $20.4\%$ in Uganda, $18.0\%$ in Nepal, $19.0\%$ in Tanzania; calcium channel blockers: $32.7\%$ in Uganda, $11.2\%$ in Nepal, $34.0\%$ in Tanzania; Aspirin: $9.9\%$ in Nepal, $74.0\%$ in Tanzania and statin: $3.1\%$ in Uganda) [21–23] that concur with the findings of this study. The consistency of these studies’ results may be attributable to the similar methodology adopted; both study utilized data from the nationally representative sample obtained by the DHS programme. Consequently, the interview questionnaires were nearly identical. Additionally, these countries have some of the same socio-economic factors because they are both LMICs. In addition, we observed a change in the availability of EM-CVD at HCFs in Bangladesh from 2014 to 2017. The availability of aspirin, calcium channel blockers, and simvastatin/atorvastatin has increased since 2014, while the availability of beta blockers, thiazide, and ACE inhibitors has decreased. The downward trend of the CVD medicines could be caused by a lack of supply, bad stock management, a lack of experienced pharmacists, medicine theft, bad transport and distribution systems, or delayed stock level ordering and monitoring.
Consistent with a prior study [22], our study revealed that private facilities are more likely to have EM-CVD than public facilities. The availability of essential medicine is typically a mandate of the government and it is less available in public facilities frequently due to a poorly funded procurement strategy. However, with several business techniques, the private sector looks to have a considerable influence on the procurement of essential medicines [27]. As a result, these medications are less likely to be available for free at public facilities, but they are readily available at a high cost at private facilities [22]. Furthermore, because private institutions are not subsidized and rely on revenues from clients, they are more interested in providing quality services and meeting clients’ healthcare demands. In doing this, they will be able to develop satisfied and devoted clients who will return to the facility in the future for their requirements and who will also act as a source of referrals for friends and family, thus ensuring the long-term survival of private hospitals [28]. Therefore, the poor availability of essential medicines, specifically in the public sector, needs more concentration and endeavor by the relevant authorities to ensure these medicines are stocked in the facilities [28–31].
Between urban and rural HCFs, previous studies observed significant inequalities in service availability and service provision of health care [29–32]. Consistent with prior studies, the present study found that rural facilities were less likely to have high EM-CVD availability than urban facilities. This finding may be because the healthcare necessities of individuals’ lives in rural areas are different from those in urban areas, and rural areas often experience a lack of access to healthcare. Furthermore, some of the HCFs located in remote and rural areas are affected by little funding and remaining resource constraints, which cause ineptitude in the management of supply logistic systems that result in the poor availability of essential medicines and equipment [33]. Hence, increased efforts are required to draw the attention of policymakers and health service planners focusing on rural HCF to improve CVD medicine availability where most individuals live in rural areas [33].
In 2014, we found that the administrative division of the facility was a strong predictor of the availability of EM-CVD. In Bangladesh, there are distinct divisions where EM-CVD is available. In 2014, facilities from all other divisions had a lower percentage of EM-CVD availability compared to those in Barishal. This may be due to the fact that Barishal had good CVD service provision, making it more probable that these facilities would have EM-CVD. Further research may be done to investigate the underlying causes of this variation among the HCFs in Bangladesh since the reasons of these differences are yet unknown. Additionally, targeted investments are required to enhance service delivery in regions where EM-CVD is less available.
A policy option that numerous developing nations have employed to collect money to fulfill the rising demand for health care services is the partial or full payment for health services by clients [34]. In line with a study [35], we found that the availability of EM-CVD was significantly linked to the facility’s regular user fees, where facilities that charge regular user fees or costs for client services are more likely to have EM-CVD than those that don’t in 2017. User fees are a way for health systems to get more money to improve the quality of health services and offer more services. This could lead to more medicines being available.
Staff scheduling is the method of forming duty timetables for its staff and is a vital portion of staff management of a HCF. Patient satisfaction improves when an HCF’s ability to have excellent personnel on duty at the proper time is an important factor in attending to the patient [36]. Our data reveals that facilities with 24-hour staff scheduling had a higher chance of having EM-CVD than facilities without 24-hour staff scheduling. Facilities with 24-hour personnel coverage may be more likely to provide high-quality services and better meet patients’ needs. As a result, these facilities are more likely to increase service availability, perhaps resulting in high EM-CVD availability. Hence, these facilities are more likely to upgrade the availability of services, which may result in their high availability of EM-CVD.
A process of ensuring and maintaining a high level of service in various HCFs is known as "quality assurance" (QA) [37]. Regular quality assurance activities were significantly associated with EM-CVD availability in 2017, consistent with studies that looked at basic emergency obstetric and newborn care in Tanzania and general services in Bangladesh [38, 39]. Both found that facilities with regular quality assurance activities had better preparedness scores. According to the results of this study, facilities that participated in routine quality assurance activities had a higher chance of having EM-CVD available than facilities that did not participate in quality assurance activities. This is because QA entails continuous monitoring and feedback to improve the delivery of services [40]. Therefore, based on recommendations from the quality assurance team, these facilities have a greater probability of improving the provision of services, which will result in the increased availability of medicines. Despite this, the uptake of quality assurance activities in Bangladeshi health care facilities remains low because only a few facilities conduct routine QA activities and maintain records of these activities [41].
## Strength and limitations
This research has several advantages. To begin with, this is the first study of its kind in Bangladesh, according to our knowledge, and it provides critical insight into the essential medicine availability for CVD patients. The findings of this study, which used a nationally representative sample of HCFs in Bangladesh, give crucial information regarding the factors that influence drug availability for treating CVD. Second, the estimates in this study have been adjusted to account for cluster effects and sample weights because SPA data is acquired using a complex sampling approaches. Finally, we did a comparative analysis utilizing the 2014 and 2017 BHFS to see how changes in EM-CVD availability have changed throughout Bangladesh’s health facilities.
Our approach has some significant drawbacks. To begin with, this study is unable to infer causality. Second, data on essential medicine cost is not collected by the BHFS. As a result, we were unable to directly address the issue of access, which is linked to both availability and cost. Third, BHFS gathered information on how many drugs a particular institution had in stock on a given day. This method is unable to account for changes in supply over time. Fourth, despite a large number of zeros in the outcome variable, we did not use zero-inflated count models in this investigation. Finally, improving NCD outcomes requires taking a close look at prescription pattern and provider behavior. Facilities might not store some of these medications because the healthcare professionals who work there aren’t writing prescriptions for them. We were unable to examine the prescription pattern because these were absent from the health facility surveys.
## Conclusions
For health systems and individuals in Bangladesh and other LMICs that will face the unprecedented burden of CVD in the coming decades, it is crucial to understand the availability of essential medicines and health facility characteristics associated with the availability of these medicines. We found that only a few health facilities possessed adequate necessary medicines for treating CVD. To increase the availability of EM-CVD and to reduce the rising burden of CVDs in LMICs like Bangladesh, relevant authorities should consider a variety of policy options, including a focus on public and rural facilities, 24-hour staff scheduling, and conducting QA activities.
## References
1. Kaptoge S, Pennells L, De Bacquer D, Cooney MT, Kavousi M. **World Health Organization cardiovascular disease risk charts: revised models to estimate risk in 21 global regions**. *The Lancet Global Health* (2019.0) **7** e1332-e1345. DOI: 10.1016/S2214-109X(19)30318-3
2. Organization WH. *Cardiovascular diseases (CVDs) Key facts* (2021.0)
3. Bloom DE, Cafiero E, Jané-Llopis E, Abrahams-Gessel S, Bloom LR. **The global economic burden of noncommunicable diseases.**. *Program on the Global Demography of Aging* (2012.0)
4. Bloom DE, Chisholm D, Jané-Llopis E, Prettner K, Stein A. **From burden to" best buys": reducing the economic impact of non-communicable disease in low-and middle-income countries.**. *Program on the Global Demography of Aging* (2011.0)
5. Ahsan Karar Z, Alam N, Kim Streatfield P. **Epidemiological transition in rural Bangladesh, 1986–2006**. *Global health action* (2009.0) **2** 1904
6. World Health O.. *National STEPS survey for non-communicable diseases risk factors in Bangladesh* (2018.0) **2018**
7. Engelgau MM, El-Saharty S, Kudesia P, Rajan V, Rosenhouse S. *Capitalizing on the demographic transition: tackling noncommunicable diseases in South Asia* (2011.0)
8. El-Saharty S, Ahsan KZ, Koehlmoos TLP, Engelgau MM. *Tackling noncommunicable diseases in Bangladesh: now is the time* (2013.0)
9. Holloway KA. *Bangladesh: Pharmaceuticals in health care delivery. Mission Report* (2010.0) 17-24
10. Kasonde L, Tordrup D, Naheed A, Zeng W, Ahmed S. **Evaluating medicine prices, availability and affordability in Bangladesh using World Health Organisation and Health Action International methodology**. *BMC health services research* (2019.0) **19** 383. DOI: 10.1186/s12913-019-4221-z
11. Manikandan S.. **Are we moving towards a new definition of essential medicines?**. *Journal of pharmacology & pharmacotherapeutics* (2015.0) **6** 123. DOI: 10.4103/0976-500X.162008
12. Hogan DR, Stevens GA, Hosseinpoor AR, Boerma T. **Monitoring universal health coverage within the Sustainable Development Goals: development and baseline data for an index of essential health services**. *The Lancet Global Health* (2018.0) **6** e152-e168. DOI: 10.1016/S2214-109X(17)30472-2
13. Penchansky R, Thomas JW. **The concept of access: definition and relationship to consumer satisfaction.**. *Medical care* (1981.0) 127-140. DOI: 10.1097/00005650-198102000-00001
14. Nascimento RCRMd, Álvares J, Guerra Junior AA, Gomes IC, Costa EA. **Availability of essential medicines in primary health care of the Brazilian Unified Health System.**. *Revista de saude publica* (2017.0) **51** 10s. DOI: 10.11606/S1518-8787.2017051007062
15. Kuwawenaruwa A, Wyss K, Wiedenmayer K, Metta E, Tediosi F. **The effects of medicines availability and stock-outs on household’s utilization of healthcare services in Dodoma region, Tanzania**. *Health policy and planning* (2020.0) **35** 323-333. DOI: 10.1093/heapol/czz173
16. van Mourik MSM, Cameron A, Ewen M, Laing RO. **Availability, price and affordability of cardiovascular medicines: a comparison across 36 countries using WHO/HAI data**. *BMC Cardiovascular disorders* (2010.0) **10** 25. DOI: 10.1186/1471-2261-10-25
17. Cameron A, Roubos I, Ewen M, Mantel-Teeuwisse AK, Leufkens HGM. **Differences in the availability of medicines for chronic and acute conditions in the public and private sectors of developing countries**. *Bulletin of the World Health Organization* (2011.0) **89** 412-421. DOI: 10.2471/BLT.10.084327
18. Khatib R, McKee M, Shannon H, Chow C, Rangarajan S. **Availability and affordability of cardiovascular disease medicines and their effect on use in high-income, middle-income, and low-income countries: an analysis of the PURE study data**. *The Lancet* (2016.0) **387** 61-69
19. Jingi AM, Noubiap JJN, Ewane Onana A, Nansseu JRN, Wang B. **Access to diagnostic tests and essential medicines for cardiovascular diseases and diabetes care: cost, availability and affordability in the West Region of Cameroon.**. *PLoS One* (2014.0) **9** e111812. DOI: 10.1371/journal.pone.0111812
20. Kaiser AH, Hehman L, Forsberg BC, Simangolwa WM, Sundewall J. **Availability, prices and affordability of essential medicines for treatment of diabetes and hypertension in private pharmacies in Zambia.**. *PloS one* (2019.0) **14** e0226169. DOI: 10.1371/journal.pone.0226169
21. Bintabara D, Ngajilo D. **Readiness of health facilities for the outpatient management of non-communicable diseases in a low-resource setting: an example from a facility-based cross-sectional survey in Tanzania.**. *BMJ open* (2020.0) **10** e040908. DOI: 10.1136/bmjopen-2020-040908
22. Armstrong-Hough M, Kishore SP, Byakika S, Mutungi G, Nunez-Smith M. **Disparities in availability of essential medicines to treat non-communicable diseases in Uganda: a Poisson analysis using the Service Availability and Readiness Assessment**. *PloS one* (2018.0) **13** e0192332. DOI: 10.1371/journal.pone.0192332
23. Ghimire U, Shrestha N, Adhikari B, Mehata S, Pokharel Y. **Health system’s readiness to provide cardiovascular, diabetes and chronic respiratory disease related services in Nepal: analysis using 2015 health facility survey.**. *BMC public health* (2020.0) **20** 1-15. PMID: 31898494
24. Biswas T, Haider MM, Gupta RD, Uddin J. **Assessing the readiness of health facilities for diabetes and cardiovascular services in Bangladesh: a cross-sectional survey**. *BMJ open* (2018.0) **8** e022817. DOI: 10.1136/bmjopen-2018-022817
25. 25National Institute of Population R, Training AfC, Population R, International ICF. (2016) Bangladesh health facility survey, 2014. NIPORT, ACPR, and ICF International Dhaka, Bangladesh.. *Bangladesh health facility survey, 2014* (2016.0)
26. 26National Institute of Population R, Training AfC, Population R, International ICF. (2019) Bangladesh health facility survey, 2017.
27. Bazargani YT, de Boer A, Leufkens HGM, Mantel-Teeuwisse AK. **Selection of essential medicines for diabetes in low and middle income countries: a survey of 32 national essential medicines lists.**. *PloS one* (2014.0) **9** e106072. DOI: 10.1371/journal.pone.0106072
28. Andaleeb SS. **Public and private hospitals in Bangladesh: service quality and predictors of hospital choice**. *Health policy and planning* (2000.0) **15** 95-102. DOI: 10.1093/heapol/15.1.95
29. Oyekale AS. **Assessment of primary health care facilities’ service readiness in Nigeria.**. *BMC health services research* (2017.0) **17** 1-12. PMID: 28049468
30. Robertson J, Macé C, Forte G, de Joncheere K, Beran D. **Medicines availability for non-communicable diseases: the case for standardized monitoring**. *Globalization and health* (2015.0) **11** 1-6. PMID: 25889826
31. Spasojevic N, Vasilj I, Hrabac B, Celik D. **Rural–urban differences in health care quality assessment**. *Materia socio-medica* (2015.0) **27** 409. DOI: 10.5455/msm.2015.27.409-411
32. Bintabara D, Mpondo BCT. **Preparedness of lower-level health facilities and the associated factors for the outpatient primary care of hypertension: Evidence from Tanzanian national survey**. *PloS one* (2018.0) **13** e0192942. DOI: 10.1371/journal.pone.0192942
33. Wiedenmayer K.. *Access to medicines. Medicine supply: lessons learnt in Tanzania and Mozambique. Swiss Tropical Institute Basel and Swiss Agency for Development and Cooperation Berne* (2004.0)
34. Kangwana BB, Njogu J, Wasunna B, Kedenge SV, Memusi DN. **Malaria drug shortages in Kenya: a major failure to provide access to effective treatment**. *The American journal of tropical medicine and hygiene* (2009.0) **80** 737. PMID: 19407116
35. Hakim S, Chowdhury MAB, Ahmed NU, Uddin MJ. **The availability of essential medicines for diabetes at health facilities in Bangladesh: evidence from 2014 and 2017 national surveys.**. *BMC Health Services Research* (2022.0) **22** 1-11. PMID: 34974828
36. Gondane MS, Zanwar DR. **Staff scheduling in health care systems**. *IOSR Journal of Mechanical and Civil Engineering* (2012.0) **1** 28-40
37. Busari JO. **Comparative analysis of quality assurance in health care delivery and higher medical education**. *Advances in Medical Education and Practice* (2012.0) **3** 121. DOI: 10.2147/AMEP.S38166
38. Yusuf SS, Acharya K, Ahmed R, Ahmed A. **Understanding general health service readiness and its correlates in the health facilities of Bangladesh: evidence from the Bangladesh Health Facility Survey 2017**. *Journal of Public Health* (2021.0) 1-12
39. Bintabara D, Ernest A, Mpondo B. **Health facility service availability and readiness to provide basic emergency obstetric and newborn care in a low-resource setting: evidence from a Tanzania National Survey**. *BMJ open* (2019.0) **9** e020608. DOI: 10.1136/bmjopen-2017-020608
40. De Jonge V, Nicolaas JS, van Leerdam ME, Kuipers EJ. **Overview of the quality assurance movement in health care**. *Best practice & research clinical gastroenterology* (2011.0) **25** 337-347. DOI: 10.1016/j.bpg.2011.05.001
41. 41National Institute of Population R, Training AfC, Population R, International ICF. (2020) Bangladesh health facility survey, 2017.
NIPORT, ACPR, and ICF International Dhaka, Bangladesh.. *Bangladesh health facility survey, 2017.* (2020.0)
|
---
title: 'Large gains in schooling and income are possible from minimizing adverse birth
outcomes in 121 low- and middle-income countries: A modelling study'
authors:
- Mia M. Blakstad
- Nandita Perumal
- Lilia Bliznashka
- Mark J. Lambiris
- Günther Fink
- Goodarz Danaei
- Christopher R. Sudfeld
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021521
doi: 10.1371/journal.pgph.0000218
license: CC BY 4.0
---
# Large gains in schooling and income are possible from minimizing adverse birth outcomes in 121 low- and middle-income countries: A modelling study
## Abstract
While the global contributions of adverse birth outcomes to child morbidity and mortality is relatively well documented, the potential long-term schooling and economic consequences of adverse birth outcomes has not been estimated. We sought to quantify the potential schooling and lifetime income gains associated with reducing the excess prevalence of adverse birth outcomes in 121 low- and middle-income countries. We used a linear deterministic model to estimate the potential gains in schooling and lifetime income that may be achieved by attaining theoretical minimum prevalence of low birthweight, preterm birth and small-for-gestational age births at the national, regional, and global levels. We estimated that potential total gains across the 121 countries from reducing low birthweight to the theoretical minimum were 20.3 million school years ($95\%$ CI: 6.0,34.8) and US$ 68.8 billion ($95\%$ CI: 20.3,117.9) in lifetime income gains per birth cohort. As for preterm birth, we estimated gains of 9.8 million school years ($95\%$ CI: 1.5,18.4) and US$ 41.9 billion ($95\%$ CI: 6.1,80.9) in lifetime income. The potential gains from small-for-gestational age were 39.5 million ($95\%$ CI: 19.1,60.3) school years and US$113.6 billion ($95\%$ CI: 55.5,174.2) in lifetime income gained. In summary, reducing the excess prevalence of low birthweight, preterm birth or small-for-gestational age births in low- and middle-income countries may lead to substantial long-term human capital gains in addition to benefits on child mortality, growth, and development as well as on risk of non-communicable diseases in adults and other consequences across the life course.
## Introduction
Globally, it is estimated that $14.6\%$ of all live births are low birthweight (LBW, birthweight <2500g), $10.6\%$ are preterm (PTB, <37 weeks of gestation), and $27.0\%$ are small-for-gestational age (SGA, birth weight below the 10th percentile of gestational age- and sex relative to the standard reference population) [1–3]. While LBW, PTB, and SGA can co-occur in the same child the population-level prevalence of these adverse birth outcomes by country and setting [4]. Complications from adverse birth outcomes are major causes of death in children under 5 years of age with preterm birth as the current leading cause of global child mortality [1–4]. In addition, prematurity and SGA (an indicator of intrauterine growth restriction) are associated with increased risk of child linear growth faltering and suboptimal development [5, 6]. LBW and PTB may also have long-term consequences including an increase in the risk of non-communicable diseases in adulthood, including heart disease and diabetes [7–9].
While the consequences of adverse birth outcomes on child health, nutrition, and development are well-documented, evidence on the potential downstream effects on human capital consequences are much more limited [7]. To the best of our knowledge, only one study has estimated the potential economic benefits of eliminating LBW in LMICs. This study estimated a gain of $US 510 per infant that moved from LBW to non-LBW status due to reductions in costs associated with infant and child mortality, illness and care, stunting, suboptimal cognitive ability, and increased risk of chronic disease [10]. However, this study assumed the same $15\%$ prevalence of LBW across all LMICs and did not estimate economic benefits of reducing other adverse birth outcomes such as PTB and SGA. In addition, this prior study modeled the scenario of reducing LBW prevalence to $0\%$ globally which is biologically implausible given that the prevalence of LBW was $3.2\%$ among births in the INTERGROWTH-21st fetal growth reference study [11]. In this modeling paper, we quantified the potential gains in school years and lifetime income that may be gained from reducing the current prevalence of LBW, SGA, and PTB in LMICs to the theoretical minimum prevalence.
## Methods
We used a population attributable risk framework to estimate the potential human capital gains in school years and lifetime income for reducing the excess prevalence of LBW, SGA, PTB from their 2015 national prevalence in 121 LMICs to the theoretical minimum prevalence levels [12]. We used a framework previously conceptualized to estimate the potential human capital benefits of scaling up prenatal nutrition interventions on schooling and lifetime income [12] *In this* study we applied a theoretical approach to estimate the potential benefits of reducing the prevalence of adverse birth outcomes to the minimum prevalence observed in a population of healthy pregnancies.
## Counterfactual exposure distribution
Risk assessment requires a reference scenario, a hypothetical counterfactual comparison group. The theoretical minimum-risk exposure distribution (TMRED) is a distribution of exposures corresponding to the lowest levels of risk. In the case of this analysis, the TMRED represents the theoretical prevalence of adverse birth outcomes in a population of healthy pregnancies in low- and middle-income countries. We used evidence from the INTERGROWTH-21st population-based study to define the theoretical minimal prevalence of LBW, SGA and PTB [11]. From a cohort of 59,137 women across eight countries, the project selected 20,486 women meeting strict individual eligibility criteria for pregnancies with low risk of impaired fetal growth to construct normative anthropometric standards for newborn size for use across multiethnic populations. In the absence of socio-economic, health, or nutritional risk factors, LBW prevalence was $3.2\%$ and PTB prevalence was $5.5\%$. In addition, by definition, the theoretical prevalence of SGA (defined by the 10th percentile) in a population of healthy pregnancies is $10\%$. Notably, these adverse birth outcomes are not mutually exclusive (i.e., infants born LBW and/or PTB can be SGA or appropriate-for-gestational age [13]) and because we lacked country-level estimates of their joint distributions, we estimated the potential human capital impacts of LBW, PTB, and SGA births independently. For countries with prevalences at or below the theoretical minimum, the expected gains in school years and lifetime income were set to 0. Therefore, for each country, we estimated the absolute percentage point difference between current excess prevalence of each adverse birth outcome and the theoretical optimal exposure risk threshold.
## Assessment of exposure distribution
This analysis uses secondary datasets that are publicly available through the referenced sources. We included 121 LMICs as defined by the World Bank July 2019 income classification [14]. We obtained the most recent country-specific prevalence estimates for LBW, PTB, and SGA from global modeling analyses in 2015 (S1 Table) [1–3]. We used 2010 prevalence estimates for LBW when newer estimates were not available [3]. For countries for which prevalence of birth outcomes were unavailable, we performed random-effects meta-analyses to impute the sub-regional level averages and variances, as defined by the Global Burden of Disease (GBD) regions [15]. Where 2010 LBW prevalences were reported without associated variance, we used the largest variance in the GBD subregion from the 2015 data [1]. Eleven Eastern European countries were excluded from the analysis because SGA prevalences were missing and no other GBD subregion prevalence estimates were available. We also excluded South Sudan because SGA prevalence estimates were not available and there was too high variation in SGA estimates of the neighboring counties within the GBD subregion to confidently impute the prevalence [3].
## Effect of adverse birth outcomes on schooling
We conducted a de novo systematic review and meta-analysis of the link between birthweight and human capital outcomes later in life [16]. The majority of studies identified in the systematic review were from high-income countries; as such, we used random-effects meta-analysis to derive a pooled estimate of the effect of LBW on attained schooling by combining evidence from the de novo systematic review and data from five birth cohorts in LMICs [7, 17]. It was estimated that on average LBW-born individuals complete -0.29 ($95\%$ CI: -0.48, -0.10) fewer school years (Table 1) [12]. For PTB and SGA, estimates were available from the COHORTS collaboration [7]. PTB-born individuals were estimated to complete an average of -0.32 ($95\%$ CI: -0.57, -0.06) fewer school years, while SGA-born individuals complete an average of -0.41 ($95\%$ CI: -0.62, -0.20) fewer school years (Table 1).
**Table 1**
| Data input | Value (Mean difference ± 95% CI) | Source | Method |
| --- | --- | --- | --- |
| Estimated effect on birth outcome on schooling attainment | | | |
| LBW | -0.2894 years (-0.483, -0.096) | Adair et al. 2013 | Pooled estimate combining evidence from de novo review of the economics literature and Adair et al. estimates |
| PTB | -0.317 years (-0.572, -0.062) | Adair et al. 2013 | Unpublished estimates from Adair et al. from COHORTS |
| SGA | -0.41 years (-0.62, -0.20) | Stein et al. 2013 | From COHORTS data |
| Estimates for returns on wage gains (%) per additional year of schooling | Varies by country. Examples: Bangladesh 7.1%, Burkina Faso 6.3%. | Fink et al., 2016 | Fink et al. for LMICs outside Europe. Peet et al. for LMICs in Europe. |
| Estimates for returns on wage gains (%) per additional year of schooling | Varies by country. Examples: Bangladesh 7.1%, Burkina Faso 6.3%. | Peet et al. 2015 | Fink et al. for LMICs outside Europe. Peet et al. for LMICs in Europe. |
We multiplied the effect sizes linking birth outcomes to educational attainment with the absolute percentage point change in each adverse birth outcome to obtain the average school years gained per child due to reductions in LBW, PTB, or SGA prevalence independently. We then used the school years gained per child, multiplied by each country’s 2015 birth cohort size and the country-specific probability that the child would survive until age 25, to quantify the estimated total gains in years of schooling completed [18].
## Effect of education on net present value of future income
We quantified the gains in adult earning potential using effect estimates of returns in income for each added year of schooling for each country [19]. To calculate the net present value of future lifetime income we summed annual income between ages 20 and 59 for each country. In line with previous estimations, annual incomes were estimated at $\frac{2}{3}$rds of gross domestic product as reported by the World Development Indicators [20, 21]. We used 2010 constant US$, corrected for a $3\%$ discounting rate, a $2\%$ real income growth rate, and the country-specific survival probabilities for each working year based on the sex-average survival probabilities estimated by the Institute for Health Metrics and Evaluation [22]. The total gains in lifetime income per birth cohort was estimated by multiplying the gains in income per child by the country-specific birth cohort size. We further estimated the percent increase in net present value of lifetime income per child by dividing the increase in per child lifetime income by the net present value of working from 20 to 59 years. Five countries (Dominica, Republic of Kosovo, Marshall Islands, Nauru, and Tuvalu) were excluded from the analysis because they lacked data on the number of live births and probability of survival.
A working example for our methods available in the S1 Text.
## Uncertainty estimation
We used a first order Monte Carlo simulation with 1,000 iterations to propagate uncertainty of each parameter of the model. Uncertainty estimates were available from published estimates for LBW, PTB, and SGA prevalences, for the effect sizes of the association between adverse birth outcomes and years of schooling, and between adverse birth outcomes and returns on income per additional year of education. All model parameters were assumed to be normally distributed and to be independent as they are derived from different studies. The $95\%$ confidence intervals were drawn from the 2.5th and 95.5th percentile of from the distribution of 1,000 iterations. All analyses were performed using Stata 16.1 Statistical Software package (StataCorp LP).
## Results
Across the 121 LMICs included in the analysis, median prevalences of LBW, PTB, and SGA were $11.5\%$ (IQR: 8.2,15.4), $10.4\%$ (9.8–12.0), and $20.4\%$ (13.6–25.1), respectively (Table 2).
**Table 2**
| Birth outcome | Birth cohort size1 | Low birthweight (<2500g) | Low birthweight (<2500g).1 | Low birthweight (<2500g).2 | Preterm birth (<37 weeks’ gestation) | Preterm birth (<37 weeks’ gestation).1 | Preterm birth (<37 weeks’ gestation).2 | Small for gestational age (<10th percentile) | Small for gestational age (<10th percentile).1 | Small for gestational age (<10th percentile).2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Birth outcome | Birth cohort size1 | Prevalence in 2015 Median, IQR2 | Years of Schooling3 (mln) | Income gains4 (US$ bln) | Prevalence Median, IQR | Years of Schooling (mln) | Income gains (US$ bln) | Prevalence Median, IQR | Years of Schooling (mln) | Income gains (US$ bln) |
| Central Asia (n = 9) | 10.0 | 5.5 (5.3–6.7) | 0.67 (0.01,0.12) | 0.3 (0.1,0.6) | 10.4 (10.2–10.4) | 0.12 (0.02,0.23) | 0.3 (0.0,0.7) | 15.4 (12.9–16.5) | 0.18 (0.07,0.33) | 1.0 (0.3,2.1) |
| Latin America and Caribbean (n = 24) | 50.6 | 9.7 (7.9–11.1) | 0.82 (0.25,1.43) | 14.1 (4.3,24.5) | 9.8 (9.0–9.8) | 0.63 (0.09,1.19) | 12.0 (1.7,23.9) | 13.4 (11.3–16.0) | 0.57 (0.25,1.06) | 8.7 (2.6,19.4) |
| North Africa and Middle East (n = 14) | 57.9 | 9.3 (6.9–14.7) | 1.15 (0.34,2.00) | 6.6 (1.9,11.5) | 10.4 (10.3–13.4) | 1.01 (0.15,2.21) | 5.8 (0.9,13.2) | 13.4 (9.8–19.7) | 1.94 (0.94,3.09) | 9.7 (4.1,18.1) |
| South Asia (n = 5) | 168.8 | 26.0 (16.8–27.7) | 10.74 (3.09,19.42) | 24.1 (6.9,44.1) | 10.4 (6.8–16.4) | 3.66 (0.55,7.37) | 8.3 (1.3,16.7) | 39.6 (30.5–47.0) | 23.03 (11.18,35.85) | 52.4 (25.5,81.7) |
| Sub-Saharan Africa (n = 47) | 188.7 | 14.4 (12.0–16.9) | 5.35 (1.59,9.21) | 13.9 (4.2,24.5) | 12.0 (12.0–12.0) | 3.06 (0.46,5.92) | 8.3 (1.2,17.0) | 24.6 (21.9–29.0) | 10.27 (4.90,15.98) | 25.4 (12.0,38.5) |
| Southeast Asia, East Asia, and Oceania (n = 22) | 146.9 | 11.0 (8.2–12.4) | 1.92 (0.60,3.35) | 8.7 (2.5,16.3) | 10.0 (10.0–10.4) | 1.27 (0.21,2.49) | 6.3 (1.1,13.6) | 20.3 (17.5–23.9) | 3.31 (1.49,5.22) | 15.3 (6.7,27.4) |
| All LMICs (n = 121) | 622.9 | 11.5 (8.2–15.4) | 20.31 (5.98,34.78) | 68.8 (20.3,117.9) | 10.4 (9.8–12.0) | 9.80 (1.45,18.38) | 41.9 (6.1,80.9) | 20.4 (13.6,25.1) | 39.52 (19.13,60.31) | 113.6 (55.5,174.2) |
The estimated schooling and income gains attributed to excess LBW, PTB, and SGA is presented by region in Table 2 with country-specific estimates in S1 Table. When each country’s prevalence of LBW was reduced to the theoretical minimum of $3.2\%$, the estimated gains in educational attainment were 20.3 million ($95\%$ CI 5.98, 34.8) school years and US$ 68.8 billion ($95\%$ CI: 20.3,117.9) in net present value of lifetime income across all 121 LMICs. The largest absolute gains in schooling years were estimated in South Asia (10.7 million; $95\%$ CI: 3.1,19.4) and sub-Saharan Africa (5.4 million; $95\%$ CI: 1.6,9.2). India had the largest total potential gains from reductions in LBW with 7.8 million ($95\%$ CI: 2.1,15.0) school years and US$ 19.8 billion ($95\%$ CI: 5.2, 38.0) per birth cohort. In addition, Fig 1 present the estimated increases in educational attainment and lifetime earnings per child (scaled by population): The percentage increase in the net present value of lifetime income per child were largest in Sudan ($0.78\%$, $95\%$ CI: 0.22, 1.40), Morocco ($0.62\%$, $95\%$ CI: 0.18, 1.12), and the Comoros ($0.58\%$, $95\%$ CI: 0.18, 1.15). Potential gains were highest in Latin America and the Caribbean, with absolute gains in net present value in lifetime income per child born at US$ 279 per child born ($95\%$ CI: 84.4,85.5; Fig 2; LBW) and per-LBW child gains at US$ 3,042 ($95\%$ CI: 919, 5,284; Fig 3).
**Fig 1:** *Absolute gains in schooling and relative increase in annual income by country.Maps were made using the spData package in R and basemaps from https://www.naturalearthdata.com/.* **Fig 2:** *Population average gains in lifetime income per child born in US$, by region.* **Fig 3:** *Gains in lifetime income per child born low birthweight, preterm, or small-for-gestational age in US$, by region.Maps were made using the spData package in R.*
When each country’s prevalence of PTB was reduced to the theoretical minimum of $5.5\%$, the estimated gains in schooling years across all 121 countries were 9.8 million ($95\%$ CI: 1.5, 18.4) years, with the largest gains seen in South Asia (6.7 million school years; $95\%$ CI: 0.6,7.4). Overall gains in net present value of lifetime income were US$ 41.9 billion ($95\%$ CI: 6.1, 80.9), with South Asia (US$ 8.3 billion; $95\%$ CI: 1.4,16.7) and sub-Saharan Africa (US$ 8.3 billion; $95\%$ CI: 1.2,17.0) each contributing US$ 8.3 billion per birth cohort. Fig 1 shows that increases in educational attainment per 1,000 children were high across Africa and South Asia, whereas percentage increases in net present value of lifetime income were highest in the Southern Africa and South Asia regions. The three countries with the highest percentage increase in the net present value of lifetime income per child were South Africa ($0.34\%$, $95\%$ CI: 0.0, 0.9), Morocco ($0.33\%$, $95\%$ CI: 0.0, 1.2), and eSwatini ($0.30\%$, $95\%$ CI: 0.0,0.6). Potential gains were highest in Latin America and the Caribbean, with absolute gains in net present value in lifetime income per child born at US$ 2236 ($95\%$ CI: 33, 472; Fig 2; PTB) and per-LBW child gains at US$ 2,392 ($95\%$ CI: 339, 4,775; Fig 3).
When each country’s prevalence of SGA was reduced to the theoretical minimum of $10\%$, the estimated gains in schooling were 39.5 million ($95\%$ CI: 19.1,60.3) school years across 121 LMICs, with the largest gains seen in South Asia (23.0 million school years; $95\%$ CI: 11.2,35.9). Overall potential gains in net present value of lifetime income across all included LMICs was US$ 113.6 billion ($95\%$ CI: 55.5,174.2). Both increase in educational attainment per 1,000 children and percentage increase in net present value of lifetime income were highest in sub-Saharan Africa and South Asia; at the country-level, the percentage increase in net present value of lifetime wages was largest in Sudan ($1.27\%$, $95\%$ CI: 0.6,2.1), Pakistan ($1.14\%$, $95\%$ CI:0.5,1.8), and India ($1.13\%$, $95\%$ CI:0.5,1.8) (Fig 1). Absolute potential gains in net present value highest in South Asia, at US$ 311 per child born ($95\%$ CI: 151,484; Fig 2), but potential gains per child born SGA were highest in Latin America at US$ 1,252 ($95\%$ CI: 373, 2,808; Fig 3).
## Discussion
Through this analysis, we found that there may be substantial potential gains in schooling and lifetime income to gain from reducing adverse birth outcomes to theoretical minimum values across 121 LMICs. The estimated total gains in additional school years and lifetime income from shifting the excess prevalence of LBW, PTB, and SGA to the theoretical minimum levels were 20.3 million, 9.8 million, and 20.3 million school years, and US$ 68.8 billion, US$ 41.9 billion, and US$113.6 billion in lifetime income gained per birth cohort, respectively. While the education and income gains were estimated to be large for each adverse birth outcome considered, the potential gains were generally largest for reducing SGA to the theoretical minimum of $10\%$. Notably, South Asia and sub-Saharan Africa had substantial potential gains in absolute and per-child terms for education attainment and lifetime earnings largely owing to the high prevalence of adverse birth outcomes and large population of these regions.
Our findings add to a very limited literature on the potential long-term economic returns of reducing birth outcomes and early life exposures and adversities in LMICs [10, 20]. To our knowledge, only one other study has looked at the potential economic gains of reducing the prevalence of LBW: Alderman and Behrman estimated that relative to infants born LBW, non-LBW infants may have a greater lifetime earning potential of US $832 (at $3\%$ discounting) [10]. Alderman and Behrman used different methods and investigated a wider range of potential gains, including not only cognitive delay, but also reduced costs from infant mortality, neonatal care, sickness, stunting, and chronic diseases. Of that, US $367 were estimated as productivity gains from increased intelligence quotients and schooling attainment, in the absence of LBW [10]. In comparison, we estimated a global average potential gain of US$ 717 per LBW case averted. Our results suggest that the returns on wage gains from increased schooling may be almost twice that estimated in their analysis.
We report the first estimates the potential education and lifetime earnings gains to reducing PTB and SGA. It is important to note that the large gains in sub-Saharan Africa and South Asia can in part be attributed to the higher prevalence of PTB and SGA in these regions. There were particularly high returns to reducing SGA to the TMRED in South Asia, where the median prevalence of SGA born babies was $40\%$. Several studies have found significant associations between prematurity and birthweight for gestational age with cognitive performance of school-aged children [23, 24].
Our study builds on data from birth cohort studies that linked adverse birth outcomes with poorer education outcomes. In terms of mechanisms, there are multiple pathways from adverse birth outcomes to reduced number of completed years of schooling, including suboptimal cognitive, executive function, socioemotional and other neurodevelopmental pathways. There is a relatively large literature in both high-income and LMIC that have linked prematurity and LBW with poorer child neurodevelopment outcomes [5, 23–25]. Animal and human studies have also found that that in-utero malnutrition is associated with reduced brain volume and with adverse effects on neuron proliferation, synaptogenesis, and myelination [26]. In addition, higher birthweight have been associated with greater lifetime earnings in data from high-income settings [16]. However, little to no data on the relationship of adverse birth outcomes with lifetime earnings exists for LMICs and therefore we needed to use schooling as a mediator.
Our analytical strategy included analyses based on not just LBW, but also PTB and SGA, and enables a comparison of potential impacts across each type of adverse birth outcome. This was done because LBW does not differentiate between prematurity and intrauterine growth restriction and each mechanism may have different long-term effects on schooling and life income. The inclusion of PTB and SGA is an important addition to the limited literature on potential long-term human capital gains of reducing adverse birth outcomes. It is also important to note that LBW, PTB, and SGA can be coexisting conditions; therefore, the impacts of reducing each on schooling and income should be interpreted independently rather than cumulatively. We were not able to calculate the joint effect of the three outcomes due to lack of country-level estimates of their joint distribution. Nevertheless, the joint effect would be larger than each individual effect, but smaller than the sum of all three. Therefore, the impacts of each birth outcome should be interpreted independent of the other adverse birth outcomes.
Our study has several limitations and strengths. Our estimates are limited by the availability and quality of relevant data. Estimates of adverse birth outcomes estimates from nationally representative surveys are not available in many LMICs, and we therefore used the latest modeled prevalence estimates of LBW, PTB and SGA, which are limited by the quality of the data used in the models. Further, five countries (Dominica, Republic of Kosovo, Marshall Islands, Nauru, and Tuvalu) were excluded from the analysis because they lacked data on the number of live births and probability of survival. Ultrasound dating to assess gestational age is not common in LMICs and therefore estimation of the national prevalences of PTB and SGA at the population-level is particularly challenging in LMICs. However, our study also has several strengths. We used a first order Monte Carlo simulation propagated uncertainty in LBW, PTB and SGA are were therefore able account for variation in parameters and compute bootstrapped confidence intervals [27]. In addition, we quantified the impact of reducing adverse birth outcomes to their theoretical minima on lifetime earnings exclusively through educational attainment. Therefore, our estimates for lifetime earnings are likely conservative since they do not take into account for additional benefits on lifetime earnings that may be accrued through pathways outside of schooling, such as reductions in morbidity, well-being, and the associated increases in adult productivity and lower healthcare costs [10, 28, 29]. In addition, reductions in adverse birth outcomes would reduce child mortality that would lead to more children surviving, enrolling in school, and leading productive lives which are additional benefits on top of the ones accounted for in our model.
The potential education and lifetime earnings gains from reducing LBW, PTB, and SGA globally may be substantial. The long-term gains documented in this analysis should be considered in cost-effectiveness evaluations of interventions that improve birth outcomes in addition to more immediate benefits on child mortality, growth, and development. As a result, greater investment in interventions and programs that improve birth outcomes may benefit individuals across the life course and provide substantial population-level human capital returns.
## References
1. Blencowe H, Krasevec J, de Onis M, Black RE, An X, Stevens GA. **National, regional, and worldwide estimates of low birthweight in 2015, with trends from 2000: a systematic analysis**. *The Lancet global health* (2019.0) **7** e849-e60. DOI: 10.1016/S2214-109X(18)30565-5
2. Chawanpaiboon S, Vogel JP, Moller A-B, Lumbiganon P, Petzold M, Hogan D. **Global, regional, and national estimates of levels of preterm birth in 2014: a systematic review and modelling analysis**. *The Lancet global health* (2019.0) **7** e37-e46. DOI: 10.1016/S2214-109X(18)30451-0
3. Lee ACC, Katz J, Blencowe H, Cousens S, Kozuki N, Vogel JP. **National and regional estimates of term and preterm babies born small for gestational age in 138 low-income and middle-income countries in 2010**. *The Lancet global health* (2013.0) **1** e26-e36. DOI: 10.1016/S2214-109X(13)70006-8
4. Katz J, Lee ACC, Kozuki N, Lawn JE, Cousens S, Blencowe H. **Mortality risk in preterm and small-for-gestational-age infants in low-income and middle-income countries: a pooled country analysis**. *The Lancet (British edition).* (2013.0) **382** 417-25. DOI: 10.1016/S0140-6736(13)60993-9
5. Sania A, Sudfeld CR, Danaei G, Fink G, McCoy DC, Zhu Z. **Early life risk factors of motor, cognitive and language development: a pooled analysis of studies from low/middle-income countries**. *BMJ Open* (2019.0) **9** e026449. DOI: 10.1136/bmjopen-2018-026449
6. Blencowe H, Lee ACC, Cousens S, Bahalim A, Narwal R, Zhong N. **Preterm birth–associated neurodevelopmental impairment estimates at regional and global levels for 2010**. *Pediatric Research* (2013.0) **74** 17. DOI: 10.1038/pr.2013.204
7. Stein AD, Barros FC, Bhargava SK, Hao W, Horta BL, Lee N. **Birth Status, Child Growth, and Adult Outcomes in Low- and Middle-Income Countries**. *The Journal of pediatrics* (2013.0) **163** 1740-6.e4. DOI: 10.1016/j.jpeds.2013.08.012
8. Markopoulou P, Papanikolaou E, Analytis A, Zoumakis E, Siahanidou T. **Preterm Birth as a Risk Factor for Metabolic Syndrome and Cardiovascular Disease in Adult Life: A Systematic Review and Meta-Analysis**. *The Journal of pediatrics* (2019.0) **210** 69-80.e5. DOI: 10.1016/j.jpeds.2019.02.041
9. Knop MR, Geng T-T, Gorny AW, Ding R, Li C, Ley SH. **Birth Weight and Risk of Type 2 Diabetes Mellitus, Cardiovascular Disease, and Hypertension in Adults: A Meta-Analysis of 7 646 267 Participants From 135 Studies**. *Journal of the American Heart Association* (2018.0) **7** e008870. DOI: 10.1161/JAHA.118.008870
10. Alderman H, Behrman JR. **Reducing the Incidence of Low Birth Weight in Low-Income Countries Has Substantial Economic Benefits**. *World Bank Research Observer* (2006.0) **21** 25-48
11. Villar JP, Ismail LCP, Victora CGP, Ohuma EOM, Bertino EP, Altman DGP. **International standards for newborn weight, length, and head circumference by gestational age and sex: the Newborn Cross-Sectional Study of the INTERGROWTH-21st Project.**. *Lancet, The* (2014.0) **384** 857-68
12. Perumal N, Blakstad MM, Fink G, Lambiris M, Bliznashka L, Danaei G. **Impact of scaling up prenatal nutrition interventions on human capital outcomes in low- and middle-income countries: a modeling analysis**. *The American Journal of Clinical Nutrition* (2021.0) **114** 1708-18. DOI: 10.1093/ajcn/nqab234
13. Lee ACC, Kozuki N, Cousens S, Stevens GA, Blencowe H, Silveira MF. **Estimates of burden and consequences of infants born small for gestational age in low and middle income countries with INTERGROWTH-21st standard: analysis of CHERG datasets**. *BMJ* (2017.0) **358** j4229. DOI: 10.1136/bmj.j4229
14. 14Nations) UWWUUNCsFWHOWBGU. Levels & Trends in Child Mortality Estimates developed by the UN Inter-agency Group for Child Mortality Estimation. New York, NY: UNICEF.; 2019.. (2019.0)
15. 15Institute for Health Metrics and Evaluation. Countries 2021 [Available from: http://ghdx.healthdata.org/countries.
16. Lambiris MJ, Blakstad MM, Perumal N, Danaei G, Bliznashka L, Fink G. **Birth weight and adult earnings: a systematic review and meta-analysis**. *Journal of Developmental Origins of Health and Disease* (2021.0) 1-8. DOI: 10.1017/S2040174421000404
17. Adair LS, Fall CHD, Osmond C, Stein AD, Martorell R, Ramirez-Zea M. **Associations of linear growth and relative weight gain during early life with adult health and human capital in countries of low and middle income: findings from five birth cohort studies**. *The Lancet (British edition).* (2013.0) **382** 525-34. DOI: 10.1016/S0140-6736(13)60103-8
18. 18United Nations Department of Economic and Social Affairs Population Division.
World Population Prospects 2019. Online Edition.; 2019.. *World Population Prospects 2019* (2019.0)
19. Peet ED, Fink G, Fawzi W. **Returns to education in developing countries: Evidence from the living standards and measurement study surveys**. *Economics of education review* (2015.0) **49** 69-90
20. Fink G, Peet E, Danaei G, Andrews K, McCoy DC, Sudfeld CR. **Schooling and wage income losses due to early-childhood growth faltering in developing countries: national, regional, and global estimates**. *The American journal of clinical nutrition* (2016.0) **104** 104-12. DOI: 10.3945/ajcn.115.123968
21. 21World Bank. World Development Indicators
2020. http://dataworldbankorg/data-catalog/world-development-indicators.. *World Development Indicators* (2020.0)
22. 22Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2019 (GBD 2019) Results.
Institute for Health Metrics and Evaluation (IHME), Seattle, United States: Available from: http://ghdx.healthdata.org/gbd-results-tool; 2020.. (2020.0)
23. Bhutta AT, Cleves MA, Casey PH, Cradock MM, Anand KJS. **Cognitive and Behavioral Outcomes of School-Aged Children Who Were Born Preterm: A Meta-analysis.**. *JAMA: the journal of the American Medical Association* (2002.0) **288** 728-37. DOI: 10.1001/jama.288.6.728
24. Christian P, Murray-Kolb LE, Tielsch JM, Katz J, LeClerq SC, Khatry SK. **Associations between preterm birth, small-for-gestational age, and neonatal morbidity and cognitive function among school-age children in Nepal**. *BMC pediatrics* (2014.0) **14** 58. PMID: 24575933
25. Perumal N, Manji KP, Darling AM, Kisenge RR, Kvestad I, Hysing M. **Gestational age, birth weight and neurocognitive development in adolescents in Tanzania**. *Journal of Pediatrics* (2021.0). DOI: 10.1016/j.jpeds.2021.04.036
26. Prado EL, Dewey KG. **Nutrition and brain development in early life**. *Nutrition reviews* (2014.0) **72** 267-84. DOI: 10.1111/nure.12102
27. Kozuki N, Katz J, Clermont A, Walker N. **New Option in the Lives Saved Tool (LiST) Allows for the Conversion of Prevalence of Small-for-Gestational-Age and Preterm Births to Prevalence of Low Birth Weight**. *The Journal of nutrition* (2017.0) **147** 2141S. DOI: 10.3945/jn.117.247767
28. He S, Stein AD. **Early-Life Nutrition Interventions and Associated Long-Term Cardiometabolic Outcomes: A Systematic Review and Meta-Analysis of Randomized Controlled Trials**. *Advances in Nutrition* (2020.0)
29. Yoshida S, Martines J, Lawn JE, Wall S, Souza JP, Rudan I. **Setting research priorities to improve global newborn health and prevent stillbirths by 2025**. *Journal of global health* **6**. DOI: 10.7189/jogh.06.010508
|
---
title: A longitudinal analysis of PM2.5 exposure and multimorbidity clusters and accumulation
among adults aged 45-85 in China
authors:
- Kai Hu
- Katherine Keenan
- Jo Mhairi Hale
- Yang Liu
- Hill Kulu
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021527
doi: 10.1371/journal.pgph.0000520
license: CC BY 4.0
---
# A longitudinal analysis of PM2.5 exposure and multimorbidity clusters and accumulation among adults aged 45-85 in China
## Abstract
While previous studies have emphasised the role of individual factors in understanding multimorbidity disparities, few have investigated contextual factors such as air pollution (AP). We first use cross-sectional latent class analysis (LCA) to assess the associations between PM2.5 exposure and multimorbidity disease clusters, and then estimate the associations between PM2.5 exposure and the development of multimorbidity longitudinally using growth curve modelling (GCM) among adults aged 45–85 in China. The results of LCA modelling suggest four latent classes representing three multimorbidity patterns (respiratory, musculoskeletal, cardio-metabolic) and one healthy pattern. The analysis shows that a 1 μg/m3 increase in cumulative exposure to PM2.5 is associated with a higher likelihood of belonging to respiratory, musculoskeletal or cardio-metabolic clusters: $2.4\%$ ($95\%$ CI: 1.02, 1.03), $1.5\%$ ($95\%$ CI: 1.01, 1.02) and $3.3\%$ ($95\%$ CI: 1.03, 1.04), respectively. The GCM models show that there is a u-shaped association between PM2.5 exposure and multimorbidity, indicating that both lower and higher PM2.5 exposure is associated with increased multimorbidity levels. Higher multimorbidity in areas of low AP is explained by clustering of musculoskeletal diseases, whereas higher AP is associated with cardio-metabolic disease clusters. The study shows how multimorbidity clusters vary contextually and that PM2.5 exposure is more detrimental to health among older adults.
## Introduction
The number of older adults living with multimorbidity, defined as the coexistence of two or more chronic diseases or conditions, is rising globally [1]. Research based on survey data shows a high prevalence of multimorbidity among older adults in low and middle income countries (LMICs), such as $63\%$ in India [2], $65\%$ in Brazil [3], $69\%$ in South Africa [4], and $46\%$ in China [5]. The increasing prevalence of multimorbidity is associated with worse functional ability, reduced healthy life expectancy, increased mortality, and a higher rate of hospitalisations [6–8], leading to a heavy burden on medical and health systems and inequalities in health outcomes [9]. This is a particular challenge in rapidly ageing societies such as China, which has a high prevalence of multimorbidity among older adults [5]. The proportion of the population with multimorbidity in China, as measured by population-based panel data, was $62\%$ for people aged 50 years and $69\%$ for those aged over 75 years [9]. In terms of the determinants of multimorbidity, current studies highlight a range of individual factors, including demographic (e.g., age, sex and race) and socio-economic (e.g., education and income) characteristics [9–12]. However, research on possible contextual and environmental determinants of multimorbidity is less common, and in particular the role of air pollution (AP) remains poorly understood [13].
Recent research on elderly health shows that older people are more susceptible to environmental factors than younger adults [14], with higher risks of living with chronic diseases due to exposure to environmental pollution [15, 16]. Furthermore, there is abundant evidence on the association between AP and individual chronic diseases, for example, cardiorespiratory disease [17], chronic obstructive pulmonary disease (COPD) [18], diabetes [19], heart disease [20], hypertension [21], and kidney diseases [22]. Although chronic diseases cluster due to shared biological or environmental risk factors [13], there is limited understanding of how AP might operate to promote accumulation of multiple chronic diseases.
Similar to many LMICs, AP is an important public health risk in China: in 2017, $81\%$ of the Chinese population lived in regions which exceed the World Health Organisation Interim Target 1 (35 μg/m3) [23]. In particular, ambient AP was estimated to be responsible for over 850,000 deaths in China in 2017 [23]. Although ambient AP has decreased markedly in the last two decades, the older population of China has spent a large proportion of the life course experiencing historically high levels of AP exposure [24, 25]. China therefore suffers from a double burden of multimorbid ageing and AP, and understanding the association between them may be beneficial for development of strategies to prevent or manage chronic diseases in later life.
Evidence shows that multimorbidity prevalence is likely higher in some social groups because chronic diseases often cluster due to common risk factors, such as socioeconomic deprivation and environmental risks [13]. These risk factors for diseases clustering make it difficult to isolate the effects of AP from other factors of socioeconomic deprivation [26]. Therefore, motivations for this study are not only to understand the relationship between historic AP exposure and changes in multimorbidity, but also to explore individual-level characteristics that are associated with multimorbidity inequalities.
In this study, we analyse the associations between cumulative, historic exposure to AP as predictive of cross-sectional multimorbidity disease clusters and multimorbidity accumulation. We use large, prospective and nationally representative survey data, the China Health and Retirement Longitudinal Study (CHARLS), linked with historical satellite data on PM2.5 exposure over 15 years. Using this novel dataset, we address a research gap for longitudinal studies of multimorbidity and provide assessment of the associations between AP and the development of multimorbidity.
## Study population
Data used in this study are from three waves of the China Health and Retirement Longitudinal Study (CHARLS 2011, 2013, 2015), which is a nationally representative longitudinal survey of the middle-aged and elderly population of China, consisting of persons aged 45 years or older, as well as their spouses when possible. CHARLS used computer-assisted in-person interviews to obtain samples through four-stage stratified sampling, with an overall response rate of $80.5\%$ at the baseline. From June 2011 and March 2012, CHARLS conducted the baseline survey (wave 1) that included assessments of the social, economic, and health circumstances of 17,705 respondents from 28 provinces, 150 cities/counties/districts, 450 communities, and 10,257 households [27]. Following wave 1, two follow-up surveys were conducted in 2013 and 2015.
Fig 1 shows the criteria of sample inclusion. In 2011, the baseline CHARLS sample size was 17,705. Between 2011 and 2013, 2,526 respondents attrited due to death ($$n = 441$$) or non-specified reasons ($$n = 2$$,125). In 2013, a refreshment sample was added of 3,425 new respondents, making the total 2013 sample consist of 18,604 individuals. The 2015 wave of CHARLS had a total sample size of 21,100, including 3,826 new joiners. Between 2013 and 2015, 689 respondents died and 1,658 attrited, and 1,017 respondents, interviewed in 2011 but missing in 2013, returned. It is noted that these refreshment samples did not participate in the first wave in 2011. Finally, the three waves of CHARLS include 24,956 respondents (57,409 observations).
**Fig 1:** *Flowchart of study inclusion criteria.Note: CHARLS, the China Health and Retirement Longitudinal Study.*
We restrict our analysis to the 19,098 respondents (45,788 observations) from 125 cities who were aged 45 to 85 years old at any wave of the study. The listwise deletion process is also shown in Fig 1.
## Main outcome: Multimorbidity
The ‘*Health status* and function’ module in the CHARLS questionnaire includes 14 self-reported doctor-diagnosed chronic diseases for each respondent, asking “Have you been diagnosed with the following conditions by a doctor”: hypertension; dyslipidaemia; diabetes or high blood sugar; cancer or malignant tumour; chronic lung disease; liver disease; heart problems; stroke; kidney disease; stomach or other digestive diseases; emotional, nervous or psychiatric problems; memory-related disease; arthritis or rheumatism [12]. In line with previous CHARLS studies, we use a disease count approach where we summed 14 binary disease indicators (range 0–14) to capture multimorbidity [12, 28].
We exclude respondents who are missing any components of these 14 indicators of chronic disease. Using this method, there are 7,469 observations with missingness on multimorbidity. Note that CHARLS 2015 contributes to over half of these (4,654 observations), partly due to survey design, because around 2,800 respondents were interviewed in the Life History Survey (in 2014) but not in previous waves (CHARLS 2011 and 2013). In these cases, the respondents were not asked if they ever had a condition and therefore their chronic disease records are missing in 2015. We explain how we deal with missingness in the statistical analysis section.
## Air pollution: PM2.5
Many studies use ground air pollutants concentration (e.g., PM2.5, PM10, NO2) from monitoring stations to measure the exposure to AP [29, 30]. In China, however, most AP monitoring stations were established by the Ministry of Ecology and Environment only after 2013, limiting the ability to study long-term exposure.
Compared with ground monitoring data, satellite data with broad spatial coverage, a long-term data record, and high spatial resolutions could support the assessment of historical AP levels in developing regions. A detailed description of the ensemble machine learning model to generate our long-term PM2.5 exposure estimates is published elsewhere [31], and summarised briefly here. Given the large modelling domain, China was first divided into seven subregions using a geographically weighted regression approach to allow the machine learning algorithms to capture different spatiotemporal patterns of PM2.5 in each subregion caused by different terrain, weather conditions, and emission source profiles. A random forest, an extreme gradient boosting (XGBoost), and a generalized additive model (GAM) were then trained in each subregion. Their predictions of daily PM2.5 concentration levels at 10 km2 spatial resolution were combined using weights determined by prediction accuracy. Compared with previous models, this ensemble model provided more accurate out-of-range predictions at the daily and monthly levels. Based on the administrative regions in China, 150 cities of CHARLS are clustered to aggregate PM2.5 concentration from satellite data at the city level. To match with CHARLS, we selected monthly PM2.5 concentration as the temporal scale (see Table G in S1 File).
As used in previous studies, the measure of cumulative exposure is operationalised using the average mean of pollutant concentrations during the exposure window [32, 33]. Chinese ambient air quality standards use two cut-offs to indicate the hazardous level of exposures, 35 μg/m3 (Level 1) and 75 μg/m3 (Level 2) [34]. In this study, we exploit our longer-term PM2.5 data and calculate a more fine-grained measure: the average concentration of monthly exposure from March 2000 until the survey date, which provides a measure of historical exposure. Additionally, we initially categorise PM2.5 exposure into six groups using 10 μg/m3 intervals: 0–35, 36–45, 46–55, 56–65, 66–75, 76+ μg/m3. Due to small numbers exposed to PM2.5 over 76+ μg/m3, we later collapse the last two categories resulting in five groups.
## Covariates: Demographic, socioeconomic status (SES), health behaviour and regional factors
In this study, the covariates include four components: demographic, socioeconomic status (SES), behavioural, and contextual factors. Demographic variables include age, age squared, gender, and marital status (single vs. partnered). Individual SES consists of education (no schooling, primary, middle or more education), occupation (agricultural, non-agricultural, and managerial), and HuKou (rural, rural-urban, urban). CHARLS life history survey records the longest occupation during the respondents’ occupational history. Agricultural jobs include farming, fishing, managing forest products or fruit trees, raising livestock, and selling these products in the market. Non-agricultural jobs include civil servants, office clerks or non-agricultural self-employment (e.g., running a restaurant or supermarket). Respondents who are in a supervisory position in their offices are considered “managers”. Respondents who have never worked (e.g., “housewife”, or disabled people) are sparse in the CHARLS (only 137 respondents) and are marked as missing.
HuKou is a special national household registration system in China that has two categories: rural and urban. People usually remain in the same HuKou as their parents, and once HuKou is registered, it is difficult to change even if people move. HuKou is related to occupational status, education, and health care access [35, 36]. Due to the urbanisation and internal migration, people who originate in a rural HuKou increasingly live in the urban areas. Thus, considering both HuKou and current residence, there are three types: rural (rural HuKou living in rural areas), rural-urban (rural HuKou living in urban areas), and urban (urban HuKou living in urban areas). HuKou is an important feature in this study as it is strongly related to personal SES, not solely housing address [35].
Smoking status is controlled for as an important health behaviour (never smoking, former smoker, and current smoker). To account for the urbanisation and industrialisation of cities, this study includes annual regional Gross Domestic Product (GDP) at the city level (logged).
## Analysis strategies
To analyse multimorbidity disease clusters, we use latent class analysis (LCA) on a cross-sectional sample of the baseline wave of CHARLS as LCA can pick up clusters of diseases shared with the common risk factors. The LCA models can identify multimorbidity patterns by assigning individuals to a set of discrete, mutually exclusive groups—latent classes—based on their responses to the 14 chronic disease indicators in the CHARLS. A sequence of 14 LCA models was estimated starting with a one-class model and increasing the number of classes in a stepwise approach. Following examples in previous studies [37, 38], the LCA model selection was based on examinations of several fit indices, including AIC, BIC and likelihood estimates. We then regressed the resulting latent class memberships on cumulative PM2.5 exposure in 2011 using multinomial regression, adjusting for the covariates discussed above.
We use growth curve modelling (GCM) to examine the relationship between PM2.5 exposure in the period of 2000–2015 and multimorbidity accumulation between 2011–2015. An important advantage of GCM is the ability to model the trajectories of individuals over time and distinguish within-individual from between-individual heterogeneities in estimating multimorbidity accumulation/changes shaped by other variables [39]. In this study, we use three waves of CHARLS across four years (2011–2015) of data collection. As multimorbidity is a count variable, we assume a Poisson distribution. As there is likely a non-linear relationship between PM2.5 exposure and elderly health, we add a quadratic term of PM2.5 exposure, as well as alternative models (detailed below) using categorisations of PM2.5. To examine heterogeneity in the associations of PM2.5 exposure among different groups, we further explore the interactions between PM2.5 exposure and age, SES (education, occupational status, HuKou-residence), and smoking status.
We conduct a number of robustness checks. First, we run the same set of LCA and GCMs models but use a categorical measure of PM2.5 exposure to allow for a more flexible estimation of the association between AP and multimorbidity (shown in Tables B and C in S1 File). Second, we run the GCMs first on the entire sample 45–85 years, then we subset the data into ages 45–64 and 65–85 years to compare middle-aged and oldest individuals (Tables D and E in S1 File). Third, given that 7,867 observations are deleted due to missingness, we apply multiple imputation (MI) using chained equations to complete our analysis samples under the missing-at-random assumption. We then use multilevel random-intercept Poisson regression to compare the results from the MI and complete datasets (Table F in S1 File).
## Descriptive analysis
At baseline, the average age of respondents is approximately 59 years old. $49\%$ are men. $39\%$ of the population attained primary education and $34\%$ had middle or higher education. $78\%$ of respondents are registered with rural HuKou but $19\%$ respondents with rural HuKou are living in urban areas; $72\%$ worked for agricultural jobs, and $88\%$ are married. Average multimorbidity is 1.5 at baseline, increasing to 2.08 by 2015. From 2011 to 2015, average PM2.5 rises only slightly from 51.27 to 52.90 μg/m3 (Table 1). In addition, Table 1 shows that descriptive statistics for both the baseline and entire period’s analytical samples are very similar to the entire sample, suggesting that sample selection, including attrition, may not substantially affect results.
**Table 1**
| Unnamed: 0 | CHARLS | CHARLS.1 | CHARLS.2 | CHARLS 2011- | CHARLS.3 |
| --- | --- | --- | --- | --- | --- |
| | 2011 (analysis) | 2013 (analysis) | 2015 (analysis) | 2015 (analysis) | 2011–2015 |
| Multimorbidity (Mean/SD) | 1.45 (1.43) | 1.60 (1.52) | 2.08 (1.75) | 1.69 (1.59) | 1.69 (1.59) |
| Cumulative PM2.5 exposure (Mean/SD) | 51.27 (15.63) | 52.10 (16.44) | 52.90 (16.84) | 52.06 (16.31) | 52.01 (16.42) |
| Age (Mean/SD) | 59.10 (9.49) | 60.01 (9.54) | 61.32 (9.26) | 60.10 (9.48) | 59.82 (9.66) |
| Gender (N/%) | | | | | |
| Men | 7,795 (49.00) | 7,582 (48.65) | 6,887 (48.18) | 9,277 (48.58) | 11,358 (49.12) |
| Women | 8,114 (51.00) | 8,003 (51.35) | 7,407 (51.82) | 9,821 (51.42) | 11,766 (50.88) |
| Education (N/%) | | | | | |
| No schooling | 4,318 (27.14) | 4,019 (25.79) | 3,621 (25.33) | 4,914 (25.73) | 5,572 (24.14) |
| Primary | 6,243 (39.24) | 6,259 (40.16) | 5,824 (40.74) | 7,548 (39.52) | 9,026 (39.11) |
| Middle + | 5,348 (33.62) | 5,307 (34.05) | 4,849 (33.92) | 6,636 (34.75) | 8,481 (36.75) |
| HuKou (N/%) | | | | | |
| Rural | 9,320 (58.58) | 9,193 (58.99) | 8,481 (59.33) | 10,852 (56.21) | 12,647 (54.18) |
| Rural-urban | 3,065 (19.27) | 2,973 (19.08) | 2,734 (19.13) | 3,777 (19.56) | 4,757 (20.38) |
| Urban | 3,524 (22.15) | 3,419 (21.94) | 3,079 (21.54) | 4,676 (24.22) | 5,938 (25.44) |
| Occupation (N/%) | | | | | |
| Agricultural | 11,413 (71.74) | 11,965 (76.77) | 10,661 (74.58) | 14,276 (67.80) | 16,649 (66.55) |
| Non-agricultural | 3,582 (22.52) | 3,4015 (21.82) | 2,917 (20.41) | 5,146 (24.44) | 6,371 (25.47) |
| Managerial | 914 (5.75) | 219 (1.41) | 716 (5.01) | 1,633 (7.76) | 1,998 (7.99) |
| Marital (N/%) | | | | | |
| Partnered | 13,976 (87.85) | 13,660 (87.65) | 12,420 (86.89) | 16,960 (85.52) | 20,589 (85.96) |
| Single | 1,933 (12.15) | 1,925 (12.35) | 1,874 (13.11) | 2,871 (14.48) | 3,362 (14.04) |
| Smoking status (N/%) | | | | | |
| Never | 9,433 (59.29) | 8,869 (56.91) | 8,317 (58.19) | 11,774 (53.62) | 14,119 (53.92) |
| Former | 1,401 (8.81) | 1,146 (7.35) | 1,964 (13.74) | 3,176 (14.46) | 3,721 (14.21) |
| Current | 5,075 (31.90) | 5,570 (35.74) | 4,017 (28.07) | 7,007 (31.91) | 8,345 (31.87) |
| Log GDP (Mean/SD) | 10.29 (0.56) | 10.47 (0.65) | 10.47 (0.63) | 10.41 (0.63) | 10.42 (0.62) |
| Number of respondents | 15909 | 15585 | 14294 | 19098 | 23124 |
## Latent class analysis
First, we use LCA to explore the association between PM2.5 exposure and multimorbidity patterns at baseline (CHARLS 2011). Based on the comparisons of AIC, BIC, and likelihood estimates, the four-class model was chosen as the final model in this study (Table A and Fig A in S1 File). Table 2 shows the distribution of the sample across the four classes. Based on the probability distribution of chronic diseases across the classes, they are labelled: respiratory (Class 1), musculoskeletal (Class 2), cardio-metabolic (Class 3) and relatively healthy (Class 4). The label takes its name from the main diseases (items) that characterise it. For example, we labelled Class 1 as respiratory because the probability of lung diseases is 0.79, which means $79\%$ of respondents in Class 1 suffered from lung diseases. Similarly, labels of class 2 and 3 are originated from high probabilities of arthritis/rheumatism (0.81) and hypertension (0.71).
**Table 2**
| Unnamed: 0 | Latent Class | Latent Class.1 | Latent Class.2 | Latent Class.3 |
| --- | --- | --- | --- | --- |
| | Class 1 | Class 2 | Class 3 | Class 4 |
| Assigned label | Respiratory | Musculoskeletal | Cardio-metabolic | Relatively healthy |
| Class Proportion | 0.075 | 0.207 | 0.172 | 0.546 |
| Items (chronic diseases) | Response probabilities | Response probabilities | Response probabilities | Response probabilities |
| Hypertension | 0.319 | 0.243 | 0.713 | 0.129 |
| Dyslipidaemia | 0.099 | 0.039 | 0.394 | 0.029 |
| Diabetes | 0.069 | 0.027 | 0.244 | 0.021 |
| Cancer/malignant tumour | 0.011 | 0.006 | 0.034 | 0.005 |
| Lung diseases | 0.787 | 0.094 | 0.053 | 0.034 |
| Liver diseases | 0.072 | 0.065 | 0.052 | 0.018 |
| Heart problems | 0.269 | 0.156 | 0.354 | 0.025 |
| Stroke | 0.044 | 0.024 | 0.102 | 0.005 |
| Kidney diseases | 0.123 | 0.122 | 0.088 | 0.023 |
| Stomach/digestive diseases | 0.323 | 0.447 | 0.204 | 0.138 |
| Emotional/nervous/psychiatric | 0.041 | 0.018 | 0.017 | 0.011 |
| Memory-related diseases | 0.059 | 0.020 | 0.051 | 0.005 |
| Arthritis/rheumatism | 0.477 | 0.813 | 0.340 | 0.155 |
| Asthma | 0.496 | 0.018 | 0.017 | 0.006 |
## Multinomial regression
Second, we present the results from the cross-sectional multinomial models analysing the associations between PM2.5 and multimorbidity patterns at baseline, controlling for a set of covariates. The findings show that higher exposure to AP is associated with a higher prevalence of the other three classes of chronic diseases, compared to those who are “relatively healthy” (Table 3). Specifically, a 1 μg/m3 increase in cumulative exposure to PM2.5, is associated with a higher likelihood of belonging to respiratory, musculoskeletal and cardio-metabolic cluster: $2.4\%$ ($95\%$ CI: 1.02, 1.03), $1.5\%$ ($95\%$ CI: 1.01, 1.02) and $3.3\%$ ($95\%$ CI: 1.03, 1.04), respectively.
**Table 3**
| Unnamed: 0 | Class 1: | Class 2: | Class 3: |
| --- | --- | --- | --- |
| | Respiratory | Musculoskeletal | Cardio-metabolic |
| PM2.5 exposure | 1.024*** | 1.015*** | 1.033*** |
| PM2.5 exposure | (1.017–1.031) | (1.011–1.020) | (1.026–1.039) |
| Age | 1.046*** | 0.971*** | 1.041*** |
| Age | (1.033–1.058) | (0.963–0.980) | (1.030–1.052) |
| Gender (ref: Men) | | | |
| Women | 0.479*** | 0.427*** | 0.682** |
| Women | (0.352–0.652) | (0.347–0.524) | (0.522–0.890) |
| Education (ref: no schooling) | | | |
| Primary | 0.963 | 0.671*** | 0.989 |
| Primary | (0.751–1.235) | (0.569–0.791) | (0.788–1.241) |
| Middle + | 0.879 | 0.995 | 1.481** |
| Middle + | (0.629–1.227) | (0.801–1.235) | (1.115–1.967) |
| HuKou (ref: Rural) | | | |
| Rural-urban | 0.829 | 1.003 | 1.294** |
| Rural-urban | (0.631–1.088) | (0.846–1.189) | (1.032–1.622) |
| Urban | 2.132*** | 1.766*** | 4.695*** |
| Urban | (1.520–2.991) | (1.352–2.307) | (3.507–6.285) |
| Occupation (ref: agricultural) | | | |
| Non- agricultural | 1.456** | 1.287* | 1.575*** |
| Non- agricultural | (1.097–1.934) | (1.038–1.596) | (1.230–2.018) |
| Managerial | 0.929 | 1.669* | 1.277 |
| Managerial | (0.490–1.763) | (1.125–2.477) | (0.782–2.086) |
| Marital (ref: partnered) | | | |
| Single | 1.408* | 1.176 | 1.000 |
| Single | (1.064–1.863) | (0.951–1.455) | (0.766–1.305) |
| Smoking (ref: never) | | | |
| Former | 2.820*** | 0.762# | 1.477* |
| Former | (1.945–4.088) | (0.556–1.045) | (1.033–2.111) |
| Current | 1.467** | 0.966 | 0.691** |
| Current | (1.100–1.956) | (0.792–1.178) | (0.536–0.892) |
| Log GDP | 1.488*** | 2.393*** | 2.686*** |
| Log GDP | (1.220–1.816) | (2.096–2.733) | (2.265–3.186) |
| Constant | 0.0001*** | 0.002*** | 3.93e-07*** |
| Constant | (0.000–0.001) | (0.0004–0.006) | (0.000–0.000) |
In terms of the other covariates, the regression model shows that women have a lower likelihood of belonging to any of three multimorbidity classes compared with the healthy class. Older age is associated with a higher likelihood of belonging to respiratory and cardio-metabolic clusters but with a lower likelihood in musculoskeletal cluster. The relationship between education and multimorbidity patterns is complex. Compared with respondents without schooling, those with primary education have a lower likelihood of belonging to the musculoskeletal cluster, and those with middle or higher education have a higher likelihood of belonging to cardio-metabolic cluster.
Respondents with urban HuKou have a higher likelihood of belonging to any of the three disease clusters, especially the cardio-metabolic cluster. Working in non-agricultural positions is associated with a higher likelihood of being in those three classes. Being single is associated with a higher likelihood of being in the respiratory cluster but is not associated with membership in the musculoskeletal and cardio-metabolic clusters. Smoking status is associated with a higher likelihood of respiratory and cardio-metabolic clusters but not associated with the musculoskeletal cluster. Compared with non-smokers, former smokers have a higher likelihood of belonging to the cardio-metabolic cluster, whereas current smokers have a lower likelihood. Higher GDP is associated with a higher likelihood of belonging to any of the three disease clusters.
## Heterogeneity in patterns
We also analyse the heterogeneity in multimorbidity cluster membership by age, gender and HuKou. Fig 2 shows the association between multimorbidity patterns and 11-year exposure to PM2.5, comparing middle-aged (45–65 years old) with older adults (66–85 years old). Generally, higher exposure to PM2.5 is associated with a higher probability of belonging to respiratory and, especially, cardiometabolic clusters. Unexpectedly, lower levels of PM2.5 are associated with higher likelihood of belonging to the musculoskeletal cluster. We can see the negative associations of PM2.5 exposure are more substantial among the group aged 66–85, because there is a lower likelihood of belonging to the relatively healthy class but a higher likelihood in the respiratory, musculoskeletal, and cardio-metabolic clusters when PM2.5 exposure levels are elevating.
**Fig 2:** *Predicted probabilities of 11-year PM2.5 exposure on latent multimorbidity patterns by age groups.Note: Models adjusted for age, age squared, gender, education, HuKou-residence, occupations, marital status, smoking status and logged GDP.*
## Growth curve models
To examine the associations between cumulative PM2.5 exposure and multimorbidity accumulation, we conduct a set of GCMs. First, we examine the linear and non-linear relationships between PM2.5 exposure and multimorbidity by adding linear and quadratic terms of PM2.5 exposure in GCMs (Table 4). The significant coefficients of both PM2.5 exposure and its quadratic term suggest a u-shaped association between PM2.5 exposure and multimorbidity (Table 4). This means that exposure to PM2.5 is positively associated with the likelihood of multimorbidity when the concentration of PM2.5 exposure is higher than 53.3 μg/m3. The inflection point of the u-shaped curve is slightly lower among older adults (Tables D and E in S1 File). For example, the inflection point is 56 μg/m3 among adults aged 45–64 years old and declines to 43 μg/m3 among adults aged 65–85 years old, suggesting that higher AP exposure has worse health effects on the older population.
**Table 4**
| Unnamed: 0 | Model 1: | Model 2: Model 1 | Model 3: Model 2 | Model 4: Model 3 |
| --- | --- | --- | --- | --- |
| | Base | + Education | + SES | + Smoking+ GDP |
| PM2.5 exposure | -0.191*** | -0.200*** | -0.206*** | -0.202*** |
| PM2.5 exposure | (-0.235 - -0.147) | (-0.244 - -0.155) | (-0.250 - -0.162) | (-0.246 - -0.158) |
| PM2.5 exposure square | 0.018*** | 0.019*** | 0.019*** | 0.019*** |
| PM2.5 exposure square | (0.014–0.022) | (0.015–0.023) | (0.015–0.023) | (0.015–0.023) |
| Age | 0.140*** | 0.139*** | 0.137*** | 0.137*** |
| Age | (0.124–0.155) | (0.123–0.154) | (0.121–0.153) | (0.122–0.153) |
| Age square | -0.0009*** | -0.0009*** | -0.0009*** | -0.0009*** |
| Age square | (-0.0011 - -0.0008) | (-0.00099 - -0.0007) | (-0.00098 - -0.0007) | (-0.00099 - -0.0007) |
| Gender (ref: Men) | | | | |
| Women | 0.152*** | 0.189*** | 0.180*** | 0.197*** |
| Women | (0.124–0.180) | (0.159–0.219) | (0.150–0.210) | (0.159–0.235) |
| Education (ref: no schooling) | | | | |
| Primary | | 0.147*** | 0.136*** | 0.135*** |
| Primary | | (0.110–0.184) | (0.0983–0.173) | (0.098–0.173) |
| Middle + | | 0.130*** | 0.0867*** | 0.0869*** |
| Middle + | | (0.089–0.171) | (0.0423–0.131) | (0.0425–0.131) |
| HuKou (ref: Rural) | | | | |
| Rural-urban | | | -0.029 | -0.025 |
| Rural-urban | | | (-0.066–0.008) | (-0.062–0.012) |
| Urban | | | 0.105*** | 0.107*** |
| Urban | | | (0.066–0.144) | (0.068–0.146) |
| Occupation (ref: agricultural) | | | | |
| Non- agricultural | | | -0.008 | -0.012 |
| Non- agricultural | | | (-0.043–0.026) | (-0.046–0.022) |
| Managerial | | | -0.042 | -0.044 |
| Managerial | | | (-0.095–0.011) | (-0.097–0.009) |
| Marital (ref: partnered) | | | | |
| Single | | | 0.014 | 0.017 |
| Single | | | (-0.024–0.052) | (-0.021–0.054) |
| Smoking (ref: Never) | | | | |
| Former | | | | 0.175*** |
| Former | | | | (0.137–0.214) |
| Current | | | | -0.024 |
| Current | | | | (-0.060–0.012) |
| Log GDP | | | | -0.014 |
| Log GDP | | | | (-0.036–0.008) |
| Constant | -4.518*** | -4.631*** | -4.521*** | -4.385*** |
| Constant | (-5.015 - -4.022) | (-5.131 - -4.131) | (-5.024 - -4.018) | (-4.931 - -3.838) |
| Random effects | | | | |
| Variance | | | | |
| Individuals (age) | 4.75e-17*** | 3.12e-17*** | 2.23e-17*** | 4.84e-17*** |
| Years | 0.608*** | 0.607*** | 0.602*** | 0.593*** |
| Years | (0.586–0.631) | (0.585–0.630) | (0.580–0.625) | (0.572–0.615) |
| Covariance | | | | |
| Individuals—Years | 3.95e-12*** | 2.85e-12*** | 2.49e-12*** | 4.87e-12*** |
| Log likelihood | -69877.261 | -69845.32 | -69824.071 | -69750.266 |
| Observations | 45788 | 45788 | 45788 | 45788 |
| Number of IDs | 19098 | 19098 | 19098 | 19098 |
In Model 4 (the full model) women have a higher prevalence of multimorbidity than men, and there is a curvilinear increase in multimorbidity over age. Unexpectedly, people with higher education and urban HuKou have a higher prevalence of multimorbidity. Occupation and partnership status is not significantly associated with the risk of multimorbidity. Compared with respondents who never smoked, former smokers have a higher prevalence of multimorbidity, but current smokers do not. There is not a significant association between GDP and multimorbidity accumulation.
## Heterogeneity in multimorbidity accumulation
To explore the trajectory of multimorbidity associated with PM2.5 exposure across age, based on Model 4 in Table 4, we interact age with PM2.5 exposure and PM2.5 squared; then, we plot the association between PM2.5 exposure and multimorbidity scores in Fig 3. Generally, respondents exposed to higher PM2.5 exposure have higher risk of multimorbidity. Over the age of 60, the respondents in the highest AP exposure categories (e.g., PM2.5 over 80+) have steeper multimorbidity trajectories than those in lower exposure categories. However, it is only at certain older ages (75 years, for example) that this is statistically significant. There is an inverted u-shaped relationship between age and multimorbidity, indicating a higher risk in multimorbidity with ageing among adults aged under 75 but a decline in multimorbidity with ageing among those between 75–85 years old. The oldest old (aged over 75) have a lower prevalence of multimorbidity.
**Fig 3:** *Predicted multimorbidity score across PM2.5 exposure by age.Note that this model controls age, age squared, gender, education, HuKou-residence, occupations, marital status, smoking status, and logged GDP.*
We conduct a set of analyses to understand heterogeneities in the association between PM2.5 and multimorbidity among different HuKou-residence groups. Generally, we can see that there is a u-shaped relationship between PM2.5 exposure and multimorbidity accumulation among all groups. Fig 4 shows that people with urban HuKou have a higher likelihood of being multimorbid than those with a rural Kukou when PM2.5 concentration is lower than 70 μg/m3, but when PM2.5 exposure is over 70 μg/m3, the associations of PM2.5 are not different between rural, rural-urban and urban residents. However, respondents with rural HuKou share a similar trajectory of multimorbidity across PM2.5 exposure regardless of their residences.
**Fig 4:** *Predicted multimorbidity score across PM2.5 exposure by HuKou-residence.Note that this model controls age, age squared, gender, education, HuKou-residence, occupations, marital status, smoking status and logged GDP.*
We conduct a number of robustness checks. We use categorical PM2.5 exposure to check the non-linear relationship between PM2.5 exposure and multimorbidity. First, we find that compared with the relatively healthy class, higher exposure to PM2.5 is associated with a higher prevalence of the other three classes of chronic diseases (Table B in S1 File). Second, higher exposure to PM2.5 is associated with a higher incidence rate ratio of multimorbidity in the longitudinal analyses (Table C in S1 File). In addition, comparing findings from complete data and MI data shows that results are consistent (Table F in S1 File). These sensitivity analyses indicate the robustness of our results.
## Discussion
By linking the CHARLS, a nationally representative dataset, with historical PM2.5 records derived from remote sensing technology, we investigate the associations between long-term exposure to PM2.5 and patterns and accumulation of multimorbidity. Incorporating 15-year PM2.5 exposure histories enables us to capture the associations between long-term exposure and chronic disease accumulation. To the best of our knowledge, this is the first study to establish a link between PM2.5 exposure and multimorbidity patterns, and to estimate associations between cumulative exposure and the accumulation of multimorbidity longitudinally. Findings from the LCA for multimorbidity patterns suggest that higher exposure to PM2.5 is associated with a higher risk of cardio-metabolic and respiratory multimorbidity (dominated by lung disease), whereas lower PM2.5 exposure is associated with a higher likelihood of musculoskeletal multimorbidity. Our longitudinal GCM findings show that both lower and higher historical AP exposure is associated with faster multimorbidity accumulation. This u-shaped association may be explained by the different multimorbidity clusters at opposite ends of AP exposure spectrum, as shown in the LCA models. These estimates suggested that for many middle-income countries such as China, more efforts to reduce PM2.5 concentrations would be associated with a substantial reduction in burden of multiple diseases.
First, our LCA analyses show that the four latent classes are differentially associated with PM2.5 exposure, which are partly in accordance with previous studies of AP and single diseases [40, 41]. For example, higher exposure to PM2.5 is associated with an increased likelihood of developing respiratory diseases and particularly, cardio-metabolic diseases (a cluster dominated by hypertension). Previous studies about associations between AP and hypertension are inconsistent, some of which find significant associations between them but others do not [32, 42–44]. These studies suggest that hypertension may be related to AP or caused and exacerbated by other cardiometabolic disorders (e.g., cardiovascular diseases) attributable to PM2.5.
One unexpected finding, partly inconsistent with previous research [45], is that higher PM2.5 exposure is associated with a reduced likelihood of developing musculoskeletal diseases such as arthritis, or to put it another way, that those suffering with musculoskeletal multimorbidity are more likely to live in areas of low air pollution. A possible explanation is that the rural-urban difference in environmental exposures (besides PM2.5 exposure)—like noise, green space, food environments—which makes urban residents more likely to be ill from metabolic conditions than achy joints and cartilages [32, 46–48]. Second, rural and urban residents may have had different occupational exposures throughout their lifetime. Our analysis shows that there are more urban residents working at non-agricultural jobs than rural residents ($65\%$ vs. $15\%$). It is likely that urban residents have led a more sedentary lifestyle (working in white-collar occupations) compared with rural residents who have been employed in farming, forest, hunting and fishing that are physically demanding [49]. This might predispose them to develop musculoskeletal conditions as there is more wear and tear on the body [50, 51]. Third, although AP exposure is higher in urban areas than in rural areas in China, there are potential offsetting benefits of urban residence, such as better access to health care [52]. Urban residents may be more likely to have better healthcare and get diagnosed for conditions that are not immediately obvious (e.g., hypertension); hence, those disorders might be undiagnosed and potentially underestimated in rural residents. Fourth, in China, rural areas consume more solid fuels (e.g., coal and wood) as the major source of energy [53], which leads to more severe indoor air pollution that is associated with increased odds of arthritis [54]. Furthermore, compared with respiratory or cardio-metabolic diseases, musculoskeletal diseases (especially arthritis) might be more influenced by health care than AP [55]. This should be explored in future research by using finer measures of rural-urban residence and migration, and by explicitly investigating these potential mechanisms for disparities.
Our results of GCMs suggest that cumulative PM2.5 exposure is associated with higher multimorbidity scores at both lower and higher levels of PM2.5 (e.g., a u-shaped association). Although there are few studies related to the u-shape association of environmental pollution with multimorbidity, the u-shape links between AP and health risks (e.g., hospital admissions and mortality) are well established [56]. Most AP in Chinese cities comes from industrial production and vehicle traffic, which is increasing in conjunction with economic development [57]. Due to a high proportion of industrial sectors and increasing traffic intensity in China, massive fossil fuels, especially coal, are consumed for economic development, and AP has been more and more severe [57]. This may explain why higher AP is associated with higher risk of chronic health diseases. Apart from the above contextual characteristics (urbanisation and economic development), the elevated multimorbidity at lower PM2.5 exposure might be attributed to higher levels of musculoskeletal multimorbidity that are related to the differences in the rural-urban context (as explained above). However, due to the strong side effects of AP, higher exposure to PM2.5 could lead to increased risk of multimorbidity once PM2.5 levels exceed the threshold (approximately 53 μg/m3 in our findings). The annual average PM2.5 exposure in China in 2015 was 55.2 μg/m3 [58], indicating that current PM2.5 exposure is harmful to human health among the majority of Chinese adults. These findings suggest that corresponding policies regarding AP should be implemented based on the strategies of sustainable development and disease prevention.
Third, the associations of multimorbidity accumulation show unexpected links with SES (the higher SES, including higher education and urban HuKou, is associated with a higher risk of multimorbidity). These unexpected results regarding SES can be understood from two perspectives. First of all, respondents with higher SES are more likely to be urban dwellers, who are exposed to higher AP as well as engaging in less physical exercise and more harmful health behaviours [59, 60]. Second, as mentioned above, this rural-urban differences in social fabric, including contextual and compositional factors associated with rural-urban residence, are perhaps fully captured by the current covariates. For example, urban residents tend to have higher educational levels and other social advantages (e.g., income, wealth, health awareness, social support etc.), as well as better access to health care. In addition, the measure of multimorbidity in this study is based on self-rated diagnosis, so respondents with higher SES might report a higher prevalence of chronic diseases [59]. The rural-urban environmental context may also contribute to these findings. Urban residents experience more environmental stressors (including noise and fast food), which lead to higher risks of multimorbidity [61, 62].
In the LCA results, we find that women have a lower likelihood of belonging to multimorbid classes, which is inconsistent with our findings in the longitudinal analyses. The reason might be that the multimorbidity classes in the LCA are each dominated by one disease. In this study, the respiratory cluster is dominated by lung diseases, and cardio-metabolic and musculoskeletal clusters are dominated by hypertension and arthritis. These findings are in line with previous studies that show that the prevalence of pulmonary diseases, hypertension and arthritis, is higher in men than women in China [63–65]. However, when considering multiple chronic diseases, women might have higher risks in multimorbidity because Chinese women have less access to medical resources than men [12].
In additional analyses (Figs 3 and 4), we further explore the associations between PM2.5 exposure and multimorbidity accumulation by age and HuKou-residence. First, these associations by age show that PM2.5 is associated with a higher number of morbidities among respondents aged 45–75, whereas for respondents aged over 75, PM2.5 is associated with lower risk of multimorbidity. This might be due to mortality selection. Older people with multiple morbidities might die younger or they are less exposed to AP. Second, we see that when PM2.5 exposure is lower than 80 μg/m3, respondents with urban HuKou have a higher risk of multimorbidity accumulation than those with rural residence. As previously discussed, urban residents have more advantageous socio-economic conditions and might report a higher prevalence of multimorbidity. However, when PM2.5 exposure exceeds 80 μg/m3, there is no significant difference in the associations between PM2.5 exposure and multimorbidity among groups with different HuKou-residence. This inconsistent finding might be due to two reasons. Firstly, the insignificant difference might be due to small sample sizes. There are only $5\%$ of respondents living in cities where PM2.5 levels are over 80 μg/m3. Secondly, due to more opportunities in major cities (e.g., more high-paying jobs, better education, and more access to health care), many people with rural HuKou decide to live and work in urban areas ($30\%$ people with rural HuKou live in urban areas). This might explain why there is no rural-urban difference in areas with high exposure (over 80 μg/m3).
Our study has several advantages over previous studies. First, we link the CHARLS with historical PM2.5 records over a 15-year period, which enables us to measure long-term exposure to AP for each respondent. Second, this is a first study to analyse associations between AP and multimorbidity clusters. Nevertheless, there are several limitations in this study. First, we were unable to obtain detailed addresses of respondents from CHARLS, and thus we use city-level exposure to predict individual multimorbidity. In the Chinese context, a city might cover a large area (e.g., Beijing city) and consist of inner-city areas (more urban, more polluted) and suburb areas (more rural, less polluted). This means that we cannot accurately compare the variations within the same city. Future research should link AP at a smaller geographic scale. Second, multimorbidity is measured by self-reported doctor’s diagnosis which might underestimate chronic diseases due to lower levels of diagnosis in some groups [66]. This might be related to individual characteristics (e.g., gender, age) making it less likely to seek treatment or related to reduced access to healthcare in some places. Third, our findings only indicate the associations of PM2.5 with multimorbidity, and thus do not have a causal interpretation. We cannot rule out that our findings might be the result of other air pollutants (e.g., indoor air pollution, NO2, O3, etc), or contextual factors that are highly correlated with PM2.5 exposure, for example, lack of green space or noise pollution. Fourth, our latent class analysis is based on 14 chronic diseases available in the CHARLS and does not therefore cover all chronic diseases. From the perspective of the association between multimorbidity and exposure to air pollution, the breadth of chronic diseases not included in the CHARLS survey makes it difficult to predict the direction of bias for each cluster. Thus, further analyses should collect a broader range of diseases to reduce the bias in disease clusters. Fifth, $13\%$ of participants (7,469 observations) did not have complete disease-reporting data. In longitudinal analyses, we used multiple imputation to complete the dataset; however, for the LCA analyses, this was not an option. It is possible that underreporting of certain types of diseases might create a bias in how disease clusters are associated with exposure to air pollution, although it is difficult to predict that bias. Future research with more complete disease data could help to solve this issue. Finally, we measure multimorbidity accumulation over a relatively short period of 4 years (2011–2015).
## Conclusion
This study provides evidence showing that higher cumulative exposure to PM2.5 is associated with increased risks of all types of multimorbidity patterns, but especially cardio-metabolic multimorbidity, and higher multimorbidity accumulation over 15 years. Notably, areas with low AP exposure still have higher rates of multimorbidity, associated with musculoskeletal disorders. Thus, our study highlights how multimorbidity clusters vary contextually and reveals that PM2.5 exposure is more detrimental to health among older adults. However, further research is needed to unpick the nexus of contextual and compositional factors associated with the development of chronic diseases in rural and urban settings and to detect their causal mechanisms.
## References
1. Prados-Torres A, Calderón-Larrañaga A, Hancco-Saavedra J, Poblador-Plou B, Van Den Akker M. **Multimorbidity patterns: A systematic review**. *Journal of Clinical Epidemiology. Pergamon* (2014) 254-266. DOI: 10.1016/j.jclinepi.2013.09.021
2. Mini GK, Thankappan KR. **Pattern, correlates and implications of non-communicable disease multimorbidity among older adults in selected Indian states: a cross-sectional study**. *BMJ Open* (2017) **7** e013529. DOI: 10.1136/BMJOPEN-2016-013529
3. Montes MC, Bortolotto CC, Tomasi E, Gonzalez MC, Barbosa-Silva TG, Domingues MR. **Strength and multimorbidity among community-dwelling elderly from southern Brazil**. *Nutrition* (2020) **71** 110636. DOI: 10.1016/j.nut.2019.110636
4. Chang AY, Gómez-Olivé FX, Payne C, Rohr JK, Manne-Goehler J, Wade AN. **Chronic multimorbidity among older adults in rural South Africa**. *BMJ Glob Heal* (2019) **4** e001386. DOI: 10.1136/BMJGH-2018-001386
5. Chen H, Cheng M, Zhuang Y, Broad JB. **Multimorbidity among middle-aged and older persons in urban China: Prevalence, characteristics and health service utilization**. *Geriatr Gerontol Int* (2018) **18** 1447-1452. DOI: 10.1111/ggi.13510
6. Singer L, Green M, Rowe F, Ben-Shlomo Y, Morrissey K. **Social determinants of multimorbidity and multiple functional limitations among the ageing population of England, 2002–2015**. *SSM—Popul Heal* (2019) **8** 100413. DOI: 10.1016/j.ssmph.2019.100413
7. Zhao YW, Haregu TN, He L, Lu S, Katar A, Wang H. **The effect of multimorbidity on functional limitations and depression amongst middle-aged and older population in China: a nationwide longitudinal study**. *Age Ageing* (2020) 1-8. DOI: 10.1093/ageing/afaa117
8. Chudasama Y V, Khunti K, Gillies CL, Dhalwani NN, Davies MJ, Yates T, Basu S. **Healthy lifestyle and life expectancy in people with multimorbidity in the UK Biobank: A longitudinal cohort study**. *PLOS Med* (2020) **17** e1003332. DOI: 10.1371/journal.pmed.1003332
9. Zhao Y, Atun R, Oldenburg B, McPake B, Tang S, Mercer SW. **Physical multimorbidity, health service use, and catastrophic health expenditure by socioeconomic groups in China: an analysis of population-based panel data**. *Lancet Glob Heal* (2020) **8** e840-e849. DOI: 10.1016/S2214-109X(20)30127-3
10. Chan MS, van den Hout A, Pujades-Rodriguez M, Jones MM, Matthews FE, Jagger C. **Socio-economic inequalities in life expectancy of older adults with and without multimorbidity: a record linkage study of 1.1 million people in England**. *Int J Epidemiol* (2019) **48** 1340-1351. DOI: 10.1093/ije/dyz052
11. Quiñones AR, Liang J, Bennett JM, Xu X, Ye W. **How does the trajectory of multimorbidity vary across black, white, and mexican americans in middle and old age?**. *Journals Gerontol—Ser B Psychol Sci Soc Sci* (2011) **66 B** 739-749. DOI: 10.1093/geronb/gbr106
12. Yao SS, Cao GY, Han L, Chen ZS, Huang ZT, Gong P, Newman A. **Prevalence and patterns of multimorbidity in a nationally representative sample of older chinese: Results from the china health and retirement longitudinal study**. *Journals Gerontol—Ser A Biol Sci Med Sci* (2020) **75** 1974-1980. DOI: 10.1093/gerona/glz185
13. Whitty CJM, Watt FM. **Map clusters of diseases to tackle multimorbidity**. *Nature. Nature Research* (2020) 494-496. DOI: 10.1038/d41586-020-00837-4
14. Wang J, Li T, Lv Y, Kraus VB, Zhang Y, Mao C. **Fine particulate matter and poor cognitive function among Chinese older adults: Evidence from a community-based, 12-year prospective cohort study**. *Environ Health Perspect* (2020) **128** 1-9. DOI: 10.1289/EHP5304
15. Liu X, Tu R, Qiao D, Niu M, Li R, Mao Z. **Association between long-term exposure to ambient air pollution and obesity in a Chinese rural population: The Henan Rural Cohort Study**. *Environ Pollut* (2020) **260** 114077. DOI: 10.1016/j.envpol.2020.114077
16. Tian Y, Liu H, Liang T, Xiang X, Li M, Juan J. **Fine particulate air pollution and adult hospital admissions in 200 Chinese cities: A time-series analysis**. *Int J Epidemiol* (2019) **48** 1142-1151. DOI: 10.1093/ije/dyz106
17. Wang L, Bai Y, Zhang F, Wang W, Liu X, Krafft T. **Spatiotemporal Patterns of Ozone and Cardiovascular and Respiratory Disease Mortalities Due to Ozone in Shenzhen**. *Sustainability* (2017) **9** 559. DOI: 10.3390/su9040559
18. Schikowski T, Ranft U, Sugiri D, Vierkötter A, Brüning T, Harth V. **Decline in air pollution and change in prevalence in respiratory symptoms and chronic obstructive pulmonary disease in elderly women**. *Respir Res* (2010) **11**. DOI: 10.1186/1465-9921-11-113
19. Liu S, Yan Z, Liu Y, Yin Q, Kuang L. **Association between air pollution and chronic diseases among the elderly in China**. *Nat Hazards* (2017) **89** 79-91. DOI: 10.1007/s11069-017-2955-7
20. Mordukhovich I, Coull B, Kloog I, Koutrakis P, Vokonas P, Schwartz J. **Exposure to sub-chronic and long-term particulate air pollution and heart rate variability in an elderly cohort: the Normative Aging Study**. *Environ Heal A Glob Access Sci Source* (2015) **14**. DOI: 10.1186/s12940-015-0074-z
21. Delfino RJ, Tjoa T, Gillen DL, Staimer N, Polidori A, Arhami M. **Traffic-related air pollution and blood pressure in elderly subjects with coronary artery disease**. *Epidemiology* (2010) **21**. DOI: 10.1097/EDE.0b013e3181d5e19b
22. Chen SY, Chu DC, Lee JH, Yang YR, Chan CC. **Traffic-related air pollution associated with chronic kidney disease among elderly residents in Taipei City**. *Environ Pollut* (2018) **234** 838-845. DOI: 10.1016/j.envpol.2017.11.084
23. Yin P, Brauer M, Cohen AJ, Wang H, Li J, Burnett RT. **The effect of air pollution on deaths, disease burden, and life expectancy across China and its provinces, 1990–2017: an analysis for the Global Burden of Disease Study 2017**. *Lancet Planet Heal* (2020) **4** e386-e398. DOI: 10.1016/S2542-5196(20)30161-3
24. Xiao Q, Geng G, Liang F, Wang X, Lv Z, Lei Y. **Changes in spatial patterns of PM2.5 pollution in China 2000–2018: Impact of clean air policies**. *Environ Int* (2020) **141** 105776. DOI: 10.1016/j.envint.2020.105776
25. Zeng Y, Cao Y, Qiao X, Seyler BC, Tang Y. **Air pollution reduction in China: Recent success but great challenge for the future**. *Sci Total Environ* (2019) **663** 329-337. DOI: 10.1016/j.scitotenv.2019.01.262
26. Tong S.. **Air pollution and disease burden**. *The Lancet Planetary Health. Elsevier B.V* (2019) e49-e50. DOI: 10.1016/S2542-5196(18)30288-2
27. Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. **Cohort profile: The China health and retirement longitudinal study (CHARLS)**. *Int J Epidemiol* (2014) **43** 61-68. DOI: 10.1093/ije/dys203
28. Zhang R, Lu Y, Shi L, Zhang S, Chang F. **Prevalence and patterns of multimorbidity among the elderly in China: A cross-sectional study using national survey data**. *BMJ Open. BMJ Publishing Group* (2019) e024268. DOI: 10.1136/bmjopen-2018-024268
29. Shang Y, Sun Z, Cao J, Wang X, Zhong L, Bi X. **Systematic review of Chinese studies of short-term exposure to air pollution and daily mortality**. *Environment International. Elsevier Ltd* (2013) 100-111. DOI: 10.1016/j.envint.2013.01.010
30. Dauchet L, Hulo S, Cherot-Kornobis N, Matran R, Amouyel P, Edmé JL. **Short-term exposure to air pollution: Associations with lung function and inflammatory markers in non-smoking, healthy adults**. *Environ Int* (2018) **121** 610-619. DOI: 10.1016/j.envint.2018.09.036
31. Xiao Q, Chang HH, Geng G, Liu Y. **An Ensemble Machine-Learning Model to Predict Historical PM2.5 Concentrations in China from Satellite Data**. *Environ Sci Technol* (2018) **52** 13260-13269. DOI: 10.1021/acs.est.8b02917
32. Liu C, Chen R, Zhao Y, Ma Z, Bi J, Liu Y. **Associations between ambient fine particulate air pollution and hypertension: A nationwide cross-sectional study in China**. *Sci Total Environ* (2017) **584–585** 869-874. DOI: 10.1016/j.scitotenv.2017.01.133
33. Mudway IS, Dundas I, Wood HE, Marlin N, Jamaludin JB, Bremner SA. **Impact of London’s low emission zone on air quality and children’s respiratory health: a sequential annual cross-sectional study**. *Lancet Public Heal* (2019) **4** e28-e40. DOI: 10.1016/S2468-2667(18)30202-0
34. Cao J, Chow JC, Lee FSC, Watson JG. **Evolution of PM2.5 measurements and standards in the U.S. And future perspectives for China**. *Aerosol Air Qual Res* (2013) **13** 1197-1211. DOI: 10.4209/aaqr.2012.11.0302
35. Hou B, Nazroo J, Banks J, Marshall A. **Are cities good for health? A study of the impacts of planned urbanization in China**. *Int J Epidemiol* (2019) **48** 1083-1090. DOI: 10.1093/ije/dyz031
36. Walder AG. **Career Mobility and the Communist Political Order**. *Am Sociol Rev* (1995) **60** 309. DOI: 10.2307/2096416
37. Larsen FB, Pedersen MH, Friis K, Glümer C, Lasgaard M, Boltze J. **A Latent Class Analysis of Multimorbidity and the Relationship to Socio-Demographic Factors and Health-Related Quality of Life. A National Population-Based Study of 162,283 Danish Adults**. *PLoS One* (2017) **12** e0169426. DOI: 10.1371/journal.pone.0169426
38. Bayes-Marin I, Sanchez-Niubo A, Egea-Cortés L, Nguyen H, Prina M, Fernández D. **Multimorbidity patterns in low-middle and high income regions: A multiregion latent class analysis using ATHLOS harmonised cohorts**. *BMJ Open* (2020) **10** 34441. DOI: 10.1136/bmjopen-2019-034441
39. Luo Y, Zhang L, Pan X. **Neighborhood Environments and Cognitive Decline among Middle-Aged and Older People in China**. *Journals Gerontol—Ser B Psychol Sci Soc Sci* (2019) **74** e60-e71. DOI: 10.1093/geronb/gbz016
40. Seposo X, Ueda K, Sugata S, Yoshino A, Takami A. **Short-term effects of air pollution on daily single- and co-morbidity cardiorespiratory outpatient visits**. *Sci Total Environ* (2020) **729** 138934. DOI: 10.1016/j.scitotenv.2020.138934
41. Jiao K, Xu M, Liu M. **Health status and air pollution related socioeconomic concerns in urban China**. *Int J Equity Health* (2018) **17** 18. DOI: 10.1186/s12939-018-0719-y
42. Adar SD, Chen YH, D’souza JC, O’neill MS, Szpiro AA, Auchincloss AH. **Longitudinal analysis of long-term air pollution levels and blood pressure: A cautionary tale from the multi-ethnic study of atherosclerosis**. *Environ Health Perspect* (2018) **126**. DOI: 10.1289/EHP2966
43. Fuks KB, Weinmayr G, Basagaña X, Gruzieva O, Hampel R, Oftedal B. **Long-termexposure to ambient air pollution and traffic noise and incident hypertension in seven cohorts of the European study of cohorts for air pollution effects (ESCAPE)**. *Eur Heart J* (2017) **38** 983-990. DOI: 10.1093/eurheartj/ehw413
44. Pope CA, Turner MC, Burnett RT, Jerrett M, Gapstur SM, Diver WR. **Relationships between fine particulate air pollution, cardiometabolic disorders, and cardiovascular mortality**. *Circ Res* (2015) **116** 108-115. DOI: 10.1161/CIRCRESAHA.116.305060
45. Hart JE, Laden F, Puett RC, Costenbader KH, Karlson EW. **Exposure to traffic pollution and increased risk of rheumatoid arthritis**. *Environ Health Perspect* (2009) **117** 1065-9. DOI: 10.1289/ehp.0800503
46. De Roos AJ, Koehoorn M, Tamburic L, Davies HW, Brauer M. **Proximity to traffic, Ambient air pollution, And community noise in relation to incident rheumatoid arthritis**. *Environ Health Perspect* (2014) **122** 1075-1080. DOI: 10.1289/ehp.1307413
47. Chang KH, Hsu CC, Muo CH, Hsu CY, Liu HC, Kao CH. **Air pollution exposure increases the risk of rheumatoid arthritis: A longitudinal and nationwide study**. *Environ Int* (2016) **94** 495-499. DOI: 10.1016/j.envint.2016.06.008
48. Liu C, Yang C, Zhao Y, Ma Z, Bi J, Liu Y. **Associations between long-term exposure to ambient particulate air pollution and type 2 diabetes prevalence, blood glucose and glycosylated hemoglobin levels in China**. *Environ Int* (2016) **92–93** 416-421. DOI: 10.1016/j.envint.2016.03.028
49. Matz CJ, Stieb DM, Brion O. **Urban-rural differences in daily time-activity patterns, occupational activity and housing characteristics**. *Environ Heal A Glob Access Sci Source* (2015) **14** 1-11. DOI: 10.1186/s12940-015-0075-y
50. Felson DT. **Do occupation-related physical factors contribute to arthritis?**. *Baillieres Clin Rheumatol* (1994) **8** 63-77. DOI: 10.1016/s0950-3579(05)80225-0
51. Brennan-Olsen SL, Solovieva S, Viikari-Juntura E, Ackerman IN, Bowe SJ, Kowal P. **Arthritis diagnosis and symptoms are positively associated with specific physical job exposures in lower- and middle-income countries: Cross-sectional results from the World Health Organization’s Study on global AGEing and adult health (SAGE)**. *BMC Public Health* (2018) **18** 1-12. DOI: 10.1186/s12889-018-5631-2
52. Song Q, Smith JP. **Hukou system, mechanisms, and health stratification across the life course in rural and urban China**. *Heal Place* (2019) **58** 102150. DOI: 10.1016/j.healthplace.2019.102150
53. Qiu Y, Yang FA, Lai W. **The impact of indoor air pollution on health outcomes and cognitive abilities: empirical evidence from China**. *Popul Environ* (2019) **40** 388-410. DOI: 10.1007/s11111-019-00317-6
54. Yamamoto SS, Yacyshyn E, Jhangri GS, Chopra A, Parmar D, Jones CA. **Household air pollution and arthritis in lowand middle-income countries: Cross-sectional evidence from the World Health Organization’s study on Global Ageing and Adult Health**. *PLoS One* (2019) **14** e0226738. DOI: 10.1371/journal.pone.0226738
55. Lutfiyya MN, McCullough JE, Saman DM, Lemieux A, Hendrickson S, McGrath CA. **Rural/Urban differences in health services deficits among U.S. adults with arthritis: A population- based study**. *J Nurs Educ Pract* (2013) **3**. DOI: 10.5430/jnep.v3n11p43
56. Zhang K, Batterman S. **Air pollution and health risks due to vehicle traffic**. *Sci Total Environ* (2013) **450–451** 307-316. DOI: 10.1016/j.scitotenv.2013.01.074
57. Cheng Z, Li L, Liu J. **Identifying the spatial effects and driving factors of urban PM2.5 pollution in China**. *Ecol Indic* (2017) **82** 61-75. DOI: 10.1016/j.ecolind.2017.06.043
58. Huang J, Pan X, Guo X, Li G. **Health impact of China’s Air Pollution Prevention and Control Action Plan: an analysis of national air quality monitoring and mortality data**. *Lancet Planet Heal* (2018) **2** e313-e323. DOI: 10.1016/S2542-5196(18)30141-4
59. Wang H, Stokes JE. **Trajectories of rural-urban disparities in biological risks for cardiovascular disease among Chinese middle-aged and older adults**. *Heal Place* (2020) **64** 102354. DOI: 10.1016/j.healthplace.2020.102354
60. Han W, Li Z, Guo J, Su T, Chen T, Wei J. **The Urban–Rural Heterogeneity of Air Pollution in 35 Metropolitan Regions across China**. *Remote Sens* (2020) **12** 2320. DOI: 10.3390/rs12142320
61. Münzel T, Sørensen M, Gori T, Schmidt FP, Rao X, Brook J. **Environmental stressors and cardio-metabolic disease: Part I-epidemiologic evidence supporting a role for noise and air pollution and effects of mitigation strategies**. *European Heart Journal. Oxford University Press* (2017) 550-556. DOI: 10.1093/eurheartj/ehw269
62. Allen L, Williams J, Townsend N, Mikkelsen B, Roberts N, Foster C. **Socioeconomic status and non-communicable disease behavioural risk factors in low-income and lower-middle-income countries: a systematic review**. *Lancet Glob Heal* (2017) **5** e277-e289. DOI: 10.1016/S2214-109X(17)30058-X
63. Lin DF, Yang WQ, Zhang PP, Lv Q, Jin O, Gu JR. **Clinical and prognostic characteristics of 158 cases of relapsing polychondritis in China and review of the literature**. *Rheumatol Int* (2016) **36** 1003-1009. DOI: 10.1007/s00296-016-3449-8
64. Lei X, Yin N, Zhao Y. **Socioeconomic status and chronic diseases: The case of hypertension in China**. *China Econ Rev* (2012) **23** 105-121. DOI: 10.1016/j.chieco.2011.08.004
65. Chan KY, Li X, Chen W, Song P, Wong NWK, Poon AN. **Prevalence of chronic obstructive pulmonary disease (COPD) in China in 1990 and 2010**. *J Glob Health* (2017) **7**. DOI: 10.7189/jogh.07.020704
66. Li HL, Fang J, Zhao LG, Liu DK, Wang J, Han LH. **Personal characteristics effects on validation of self-reported type 2 diabetes from a cross-sectional survey among Chinese adults**. *J Epidemiol* (2020) **30** 516-521. DOI: 10.2188/jea.JE20190178
|
---
title: 'Non-communicable disease policy implementation in Libya: A mixed methods assessment'
authors:
- Luke N. Allen
- Cervantée E. K. Wild
- Giulia Loffreda
- Mohini Kak
- Mohamed Aghilla
- Taher Emahbes
- Atousa Bonyani
- Arian Hatefi
- Christopher Herbst
- Haider M. El Saeh
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021530
doi: 10.1371/journal.pgph.0000615
license: CC BY 4.0
---
# Non-communicable disease policy implementation in Libya: A mixed methods assessment
## Abstract
The Libyan Ministry of *Health is* keen to understand how it can introduce policies to protect its population from non-communicable diseases (NCDs). We aimed to perform an implementation research assessment of the current situation, including challenges and opportunities. We used an explanatory sequential mixed methods design. We started with a quantitative assessment of NCD policy performance based on review of the WHO NCD Progress Monitor Reports. Once we had identified Libya’s NCD policy gaps we performed a systematic review to identify international lessons around barriers and successful strategies for the policies Libya has not yet implemented. Finally, we performed a series of key stakeholder interviews with senior policymakers to explore their perspectives around promising policy actions. We used a realist paradigm, methods triangulation, and a joint display to synthesise the interpretation of our findings and develop recommendations. Libya has not fully implemented any of the recommended policies for diet, physical activity, primary care guidelines & therapeutics, or data collection, targets & surveillance. It does not have robust tobacco policies in place. Evidence from the international literature and policymaker interviews emphasised the centrality of according strong political leadership, governance structures, multisectoral engagement, and adequate financing to policy development activities. Libya’s complex political and security situation are major barriers for policy implementation. Whilst some policies will be very challenging to develop and deploy, there are a number of simple policy actions that could be implemented with minimum effort; from inviting WHO to conduct a second STEPS survey, to signing the international code on breast-milk substitutes. Like many other fragile and conflict-affected states, Libya has not accorded NCDs the policy attention they demand. Whilst strong high-level leadership is the ultimate key to providing adequate protections, there are a range of simple measures that can be implemented with relative ease.
## Background
Non-communicable diseases (NCDs) have risen to become the leading cause of death and disability in Libya, as they are worldwide [1]. In 2019, $79\%$ of all Libyan deaths and $78\%$ of DALYS were caused by NCDs [2]. The ‘big four’ (cardiovascular diseases, cancers, chronic respiratory diseases, and diabetes mellitus) collectively account for two thirds of Libyan mortality. Ischemic heart disease and stroke have been the two top causes of deaths since 2009, and the prevalence of hypertensive heart disease and diabetes are increasing [3].
The 2009 WHO ‘STEPS’ survey found that $99.8\%$ of the population had at least one of the following NCD risk factors; daily smoking; consuming fewer than five servings of fruits & vegetables per day; low levels of physical activity; overweight; or raised blood pressure [4, 5]. Dietary risks are much larger contributors to morbidity than tobacco or alcohol [2]. Obesity has more than doubled during the last three decades and over two thirds of Libyan adults are now overweight or obese [6, 7].
To address this high NCD burden, the Libyan government has endorsed the full set of WHO-recommended ‘Best Buy’ policies outlined in the WHO NCD Global Action Plan [8] (and summarised in Box 1), however they have not been fully implemented. These population-level interventions have a demonstrated effect size, and have been assessed for cost effectiveness, feasibility, as well as non-financial considerations in low- and middle-income countries.
Libya is emerging from a decade of conflict that followed the Arab Spring protests of 2011, which has left the health system fragile and deeply impacted in terms of access, service delivery and quality. For several years the country has been divided and governed by the Tripoli-based Government of National Accord (GNA) in the West and the Libyan National Army (LNA) in the East. In March 2021, through a UN facilitated peace process, a unified Government of National Unity (GNU) was established and will lead Libya until elections scheduled for 2022. The possibility of peace and stability provides tremendous opportunity for the new Government to work systematically towards strengthening and reforming health programs and policies to better align them with the current needs of the population.
To better understand the barriers, opportunities, and most appropriate approaches for implementing effective NCD policies in Libya, the Ministry of Health initiated an assessment in partnership with the World Bank, reported here. The threefold aims of this project were to understand: This research was funded by the Middle East and North Africa (MENA) Transition Fund managed by the World Bank and implemented/conducted in collaboration with the Libyan Ministry of Health (MoH). Methods specialists from London School of Hygiene and Tropical Medicine, the University of Oxford, and Queen Margaret University’s Institute for Global Health and Development were employed to lead the project as World Bank consultants. Professors from the University of Tripoli with key high-level policymaking experience completed the research team.
## Box 1: WHO ‘Best Buy’ NCD Policies
Source: WHO 2020 NCD Progress Monitor Report
## Ethics statement
This project was initiated by the World Bank in collaboration with the Libyan Ministry of Health. The project had a very tight deadline (three months). As soon as the research team had been put together and the study protocol had been developed, our research lead (LA), World Bank team coordinator (MK) and Libyan university-based co-investigators (MES, TE) sought to obtain ethical approval from a Libyan ethics board. Very few–if any–research ethics committees are currently functioning in Libya, and none were able to review our protocol. We took advice from the Oxford Tropical Research Ethics Committee. After submitting correspondence from the World Bank (the study sponsor) confirming that it was not possible to obtain ethical approval from a Libyan board, they agreed to review the study, noting that the small number of interviews and our ethical safeguards rendered it “very low-risk”. The Oxford Tropical Research Ethics Committee approved the study in July 2021 (OxTREC 541–21).
## Approach
We aimed to identify Libya’s NCD Best Buy policy gaps, assess how other countries had approached similar implementation challenges, and then identify the unique policy challenges and opportunities facing Libyan policymakers. As such, we used an explanatory sequential mixed methods approach [9, 10].
## Quantitative policy review
First, we performed a quantitative assessment of Libya’s NCD policy scores as reported in the 2015, 2017, and 2020 WHO NCD Progress Monitor Reports [11–13]. These reports present country-level data, scoring the level of implementation of each of 19 NCD policies (‘full’ implementation = 1 point; ‘partial’ = 0.5 points, and ‘not implemented’ = 0). Analysing Libya’s NCD policy scores over the three years allowed us to identify four distinct groupings: We used simple descriptive statistics to summarise the findings and compare Libya’s performance to global means from the remaining 193 WHO Member States.
By 2020, Libya had fully implemented five of the 19 WHO NCD Best Buy policies ($26\%$): smoke-free places, tobacco advertising restrictions, alcohol sales restrictions, alcohol advertising restrictions, and alcohol taxation. Regular risk factor surveys were partially implemented in 2015 and 2017 but dropped in 2020. Clinical guidelines were partially introduced in 2020. Table 1 summarizes Libya’s performance over time:
**Table 1**
| Best Buy Policies, sorted by cluster | 2015 | 2017 | 2020 |
| --- | --- | --- | --- |
| Targets, data collection, and plans | Targets, data collection, and plans | Targets, data collection, and plans | Targets, data collection, and plans |
| National NCD targets | 0 | 0 | 0 |
| Routine mortality data collection | 0 | 0 | 0 |
| Regular risk factor surveys | 0.5 | 0.5 | 0 |
| Multisectoral NCD plan | 0 | 0 | 0 |
| Tobacco | Tobacco | Tobacco | Tobacco |
| Tobacco tax | 0 | 0 | 0 |
| Smoke-free places | 1 | 1 | 1 |
| Tobacco graphic warnings | 0 | 0 | 0 |
| Tobacco advertising restrictions | 1 | 1 | 1 |
| Tobacco mass media campaigns | | 0 | 0 |
| Alcohol | Alcohol | Alcohol | Alcohol |
| Alcohol sale restrictions | 1 | 1 | 1 |
| Alcohol advertising restrictions | 1 | 1 | 1 |
| Alcohol tax | 1 | 1 | 1 |
| Diet | Diet | Diet | Diet |
| Salt reduction policies | 0 | 0 | 0 |
| Fat reduction policies | 0 | 0 | 0 |
| Child food marketing policies | 0 | 0 | 0 |
| Breast-milk substitute marketing | 0 | 0 | 0 |
| Physical activity | Physical activity | Physical activity | Physical activity |
| Physical activity mass media campaigns | 0 | 0 | 0 |
| Primary care guidelines and therapeutics | Primary care guidelines and therapeutics | Primary care guidelines and therapeutics | Primary care guidelines and therapeutics |
| Clinical guidelines | 0 | 0 | 0.5 |
| Cardiovascular therapies | 0 | 0 | 0 |
| Total | 5.5 | 5.5 | 5.5 |
In terms of international context, Libya’s overall implementation score of 5.5 places it in the lower third globally. The mean implementation score is $\frac{8.9}{19}$ for the WHO eastern Mediterranean region, $\frac{7.6}{19}$ for lower-middle-income countries, and $\frac{9.5}{19}$ for upper-middle-income countries. Libya bucks the global trends for alcohol policies, with all Best Buys in place. This is a real area of strength, shared with other Middle Eastern countries with large Muslim populations.
Libya has not implemented targets or plans. These policies are among the most widely implemented policies worldwide—possibly because they require relatively low expenditures and because WHO has produced guidance, template targets, and technical support on these areas [29].
Box 2 highlights the 12 policies that have never been implemented and the single lapsed policy; together representing the country’s NCD policy gaps.
## Systematic review
Once we had identified Libya’s NCD policy gaps we conducted a systematic review to synthesise international evidence of the mechanisms used to successfully implement these policies in other settings.
Our review was registered (PROSPERO: 42020153895) and conducted according to PRISMA and Synthesis Without Meta-analysis (SWIM) guidelines [14]. The search was conducted on 5th July 2021 using a combination of terms for NCDs and NCD Best Buy policies, on Web of Science, MEDLINE (through PubMed) Scopus, Google scholar (first 30 pages), and the World Bank and WHO IRIS databases. Full search terms are presented in S1 Text. Results were restricted to >2011, the year when the Best Buys policies were introduced [15].
Studies that analysed adoption and implementation issues from a political economy perspective and provided empirical evidence were eligible. We excluded editorials, reviews, and conference abstracts, but performed bibliographic searches to uncover further relevant papers. Full inclusion and exclusion criteria are presented in S1 Text. One reviewer performed title and abstract screening with $20\%$ of records independently screened by a second reviewer, using an excel spreadsheet. Dual independent review was used for full text screening, extraction, and risk of bias assessment. The reviewers resolved disagreements by consensus, with recourse to a third reviewer if necessary.
We developed an excel-based data extraction form, based on Cochrane guidance [16], which was piloted by the review team. We extracted bibliometric information, political economy factors (actors, institutions, interests, ideas and network, context, governance, power, policy dynamics, and implementation aspects), general challenges and facilitators, and specific barriers and facilitators to policy implementation relating to adoption, implementation, and adaptation to the local context [17, 18]. Our data extraction form is reproduced in S1 Text. Two independent reviewers used CERQual [19] to assess the risk of bias for each included study.
We developed a bespoke conceptual analytic framework to synthesise and analyse our findings: we set aside a series of virtual meetings between the team members and iteratively developed the key domains that emerged from the review findings. We continually refined the model until we felt that it accurately captured interrelations between the different types of outcomes. We used the model to structure our findings, and they also formed the subheadings for the interview guide used with key stakeholders.
In assessing the approaches used by other countries to address NCD policy gaps we paid particular attention to fragile and conflict-affected States [20] that may share contextual similarities with Libya. The findings from the review informed a series of targeted interviews with key NCD policymakers. We aimed to explore whether approaches used in other settings could be used domestically, and to identify major contemporary barriers and opportunities for policy implementation in Libya.
## Qualitative interviews
We used purposive sampling to recruit key policymakers and NCD policy stakeholders that had been previously identified in a separate mapping exercise conducted by World Bank and MoH staff [21]. All potential interviewees were sent information about the study via email. Participants were senior members of government and international NGOs, including representatives from the WHO, the International Rescue Committee, the National Center for Disease Control (NCDC), and the Libyan Food and Drug Control Center (FDCC). All interviewees spoke fluent English and did not require translators. Informed verbal consent was obtained from all participants and recorded using Microsoft Teams.
Five interviews were conducted by CW, MK and LA in July–September 2021 via Microsoft Teams. Open-ended and clarifying questions were used, based on a semi-structured interview guide. All interviews were audio recorded and lasted approximately 50–60 minutes. Notes were taken during the interviews and when reviewing the audio recordings afterwards.
Based on the framework constructed from the systematic review, we conducted a deductive framework analysis [22] to identify policy landscape factors and implementation challenges and opportunities under each domain, using nVivo v1.2 (QSR International Pty Ltd, Melbourne, 2020).
## Reflexivity
Our research team is multinational, with a combination of senior Libyan academic policymakers living and working in Libya and research specialists with experience of conducting mixed-methods assessments in numerous countries but no previous links to Libya. For further detail see the COREQ checklist (S1 Checklist).
## Integration of mixed methods data
We integrated our findings from each of the three research elements using a mixed-methods joint display [23], following Guetterman and colleagues’ best practice recommendations [24]. We used methods triangulation [25, 26] to explore areas of similarity and dissonance between datasets in the processes of interpretation and developing final recommendations, undergirded by a pragmatist philosophical paradigm [27, 28].
## Box 2: Libya’s NCD policy gaps
Regressed since 2015 Never implemented Targets, data collection, and plans Tobacco
Diet Physical activity Primary care guidelines and therapeutics
## Systematic review findings
Our search identified 9,659 records, of which 186 were included (Fig 1). Most studies were conducted in Europe/Americas ($22\%$, $$n = 41$$), $6\%$ were conducted in the Eastern Mediterranean region ($$n = 12$$), $21\%$ in Sub-Saharan Africa ($$n = 40$$), $26\%$ in Asia/Pacific ($$n = 50$$), and the remaining had a global or transcontinental focus. Two thirds of studies used qualitative methods ($$n = 120$$, $64\%$) and a quarter focused on fragile and conflict-affected settings (henceforth ‘FCAS’) ($23\%$, $$n = 43$$). Almost a third of the studies focused on diet-related policies ($$n = 58$$, $31\%$). A full list of references included in systematic review in provided in S1 Text.
**Fig 1:** *PRISMA flow diagram.*
## Conceptual analytic framework
The framework that we iteratively developed to guide the organisation and analysis of our data is presented in Fig 2. Its design reflects the fact that there are a series of shared issues that relate to all Best Buy policies; then there are subsets that are unique to four different clusters: targets, data collection, and plans; primary care guidelines and therapeutics; physical activity; and a commercial determinants [30, 31] sub-cluster that included shared lessons for diet, alcohol, and tobacco. Each of these also had their own unique implementation lessons, and we also found a final subset of shared lessons that applied to diet and alcohol.
**Fig 2:** *Conceptual analytic framework developed to guide analysis and synthesis.*
## Targets, data collection, and plans
The literature suggests that functioning NCD surveillance systems are critical for target setting, monitoring, planning, and to raise awareness and reinforce political commitments; however, these systems are often inadequate in FCAS, undermining policy planning [32–38]. Strong governance systems that facilitate multisectoral collaboration, partnership building, community mobilization, social participation, and advocacy are critical for the development of national plans, as is strong leadership to coordinate national and regional action across departments and sectors [39–45]. Under-prioritization of NCDs has resulted in insufficient resource allocation and the inability to finance national activities and plans [39, 43, 46–48]. Two well-conducted qualitative studies from Kenya, and Ghana and two mixed methods studies from Uganda and Samoa noted difficulties in shifting MoH financial resources away from historically well-funded areas (e.g. HIV/AIDS) to NCDs, making it difficult to fund surveillance activities and the development of targets and plans [33, 34, 49, 50].
Financing for NCD prevention and control is a challenge and the national budget does not include a specific NCD line-item, undermining both NCD surveillance and plan development. Potential funds have been diverted to the COVID-19 response and securing vaccines. Although there may be technical capacity at the ministry level to design and implement policies, there is lack of buy-in from senior decision makers. Virtually no action can be taken without approval from the very highest levels of government. One interviewee noted that even if policy makers were to pass NCD legislation, lack of enforcement capability is likely to neuter implementation.
Few plans relating to the Best Buys have been developed; most are in drafting stage and progress has been impeded by a lack of multi-sectoral collaboration. Much of this appears to be related to the lack of NCD surveillance.
In addition, ongoing conflict prevents consistent NCD surveillance. Undertaking a second STEPS survey is considered critical to persuading decision makers of the magnitude of the impact of NCDs in Libya.
## Commercial determinants of health
There were several recurring cross-cutting issues for policies that target commercial determinants. These include legal and illegal lobbying efforts, national and transnational legal challenges, and the industry-sponsored dissemination messaging to undermine public support for new NCD policies [51–59]. Suggested actions include strategic framing of policy issues [53, 60], strengthening legal capacity to identifying ways to minimize potential ‘practical’ trade concerns [54, 55], and developing clear strategies and codes of conduct to manage engagement with the private sector [61, 62].
Tobacco. The introduction of tobacco legislation has been heavily shaped by political, historical, social, and economic contexts [63–67]. The tobacco industry often raises considerable barriers to legislative restrictions, for instance by arguing that restrictions will increase illicit trade, create problems for retailers, harm the economy (especially in tobacco-producing countries), and violate domestic laws and international treaties on intellectual property and investments [67–71]. The Framework Convention for Tobacco Control (FCTC)—to which *Libya is* a signatory—has been a key tool to catalyse the process of policy formulation, adoption, and implementation [64, 72–74]. Several studies found that governance models that ensure transparent management of conflicts of interest (in line with FCTC) [67, 75, 76] and adequate funding for implementation [68, 77, 78] have been associated with implementation of tobacco policies. Caribbean countries that proposed harmonization of excise taxes across the region (i.e., all countries levy a minimum specific excise tax at the same rate) provided a good example of tools that enable political processes to facilitate implementation of tax policy [79].
Diet. In many LMICs and fragile states experiencing epidemiological and nutritional transitions, policy planning is made challenging by the fact that overweight and malnutrition can coexist [80–84]. Another commonly reported challenge is the complexity of defining clear dietary and macronutrient targets and indicators, especially given uncertainty regarding ingredients and composition of food [54, 85–90]. A perceived lack of local evidence to support links between salt or fat and NCDs are described as detrimental to the policy course [54, 85–90]. Trade challenges at the World Trade Organization or from regional agreements can exert a regulatory chill that discourages countries from implementing NCD policies [51, 91–95]. Industry also commonly seeks to forestall legislation by pushing for self-regulation [53, 61, 81, 87, 96–103]. Where adopted, effective self-regulation requires strong government leadership, multisectoral stakeholder engagement, and independent monitoring and evaluation [81, 85, 90, 96–100, 103–110].
Alcohol. Reported barriers to implementation include incoherent policy messaging and failure to frame the problem in a way that engenders aligned political and social understanding of the problem. In this regard, the media plays an important role in shaping public attitudes [52, 92, 93]. Two high quality country case studies reported that delayed action can stem from alcohol not being perceived as a national priority [111, 112]. Countries ran into difficulties when specific directions were not provided about how all information should be presented on labels (size, font, position, wording etc) or where WHO recommendations were not followed [113]. We note that Libya has implemented all alcohol Best Buy policies.
Diet and alcohol. In contrast to tobacco, we found that diet and alcohol policies were similarly undermined by a lack of global accountability instruments [51, 91–95], a lack of globally agreed targets and evidence, and underdeveloped mechanisms for safe industry engagement [39, 41–43, 114].
Physical Activity (PA). Numerous social, cultural, and environmental barriers prevent engagement with PA, including fear of violence and crime in outdoor areas, air pollution, and cultural restrictions–all of which tend to disproportionately affect women and girls [115–118]. A WHO report identified that conflict, political instability, and epidemics can make it difficult to generate sufficient political attention for PA policies in many African countries [117]. Mass media campaigns are reported to be an underused tool, based on a high-quality case study from Nigeria [115]. Multisectoral collaboration was central to progressing PA policies in many settings, with successes seen when governments, NGOs, academia, transport, urban planners, and other stakeholders involved in PA promotion were brought together around shared policy goals [116, 118–120].
Interviewees disagreed over how much corporate influence was limiting progress. Almost all processed foods, tobacco, and alcohol is imported (illegally in the case of alcohol and most of the country’s tobacco) rather than produced domestically. Opportunities for regulating packaging, sales, and reformulation are perceived as minimal. One interviewee felt that health standards were not being applied consistently to imported goods. Overall, there has been little opportunity for transnational corporate interests to undermine policy because of the lack of population-level policies being proposed.
## Primary care guidelines and therapeutics
Fragmented health systems with a mix of private and public health provisions tend to generate a complex environment for development of national plans, particularly in fragile settings [121]. Additionally, many LMICs report competing needs from multiple disease burdens [122, 123]. Human resource capacity influences implementation: countries must train primary health care professionals and provide them with supportive materials and tools throughout the implementation period [124]. Availability of medicines and diagnostics is another major concern; this is due to limitations in procurement systems and budget allocations for medicines and blood pressure devices in some programs [50, 125]. Some population groups prefer informal providers, which adds another layer of complexity and fragmentation in health-seeking pathways [126]. Evidence on the use of traditional medicine and locally driven policies that are evidence based, cost effective, and culturally sensitive should be developed [121, 127]. Two qualitative studies conducted in Africa highlighted that the use of technology, community health workers, and task-sharing provide important frameworks to improve CVD care [128, 129].
While PHC guidelines are being developed for diabetes, cardiovascular disease, hypertension, obesity, nutritional guidelines and mental health, there is little regulation across private providers. The fragmented primary health care system makes consistent implementation difficult across facilities. Conflict, political instability, and the overall lack of coordinated governance have hampered health service delivery and provision of recommended cardiovascular therapies.
## Cross-cutting issues
Interviewees told us that implementation of the missing Best Buy policies will only be possible with backing of political leadership at the highest level. Until new elections are held, senior policymakers are reluctant to pass major health reforms as they feel they do not have a democratic mandate. All interviewees agreed that NCD prevention and control is not a priority on the national political agenda. Persistent conflict, political instability, and the recent COVID-19 pandemic have crowded out political attention. Given the fragile security and political context, the government has limited ability to conduct any major policy reforms. Virtually all health-related policy action has focused on providing day-to-day clinical services. Coordinated responses to NCD prevention and control across the country have been challenging due to the dual governments and fragmented civil and clinical services. Overall, the security situation prevents formal engagement with national policy-making processes, impedes coordination with other sectors, and forces attention toward downstream biomedical NCD actions.
## Tobacco
While the WHO and MoH have issued a number of decrees regarding tobacco control, none have been implemented. The relevance of tobacco control to Libyan context is somewhat disputed, with some interviewees asserting that importation and domestic supply is illegal when this is not the case. This can lead to the perception is that there is no need for tobacco-related action, despite evidence from the 2009 STEPS survey that one-quarter of the adult population smoke, and an increasing number of young people are consuming tobacco. Much of the tobacco is sold in *Libya is* illegal, and sold by small street vendors in the informal economy. This makes it hard to regulate.
Responsibility for tobacco control lies at the parliamentary level, where there are multiple competing priorities. Enforcement is the responsibility of the Ministry for the Interior and the police. One interview felt that tobacco control is unlikely to be strongly enforced by police officers who smoke themselves.
The overarching sense is that the illegal tobacco market is too large and complex to address effectively with the resources currently at the government’s disposal. Furthermore, the WHO Best Buys and other policy tools are not fit for purpose as they generally do not tackle black market supply.
## Diet
Libya relies on food imports, with little domestic production. The FDCC applies standards that are developed by the National Center for Standardization and Metrology, however one senior policymaker stated there are no food labelling guidelines, targets or maximum thresholds for salt, fat, or sugar content in imported or domestically produced foods and beverages, which indicates that awareness of these guidelines is low. The FDCC has little capacity to develop or implement regulation of labelling guidelines or maximum thresholds for salt, fat or sugar content of foods.
There is also a paucity of country-level evidence on national diet risk factors such as salt, trans fats, and highly processed foods, which impedes action–regional or international-level evidence is perceived to be insufficient in persuading policy makers.
The government has previously considered implementing the International Breastmilk code, but progress has stalled. It appears that responsibility for child health promotion is unofficially devolved to the Ministry of Education. Each time children were mentioned during the interviews, participants spoke about working in schools alongside the Ministry of Education, using education budgetary resources.
There was little concern about whether delegating all child-related issues to the education sector might weaken action to restrict the marketing of junk food to children.
## Alcohol
Libya has fully implemented all Best Buy alcohol policies. Whilst our systematic review suggested that diet and alcohol share similar challenges in many settings, this is not the case in Libya–one of the few countries where alcohol is banned. Interviewees acknowledged that just because alcohol is banned does not mean that it is not consumed. However, WHO surveys conducted in 2008 and 2009 indicate that the bans are strictly enforced, domestic production and imports are negligible, and <$1\%$ of the population consumes alcohol. As such, alcohol is not perceived to be a legitimate risk factor of concern.
Among interviewees, there was a lack of knowledge or awareness of steps that could be taken to address informally produced alcohol.
## Physical activity
One interview participant identified Libya’s lack of sports infrastructure as an impediment to state-organised physical activity promotion.
Developing a mass media physical activity promotional campaign through the NCD office with support from aligned NGOs would require securing a new budgetary line, which appears to be extremely difficult. A draft plan is in the early stages of development.
## Synthesis and recommendations
We used a joint display (Table 2) to synthesise our findings, focusing on the policies that have not been implemented. Core cross-cutting themes included the strategic importance of strong governance—which equates to buy-in at the highest level of leadership for Libya, as well as adequate legislative and ministerial attention–as well as mechanisms and venues for cross-government multisectoral action that is insulated from industry interference, and dedicated funding for NCD policy development, implementation, and enforcement. The fragile security situation and fragmented health service delivery model represent major challenges for national surveillance efforts. This undermines the ‘data collection’ target, but also impacts other areas as policymakers feel that domestic data are required before dietary policies can be implemented. Health system fragmentation also makes it very difficult to provide consistent access to recommended cardiovascular therapies. A number of initiatives are in development, and there is a growing appreciation of the need for collaboration across government departments, however the policy stasis is unlikely to thaw until a leader with a clear electoral mandate accords NCDs the legislative priority they deserve.
**Table 2**
| Unnamed: 0 | Targets, data collection, and plans | Tobacco | Diet | Physical Activity | Primary Care guidelines and therapies |
| --- | --- | --- | --- | --- | --- |
| Findings from quantitative policy review | Findings from quantitative policy review | Findings from quantitative policy review | Findings from quantitative policy review | Findings from quantitative policy review | Findings from quantitative policy review |
| Policy gaps | • Regular risk factor surveys• NCD targets• Routine mortality data collection• Multisectoral action plan | • Taxation• Packaging graphic warnings• Mass media campaigns | • Salt reduction policies• Fat reduction policies• Child food marketing policies• Breast-milk substitute marketing policies | • Mass media campaigns | • Cardiovascular therapies |
| Findings from the systematic review | Findings from the systematic review | Findings from the systematic review | Findings from the systematic review | Findings from the systematic review | Findings from the systematic review |
| Facilitators for policy implementation | • Strong governance• Multisectoral action• Dedicated financing | • Dedicated financing• FCTC ratification• Clear governance of conflicts of interest | • Clear macronutrient & dietary targets• Strong governance and multisectoral engagement• Independent monitoring of voluntary industry reform/action | • Multisectoral collaboration and action | • Deployment of technology• Effective exploitation of community health workers• Task-sharing |
| Challenges for policy implementation | • Competing priorities• Conflict undermining surveillance systems | • Industry opposition• Legal challenges• Public messaging/framing | • Industry opposition• Industry self-regulation forestalling effective action• Double burden of disease• Insufficient local evidence on the burden of salt & fat | • Low priority issue in many settings• Conflict, political instability, and epidemics all draw attention away | • Mixed public/private/traditional providers• Weak health systems in fragile settings• Competing priorities (HIV/Covid)• Limited human resources• Inadequate diagnostics Logistical and stocking issues |
| Findings from the qualitative interviews | Findings from the qualitative interviews | Findings from the qualitative interviews | Findings from the qualitative interviews | Findings from the qualitative interviews | Findings from the qualitative interviews |
| Challenges for policy implementation | • Few plans have been developed; most are in drafting stage• Conflict and lack of funding prevents consistent NCD surveillance• No action can be taken without sign-off from the prime minister (highest level of political leadership) | • Disputed significance; (some interviewees thought tobacco import & sale was illegal• No official national tobacco strategy, although one is being drafted• WHO and MoH have issued decrees regarding tobacco control; none have been implemented• Counterfeit tobacco is cheap, plentiful, and sold in an informal, unregulated economy• Enforcement of tobacco control challenging• WHO Best Buys and other policy tools do not focus on the black market | • Reliance on imports perceived as an issue; ‘it’s out of our control’• FDCC has low capacity to develop or implement dietary targets, labelling guidelines or maximum thresholds for salt, fat or sugar content of foods• Little country-level evidence on national diet risk factors to guide action• Responsibility for child health is effectively unofficially devolved to the Ministry of Education | • Paucity of sports infrastructure• New promotional campaigns require new budgetary lines which are difficult to secure | • PHC guidelines being developed for diabetes, cardiovascular disease, hypertension, obesity, nutritional guidelines, and mental health• Close working relationship between NCDC and primary health care division of the MoH• Fragmented primary health care system makes consistent implementation difficult across facilities• Little regulation or enforcement of private providers• Conflict, political instability, and lack of coordinated governance have hampered health service delivery and provision of medicines |
| Opportunities for policy implementation | • Second STEPS survey is considered critical to persuading decision makers of the magnitude of the impact of NCDs | • [None identified] | • Functioning FDCC with good links to the NCDC• Recognition that high-level leadership and multisectoral action is required | • Plan is in early stages of development | • Institute of Primary Health Care is taking responsibility for formalising treatment pathways |
## Discussion
In this explanatory sequential mixed-methods study we used quantitative policy review to identify Libya’s NCD policy gaps, systematic review of the global literature to identify lessons from other settings, and key stakeholder interviews to explore specific challenges and opportunities for progress.
Libya has implemented a quarter of the 19 recommended Best Buy policies, including all alcohol policies and two tobacco policies. In reality, these tobacco measures are not enforced, smoking rates are high, and the vast majority of tobacco comes from the black market. Libya’s policymakers face a unique constellation of challenges to and opportunities for introducing the NCD Best Buys. Negligible physical activity infrastructure, a world-leading illicit tobacco market, a fractured primary care system, and a decade of conflict are significant challenges. At the same time, Libya enjoys near total alcohol abstinence and reasonable political alignment regarding the need for effective NCD legislation. Perhaps the largest barrier is that new legislation requires leadership from the highest levels of government and a stable deliberative body with space on the legislative agenda. In the absence of these circumstances, policy makers have been forced to make do with downstream clinical interventions because they feel that any other form of implementation requires sign-off from an elected official.
Based on our findings, we have developed a set of policy recommendations for Libya. These fall into three groups according to feasibility. Firstly, simple, discrete actions that require very modest resource allocation: adding an NCD line-item to the national health budget to signal that NCDs will require dedicated resources; recognising and encouraging nascent multisectoral initiatives aimed at developing plans for physical activity via an official announcement, officially adopting the nine WHO-recommended NCD targets from the Global Monitoring Framework for NCDs (these can be updated/tailored at a later date); inviting WHO to unilaterally conduct a follow-up STEPS survey; and signing up to the International Code of Marketing of Breastmilk Substitutes. These six actions could be implemented relatively easily, particularly with the support of existing development partners in the country. They would protect lives, make good on pre-existing international commitments, and encourage encouraging policy work that is already underway. Libya’s Best Buy implementation score would almost double, placing it on par with the mean for upper-middle income countries. The main barrier to action seems to be the dependence on very high-level political buy-in. Many of the interviewees noted that progress is predicated on approval at the highest level of political leadership, however, a large number of competing priorities exist in the country, many of which appear more urgent than NCD risk factor control. A further proposed issue is that the current political leadership may not feel able to introduce health policy reforms until the new elections deliver a democratic mandate to act.
The second set of recommendations require dedicated policy development: developing Libya-specific NCD targets and obtaining high-level political endorsement; drafting plain packaging legislation for legal tobacco products; setting up a commission to devise a strategy to tackle illegal tobacco; raising the official tax rate on tobacco to $75\%$ of the total price; running mass media campaigns for a physical activity and tobacco; devising reduction and reformulation strategies for salt and fats; and developing a national policy to reduce child marketing of junk foods. The first four could be led by the NCDC. Indeed, work is already underway for many elements. The latter three (dietary) actions require ongoing collaboration between the FDCC and the NCDC, with greater involvement of the Ministry of Education. Libya already has the relevant institutional capacity, inter-sectoral relationships, and access to external expertise to deliver on all eight Best Buy domains covered above.
The third set of actions centre on structural reform. Harmonised and well-functioning national data collection systems, the reliable provision of therapeutics, and the development of a robust multisectoral action plan all require a stable security situation and political legitimacy. Our review and interviews highlight the importance of high-level political leadership for NCD policy implementation. It is clear that the role of the head of the *Government is* crucial for progressing nascent policy developments including to allocate hypothecated resources and engender multisectoral collaboration. The backdrop of conflict, weak governance, competing health issues (including COVID-19) and a fractured health system was found to be an important limitation in the global literature, and these issues were also felt to inhibit action in a range of areas by Libyan interviewees. These contextual issues were believed to draw legislative and fiscal attention away from NCDs as well as undermine national data collection, enforcement, and harmonised health service delivery of primary care guidelines and therapies. On one level this is understandable, however NCDs kill far more Libyans than conflict every year, and risk factor prevalence is alarmingly high. Sadly, there is little that health advocates can do to influence these macro determinants.
Our work falls under the banner of implementation research; concerned with understanding “what, why and how” interventions work for a given population in a particular setting [130, 131]. Implementation research has been defined as “the scientific study of the processes used to implement policies and interventions and the contextual factors that affect these processes” [132]. There have been increasing calls to use IR approaches to understand how the WHO Best Buy policies can be taken to scale in low and middle income countries [131, 133–135].
## Strengths and limitations
A major strength of this study is the use of mixed-methods to gain a holistic understanding of the challenges Libya faces and the opportunities for policy progress. Our quantitative analysis followed the approach previously used by WHO and Allen et al. [ 29, 135, 136]. The findings are limited by the quality of the underlying Libyan Country Capacity surveys, and not all of the recommended Best Buy policies are relevant, for instance they do not capture measures to tackle illegal tobacco.
Our review was conducted according to Cochrane guidance and the PRISMA checklist. We had to relegate many details to S1 Text, however we used a robust and reproducible approach. We only used dual review for a fifth of titles & abstracts, however we erred on the side of overinclusion and used dual review for every other step.
We used purposive sampling for our interviews and were able to access all of the senior policymakers that we approached. Again, we used a robust approach for conducting, analysing and interpreting the interviews. Ideally, we would have spoken to the higher-level political leadership (such as the health minister, finance minister, and prime minister) as they are so fundamental to the policy implementation process, but this was not thought to be feasible.
An important an overarching limitation of this work is the shifting political landscape. Whilst we deliberately spoke to policymakers who were in established positions that had survived multiple regime changes, it is possible that the institutional and legislative landscape will shift in a way that nullifies some of our recommendations. A further limitation is the narrow focus of the Best Buy policies. We sought to strike a balance between focusing on the cost-effective solutions espoused by WHO, and other NCD policies that may be more pertinent for Libya. In terms of our positionality, our research team is comprised of a mix of insiders and outsiders which provides a decent epistemological balance. The research and project leaders have clinical, policy, and research experience.
A final critical issue that our study raises is the challenge of obtaining local ethical review in fragile and conflict affected settings for short term projects that often have to deliver on very tight deadlines. We note that virtually all global health programmes run by WHO, the World Bank, and other international partner organisations operate without any form of independent ethical review. The only incentive seems to be the option of publishing findings in the peer-reviewed literature, however in an ideal world, rapid, routine processes should be in place for every project that potentially exposes participants to risk. We were unable to obtain local approval and so sought the advice of the Oxford Tropical Research Ethics Committee. They were able to provide a degree of external scrutiny. We would argue that there is a potential role for WHO to run a rapid ethics review service via a central secretariat and country representatives attached to every WHO county office. This would enable a greater proportion of non-research projects to get independent feedback on their proposed approach.
## Conclusion
NCDs are the leading cause of death and disability in Libya, however they have not been accorded adequate political attention, in part because years of conflict and constantly changing leadership inhibits action. If/when stable and legitimate national leadership emerges, it is imperative that the Government signal their blessing for greater action on NCDs. This would allow nascent collaborations and policy plans to come to fruition. Failing that, there are still a large number of simple actions that could be approved and implemented with negligible effort. NCDs may not be as attention-grabbing as armed conflict or Covid-19, but NCDs levy an even greater burden than the two combined. What is more, the government has a range of effective tools just waiting to be used.
## References
1. Malkin J, Rakic S. *Mortality and Morbidity Due to NCDs and Transport Injuries*
2. 2Institute for Health Metrics and Evaluation. Viz Hub. 2020 [cited 19 Aug 2021]. Available: http://vizhub.healthdata.org/gbd-compare
3. 3GBD Results Tool: GHDx. [cited 27 Aug 2021]. Available: http://ghdx.healthdata.org/gbd-results-tool
4. 4WHO and Libyan Ministry of Health. NCD STEPS Survey: Libya. 2009. Available: https://www.who.int/ncds/surveillance/steps/Libya_2009_STEPS_FactSheet.pdf
5. **STEPwise Approach to NCD Risk Factor Surveillance (STEPS).**
6. Lemamsha H, Randhawa G, Papadopoulos C. **Prevalence of Overweight and Obesity among Libyan Men and Women**. *Biomed Res Int* (2019.0) **2019** 8531360. DOI: 10.1155/2019/8531360
7. 7WHO. Global Health Observatory: UHC index of service coverage (Universal health coverage). In: WHO [Internet]. World Health Organization; [cited 12 Feb 2021]. Available: https://apps.who.int/gho/data/node.imr.UHC_AVAILABILITY_SCORE?lang=en
8. 8World Health Organization. Tackling NCDs: “best buys” and other recommended interventions for the prevention and control of noncommunicable diseases. Updated (2017) appendix 3 of the global action plan for the prevention and control of noncommunicable diseases 2013–2020. Endorsed at the 70th World Health Assembly, 2017. Geneva: World Health Organization; 2017. Report No.: WHO/NMH/NVI/17.9. Available: https://apps.who.int/iris/handle/10665/259232. *Endorsed at the 70th World Health Assembly, 2017* (2017.0)
9. Curry L, Nunez-Smith M. *Mixed Methods in Health Sciences Research: A Practical Primer* (2015.0). DOI: 10.4135/9781483390659
10. Cresswell J.. *A Concise Introduction to Mixed Methods Research* (2015.0)
11. **Noncommunicable Diseases Progress Monitor 2015**. (2015.0)
12. **Noncommunicable Diseases Progress Monitor 2017**. (2017.0)
13. **Noncommunicable Diseases Progress Monitor 2020**. (2020.0)
14. Moher D, Liberati A, Tetzlaff J, Altman DG. **Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement.**. *PLOS Medicine* (2009.0) **6** e1000097. DOI: 10.1371/journal.pmed.1000097
15. **“Best Buys” and other recommended intervention for the prevention and control of Noncommunicable diseases**. (2011.0)
16. Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ. *Cochrane Handbook for Systematic Reviews of Interventions.* (2019.0)
17. Reich MR. **Political economy analysis for health**. *Bull World Health Organ* (2019.0) **97** 514-514. DOI: 10.2471/BLT.19.238311
18. Shearer JC, Abelson J, Kouyaté B, Lavis JN, Walt G. **Why do policies change? Institutions, interests, ideas and networks in three cases of policy reform**. *Health Policy and Planning* (2016.0) **31** 1200-1211. DOI: 10.1093/heapol/czw052
19. Lewin S, Booth A, Glenton C, Munthe-Kaas H, Rashidian A, Wainwright M. **Applying GRADE-CERQual to qualitative evidence synthesis findings: introduction to the series**. *Implementation Science* (2018.0) **13** 2. DOI: 10.1186/s13012-017-0688-3
20. 20World bank. Classification of Fragile and Conflict-Affected Situations. 2021.. *Classification of Fragile and Conflict-Affected Situations* (2021.0)
21. Kak M, Rakic S, Shikhy A. *Noncommunicable Diseases in Libya*
22. Ritchie J, Spencer L. *Bryman A and Burgess R (Eds) Analysis of Quantitative Data* (1994.0) 173-194
23. Bazeley P.. **Integrative Analysis Strategies for Mixed Data Sources**. *American Behavioral Scientist* (2012.0) **56** 814-828. DOI: 10.1177/0002764211426330
24. Guetterman TC, Fetters MD, Creswell JW. **Integrating Quantitative and Qualitative Results in Health Science Mixed Methods Research Through Joint Displays.**. *Ann Fam Med* (2015.0) **13** 554-561. DOI: 10.1370/afm.1865
25. Denzin NK. (1978.0)
26. Patton MQ. **Enhancing the quality and credibility of qualitative analysis**. *Health Serv Res* (1999.0) **34** 1189-1208. PMID: 10591279
27. Denscombe M.. **Communities of Practice: A Research Paradigm for the Mixed Methods Approach**. *Journal of Mixed Methods Research* (2008.0) **2** 270-283. DOI: 10.1177/1558689808316807
28. Johnson RB, Onwuegbuzie AJ. **Mixed Methods Research: A Research Paradigm Whose Time Has Come.**. *Educational Researcher* (2004.0) **33** 14-26. DOI: 10.3102/0013189X033007014
29. Allen LN, Wigley S, Holmer H. **Implementation of non-communicable disease policies from 2015 to 2020: a geopolitical analysis of 194 countries**. *The Lancet Global Health* (2021.0) **9** e1528-e1538. DOI: 10.1016/S2214-109X(21)00359-4
30. Allen LN, Haring R, Kickbusch I, Ganten D, Moeti M. *Handbook of Global Health* (2020.0) 1-37. DOI: 10.1007/978-3-030-05325-3_57–1
31. Kickbusch I, Allen L, Franz C. **The commercial determinants of health**. *The Lancet Global Health* (2016.0) **4** e895-e896. DOI: 10.1016/S2214-109X(16)30217-0
32. Biswas T, Pervin S, Tanim MdIA, Niessen L, Islam A. **Bangladesh policy on prevention and control of non-communicable diseases: a policy analysis**. *BMC Public Health* (2017.0) **17** 582. DOI: 10.1186/s12889-017-4494-2
33. Essue BM, Kapiriri L. **The unfunded priorities: an evaluation of priority setting for noncommunicable disease control in Uganda.**. *Globalization and Health* (2018.0) **14** 22. DOI: 10.1186/s12992-018-0324-2
34. Juma PA, Mohamed SF, Wisdom J, Kyobutungi C, Oti S. **Analysis of Non-communicable disease prevention policies in five Sub-Saharan African countries: Study protocol**. *Arch Public Health* (2016.0) **74** 25-25. DOI: 10.1186/s13690-016-0137-9
35. Sithey G, Li M, Thow AM. **Strengthening non-communicable disease policy with lessons from Bhutan: linking gross national happiness and health policy action**. *Journal of Public Health Policy* (2018.0) **39** 327-342. DOI: 10.1057/s41271-018-0135-y
36. 36WHO. Regional Office for the Western Pacific. 2017. Progress on the Prevention and Control of Noncommunicable Diseases in the Western Pacific Region: Country Capacity Survey. Regional Office for the Western Pacific, World Health Organization, Manila. http://iris.wpro.who.int/handle/10665.1/14162.. *Progress on the Prevention and Control of Noncommunicable Diseases in the Western Pacific Region: Country Capacity Survey*
37. Nasiri T, Yazdani S, Shams L, Takian A. **Stewardship of noncommunicable diseases in Iran: a qualitative study**. *International Journal of Health Governance* (2021.0) **26** 179-198. DOI: 10.1108/IJHG-07-2020-0074
38. 38WHO. Regional Office for the Eastern Mediterranean. 2020. Noncommunicable Diseases in the Eastern Mediterranean Region.
Regional Office for the Eastern Mediterranean, World Health Organization, Cairo, Egypt. https://apps.who.int/iris/handle/10665/250371.. *Noncommunicable Diseases in the Eastern Mediterranean Region.* (2020.0)
39. Lin V, Carter B, McQueen DV. *Global Handbook on Noncommunicable Diseases and Health Promotion* (2013.0) 189-201. DOI: 10.1007/978-1-4614-7594-1_12
40. Lin Vivian, Jones Catherine, Wang Shiyong, Baris Enis. (2014.0)
41. Magnusson RS, Patterson D. **The role of law and governance reform in the global response to non-communicable diseases**. *Globalization and Health* (2014.0) **10** 44. DOI: 10.1186/1744-8603-10-44
42. Manson H, Sullivan T, Ha P, Navarro C, Martín-Moreno JM. **Goals are Not Enough: Building Public Sector Capacity for Chronic Disease Prevention**. *Public Health Reviews* (2013.0) **35** 11. DOI: 10.1007/BF03391696
43. Mendis S, Chestnov O. **Policy reform to realize the commitments of the Political Declaration on noncommunicable diseases**. *British Medical Bulletin* (2013.0) **105** 7-27. DOI: 10.1093/bmb/ldt001
44. Casey M, Hayes PS, Heaney D, Dowie L, Ólaighin G, Matero M. **Implementing transnational telemedicine solutions: a connected health project in rural and remote areas of six Northern Periphery countries Series on European collaborative projects**. *Eur J Gen Pract* (2013.0) **19** 52-58. DOI: 10.3109/13814788.2012.761440
45. 45Regional Office for South-East Asia. 2020. National Capacity for Prevention and Control of Non-Communicable Diseases in WHO SEAR—Results from NCD Country Capacity Survey 2019. Regional Office for South-East Asia, World Health Organization, New Delhi, India. Available: https://apps.who.int/iris/handle/10665/334223. *National Capacity for Prevention and Control of Non-Communicable Diseases in WHO SEAR—Results from NCD Country Capacity Survey 2019* (2020.0)
46. Chimeddamba O, Peeters A, Walls HL, Joyce C. **Noncommunicable Disease Prevention and Control in Mongolia: A Policy Analysis**. *BMC Public Health.* (2015.0) **15** 660. DOI: 10.1186/s12889-015-2040-7
47. Dodd R, Reeve E, Sparks E, George A, Vivili P, Win Tin ST. **The politics of food in the Pacific: coherence and tension in regional policies on nutrition, the food environment and non-communicable diseases**. *Public Health Nutrition.* (2020.0) **23** 168-180. DOI: 10.1017/S1368980019002118
48. Tuangratananon T, Wangmo S, Widanapathirana N, Pongutta S, Viriyathorn S, Patcharanarumol W. **Implementation of national action plans on noncommunicable diseases, Bhutan, Cambodia, Indonesia, Philippines, Sri Lanka, Thailand and Viet Nam**. *Bull World Health Organ* (2019.0) **97** 129-141. DOI: 10.2471/BLT.18.220483
49. **Implementing a National Non-Communicable Disease Policy in Sub-Saharan Africa: Experiences of Key Stakeholders in Ghana**. *Health Policy Open* **1** 100009. DOI: 10.1016/j.hpopen.2020.100009
50. Nicole Fraser-Hurt. *Care for Hypertension and Other Chronic Conditions in Samoa: Understanding the Bottlenecks and Closing the Implementation Gaps.* (2020.0)
51. Lawhon M, Herrick C. **Control in the News: The Politics of Media Representations of Alcohol Policy in South Africa.**. *Alcohol* (2013.0) **38** 987-1021. DOI: 10.1215/03616878-2334683
52. Matanje Mwagomba BL, Nkhata MJ, Baldacchino A, Wisdom J, Ngwira B. **Alcohol policies in Malawi: inclusion of WHO “best buy” interventions and use of multi-sectoral action**. *BMC Public Health* (2018.0) **18** 957. DOI: 10.1186/s12889-018-5833-7
53. Thow AM, Jones A, Hawkes C, Ali I, Labonté R. **Nutrition labelling is a trade policy issue: lessons from an analysis of specific trade concerns at the World Trade Organization**. *Health Promotion International* (2018.0) **33** 561-571. DOI: 10.1093/heapro/daw109
54. Thow AM, Snowdon W, Labonté R, Gleeson D, Stuckler D, Hattersley L. **Will the next generation of preferential trade and investment agreements undermine prevention of noncommunicable diseases? A prospective policy analysis of the Trans Pacific Partnership Agreement.**. *Health Policy* (2015.0) **119** 88-96. DOI: 10.1016/j.healthpol.2014.08.002
55. von Tigerstrom B.. **How Do International Trade Obligations Affect Policy Options for Obesity Prevention? Lessons from Recent Developments in Trade and Tobacco Control.**. *Canadian Journal of Diabetes* (2013.0) **37** 182-188. DOI: 10.1016/j.jcjd.2013.03.402
56. Barlow P, Labonte R, McKee M, Stuckler D. **Trade challenges at the World Trade Organization to national noncommunicable disease prevention policies: A thematic document analysis of trade and health policy space**. *PLOS Medicine* (2018.0) **15** e1002590. DOI: 10.1371/journal.pmed.1002590
57. Greenhalgh S.. **Soda industry influence on obesity science and policy**. *China. Journal of Public Health Policy* (2019.0) **40** 5-16. DOI: 10.1057/s41271-018-00158-x
58. Myers A, Fig D, Tugendhaft A, Myers JE, Hofman KJ. **The history of the South African sugar industry illuminates deeply rooted obstacles for sugar reduction anti-obesity interventions**. *null* (2017.0) **76** 475-490. DOI: 10.1080/00020184.2017.1311515
59. Jarman H.. **Attack on Australia: Tobacco industry challenges to plain packaging**. *Journal of Public Health Policy* (2013.0) **34** 375-387. DOI: 10.1057/jphp.2013.18
60. Colón-Ramos U, Monge-Rojas R, Campos H. **Impact of WHO recommendations to eliminate industrial trans-fatty acids from the food supply in Latin America and the Caribbean.**. *Health Policy and Planning* (2014.0) **29** 529-541. DOI: 10.1093/heapol/czt034
61. Juma PA, Mohamed SF, Matanje Mwagomba BL, Ndinda C, Mapa-tassou C, Oluwasanu M. **Non-communicable disease prevention policy process in five African countries authors**. *BMC Public Health* (2018.0) **18** 961. DOI: 10.1186/s12889-018-5825-7
62. 62WHO Regional Office for Western Pacific. Regional Meeting on National Multisectoral Plans for NCD Prevention and Control. Manila: WHO Regional Office for the Western Pacific. 2012. Available: http://iris.wpro.who.int/handle/10665.1/12558. *Regional Meeting on National Multisectoral Plans for NCD Prevention and Control* (2012.0)
63. Labonté R, Lencucha R, Goma F, Zulu R, Drope J. **Consequences of policy incoherence: how Zambia’s post-FCTC investment policy stimulated tobacco production**. *J Public Health Policy* (2019.0) **40** 286-291. DOI: 10.1057/s41271-019-00171-8
64. Mbulo L, Ogbonna N, Olarewaju I, Musa E, Salandy S, Ramanandraibe N. **Preventing tobacco epidemic in LMICs with low tobacco use—Using Nigeria GATS to review WHO MPOWER tobacco indicators and prevention strategies**. *Preventive Medicine* (2016.0) **91** S9-S15. DOI: 10.1016/j.ypmed.2016.04.005
65. Cussen A, McCool J. **Tobacco Promotion in the Pacific: The Current State of Tobacco Promotion Bans and Options for Accelerating Progress.**. *Asia Pac J Public Health* (2011.0) **23** 70-78. DOI: 10.1177/1010539510390925
66. Poorolajal Jalal, Mohammadi Younes, Mahmoodi Azam. **Challenges of Tobacco Control Program in Iran**. *Archives of Iranian Medicine* (2017.0) **20** 229-234. PMID: 28412827
67. Yang G, Wang Y, Wu Y, Yang J, Wan X. **The road to effective tobacco control in China**. *The Lancet* (2015.0) **385** 1019-1028. DOI: 10.1016/S0140-6736(15)60174-X
68. Crosbie E, Thomson G, Freeman B, Bialous S. **Advancing progressive health policy to reduce NCDs amidst international commercial opposition: Tobacco standardised packaging in Australia.**. *null* (2018.0) **13** 1753-1766. DOI: 10.1080/17441692.2018.1443485
69. Cavalcanti Erica. **The Decision-Making Process in Brazil’s Ratification of the World Health Organization Framework Convention on Tobacco Control**. *Cadernos de Saude Publica* (2017.0) **33** e00126115. DOI: 10.1590/0102-311X00126115
70. Satterlund TD, Cassady D, Treiber J, Lemp C. **Barriers to adopting and implementing local-level tobacco control policies**. *J Community Health* (2011.0) **36** 616-623. DOI: 10.1007/s10900-010-9350-6
71. Beaglehole R, Bonita R, Yach D, Mackay J, Reddy KS. **A tobacco-free world: a call to action to phase out the sale of tobacco products by 2040**. *The Lancet* (2015.0) **385** 1011-1018. DOI: 10.1016/S0140-6736(15)60133-7
72. Mapa-Tassou C, Bonono CR, Assah F, Wisdom J, Juma PA, Katte J-C. **Two decades of tobacco use prevention and control policies in Cameroon: results from the analysis of non-communicable disease prevention policies in Africa.**. *BMC Public Health* (2018.0) **18** 958. DOI: 10.1186/s12889-018-5828-4
73. Oladepo O, Oluwasanu M, Abiona O. **Analysis of tobacco control policies in Nigeria: historical development and application of multi-sectoral action**. *BMC Public Health* (2018.0) **18** 959. DOI: 10.1186/s12889-018-5831-9
74. 74World Health Organization and WHO Framework Convention on Tobacco Control. 2020. WHO FCTC Implementation Review in Pacific Island Countries. Geneva: World Health Organization. https://apps.who.int/iris/handle/10665/337194.. *WHO FCTC Implementation Review in Pacific Island Countries*
75. Charoenca N, Mock J, Kungskulniti N, Preechawong S, Kojetin N, Hamann SL. **Success Counteracting Tobacco Company Interference in Thailand: An Example of FCTC Implementation for Low- and Middle-income Countries.**. *International Journal of Environmental Research and Public Health* (2012.0) 9. DOI: 10.3390/ijerph9041111
76. Gilmore AB, Fooks G, Drope J, Bialous SA, Jackson RR. **Exposing and addressing tobacco industry conduct in low-income and middle-income countries**. *The Lancet* (2015.0) **385** 1029-1043. DOI: 10.1016/S0140-6736(15)60312-9
77. **Bridging the Gap Between Science and Public Health: Taking Advantage of Tobacco Control Experience in Brazil to Inform Policies to Counter Risk Factors for Non-Communicable Diseases.”**. *Addiction* (2013.0) **108** 1360-1366. DOI: 10.1111/add.12203
78. Lencucha R, Thow AM. **How Neoliberalism Is Shaping the Supply of Unhealthy Commodities and What This Means for NCD Prevention.**. *International Journal of Health Policy and Management* (2019.0) **8** 514-520. DOI: 10.15171/ijhpm.2019.56
79. 79World Bank Group. Advancing Action on the Implementation of Tobacco Tax Harmonization in the Organization of Eastern Caribbean States Countries. World Bank, Washington, DC. https://openknowledge.worldbank.org/handle/10986/30034. 2018.. (2018.0)
80. Thow AM, Sanders D, Drury E, Puoane T, Chowdhury SN, Tsolekile L. **Regional trade and the nutrition transition: opportunities to strengthen NCD prevention policy in the Southern African Development Community**. *null* (2015.0) **8** 28338. DOI: 10.3402/gha.v8.28338
81. Gupta P, Mohan S, Johnson C, Garg V, Thout SR, Shivashankar R. **Stakeholders’ perceptions regarding a salt reduction strategy for India: Findings from qualitative research**. *PLOS ONE* (2018.0) **13** e0201707. DOI: 10.1371/journal.pone.0201707
82. Rafieifar S. **Strategies and Opportunities Ahead to Reduce Salt Intake**. *Archives of Iranian Medicine* (2016.0) **19** 729-734. PMID: 27743439
83. Edalati S. **Development and Implementation of Nutrition Labelling in Iran: A Retrospective Policy Analysis.**. *International Journal of Health Planning and Management* (2020.0) **35** e28-e44. DOI: 10.1002/hpm.2924
84. Carriedo A, Koon AD, Encarnación LM, Lee K, Smith R, Walls H. **The political economy of sugar-sweetened beverage taxation in Latin America: lessons from Mexico, Chile and Colombia.**. *Globalization and Health* (2021.0) **17** 5. DOI: 10.1186/s12992-020-00656-2
85. Reeve E, Thow AM, Bell C, Engelhardt K, Gamolo-Naliponguit EC, Go JJ. **Implementation lessons for school food policies and marketing restrictions in the Philippines: a qualitative policy analysis**. *Globalization and Health* (2018.0) **14** 8. DOI: 10.1186/s12992-017-0320-y
86. Waqa G, Moodie M, Snowdon W, Latu C, Coriakula J, Allender S. **Exploring the dynamics of food-related policymaking processes and evidence use in Fiji using systems thinking**. *Health Research Policy and Systems* (2017.0) **15** 74. DOI: 10.1186/s12961-017-0240-6
87. Rutkow L, Jones-Smith J, Walters HJ, O’Hara M, Bleich SN. **Factors that encourage and discourage policy-making to prevent childhood obesity: Experience in the United States**. *Journal of Public Health Policy* (2016.0) **37** 514-527. DOI: 10.1057/s41271-016-0035-y
88. Studlar D, Cairney P. **Multilevel governance, public health and the regulation of food: is tobacco control policy a model**. *Journal of Public Health Policy* (2019.0) **40** 147-165. DOI: 10.1057/s41271-019-00165-6
89. Laar A, Barnes A, Aryeetey R, Tandoh A, Bash K, Mensah K. **Implementation of healthy food environment policies to prevent nutrition-related non-communicable diseases in Ghana: National experts’ assessment of government action**. *Food Policy* (2020.0) **93** 101907. DOI: 10.1016/j.foodpol.2020.101907
90. Wanjohi MN, Thow AM, Abdool Karim S, Asiki G, Erzse A, Mohamed SF. **Nutrition-related non-communicable disease and sugar-sweetened beverage policies: a landscape analysis in Kenya.**. *null* (2021.0) **14** 1902659. DOI: 10.1080/16549716.2021.1902659
91. Angus C, Holmes J, Meier PS. **Comparing alcohol taxation throughout the European Union**. *Addiction* (2019.0) **114** 1489-1494. DOI: 10.1111/add.14631
92. Keatley DA, Hardcastle SJ, Carragher N, Chikritzhs TN, Daube M, Lonsdale A. **Attitudes and beliefs towards alcohol minimum pricing**. *Western Australia. Health Promotion International* (2018.0) **33** 400-409. DOI: 10.1093/heapro/daw092
93. Li J, Lovatt M, Eadie D, Dobbie F, Meier P, Holmes J. **Public attitudes towards alcohol control policies in Scotland and England: Results from a mixed-methods study.**. *Social Science & Medicine* (2017.0) **177** 177-189. DOI: 10.1016/j.socscimed.2017.01.037
94. Limaye RJ, Rutkow L, Rimal RN, Jernigan DH. **Informal alcohol in Malawi: Stakeholder perceptions and policy recommendations**. *Journal of Public Health Policy* (2014.0) **35** 119-131. DOI: 10.1057/jphp.2013.43
95. Park K.. **New law on prevention and control of alcohol related harms in Vietnam**. *J Glob Health Sci* (2019.0) 1. DOI: 10.35500/jghs.2019.1.e49
96. Gesser-Edelsburg A.. **Nutrition Labelling and the Choices Logo in Israel: Positions and Perceptions of Leading Health Policy Makers. Journal of Human Nutrition and Dietetics**. *The Official Journal of the British Dietetic Association* (2014.0) **27** 58-68. DOI: 10.1111/jhn.12050
97. Mialon M, Mialon J, Calixto Andrade G, Jean-Claude M. **We must have a sufficient level of profitability’: food industry submissions to the French parliamentary inquiry on industrial food**. *null* (2020.0) **30** 457-467. DOI: 10.1080/09581596.2019.1606418
98. Mialon M, Swinburn B, Wate J, Tukana I, Sacks G. **Analysis of the corporate political activity of major food industry actors in Fiji.**. *Globalization and Health* (2016.0) **12** 18. DOI: 10.1186/s12992-016-0158-8
99. Pérez-Escamilla R.. **Prevention of Childhood Obesity and Food Policies in Latin America: From Research to Practice.**. *Obesity Reviews: An Official Journal of the International Association for the Study of Obesity* (2017.0) **18** 28-38. DOI: 10.1111/obr.12574
100. Phillips T, Ravuvu A, McMichael C, Thow AM, Browne J, Waqa G. **Nutrition policy-making in Fiji: working in and around neoliberalisation in the Global South**. *null* (2021.0) **31** 316-326. DOI: 10.1080/09581596.2019.1680805
101. Phulkerd S, Vandevijvere S, Lawrence M, Tangcharoensathien V, Sacks G. **Level of implementation of best practice policies for creating healthy food environments: assessment by state and non-state actors in Thailand.**. *Public Health Nutrition.* (2017.0) **20** 381-390. DOI: 10.1017/S1368980016002391
102. He FJ, Brinsden HC, MacGregor GA. **Salt reduction in the United Kingdom: a successful experiment in public health**. *Journal of Human Hypertension* (2014.0) **28** 345-352. DOI: 10.1038/jhh.2013.105
103. Thow AM, Abdool Karim S, Mukanu MM, Ahaibwe G, Wanjohi M, Gaogane L. **The political economy of sugar-sweetened beverage taxation: an analysis from seven countries in sub-Saharan Africa**. *Glob Health Action* (2021.0) **14** 1909267-1909267. DOI: 10.1080/16549716.2021.1909267
104. Breda J, Castro LSN, Whiting S, Williams J, Jewell J, Engesveen K. **Towards better nutrition in Europe: Evaluating progress and defining future directions**. *Food Policy* (2020.0) **96** 101887. DOI: 10.1016/j.foodpol.2020.101887
105. Magnusson R, Reeve B. **Food Reformulation, Responsive Regulation, and “Regulatory Scaffolding”: Strengthening Performance of Salt Reduction Programs in Australia and the United Kingdom**. *Nutrients* (2015.0) **7** 5281-5308. DOI: 10.3390/nu7075221
106. Reeve B, Magnusson R. **Food reformulation and the (neo)-liberal state: new strategies for strengthening voluntary salt reduction programs in the UK and USA**. *Public Health* (2015.0) **129** 351-363. DOI: 10.1016/j.puhe.2015.01.007
107. Trieu K, Webster J, Jan S, Hope S, Naseri T, Ieremia M. **Process evaluation of Samoa’s national salt reduction strategy (MASIMA): what interventions can be successfully replicated in lower-income countries?**. *Implementation Science* (2018.0) **13** 107. DOI: 10.1186/s13012-018-0802-1
108. Charlton K, Webster J, Kowal P. **To legislate or not to legislate? A comparison of the UK and South African approaches to the development and implementation of salt reduction programs**. *Nutrients* (2014.0) **6** 3672-3695. DOI: 10.3390/nu6093672
109. Onagan FCC, Ho BLC, Chua KKT. **Development of a sweetened beverage tax, Philippines**. *Bull World Health Organ* (2019.0) **97** 154-159. DOI: 10.2471/BLT.18.220459
110. Gostin LO, Abou-Taleb H, Roache SA, Alwan A. **Legal priorities for prevention of non-communicable diseases: innovations from WHO’s Eastern Mediterranean region**. *Public Health* (2017.0) **144** 4-12. DOI: 10.1016/j.puhe.2016.11.001
111. Abiona O, Oluwasanu M, Oladepo O. **Analysis of alcohol policy in Nigeria: multi-sectoral action and the integration of the WHO “best-buy” interventions**. *BMC Public Health* (2019.0) **19** 810. DOI: 10.1186/s12889-019-7139-9
112. Wallace K, Roberts B. **An Exploration of the Alcohol Policy Environment in Post-Conflict Countries.**. *Alcohol and Alcoholism* (2014.0) **49** 356-362. DOI: 10.1093/alcalc/agt142
113. Stockwell T, Giesbrecht N, Vallance K, Wettlaufer A. **Government Options to Reduce the Impact of Alcohol on Human Health: Obstacles to Effective Policy Implementation.**. *Nutrients* (2021.0) 13. DOI: 10.3390/nu13082846
114. Magnusson RS, McGrady B, Gostin L, Patterson D, Abou Taleb H. **Legal capacities required for prevention and control of noncommunicable diseases**. *Bull World Health Organ* (2019.0) **97** 108-117. DOI: 10.2471/BLT.18.213777
115. Oluwasanu M, Oladunni O, Oladepo O. **Multisectoral approach and WHO ‘Bestbuys’ in Nigeria’s nutrition and physical activity policies**. *Health Promotion International* (2020.0) **35** 1383-1393. DOI: 10.1093/heapro/daaa009
116. Bull F, Milton K, Kahlmeier S, Arlotti A, Juričan AB, Belander O. **Turning the tide: national policy approaches to increasing physical activity in seven European countries**. *Br J Sports Med* (2015.0) **49** 749. DOI: 10.1136/bjsports-2013-093200
117. 117Regional Committee for Africa, 70. 2020. Framework for the Implementation of the Global Action Plan on Physical Activity 2018–2030 in the WHO African Region: Report of the Secretariat. Regional Office for Africa, World Health Organization, Brazzaville, Republic of Congo. https://apps.who.int/iris/handle/10665/333737.. *Framework for the Implementation of the Global Action Plan on Physical Activity 2018–2030 in the WHO African Region: Report of the Secretariat*
118. Klepac Pogrmilovic B, Ramirez Varela A, Pratt M, Milton K, Bauman A, Biddle SJH. **National physical activity and sedentary behaviour policies in 76 countries: availability, comprehensiveness, implementation, and effectiveness.**. *International Journal of Behavioral Nutrition and Physical Activity* (2020.0) **17** 116. DOI: 10.1186/s12966-020-01022-6
119. Madhuvanti M.. **Evaluating Policy Responses to Noncommunicable Diseases in Seven Caribbean Countries: Challenges to Addressing Unhealthy Diets and Physical Inactivity.**. *Pan American Journal of Public Health* **42** e174. DOI: 10.26633/RPSP.2018.174
120. Politis CE, Mowat DL, Keen D. **Pathways to policy: Lessons learned in multisectoral collaboration for physical activity and built environment policy development from the Coalitions Linking Action and Science for Prevention (CLASP) initiative**. *Can J Public Health* (2017.0) **108** e192-e198. DOI: 10.17269/cjph.108.5758
121. **Gaps and Challenges to Integrating Diabetes Care in Myanmar.”**. *WHO South-East Asia Journal of Public Health* (2020.0) **5** 48-52. DOI: 10.4103/2224-3151.206553
122. 122World Health Organization. Countdown to 2023: WHO Report on Global Trans-Fat Elimination 2019. Washington, DC: World Health Organization. https://apps.who.int/iris/handle/10665/331300. 2019.. *Countdown to 2023: WHO Report on Global Trans-Fat Elimination 2019* (2019.0)
123. 123World Health Organization. Improving Hypertension Control in 3 Million People: Country Experiences of Programme Development and Implementation. World Health Organization, Geneva. https://apps.who.int/iris/handle/10665/336019. 2020.. (2020.0)
124. Collins D, Laatikainen T, Farrington J. **Implementing essential interventions for cardiovascular disease risk management in primary healthcare: lessons from Eastern Europe and Central Asia**. *BMJ Global Health* (2020.0) **5** e002111. DOI: 10.1136/bmjgh-2019-002111
125. Witter S, Zou G, Diaconu K, Senesi RGB, Idriss A, Walley J. **Opportunities and challenges for delivering non-communicable disease management and services in fragile and post-conflict settings: perceptions of policy-makers and health providers in Sierra Leone.**. *Conflict and Health* (2020.0) **14** 3. DOI: 10.1186/s13031-019-0248-3
126. 126Technical Package for Cardiovascular Disease Management in Primary Health Care: Healthy-Lifestyle Counselling. World Health Organization, Geneva. https://apps.who.int/iris/handle/10665/260422.
127. Moore MA. **Cancer Control Programs in East Asia: Evidence From the International Literature.**. *J Prev Med Public Health* (2014.0) **47** 183-200. DOI: 10.3961/jpmph.2014.47.4.183
128. Laar AK, Adler AJ, Kotoh AM, Legido-Quigley H, Lange IL, Perel P. **Health system challenges to hypertension and related non-communicable diseases prevention and treatment: perspectives from Ghanaian stakeholders**. *BMC Health Services Research* (2019.0) **19** 693. DOI: 10.1186/s12913-019-4571-6
129. Ndejjo R, Wanyenze RK, Nuwaha F, Bastiaens H, Musinguzi G. **Barriers and facilitators of implementation of a community cardiovascular disease prevention programme in Mukono and Buikwe districts in Uganda using the Consolidated Framework for Implementation Research**. *Implementation Science* (2020.0) **15** 106. DOI: 10.1186/s13012-020-01065-0
130. Peters DH, Adam T, Alonge O, Agyepong IA, Tran N. **Implementation research: what it is and how to do it**. *BMJ* (2013.0) **347** f6753. DOI: 10.1136/bmj.f6753
131. Collins T, Akselrod S, Berlina D, Allen LN. **Unleashing implementation research to accelerate national noncommunicable disease responses**. *Globalization and Health* (2022.0) **18** 6. DOI: 10.1186/s12992-021-00790-5
132. Peters D, Tran N, Adam T. (2021.0)
133. Isaranuwatchai W, Archer RA. (2019.0). DOI: 10.11647/OBP.0195
134. Marten R, Mikkelsen B, Shao R, Zennaro LD, Berdzuli N, Fernando T. **Committing to implementation research for health systems to manage and control non-communicable diseases**. *The Lancet Global Health* (2021.0) **9** e108-e109. DOI: 10.1016/S2214-109X(20)30485-X
135. 135WHO. Mid-point evaluation of the implementation of the WHO global action plan for the prevention and control of noncommunicable diseases 2013–2020 (NCD-GAP). 2020. Available: https://cdn.who.int/media/docs/default-source/documents/about-us/evaluation/ncd-gap-final-report.pdf?sfvrsn=55b22b89_22&download=true
136. Allen LN, Nicholson BD, Yeung BYT, Goiana-da-Silva F. **Implementation of non-communicable disease policies: a geopolitical analysis of 151 countries**. *The Lancet Global Health* (2020.0) **8** e50-e58. DOI: 10.1016/S2214-109X(19)30446-2
|
---
title: 'Barriers to effective hypertension management in rural Bihar, India: A cross-sectional,
linked supply- and demand-side study'
authors:
- Michael A. Peters
- Olakunle Alonge
- Anbrasi Edward
- Yvonne Commodore-Mensah
- Japneet Kaur
- Navneet Kumar
- Krishna D. Rao
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021531
doi: 10.1371/journal.pgph.0000513
license: CC BY 4.0
---
# Barriers to effective hypertension management in rural Bihar, India: A cross-sectional, linked supply- and demand-side study
## Abstract
Effective management of hypertension in low- and middle-income settings is a persistent public health challenge. This study examined supply- and demand-side barriers to receiving quality care and achieving effective hypertension management in rural Bihar, India. A state-representative household survey collected information from adults over 30 years of age on characteristics of the hypertension screening, diagnosis, and management services they received. A linked provider assessment determined the percent of providers who provided quality hypertension care (i.e., had a functioning BP measurement device, measured a patient’s BP, could correctly diagnose hypertension, had at least one first-line antihypertension medication, and could prescribe correctly based on standard guidelines). Patients were linked with their provider to determine the quality-adjusted coverage of hypertension management and logistic regression analysis was conducted to determine characteristics associated with receiving quality care. A total of 14,386 patients and 390 providers were studied. Nearly a quarter ($22.5\%$) of adults had never had their BP measured before and $8.1\%$ of adults reported a previous hypertension diagnosis. Less than one third ($31.0\%$) of all interviewed providers demonstrated ability to provide quality hypertension care, and quality varied between provider types ($14.8\%$ of private homeopathic, $25.2\%$ of informal, $40.0\%$ of private modern medicine, and $60.0\%$ of public providers gave quality care). While $95.8\%$ of diagnosed individuals received some treatment, only $10.9\%$ of patients received care from quality local providers. Nearly $45\%$ of individuals with hypertension received care from non-local providers. Individuals from the general caste with comorbidities living in villages with more high-quality providers were most likely to receive quality care from a local provider. Whereas the coverage of services for individuals diagnosed with hypertension is high, the quality of these services is suboptimal for economically and socially vulnerable populations, which limits effective management and control of hypertension in rural Bihar. Efforts should be targeted towards providers to initiate quality treatment upon diagnosis, including correct prescription of antihypertensives.
## Introduction
Non-communicable diseases (NCDs) are responsible for two-thirds of all mortality in low- and middle-income countries (LMICs), including nearly 18 million deaths from cardiovascular disease (CVD) annually [1]. Hypertension is the leading CVD risk factor, and improving hypertension management in primary care is key for achieving universal health coverage [1–4]. With the rising burden of hypertension and other NCDs, health systems in LMICs are challenged to adapt care models to address these lifelong conditions, while also maintaining services to reduce the impact of infectious diseases and maternal and childhood conditions [5]. To date, health systems in LMICs have largely been unable to provide health services at sufficient coverage or quality to prevent deaths from NCDs [6]. Each year, 2.4 million preventable CVD deaths are caused by poor-quality health services, more than five times the CVD deaths attributed to the lack of access to health services [6]. Research is needed to inform ways of improving the quality of health services in LMICs, especially services for preventing and managing NCDs, such as hypertension-related CVD. Efforts to identify gaps in the provision of quality hypertension management along the care continuum (from prevention, to screening and diagnosis, to management) can inform the design of health systems and programs that better respond to NCDs.
One way of assessing disease-specific health system functioning is to separate performance in a stepwise fashion along the continuum of care into a care cascade. The traditional hypertension cascade of care has steps for the prevalence of hypertension, the percent of those people who are aware of their diagnosis, the percent who are receiving treatment, and finally, those who have controlled blood pressure (BP). This hypertension care cascade framework has been used to describe BP control for decades and demonstrates that the largest barriers to hypertension management in LMICs are the diagnosis of individuals with elevated BP, and the ability to achieve non-elevated BP among those who have initiated treatment [7–10]. Most research on the hypertension care cascade relies exclusively on household surveys and population estimates, and does not provide insight into the quality of care received, therefore quality-adjusted coverage (or effective coverage) of hypertension services is largely unknown.
Without adjusting for a measure of quality, service coverage is only weakly linked to the health benefits experienced by a population [11]. There are several methods for calculating quality-adjusted coverage, including linking the content of care with population estimates of coverage, yet few studies have applied this approach to studying hypertension management [12]. A recent literature review found 12 studies which examined processes of quality care in population-based studies of hypertension management and only 8 studies which attempted to describe effective coverage of hypertension management in LMICs [13]. Estimates of quality-adjusted coverage of hypertension management services have immediate policy relevance for contexts experiencing epidemiological transition, like Bihar, India, which must prepare for an increasing burden of hypertension.
Bihar is facing a double burden of persistently high prevalence of infectious diseases and rapidly increasing incidence of chronic illnesses [14]. Hypertension is the leading CVD risk factor for death in the state, and was responsible for $6.4\%$ of disability-adjusted life years lost in 2016, more than a two-fold increase from $2.8\%$ in 1990 [14]. While studies have examined the quality of care provided by India’s pluralistic health systems, little is known about how effective public and private primary care providers are in the context of diagnosing and treating hypertension [15, 16]. The National Programme for prevention and control of cancer, diabetes, cardiovascular diseases and stroke (NPCDCS) is the main program to address NCDs across India, and a recent review highlighted the need to strengthen public health facilities to provide screening, early diagnosis, and treatment [17]. In Bihar, the coverage of screening and treatment services is low: less than half ($46.8\%$) of hypertensive individuals were recently screened and aware of their high blood pressure status, and only $17.3\%$ were receiving treatment [18]. On the other hand, the quality of hypertension management services in *Bihar is* largely unknown, especially in rural areas where informal providers (IPs) are more prevalent. This is significant because perceived quality of care is a major driver of provider choice, or switch, among Indian adults with hypertension [19]. This often leads to patients bypassing local care options to receive services that may be far away, but are more acceptable [20]. Understanding who is bypassing care is essential for making health systems more people-centered, equitable, and efficient.
Quantifying supply- and demand-side factors that prevent effective hypertension management can inform future policy for improving BP control in primary health care settings in Bihar and other LMICs. This study describes population-level (demand-side) deficits along the care cascade in Bihar, and (supply-side) gaps in the quality of hypertension care across provider types in India’s pluralistic health system. It estimates the quality-adjusted coverage of hypertension management services in Bihar and determines factors associated with receiving high quality care from local providers and with bypassing local care options. Taken together, this study helps to advance the development of high-quality health systems for addressing hypertension and NCDs in LMICs.
## Study context
With a population of 104 million people at the 2011 Census, *Bihar is* India’s third most populous state [21]. The health system is characterized by unequal access to health care, insufficient human resources and institutional capacity, poor quality care, and high out-of-pocket health expenditure: the state has the second highest ratio of private to public spending on health care in India [22]. This is partly caused by Bihar’s pluralistic primary health care system, in which nearly $75\%$ of care is provided by the private sector. The private sector includes (i) providers practicing modern medicine with a Bachelor of Medicine, Bachelor of Surgery, or MBBS degree, (ii) providers trained in Indian systems of medicine with a degree in Ayurveda (Bachelor of Ayurvedic Medicine and Surgery), Yoga and Naturopathy (Bachelor of Yoga), Unani (Bachelor of Unani Medicine and Surgery), Siddha (Bachelor of Siddha Medicine and Surgery), and Homeopathy (Bachelor of Homeopathic Medicine and Surgery), collectively known as AYUSH providers, and (iii) informally trained providers (IPs) [23]. IPs are often trusted members from the community in which they serve, with limited health training or experience, charge nominal fees (generally 100 Rupees, or $1.37 USD per consultation), and are an important source of primary care services in rural Bihar and other parts of India, though their role has evolved differently in various market settings [24]. A recent nationwide study showed that in Bihar, $3.9\%$ of villages had access to a public MBBS provider, $7.6\%$ had access to any MBBS provider (public or private), and $96.2\%$ of villages had access to any provider, including AYUSH or informal providers [16].
The public sector infrastructure includes about 1,800 Primary Health Centers (PHC), which represents about $60\%$ of the PHCs required to serve Bihar’s population based national government estimates for infrastructure allocation [22]. PHCs are staffed by MBBS or AYUSH doctors, but weaknesses in the public sector caused by staff shortages and a lack of basic infrastructure and supplies further reinforce private health care-seeking behavior. Among people who seek care for chronic conditions in Bihar, $86\%$ receive care from the private sector ($39\%$ from private MBBS and AYUSH providers, $30\%$ from IPs, and $17\%$ from pharmacists and other private sources), leaving only $14\%$ who receive care from the public sector [25].
## Study design
This cross-sectional study used primary data collected from a household survey and a provider quality assessment under the Assessment of Bihar’s Primary Health Care System (ABPHC) study. The study’s sample size was designed to detect a difference in the proportion of patients who visit PHCs with and without competent clinicians, and required a total of 9,798 households, accounting for design effect, the prevalence of illness and care-seeking, and average household size (S1 Annex). The ABPHC household survey employed a multi-stage sampling design, with rural PHCs as the primary sampling unit, villages within the PHC catchment area as the secondary sampling unit, and households within the village as the tertiary sampling unit. Rural PHCs were randomly selected from a census using stratified sampling proportional to the number of PHCs in Bihar’s nine divisions. Villages were randomly sampled by probability proportional to population size from a census of villages within each selected PHC’s catchment area, and 30 households were randomly sampled from each selected village using a complete listing of households. The total envisioned sample size was 10,500 households. All consenting members of selected households were included in the study and were administered a standard questionnaire in the local language by trained enumerators. The questionnaire included sections on demographic information, illness, and care-seeking history and experiences. Individuals aged 30 and older were asked about care-seeking related to specific chronic diseases (hypertension, diabetes, chronic heart disease, asthma, and chronic obstructive pulmonary disease). Data was collected between November 2019 and March 2020. Responses from the household survey informed sampling for the provider assessments.
Local public and private care providers (including MBBS, AYUSH, and IPs) visited by household survey respondents from 70 randomly selected villages (1 from each PHC catchment area) were located and included in the supply-side provider quality assessment if they were within five kilometers of the village. The study sample size was calculated to detect differences in the quality of public and private providers, and required 67 providers in each group (S1 Annex). The provider assessment was administered by enumerators with nursing degrees and consisted of three parts: a facility readiness assessment, responses to four clinical vignettes, and direct patient observations. The facility readiness assessment was modified from the validated Demographic Health Survey Service Provision Assessment survey, and was designed to understand the medicines, equipment, and human resources available to providers as compared to Bihar’s Essential Medicines List [26, 27]. In the clinical vignettes, hypothetical patients and scenarios were described to providers to assess knowledge on how to treat certain conditions (hypertension diagnosis, child diarrhea, child pneumonia, and angina). Providers were prompted to ask questions about the patient’s history, list the tests they would conduct on the patient, make a diagnosis, describe the advice they would give the patient, and if necessary, write a prescription for the patient. To understand provider practice, nurse enumerators observed patient-provider interactions and recorded provider actions using a standardized form. Enumerators stayed with the provider for three hours, observed up to five new patient consultations, and collected information on the same categories used in the clinical vignettes (i.e., patient history, tests conducted, diagnosis, advice given, and prescription). Data collection for the provider assessment started in January 2020, was interrupted in March due to the SARS-CoV-2 (COVID-19) outbreak, and then completed between February and March 2021. Due to the continuing spread of COVID-19, when data collection resumed in 2021, the patient observation component of the provider assessment was eliminated from the study to ensure safety of enumerators and patients.
All tools were pilot tested in non-sampled villages in rural Bihar to ensure that questions captured intended constructs. Wherever possible, the household and provider assessment tools used validated questions from surveys implemented in the same context, including the National Family Health Survey, the National Sample Survey, and World Bank Health Care Provider Surveys [15, 28, 29]. The hypertension vignette was designed in consultation with cardiologists and pilot tested with local primary care providers in Bihar for psychometric validation. Data for the household survey and provider assessment was collected on tablets using SurveySolutions, a free Computer Assisted Personal Interviewing (CAPI) software to reduce information bias [30]. Enumerators had previous experience in household survey data collection and participated in a week-long training to familiarize themselves with the tool, the CAPI platform, and to participate in large scale field tests. Direct supervision and random response reliability checks were conducted in the field to ensure data quality and reduce the risk of measurement bias. All data was passed through automatic logic checks and implausible responses were discussed and resolved between supervisors and enumerators.
Data from the household survey and the provider assessment was linked through a direct matching process. Adults with hypertension were matched with the providers from whom they received care. Individuals were not linked to their providers if they (i) sought care more than five kilometers away from their village (because providers were not eligible for the provider assessment), (ii) sought care from medical shops (because these providers were unable to complete the provider assessment during pretesting), or (iii) sought care from a provider within five kilometers of the village but the provider was not located by the study team.
The ABPHC study is approved for human subjects research by the Johns Hopkins University Institutional Review Board (IRB00009563) and by the Sigma Institutional Review Board in India (Doc#’s 1910787106 / 1910789424). Written consent to participate was obtained from all study participants.
## Study measures
Since the ABPHC study did not measure respondent BP, a traditional care cascade with a common denominator of all people with high BP could not be created [31]. Instead, population-based estimates of hypertension care cascade steps were calculated using a previously defined methodology [18]. Respondents were considered to have been “screened” if they responded “yes” to the question, “Have you ever had your blood pressure measured by a doctor or other health worker?”. Respondents were “aware” if they responded “yes” to the question, “have you been told by a doctor or other health worker that you have high blood pressure or hypertension?”. Participants were “treated” if they responded “yes” to the question, “Have you taken any medications or other treatment for high blood pressure during the last 12 months?”. Additional supply-side measures included the source of care for hypertension management, reasons for choosing a provider, self-reported distance traveled to seek hypertension care, monthly amount spent on hypertension care, annual number of visits to a provider for hypertension management, and hypertension-related hospitalizations.
Measures of provider quality were described chronologically in terms of their impact on a patient’s ability to progress through the hypertension care cascade, from screened to aware to treated and controlled BP. The definition of controlled blood pressure varies based on context and clinical guidelines. For the purposes of this study, hypertension is defined as systolic blood pressure of 140 mm Hg or more and/or diastolic blood pressure of 90 mm Hg or more and/or taking antihypertensive medication after a previous diagnosis, in accordance with International Society of Hypertension guidelines and the Fourth Indian Guidelines on Hypertension. For example, a provider’s ability to appropriately diagnose hypertension based on a high blood pressure reading contributes to a patient’s ability to progress from “screened” to “aware”. A cascade of provider ability to manage a patient with hypertension was described based on measures from hypertension vignettes and facility assessments (S2 Annex). Providers who made it to the final cascade step (had a functioning BP measurement device, would check a hypothetical patient’s BP at least once, made a correct hypertension diagnosis, initiated medical treatment upon diagnosis of stage two hypertension, had first line anti-hypertensive medication available, and wrote an appropriate prescription) are considered to deliver high quality hypertension care. The appropriateness of prescriptions was assessed by three clinicians who have practiced primary care medicine in India and other middle-income settings. Two clinicians reviewed each prescription and rated their quality on a scale including (i) inappropriate and harmful, (ii) inappropriate but not harmful, and (iii) appropriate for safely lowering BP using a standardized protocol (S3 Annex). Disagreements in ratings were discussed as a team and resolved by the third clinician. Additionally, the percent of adults over 30 who were screened by a health worker in patient observations is presented to demonstrate provider actions in practice.
In the linked supply- and demand-side analysis, we described the effective coverage of hypertension management services, defined by the proportion of hypertension-aware respondents who received high-quality care. We examined how the probability of receiving quality care and bypassing varied by the following characteristics: age, sex, caste, wealth index, education, the presence of a comorbidity, and the number of high-quality providers of hypertension care in a village. Hypertension-aware patients were considered to have bypassed local options if they reported traveling more than one kilometer farther than the local PHC to receive hypertension care. Age was grouped into 10-year categories starting at age 30, until all individuals over age 70. Caste was grouped by self-reported membership of head of household according to the Government of India’s categorization scheme: scheduled castes (SC), scheduled tribes (ST), other backward castes (OBC) and general. Household wealth quintile was calculated using principal components analysis on ten household assets following a standard methodology [32]. Education was categorized into three groups: individuals with no schooling completed, individuals with some schooling completed but less than secondary high schooling, and secondary high schooling completed and above. Individuals were determined to have a comorbidity if they responded that they had been told by a health worker that they had diabetes, asthma, chronic heart disease, or cancer in addition to hypertension.
## Statistical analysis
Normalized weights were calculated for estimates from the household survey to account for probability of selection through the ABPHC study’s design and to account for non-response rates. All descriptive statistics for population estimates from the ABPHC household survey incorporated these weights, while estimates from the provider surveys were unweighted due to the lack of a complete sampling frame of private providers. To account for the variability in private provider qualifications, supply-side results from the private sector are presented disaggregated by provider training (i.e., MBBS, AYUSH, and IP).
Logistic regression analyses were employed to determine predictors of the probability of receiving quality care in the village and the probability of bypassing local care options. Individuals with missing data are excluded from analysis and sample sizes for each regression are discussed further in S4 Annex. For each regression analysis, collinearity of explanatory variables was assessed, a likelihood ratio test was conducted for each variable, effect modification was considered and tested, and the final model was compared with stepwise model selections. Hosmer-Lemeshow tests for goodness of fit were conducted and regression diagnostics were conducted to assess final model fit. Data analysis were performed in StataSE version 14.1 (College Station Texas) [33].
## Sample characteristics
In total, 39,486 individuals were included in the household survey, including 14,386 individuals who were 30 years or older and answered the chronic disease module and 950 individuals who reported a previous hypertension diagnosis (Table 1).
**Table 1**
| Unnamed: 0 | All respondents (N = 39,486) | Respondents over 30 (N = 14,386) | Respondents aware of hypertension diagnosis (N = 950) |
| --- | --- | --- | --- |
| Sex | Sex | Sex | Sex |
| Female | 20,767 (52.6%) | 7,742 (53.8%) | 551 (58.0%) |
| Male | 18,719 (47.4%) | 6,644 (46.1%) | 339 (42.0%) |
| Missing | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) |
| Age | Age | Age | Age |
| Average (SD) | 25.8 (20.2) | 48.6 (13.9) | 58.0 (13.4) |
| Religion | Religion | Religion | Religion |
| Hindu | 35,541 (90.0%) | 13,066 (90.8%) | 839 (88.3%) |
| Muslim | 3,909 (9.9%) | 1,307 (9.1%) | 110 (11.6%) |
| Other (Christian, Buddhist Jain) | 26 (0.1%) | 8 (0.1%) | 0 (0.0%) |
| Missing | 10 (<0.1%) | 5 (<0.1%) | 1 (<0.1%) |
| Caste | Caste | Caste | Caste |
| Scheduled Caste | 10,252 (26.0%) | 3,453 (24.0%) | 139 (14.7%) |
| Scheduled Tribe | 544 (1.4%) | 180 (1.3%) | 10 (1.0%) |
| Other Backward Caste | 24,901 (63.1%) | 9,094 (63.2%) | 596 (62.8%) |
| General | 3,779 (9.6%) | 1.654 (11.5%) | 204 (21.5%) |
| Missing | 12 (<0.1%) | 12 (<0.1%) | 0 (0.0%) |
| Highest level of Education Achieved | Highest level of Education Achieved | Highest level of Education Achieved | Highest level of Education Achieved |
| No schooling | 13,115 (33.2%) | 8,072 (56.1%) | 504 (53.1%%) |
| Some schooling | 18,708 (47.4%) | 5,034 (35.0%) | 349 (36.7%) |
| Higher secondary schooling or above | 4,039 (10.2%) | 1,279 (8.9%) | 97 (10.2%) |
| Missing + | 3,624 (9.2%) | 1 (<0.1%) | 0 (0.0%) |
| Occupation | Occupation | Occupation | Occupation |
| Not employed | 30,225 (76.6%) | 7,350 (51.1%) | 609 (64.1%) |
| Agriculture | 3,111 (7.9%) | 2,727 (18.9%) | 176 (18.5%) |
| Labor | 3,346 (8.5%) | 2,282 (15.9%) | 45 (4.7%) |
| Self-employed | 1,920 (4.9%) | 1,453 (10.1%) | 81 (8.5%) |
| Service/salaried | 875 (2.2%) | 572 (4.0%) | 39 (4.1%) |
| Missing | 9 (<0.1%) | 2 (<0.1%) | 0 (0.0%) |
The provider assessment included 390 providers from 70 villages, 368 of whom ($94.4\%$) said they would treat the patient in the hypertension vignette (Table 2). In total, our study administered the provider assessment to $83.0\%$ of the local providers who provided care to individuals in the household survey. This compares favorably with a previous study in India with a similar study design, which reported a response rate of $41.9\%$ among local providers [16].
**Table 2**
| Unnamed: 0 | Private Providers (N = 319) | Public Providers (N = 71) |
| --- | --- | --- |
| Sex of Provider | Sex of Provider | Sex of Provider |
| Male | 311 (97.5%) | 62 (87.3%) |
| Female | 7 (2.2%) | 9 (12.7%) |
| Missing | 1 (0.3%) | 0 (0.0%) |
| Age of Provider | Age of Provider | Age of Provider |
| Average (SD) | 45 (14) | 48 (11) |
| Caste of Provider | Caste of Provider | Caste of Provider |
| Scheduled Caste | 31 (9.7%) | 7 (9.9%) |
| Scheduled Tribe | 11 (3.5%) | 4 (5.6%) |
| Other Backwards Caste | 165 (51.7%) | 31 (43.7%) |
| General | 110 (34.5%) | 28 (39.4%) |
| Other | 2 (0.6%) | 1 (1.4%) |
| Place of Residence | Place of Residence | Place of Residence |
| Same Town/Village as clinic | 238 (74.6%) | 47 (66.2%) |
| Other town/village | 81 (25.4%) | 24 (33.8%) |
| Provider Training | Provider Training | Provider Training |
| MBBS | 15 (4.7%) | 47 (66.2%) |
| AYUSH | 54 (16.9%) | 18 (25.4%) |
| Informal training | 250 (78.4%) | 0 (0.0%) |
| Other modern medicine* | 0 (0.0%) | 6 (8.5%) |
| Years worked at facility | Years worked at facility | Years worked at facility |
| Average (SD) | 15 (13) | 6 (6) |
## Population measures of hypertension management
Of the 14,386 individuals in the survey aged 30 or older, $77.5\%$ had ever had their BP measured by a health worker before the survey date ($95\%$ CI: $76.3\%$ to $78.6\%$). Among adults who had ever had their blood pressure measured before, $8.1\%$ reported a previous hypertension diagnosis. The prevalence of a previous hypertension diagnosis increased with age with the highest prevalence ($15.2\%$) among those 70 and older. The majority ($95.7\%$) of the hypertension-aware population was treated. Most treated individuals ($94.4\%$) availed services from the private sector ($46.4\%$ from private doctors and clinics, $18.9\%$ from pharmacies or compounders, $14.5\%$ from private hospitals, $14.3\%$ from traditional healers, and $0.3\%$ from other private providers). Among the $5.6\%$ of individuals who sought care from the public sector, the majority ($58.8\%$) sought care from district hospitals. The perceived quality of hypertension management services was an important consideration for individuals in their care-seeking decisions. Of the individuals who sought hypertension management services from the private sector, $62.2\%$ did not seek hypertension treatment from their local PHC due to the low perceived quality of care. This contributed to people traveling an average of 29.2 kilometers ($95\%$ CI: 20.9 to 37.5 kilometers) to seek hypertension management services. On average, people who sought care outside of the home to manage their hypertension made 4.8 visits per year to a provider. The median monthly amount spent on hypertension management was 250 rupees (US$3.35). The median monthly spending is slightly higher for those who receive treatment from private providers than public providers (255 vs 200 rupees). Among respondents who were aware of their high BP, $9.5\%$ had ever been hospitalized due to their hypertension. This includes $4.6\%$ of patients with hypertension who were hospitalized in the past year for chest pain, cardiovascular problems, or hypertension.
## Provider measures of hypertension management
Among the 390 sampled primary care providers, 368 ($94.4\%$) treated hypertension, and 121 ($39.0\%$) could provide high-quality hypertension management. The largest system-wide gaps in the provision of high-quality hypertension management were the ability of providers to make a correct hypertension diagnosis, the availability of at least one first line antihypertensive medication, and the ability to write an appropriate prescription to safely lower BP (Fig 1). Across provider types, public providers were most frequently able to provide quality hypertension care ($62\%$), followed by private MBBS-trained providers ($40\%$), IPs ($25\%$), and private AYUSH providers ($15\%$) (Table 3). The largest gap was the lack of any first-line antihypertensive medication at AYUSH provider-run clinics. Among IPs, the largest single barrier to providing quality care was the ability to make a correct hypertension diagnosis. For both public and private MBBS providers, the largest gap in quality care was the ability to write an appropriate prescription to manage hypertension (drop of $15\%$ and $20\%$, respectively). Among providers who correctly diagnosed the hypothetical patient as hypertensive, $16.6\%$ of private MBBS doctors, $13.2\%$ of private AYUSH doctors, $10.6\%$ of IPs and, and $9.7\%$ of public providers wrote prescriptions that were assessed to cause harm to the patient. Prescriptions were characterized as harmful because they included inappropriate drugs (e.g., Ampicillin, Betnisole, or Paracetemol), excessive doses, and/or harmful drug interactions.
**Fig 1:** *Provider cascade of quality hypertension care.BP = blood pressure; HTN = hypertension.* TABLE_PLACEHOLDER:Table 3 Across patient observations, $30.1\%$ of patients over the age of 30 had their BP measured during a consultation. Patients presenting to private primary care providers were significantly more likely to have their BP measured than those presenting to public providers ($36.5\%$ versus $19.7\%$, $$p \leq 0.02$$). Among private providers, formally trained providers (MBBS providers checked BP in $70.0\%$ of visits, AYUSH providers checked BP in $42.9\%$ of visits) checked adult BP more frequently than IPs ($31.0\%$ of visits).
## Linked assessment of hypertension management
In the 70 villages where the provider assessment was conducted, 192 individuals self-reported a previous hypertension diagnosis. Of these 192 individuals, 85 sought care from providers outside the village (more than 5 kilometers away), 14 sought care from medical shops, and 19 sought care from local providers who were not assessed. Local providers were not included in the assessment because (iii) the provider was out of town for an extended period ($$n = 9$$), (ii) the provider was not located from information provided by the individual ($$n = 7$$), and (iii) the individual did not remember the name of their provider ($$n = 3$$). In total, the quality of care received is known for 74 individuals who received care from 56 local providers (S4 Annex). Of these individuals, $83.8\%$ sought care from IPs, $8.1\%$ sought care from private AYUSH doctors, $4.0\%$ sought care from private MBBS doctors, and $4.0\%$ sought care from public providers. By caste, $93.3\%$ of SC/ST individuals sought care from IPs compared with $80.0\%$ of OBC and $85.7\%$ of General caste individuals.
Over one-third of hypertension-aware patients ($34.9\%$) received care from a local provider who did not provide quality care, including $7.3\%$ who sought care from medical stores whose operators were unable to complete the provider assessment due to a lack of medical knowledge. Of the 192 adults with hypertension, $10.9\%$ received quality care from local providers, and another $44.3\%$ received care from non-local providers. The quality-adjusted coverage of hypertension management services in rural Bihar ranges from $10.9\%$ to $65.1\%$ if unlocated local and out of market providers all provide quality care.
Among the 73 individuals whose quality of care was known and were included in the regression analysis, having a comorbidity and the number of quality providers in a village were statistically significantly associated with a higher probability of receiving quality care from a local provider (Table 4). Additionally, people from the general caste were significantly more likely to receive quality care from a local provider than those from the SC caste. Among individuals in the villages where quality of care was known, the number of quality providers in a village was not associated with the probability of bypassing local care options. In other words, there was no association between the number of local high-quality providers and the probability of bypassing local care. Among all hypertensive individuals, those with a comorbidity were most likely to bypass local care options. Additionally, individuals in the OBC or general caste, wealthier individuals, and people with higher secondary schooling or above were significantly more likely to bypass local care options than those in the ST caste, the poorest wealth quintile, and with no education, respectively.
**Table 4**
| Unnamed: 0 | (1) | (2) | (3) |
| --- | --- | --- | --- |
| Outcome variable | Receiving local quality care | Bypassing local care in villages with quality assessment | Bypassing local care in all villages |
| Sex | | | |
| Female | Ref | Ref | Ref |
| Male | 0.168 | 0.682 | 0.987 |
| | [0.0195,1.447] | [0.313,1.484] | [0.712,1.368] |
| Age Group | | | |
| 30–39 | - | Ref | Ref |
| 40–49 | - | 0.189* | 0.883 |
| | | [0.0408,0.881] | [0.491,1.589] |
| 50–59 | - | 0.404 | 0.992 |
| | | [0.0926,1.767] | [0.561,1.757] |
| 60–69 | - | 0.502 | 1.289 |
| | | [0.120,2.100] | [0.750,2.214] |
| 70 and older | - | 0.715 | 1.238 |
| | | [0.164,3.123] | [0.695,2.203] |
| Caste | | | |
| SC | Ref | Ref | Ref |
| ST | NA (omitted) | 6.354 | 1.925 |
| | | [0.675,59.78] | [0.494,7.504] |
| OBC | 39.88 | 3.258* | 1.581* |
| | [0.840,1893.3] | [1.080,9.833] | [1.031,2.424] |
| General | 83.05* | 3.292 | 1.972** |
| | [1.077,6407.1] | [0.900,12.04] | [1.186,3.279] |
| Wealth Category | | | |
| Q1 (Poorest) | Ref | Ref | Ref |
| Q2 | 21.27 | 2.223 | 1.638* |
| | [0.761,594.7] | [0.673,7.343] | [1.028,2.610] |
| Q3 | 1.711 | 0.765 | 1.102 |
| | [0.157,18.69] | [0.261,2.240] | [0.694,1.750] |
| Q4 | 2.679 | 1.193 | 1.180 |
| | [0.137,52.24] | [0.388,3.672] | [0.729,1.911] |
| Q5 (Richest) | 1.855 | 1.157 | 1.860** |
| | [0.152,22.60] | [0.416,3.218] | [1.187,2.915] |
| Education | | | |
| No schooling | Ref | Ref | Ref |
| Some schooling | 0.919 | 2.572* | 1.393 |
| | [0.132,6.403] | [1.111,5.954] | [0.992,1.957] |
| Higher secondary schooling or above | 1 | 10.83** | 1.941* |
| Higher secondary schooling or above | [1,1] | [1.935,60.59] | [1.129,3.338] |
| Comorbidity | 16.08* | 2.000 | 2.358*** |
| | [1.249,207.2] | [0.953,4.197] | [1.722,3.231] |
| Number of quality hypertension providers in village | 8.128** | 0.933 | |
| Number of quality hypertension providers in village | [2.121,31.15] | [0.664,1.312] | |
| N | 73+ | 192 | 949 |
## Discussion
This study describes the landscape of hypertension management in Bihar, India using linked supply- and demand-side estimates of quality care and receipt of care. Our findings reinforce previous population-based surveys by suggesting that most individuals in India with high BP have either never been screened or are not aware of their condition, (i.e., they have not been diagnosed as hypertensive) (S5 Annex) [18, 28]. Nearly one-quarter of adults in rural Bihar have never had their BP measured and just over eight percent of the adult population reported a previous hypertension diagnosis, compared to previous studies which reported a population prevalence of $12.8\%$ in Bihar [18]. This screening gap within the health system is confirmed by patient observations, during which fewer than a third of adult patients had their BP measured by a provider. According to NPCDCS and other screening guidelines, all adults over the age of 30 are supposed to have their BP measured opportunistically and during each interaction with the public health system [34, 35]. Across provider types, the ability to make a correct hypertension diagnosis was the largest gap in the ability to provide quality hypertension management, which also contributes to the “awareness” gap in the population. Increasing the percent of individuals who are screened and aware is critical but is not sufficient to improve hypertension control in rural Bihar.
Most “aware” individuals received care, however the quality of this care was generally low. While $95.7\%$ of adults who were aware of their condition received some treatment, the quality-adjusted coverage of hypertension management services by local providers was as low as $10.9\%$. Nearly half of individuals with reported hypertension bypassed local providers, traveling between 20 and 40 kilometers to receive care. Richer, more educated individuals from OBC and general castes and those with comorbidities were most likely to bypass local providers. Public providers were most frequently able to provide quality care across local provider types, yet most individuals perceived the quality of care at PHCs to be poor. Taken together, these results suggest that while access to hypertension management services is generally good, the quality of these services is inadequate in rural Bihar, and people who have the means to travel great distances to receive care prefer to do so rather than receive the more convenient care available in their villages. In short, local health systems are not providing high-quality, equitable services to manage hypertension in rural Bihar.
Findings from this study are subject to some important limitations. First, the ABPHC household survey design was cross-sectional and did not include direct measurements of BP. This means that all estimates among hypertensive individuals are based on self-reported hypertension diagnosis, likely resulting in underestimates of the true prevalence of hypertension. Previous studies have demonstrated validity between self-reported and true hypertension status, suggesting that this is an acceptable method for data collection [36]. Second, the clinical vignettes were not measurements of what providers did in practice and only five total consultations to manage hypertension were directly observed due to COVID-19 related interruptions to the study implementation. Information on the percent of providers that would measure BP twice during the vignette and give lifestyle advice to patients in the vignette were collected but were not included in the final definition of quality care because too few providers followed these measures. Studies have demonstrated correlations between provider competence and provider actions–there is typically a gap between the two–however provider knowledge is generally regarded as the upper limit of what a provider can do, so our estimates of poor provider quality are likely to be conservative in nature, and the true quality is likely to be worse [37, 38]. Additionally, the study definition of delivering quality hypertension care may have influenced results, as AYUSH providers and IPs are not legally allowed to stock or prescribe medications. Given the importance of immediate antihypertensive therapy in improving health outcomes, we believe that the availability of antihypertensive medications is an appropriate indicator of structural quality of care [39]. Third, the study design was limited to assessments of providers utilized in local rural markets (within five kilometers of a sampled village). While most individuals with hypertension sought care from providers within five kilometers of their home ($55.7\%$), many individuals also traveled long distances to receive treatment. The linked regression analysis had a small sample size resulting in large point estimates and wide confidence intervals; however, we were still able to understand characteristics of individuals who bypassed local care options for hypertension management.
The primary strength of this study is that it reveals new characteristics of hypertension management through a linked supply- and demand-study design which adds important information about the quality of hypertension care received by populations. This study adds insights about the health systems bottlenecks that prevent people with hypertension from achieving optimal health outcomes. Namely, poor structural quality (lack of front-line antihypertensive medications) and insufficient provider knowledge of how to properly diagnose and treat patients, combined with a lack of provider action to opportunistically screen individuals are contributing to the large number of hypertensives that are either unscreened or undiagnosed in rural Bihar. Interestingly, this study suggests that once individuals are aware of their high BP status, the vast majority access care. However, the quality of care received by individuals in rural *Bihar is* generally quite low across local provider types, which contributes to reduced rates of effective BP management and excess burden on patients who feel the need to travel great distances to receive quality care. Previous studies in similar contexts have demonstrated that hypertension can be diagnosed and treated at the community level [40]. The capacity of providers in rural Bihar, including IPs, to measure BP, diagnose hypertension, and stock and prescribe appropriate treatments must be improved within the existing legal constraints. Further, the role of community health workers, such as community health officers and accredited social health activists can be expanded in Bihar to bring services to populations. Future studies should seek to understand reasons for the perception of poor quality of care at public providers, even though PHCs often provide the highest quality of care locally available to patients. It is promising that coverage of hypertension management appears to be high in Bihar, now policymakers must urgently implement measures to improve the quality of these services in Bihar and similar contexts [41].
## Conclusion
Presenting supply-side, demand-side, and linked information about hypertension management is effective for demonstrating systemic gaps in disease control and helps to prioritize intervention design. This study’s methodology should be considered for broader use in health systems research in other contexts and for other diseases. Findings suggest that while improving screening and diagnosis services would have the largest impact on retaining people in the care cascade, the quality of care provided in rural *Bihar is* low. With the rising burden of NCDs and hypertension in Bihar, increasing screening practices without first adequately equipping providers (of all types) would be futile to effectively manage the hypertension burden. Without first improving quality of care, patients newly linked to care are not likely to experience health gains and may even suffer from adverse effects from the large proportion of providers who are unable to effectively provide hypertension management services.
## References
1. 1World Health Organization. Global Health Estimates 2020: Disease burden by Cause, Age, Sex, by Country and by Region, 2000–2019. Geneva, Switzerland; 2020. Available: https://www.who.int/data/global-health-estimates. *Global Health Estimates 2020: Disease burden by Cause, Age, Sex, by Country and by Region, 2000–2019* (2020.0)
2. **Cardiovascular disease, chronic kidney disease, and diabetes mortality burden of cardiometabolic risk factors from 1980 to 2010: a comparative risk assessment**. *Lancet Diabetes Endocrinol* (2014.0) **2** 634-647. DOI: 10.1016/S2213-8587(14)70102-0
3. Frieden TR, Varghese C V, Kishore SP, Campbell NRC, Moran AE, Padwal R. **Scaling up effective treatment of hypertension—A pathfinder for universal health coverage**. *J Clin Hypertens* (2019.0) **21** 1442-1449. DOI: 10.1111/jch.13655
4. Varghese C, Nongkynrih B, Onakpoya I, McCall M, Barkley S, Collins TE. **Better health and wellbeing for billion more people: integrating non-communicable diseases in primary care**. *BMJ* (2019.0) **364** l327. DOI: 10.1136/bmj.l327
5. Beaglehole R, Epping-Jordan J, Patel V, Chopra M, Ebrahim S, Kidd M. **Improving the prevention and management of chronic disease in low-income and middle-income countries: a priority for primary health care**. *Lancet* (2008.0) **372** 940-9. DOI: 10.1016/S0140-6736(08)61404-X
6. Kruk ME, Gage AD, Joseph NT, Danaei G, García-Saisó S, Salomon JA. **Mortality due to low-quality health systems in the universal health coverage era: a systematic analysis of amenable deaths in 137 countries**. *Lancet* (2018.0) **392** 2203-2212. DOI: 10.1016/S0140-6736(18)31668-4
7. Cummings KM, Kirscht JP. *Prevalence, awareness, treatment, and control of hypertension in the inner city* (1982.0) **582** 571-582
8. **Hypertension prevalence and the status of awareness, treatment, and control in the United States**. *Hypertension* (1985.0) 7
9. Chadha SL, Radhakrishnan S, Ramachandran K, Kaul U, Gopinath N. **Prevalence, awareness & treatment status of hypertension in urban population of Delhi**. *Indian J Med Res* (1990.0) **92** 233-240. PMID: 2228067
10. Geldsetzer P, Manne-Goehler J, Marcus M-EME, Ebert C, Zhumadilov Z, Wesseh CS. **The state of hypertension care in 44 low-income and middle-income countries: a cross-sectional study of nationally representative individual-level data from 1·1 million adults**. *Lancet* (2019.0) **394** 652-662. DOI: 10.1016/S0140-6736(19)30955-9
11. Amouzou A, Leslie HH, Ram M, Fox M, Jiwani SS, Requejo J. **Advances in the measurement of coverage for RMNCH and nutrition: from contact to effective coverage**. *BMJ Glob Heal* (2019.0) **4** e001297. DOI: 10.1136/bmjgh-2018-001297
12. Ng M, Fullman N, Dieleman JL, Flaxman AD, Murray CJL, Lim SS. **Effective coverage: a metric for monitoring Universal Health Coverage**. *PLoS Med* (2014.0) **11** e1001730. DOI: 10.1371/journal.pmed.1001730
13. Peters MA, Noonan C, Edward A, Rao KD, Commodore-Mensah Y, Alonge OO. **The expanded hypertension care cascade: a scoping review to improve effective coverage of hypertension management services in low- and middle-income countries**. *BMC Health Serv Res*
14. 14Indian Council of Medical Research, Public Health Foundation of India, Institute for Health Metrics and Evaluation. India: Health of the Nation’s States—The India State-level Disease Burden Initiative. New Delhi, India; 2017.
15. Das J, Hammer J. **Which doctor? Combining vignettes and item response to measure clinical competence**. *J Dev Econ* (2005.0) **78** 348-383. DOI: 10.1016/j.jdeveco.2004.11.004
16. Das J, Daniels B, Ashok M, Shim E-Y, Muralidharan K. **Two Indias: The structure of primary health care markets in rural Indian villages with implications for policy**. *Soc Sci Med* (2020.0) 112799. DOI: 10.1016/j.socscimed.2020.112799
17. Thakur JS, Paika R, Singh S. **Burden of noncommunicable diseases and implementation challenges of National NCD Programmes in India**. *Med journal, Armed Forces India* (2020.0) **76** 261-267. DOI: 10.1016/j.mjafi.2020.03.002
18. Prenissl J, Manne-Goehler J, Jaacks LM, Prabhakaran D, Awasthi A, Bischops AC, Kruk ME. **Hypertension screening, awareness, treatment, and control in India: A nationally representative cross-sectional study among individuals aged 15 to 49 years**. *PLoS Med* (2019.0) **16** e1002801. DOI: 10.1371/journal.pmed.1002801
19. Kujawski SA, Leslie HH, Prabhakaran D, Singh K, Kruk ME. **Reasons for low utilisation of public facilities among households with hypertension: analysis of a population-based survey in India**. *BMJ Glob Heal* (2018.0) **3** e001002. DOI: 10.1136/bmjgh-2018-001002
20. Rao KD, Sheffel A. **Quality of clinical care and bypassing of primary health centers in India**. *Soc Sci Med* (2018.0) **207** 80-88. DOI: 10.1016/j.socscimed.2018.04.040
21. 21Office of the Registrar General & Census Commissioner. Census of India. 2011 [cited 3 Dec 2019]. Available: http://www.censusindia.gov.in/2011census/hh-series/hh01.html
22. 22Berman P, Bhawalkar M, Jha R. Tracking financial resource for Primary Health Care in Bihar. 2017. Available: https://cdn1.sph.harvard.edu/wp-content/uploads/sites/2031/2017/01/Tracking-financial-resources-for-primary-health-care-in-BIHAR-India.pdf
23. Mackintosh M, Channon A, Karan A, Selvaraj S, Cavagnero E, Zhao H. **What is the private sector? Understanding private provision in the health systems of low-income and middle-income countries**. *The Lancet* (2016.0) 596-605. DOI: 10.1016/S0140-6736(16)00342-1
24. Gautham M, Shyamprasad KM, Singh R, Zachariah A, Singh R, Bloom G. **Informal rural healthcare providers in North and South India**. *Health Policy Plan* (2014.0) **29** i20-i29. DOI: 10.1093/heapol/czt050
25. Raza WA, Van de Poel E, Panda P, Dror D, Bedi A. **Healthcare seeking behaviour among self-help group households in Rural Bihar and Uttar Pradesh, India**. *BMC Heal Serv Res* (2016.0) **16** 1-13. DOI: 10.1186/s12913-015-1254-9
26. 26Government of Bihar. List of Essential Drugs, Medical Devices and Consumables (Health Facility Wise). Patna; 2018. Available: http://www.bmsicl.gov.in/uploads/Drugs/finalEssentialDrugList_2018.pdf
27. 27The DHS Program. SPA Overview. [cited 5 Oct 2020]. Available: https://dhsprogram.com/What-We-Do/Survey-Types/SPA.cfm
28. 28International Institute for Population Sciences (IIPS), ICF. National Family Health Survey (NFHS-4), India, 2015–2016, Bihar. Minist Heal Fam Welfare, Gov India. 2017. Available: http://rchiips.org/nfhs/NFHS-4Reports/Bihar.pdf
29. 29Ministry of Statistics and Programme Implementation. India National Sample Survey (NSS) Round 71. New Delhi, India; 2014.
30. 30World Bank. SurveySolutions. Washington DC; Available: https://mysurvey.solutions/en/
31. Haber N, Pillay D, Porter K, Bärnighausen T. **Constructing the cascade of HIV care: methods for measurement**. *Curr Opin HIV AIDS* (2016.0) 11. DOI: 10.1097/COH.0000000000000212
32. Filmer D, Pritchett LH. **Estimating wealth effects without expenditure data—or tears: An application to educational enrollments in states of India**. *Demography* (2001.0) **38** 115-132. DOI: 10.1353/dem.2001.0003
33. 33StataCorp. Stata Statistical Software Release 14. College Station, TX.; 2014.
34. 34Ministry of Health and Family Welfare, National Health System and Resource Centre, Government of India. AYUSHMAN BHARAT Comprehensive Primary Health Care through Health and Wellness Centers Operational Guidelines. 2018; 96.
35. 35Directorate General of Health Services, Ministry of Health & Family welfare, Government Of India. National Programme for Prevention and Control of Cancer, Diabetes, Cardiovascular Diseases & stroke (npcdcs) operational guidelines (revised: 2013–17) directorate general of health services ministry of health & family welfare government of india. 2013; 78. Available: https://mohfw.gov.in/sites/default/files/Operational Guidelines of NPCDCS (Revised—2013–17).pdf
36. Najafi F, Pasdar Y, Shakiba E, Hamzeh B, Darbandi M, Moradinazar M. **Validity of Self-reported Hypertension and Factors Related to Discordance Between Self-reported and Objectively Measured Hypertension: Evidence From a Cohort Study in Iran**. *J Prev Med public Heal* (2019.0) **52** 131-139. DOI: 10.3961/jpmph.18.257
37. Leonard KL, Masatu MC. **Professionalism and the know-do gap: exploring intrinsic motivation among health workers in Tanzania**. *Health Econ* (2010.0) **19** 1461-1477. DOI: 10.1002/hec.1564
38. Das J, Hammer J. **Quality of Primary Care in Low-Income Countries: Facts and Economics**. *Annu Rev Econom* (2014.0) **6** 525-553. DOI: 10.1146/annurev-economics-080213-041350
39. Staessen JA, Thijisq L, Fagard R, Celis H, Birkenhäger WH, Bulpitt CJ. **Effects of immediate versus delayed antihypertensive therapy on outcome in the Systolic Hypertension in Europe Trial**. *J Hypertens* (2004.0) **22** 847-857. DOI: 10.1097/00004872-200404000-00029
40. Neupane D, McLachlan CS, Mishra SR, Olsen MH, Perry HB, Karki A. **Effectiveness of a lifestyle intervention led by female community health volunteers versus usual care in blood pressure reduction (COBIN): an open-label, cluster-randomised trial**. *Lancet Glob Heal* (2018.0) **6** e66-e73. DOI: 10.1016/S2214-109X(17)30411-4
41. Vedanthan R, Bernabe-Ortiz A, Herasme OI, Joshi R, Lopez-Jaramillo P, Thrift AG. **Innovative Approaches to Hypertension Control in Low- and Middle-Income Countries**. *Cardiol Clin* (2017.0) **35** 99-115. DOI: 10.1016/j.ccl.2016.08.010
|
---
title: 'Physical activity and functional disability among older adults in Ghana: The
moderating role of multi-morbidity'
authors:
- Kofi Awuviry-Newton
- Mary Amponsah
- Dinah Amoah
- Pablo Villalobos Dintrans
- Adjeiwa Akosua Afram
- Julie Byles
- Jacob Rugare Mugumbate
- Paul Kowal
- Nestor Asiamah
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021534
doi: 10.1371/journal.pgph.0001014
license: CC BY 4.0
---
# Physical activity and functional disability among older adults in Ghana: The moderating role of multi-morbidity
## Abstract
Knowledge about how physical activity levels relate to functional disability is essential for health promotion and planning older adults’ care or rehabilitation. The risk of living with one or more chronic health conditions increases with increasing age in lower and higher income countries–many of which are associated with physical inactivity. We conducted a cross-sectional study to examine the moderating role of multimorbidity on physical activity and its measures on functional disability among older adults in Ghana. Data from WHO’s Study on global AGEing and adult health Ghana Wave 2 with a sample of 4,446 people aged 50+ years was used for this study. Functional disability was assessed using the 12-item WHO Disability Assessment Schedule 2.0. Three categories of physical activity levels were used: vigorous intensity, moderate intensity, and walking. Past month diagnosis by a doctor was used to assess the presence of a chronic condition, and the presence of two or more conditions was used to define multi-morbidity. Logistic regressions with a post hoc interactional tests were used to examine the associations. Overall, physical activity had a significant association with functional disability (OR = 0.25, $95\%$CI; 0.12, 0.32). A similar relationship was found for vigorous-intensity (OR = 0.19, $95\%$CI: 0.12, 0.29), moderate-intensity (OR = 0.19, $95\%$CI: 0.15, 0.25) and walking (OR = 0.41, $95\%$CI: 0.33, 0.51). Older adults living with one condition and physically active were $47\%$ less likely to experience functional disability compared with the less active counterparts living with at least two chronic conditions. Among the three measures of physical activity, multimorbidity moderated the relationship between walking and functional disability. Future strategies for meeting the health and long-term care needs of older adults, particularly those living with only one chronic condition in Ghana should consider encouraging walking. Policies, financial assistance, family, and community level interventions aimed to promote and sustain physical activity among older adults should be a priority for stakeholders in Ghana.
## Introduction
The call for implementing long-term care systems for older adults in every country, including Ghana [1], requires an understanding of the relationship, among others, between physical activity and functional disability. Physical activity has proven to be essential in maintaining health and addressing older adults’ health and long-term care needs in western countries [2, 3]. Physical activity is defined as any bodily movement produced by skeletal muscles that require energy expenditure either through moderate or vigorous-intensity activities or walking [4]. Related to this, functional disability refers to the difficulty in completing activities relating to cognition, mobility, self-care, getting along with others, life activities and participation in society [5–7].
Although evidence on the effects of physical activity on functional disability in African countries such as *Ghana is* limited, there is more information from high-income countries. These studies have demonstrated how physical activity leads to a reduction in functional disability [8–11] by reducing the occurrence of chronic diseases [12]. For example, Kim, Park [13] revealed that engaging in a form of physical activity can lead to decrease in the incidence of depression later in life for adults in South Korea. Also, regular physical activity improves the gait and balance of older adults, consequently reducing the incidence of falls [14] as well as improving motor and auditory attention [15]. A meta-analysis by Tak et al. [ 16] revealed that an increase in physical activity led to delayed progression of functional disability in older adults; however, they did not find any difference in the rate of decline in functional disability following an increase in physical activity by either an older adult with or without disability [16]. Finally, evidence from 47 low-and middle-income countries reports that high physical activity was associated with less severe subjective memory and learning difficulties [17].
In both lower and higher income countries, multiple factors are known to moderate the relationship between physical activity and functional disability among older adults. Chronic pain is a key factor responsible for physical inactivity [18, 19], and a risk factor for functional disability in older adults [20]. Other known moderating factors include poor health status [21, 22], residing in urban areas [21, 23–25], being an older female [26, 27], educational level [22, 28], and marital status [18, 29, 30].
In sub-Saharan African countries, evidence on how these moderators influence the relationship between physical activity measures and functional disability is scant, despite its relevance in determining healthy ageing. Available evidence from a study conducted in low-and middle-income countries suggests that multi-morbidity may moderate the association between physical activity and functional disability [17]. This study aims to evaluate the interactional role of multi-morbidity on the association between physical activity and functional disability, controlling for a number of confounding variables. The novelty of the present study lies in its intention to provide baseline evidence on the relationship between physical activity, functional disability and multimorbidity among older adults in Ghana.
## Ethics statement
Ethical approval for this study was obtained from the WHO Ethical Research Committee (#ID3925). Consent for participants were obtained before the commencement of the study [31].
## Study sample
We used data from the Study on global AGEing and adult health (SAGE) Ghana Wave 2 conducted between $\frac{2014}{2015.}$ SAGE is a Multi-Country (Ghana, South Africa, China, India, Mexico, and Russia) longitudinal study that employed multistage cluster sampling strategies. The University of Ghana Medical School through the Department of Community Health, and in collaboration with the World Health Organization (WHO), implemented the SAGE Wave 2 in Ghana. The current study used a sample size of 4,446 participants (+50 years) who answered all 12 questions on functional disability. Details about the methodology and other relevant information on the study are available elsewhere [32].
## Functional disability
Functional disability was defined using the 12-item version of the WHO Disability Assessment Schedule (WHODAS 2.0), which classifies responses into five disability categories: none, mild, moderate, severe, and extremely severe. In its full version, the WHODAS 2.0 contains 12 questions from six domains: cognition, mobility, self-care, getting along, life activities, and participation in society [33]. S1 Appendix contains the questions included in the analysis. WHODAS 2.0 was scored on a scale of 0 to 100, with a lower score implying point for determining the severity of the disability [11, 34, 35]. Participants scoring <$90.18\%$ were denoted as “no disability” and participants who score > = $90.18\%$ were denoted as “with a disability.”
## Physical activity
Physical activity was measured with three separate items including vigorous activity, moderate activity, and walking. Vigorous-intensity activity was measured by the question “Does your work involve vigorous-intensity activity that causes large increases in breathing or heart rate, [like heavy lifting, digging, or chopping wood] for at least 10 minutes continuously (Yes/No)? The question “Does your work involve moderate-intensity activity that causes small increases in breathing or heart rate [such as brisk walking, carrying light loads, cleaning, cooking, or washing clothes] for at least 10 minutes continuously?” ( Yes/No) was used to measure work-related moderate-intensity activity. We used the question “Do you walk or use a bicycle (pedal cycle) for at least 10 minutes continuously to get to and from places?” ( Yes/ No) to measure older adults’ engagement in walking.
Additionally, these three measures were scored and aggregated into two response categories; yes (engages in at least one of the three measures of physical activity) and no (engages in none of the three measures). The Cronbach’s α of the three physical activity items combined was 0.61.
## Multimorbidity
The question “Have you ever been diagnosed with/told you have … in the past month? ( Yes/No) was used to identify the presence of each of the 11 chronic conditions including stroke, hypertension, injuries, depression, diabetes, angina, arthritis, chronic lung disease, asthma, cataract, and oral health among older adults. Responses were combined and a variable capturing the presence of different conditions was generated to measure multimorbidity (1 = no condition, 2 = one chronic condition and 3 = at least two conditions).
## Covariates
Sociodemographic and health confounding variables included in the analysis were age (continuous), sex (1 = male, 2 = female), marital status (1 = never married, 2 = married/cohabiting, 3 = separated/divorced, 4 = widowed), education (1 = less than primary school, 2 = primary education completed, 3 = senior high completed, 4 = university degree/post), location of residence (1 = rural, 2 = urban) and self-reported health status (1 = good, 2 = moderate, 3 = bad). The health status variable was excluded from the model as it was not statistically significant at $p \leq 0.05.$
## Data analysis
Using STATA version 16, frequency, and percentages, and means and standard deviations were used to describe the variables in the study. Second, bivariate analyses were performed through chi-square, Fisher’s test, and t-test to test relationships between independent variables and dependent variables. Finally, bivariate, and multivariate logistic regression were performed to estimate the crude and adjusted odds ratios (OR) and $95\%$ confidence intervals (CI) for the associations between physical activity and functional disability, and the moderation effect (interaction) of multimorbidity in the relationship.
## ‘Inclusivity in global research’
Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information.
## Descriptive statistics
The results in Table 1 show the univariate and bivariate analysis of independent variables in relationship to functional disability. Most of the participants were females ($58.9\%$). The mean age of participants was 57.6 years but the sample of people with functional disability was older, reaching a mean of 74 years (SD,12.2). The prevalence of functional disability among females was higher than in males ($64.5\%$ vs $35.5\%$). A high proportion of functional disability was reported among widowed older adults ($46.5\%$), higher in rural compared to urban areas. A higher prevalence of functional disability was reported among older adults who completed senior high education ($39.5\%$). Regarding multimorbidity, a high prevalence of functional disability was reported among older adults living with at least two chronic conditions ($45.9\%$, $p \leq 0.001$). Similarly, older adults with self-reported bad health experienced a high prevalence of functional disability ($68.2\%$, $p \leq 0.001$).
**Table 1**
| Independent variables | Overall | Functional disability | Functional disability.1 | p-value |
| --- | --- | --- | --- | --- |
| Independent variables | N (%) | No disability, N (%) | With disability, N (%) | p-value |
| Age (Mean, SD) | 57.6±16.7 | 55.0±16.0 | 74.1±12.2 | <0.001 |
| Sex | | | | <0.01 |
| Male | 1,826 (41.1) | 1,658 (41.7) | 168 (35.5) | |
| Female | 2,620 (58.9) | 2,315 (58.3) | 305 (64.5) | |
| Marital status | | | | <0.001 |
| Never married | 416 (9.36) | 409 (10.3) | 7 (1.48) | |
| Married/cohabiting | 2,555 (57.5) | 2,366 (59.6) | 189 (40.0) | |
| Separated/divorce | 499 (11.2) | 442 (11.1) | 57 (12.1) | |
| Widowed | 976 (21.9) | 756 (19.0) | 220 (46.5) | |
| Location of residence | | | | 0.331 |
| Rural | 2,624 (59.0) | 2,335 (58.8) | 289 (61.1) | |
| Urban | 1822 (41.0) | 1,638 (41.2) | 184 (38.9) | |
| Education | | | | 0.056 |
| Less than primary school | 610 (23.6) | 559 (23.0) | 51 (32.5) | |
| Primary education completed | 664 (25.7) | 629 (25.9) | 35 (22.3) | |
| Senior high completed | 1,168 (45.2) | 1,106 (45.6) | 62 (39.5) | |
| University degree/post | 142 (5.50) | 133 (5.48) | 9 (5.73) | |
| Health status | | | | <0.001* |
| Good | 627 (18.4) | 625 (19.8) | 2 (0.76) | |
| Moderate | 2,448 (71.7) | 2,366 (75.1) | 82 (31.1) | |
| Bad | 341 (9.98) | 161 (5.11) | 180 (68.2) | |
| Multimorbidity | | | | <0.001 |
| No morbidity | 2,365 (53.2) | 2,193 (55.2) | 172 (36.4) | |
| Only one morbidity | 529 (11.9) | 445 (11.2) | 84 (17.8) | |
| 2 or more morbidities | 1,335 (33.6) | 1,335 (33.6) | 217 (45.9) | |
| Physical activity (PA) | | | | |
| Vigorous-intensity activity | | | | <0.001 |
| Yes | 1,346 (30.5) | 1,324 (33.6) | 22 (4.69) | |
| No | 3,069 (69.5) | 2,622 (66.5) | 447 (95.3) | |
| Moderate-intensity activity | | | | <0.001 |
| Yes | 2,576 (58.4) | 2,489 (63.1) | 87 (18.6) | |
| No | 1,839 (41.7) | 1,457 (36.9) | 382 (81.5) | |
| Walk | | | | <0.001 |
| Yes | 2,878 (65.2) | 2,684 (68.0) | 194 (41.4) | |
| No | 1,537 (34.8) | 1,262 (32.0) | 275 (58.6) | |
| PA (Overall) | | | | <0.001 |
| Yes | 3,392 (76.3) | 3,182 (80.1) | 210 (44.4) | |
| No | 1,054 (23.7) | 791 (19.9) | 263 (55.6) | |
| Functional disability | | | | |
| No | 3,973 (89.4) | - | - | - |
| Yes | 473 (10.6) | - | - | - |
Older adults who engaged in physical activity of any kind reported lower functional disability compared to those who were not active ($44.4\%$ vs 58.6, $p \leq 0.001$). A similar result was revealed in the engagement of three specific measures of physical activity (vigorous ($4.69\%$), moderate ($18.6\%$) or walking ($41.4\%$) compared to those who do not engage in these activities ($p \leq 0.001$). The prevalence of functional disability among older adults was approximately $11\%$.
## Primary results
In this section, we discuss the relationship between overall physical activity and functional disability, while controlling for potential confounding variables including age, sex, marital status and multimorbidity (Table 2). In the unadjusted model (column 1), older adults who were physically active were $80\%$ less likely to experience functional disability (OR = 0.20, $95\%$CI; 0.16, 0.24). This effect is still observed in the adjusted models (columns 2–6), showing robust evidence of a negative significant relationship between physical activity and functional disability (OR = 0.25, $95\%$CI; 0.12, 0.32), with minimal changes in the odds ratios. It is important to note that physical activity is measured in its aggregate form, hence the results do not show the potential different effects of different measures of physical activity.
**Table 2**
| Unnamed: 0 | Model 1 | Model 2 | Model 03 | Model 4 | Model 5 | Model 6 |
| --- | --- | --- | --- | --- | --- | --- |
| Physical activity (PA) | 0.20 (0.16, 0.24)*** | 0.26 (0.21, 0.32)*** | 0.20 (0.16, 0.24)*** | 0.19 (0.16, 0.24)*** | 0.21 (0.17, 0.25)*** | 0.25 (0.12, 0.32)*** |
| Age | | 1.10 (1.09, 1.11)*** | | | | 1.10 (1.09, 1.11)*** |
| Sex | | | | | | |
| Male | | | 1 | | | 1 |
| Female | | | 1.24 (1.01, 1.52)* | | | 1.35 (1.03, 1.76)* |
| Marital status | | | | | | |
| Never married | | | | 0.18 (0.08, 0.38)*** | | 0.41 (0.18, 0.93)* |
| Married/cohabiting | | | | 1 | | 1 |
| Separated/divorce | | | | 1.78 (1.28, 2.46)*** | | 1.54 (1.08, 2.21)* |
| Widowed | | | | 3.48 (2.79, 4.34)*** | | 1.37 (1.03, 1.83)* |
| Multimorbidity | | | | | | |
| No morbidity | | | | | 1 | 1 |
| Any one morbidity | | | | | 2.13 (1.59, 2.85)*** | 1.02 (0.91, 1.41) |
| 2 or more morbidities | | | | | 1.92 (1.54, 2.39)*** | 1.39 (1.09, 1.77)*** |
## Interactions of multimorbidity on the association between physical activity and functional disability
As part of the objective of this study, we estimated the interactions of the impact of morbidity on the relationship between physical activity and functional disability. The results revealed that having at least one chronic condition was associated with more disability; while the interaction between physical activity and multi-morbidity showed no significant effect on functional disability, except for people who experienced one morbidity: for this group, older people who engaged in overall physical activity were $47\%$ less likely to experience functional disability (OR = 0.53, $95\%$CI; 0.29, 0.96) compared to their counterparts (Tables 3 and 4).
## Sensitivity analysis
This analysis was conducted to determine which type of work-related physical activity (vigorous-intensity activity, moderate-intensity activity, and walking) is more likely to affect functional disability. Table 3 reports the association of these activities on functional disability while Table 4 reports the interaction between multimorbidity and physical activity association on functional disability.
Work-related vigorously intense activity was significantly associated with functional disability (Table 3, columns 1–6). Older adults who engaged in vigorously intense activity were $90\%$ less likely to experience functional disability (OR = 0.10, $95\%$CI: 0.06, 0.15) compared to their counterparts. When adjusted for all confounding included in the study (Table 3, column 6), vigorously intense activity significantly and independently was associated with functional disability (OR = 0.19, $95\%$CI: 0.12, 0.29).
For moderate intensity (Table 3, columns 7–12), there was a significant association with functional disability (OR = 0.13, $95\%$CI: 0.11, 0.17). After controlling for confounding variables, the significant association remained (OR = 0.19, $95\%$CI: 0.15, 0.25).
Similarly, older adults who engaged in walking were $67\%$ less likely to experience functional disability (OR = 0.33, $95\%$CI: 0.27, 0.40) compared to their counterparts (unadjusted column 13 in Table 3). A similar association was found when controlling for all confounders in Model 18 in Table 3 (OR = 0.41, $95\%$CI: 0.33, 0.51).
Table 4 presents the results of including multimorbidity as a moderator in the relationship between functional disability and physical activity. While no evidence existed that multimorbidity moderated the relationships between functional disability associations with vigorous- moderate activity, we found significant evidence of the impact of multimorbidity on the association between walking and functional disability. That is, older adults who experience only one chronic condition and engaged in walking were $58\%$ less likely to experience functional disability (OR = 0.42, $95\%$CI: 0.23, 0.76) compared to their counterparts. This is evidence that engaging in a physical activity, particularly walking, when having any one chronic condition, decreases the likelihood of experiencing functional disability.
## Discussion
The current study revealed that older adults with physical activity engagement were $75\%$ less likely to experience functional disability. Physical activity measures such as vigorous-intensity, moderate-intensity and walking were independently significantly associated with functional disability among older adults. Older adults living with only one chronic condition and engaging in walking were $58\%$ less likely to experience functional disability compared to those with no or at least two chronic conditions. The current study adds to the available literature that multimorbidity moderates the relationship between physical activity (walking) and functional disability. This finding extends the existing evidence that significant number of older adults in low- and middle-income countries live with functional disability [11], by adding that engaging in specific physical activity such as vigorous, moderate or walking can contribute to the improvement of the functional ability of older adults. The result is consistent with previous studies from developed countries on how engagement in physical activity reduces the incidence of chronic diseases and depression later in life [12, 13], ultimately reducing premature mortalities [36–38], and improving quality of life [39].
The current study revealed a negative relationship between physical activity and functional disability. The examination of the intensity of physical activity that yielded a reduced functional disability revealed that all measures (vigorous, moderate, or walking) resulted in an improvement in older adults ‘functional ability, implying that engagement of any form of physical activity is more likely to lead to reduced functional disability among older adults [13, 15, 20].
Even though evidence has shown that poor health conditions are barrier to physical activity among older adults [21, 22], the current study adds that engagement of physical activity is effective when recommended for older adults with only one chronic condition as they are $47\%$ less likely to experience functional disability as compared to their counterparts. With regards to the intensity of physical activity recommended to achieve a reduction in functional disability, the current study revealed that older adults who experienced only one chronic condition and engaged in walking have better chance of minimizing incidence of functional disability. This finding partly supports a recent study [40] which recommends physical activity for older adults with chronic conditions as it improves functional ability, however, the intensity effective to achieve a reduced functional disability was proposed to be moderate to vigorous physical activity as opposed to light physical activity. The disparity could be due to the nature, duration, and severity of the condition which was not specified. It can therefore be explained that poor health status is not absolutely a barrier to physical activity among older people but the need for sensitization of the benefits of physical activity is paramount [28]. If the benefits of physical activity are made known, older adults may tend to see poor health as a motivation to engage in physical activity than taking medications as reported in a previous study [41]. In this study, the finding that older adults who participated in walking and living with at least two chronic conditions had no association with functional disability is understandable as pains associated with these conditions is usually intense making impact of physical activity such as walking on physical activity insignificant.
Despite the benefits outlined by this study, there are however some limitations. First, data on the type of morbidities were not available as its inclusion would have provided an in-depth understanding of the relationship. Third, the cross-sectional nature of the study limits it in determining the causality of the relationship; future studies using longitudinal research methods can help in understanding how this relationship change over time.
## Conclusions and implications
Physical activity particularly, walk, should be encouraged among older adults, even for those with some health issues, as evidence proves it associates with a decrease in functional disability. This can be achieved through a regular creation of awareness through media, campaigns, commemoration of health events on the need for older adults with one form of morbidity to engage in physical activity. In addition, medical personnel attending to older adults with morbidities should be sensitized on the need to incorporate physical activity into the routine care or post discharge care in a bid to reduce functional dysfunction. Policy interventions should therefore incorporate a built environment (sidewalks) that will facilitate walking, organize regular community activities or programs for elderly adults that involves physical activities. Moreover, caregivers should encourage older adults to take part in community meetings or family functions as this can encourage walking to improve the quality of life of older adult. Policies, public health interventions, and financial assistance aimed to promote and sustain physical activity among older adults should be a priority for stakeholders of rehabilitation of older adults receiving long-term care services of any kind in Ghana. The current finding does not only benefit Ghana, but also other low-and middle-income countries seeking to promote healthy ageing.
## References
1. 1World Health Organisation. Decade of Healthy Ageing.
30
August
2020. https://www.who.int/publications/i/item/9789240017900.. *Decade of Healthy Ageing.* (2020)
2. Lau RS, Ohinmaa A, Johnson JA. **Predicting the future burden of diabetes in Alberta from 2008 to 2035.**. *Canadian Journal of Diabetes.* (2011) **35** 274-81. DOI: 10.1016/S1499-2671(11)53011-4
3. Paterson DH, Warburton DE. **Physical activity and functional limitations in older adults: a systematic review related to Canada’s Physical Activity Guidelines.**. *Int J Behav Nutr Phys Act* (2010) **7** 38. DOI: 10.1186/1479-5868-7-38
4. 4WHO. Physical Activity
2020 [Available from: https://www.who.int/news-room/fact-sheets/detail/physical-activity.. *Physical Activity* (2020)
5. 5World Health Organisation. Measuring health and disability: manual for WHO Disability Assessment Schedule (WHODAS 2.0).
Geneva: World Health Organisation; 2012. https://www.who.int/publications/i/item/measuring-health-and-disability-manual-for-who-disability-assessment-schedule-(-whodas-2.0).. *Measuring health and disability: manual for WHO Disability Assessment Schedule (WHODAS 2.0).* (2012)
6. Connolly D, Garvey J, McKee G. **Factors associated with ADL/IADL disability in community dwelling older adults in the Irish longitudinal study on ageing (TILDA).**. *Disability and rehabilitation.* (2017) **39** 809-16. DOI: 10.3109/09638288.2016.1161848
7. Chen SW, Chippendale T. **Factors associated with IADL independence: implications for OT practice.**. *Scand J Occup Ther* (2017) **24** 109-15. DOI: 10.1080/11038128.2016.1194464
8. Miller ME, Rejeski WJ, Reboussin BA, Ten Have TR, Ettinger WH. **Physical activity, functional limitations, and disability in older adults**. *Journal of the American Geriatrics Society* (2000) **48** 1264-72. DOI: 10.1111/j.1532-5415.2000.tb02600.x
9. Taylor D.. **Physical activity is medicine for older adults**. *Postgraduate Medical Journal* (2014) **90** 26-32. DOI: 10.1136/postgradmedj-2012-131366
10. Van Rossum M, Koek H. **Predictors of functional disability in mild cognitive impairment and dementia.**. *Maturitas* (2016) **90** 31-6. DOI: 10.1016/j.maturitas.2016.05.007
11. Biritwum R, Minicuci N, Yawson A, Theou O, Mensah G, Naidoo N. **Prevalence of and factors associated with frailty and disability in older adults from China, Ghana, India, Mexico, Russia and South Africa.**. *Maturitas.* (2016) **91** 8-18. DOI: 10.1016/j.maturitas.2016.05.012
12. Oguma Y, Shinoda-Tagawa T. **Physical activity decreases cardiovascular disease risk in women: review and meta-analysis**. *American Journal of Preventive Medicine* (2004) **26** 407-18. DOI: 10.1016/j.amepre.2004.02.007
13. Kim S-Y, Park J-H, Lee MY, Oh K-S, Shin D-W, Shin Y-C. **Physical activity and the prevention of depression: A cohort study.**. *General Hospital Psychiatry* (2019) **60** 90-7. DOI: 10.1016/j.genhosppsych.2019.07.010
14. Paterson DH, Jones GR, Rice CL. **Ageing and physical activity: evidence to develop exercise recommendations for older adults**. *Can J Public Health* (2007) **2** S69-108. DOI: 10.1139/H07-111@apnm-vis.issue01
15. Angevaren MA, Geert HJJ, Verhaar A, Aleman L. **Physical activity and enhanced fitness to improve cognitive function in older people withoutknown cognitive impairment.**. *Cochrane Database of Systematic Reviews* (2008). DOI: 10.1002/14651858.CD005381.pub2
16. Tak E, Kuiper R, Chorus A, Hopman-Rock M. **Prevention of onset and progression of basic ADL disability by physical activity in community dwelling older adults: a meta-analysis.**. *Ageing research reviews.* (2013) **12** 329-38. DOI: 10.1016/j.arr.2012.10.001
17. Felez-Nobrega M, Haro JM, Erickson KI, Koyanagi A. **Physical activity is associated with fewer subjective cognitive complaints in 47 low-and middle-income countries**. *Journal of the American Medical Directors Association* (2020) **21** 1423-9. DOI: 10.1016/j.jamda.2020.02.014
18. Gureje O, Ogunniyi A, Kola L, Afolabi E. **Functional disability in elderly Nigerians: Results from the Ibadan Study of Aging**. *Journal of the American Geriatrics Society* (2006) **54** 1784-9. DOI: 10.1111/j.1532-5415.2006.00944.x
19. Dansie EJ, Turk DC, Martin KR, Van Domelen DR, Patel KV. **Association of chronic widespread pain with objectively measured physical activity in adults: findings from the National Health and Nutrition Examination survey**. *The Journal of Pain* (2014) **15** 507-15. DOI: 10.1016/j.jpain.2014.01.489
20. Eggermont LH, Leveille SG, Shi L, Kiely DK, Shmerling RH, Jones RN. **Pain characteristics associated with the onset of disability in older adults: the maintenance of balance, independent living, intellect, and zest in the Elderly Boston Study.**. *Journal of the American Geriatrics Society.* (2014) **62** 1007-16. DOI: 10.1111/jgs.12848
21. Gureje O, Kola L, Afolabi E, Olley BO. **Determinants of quality of life of elderly Nigerians: results from the Ibadan study of ageing.**. *Afr J Med Med Sci* (2008) **37** 239-47. PMID: 18982816
22. Liu H, Jiao J, Zhu C, Zhu M, Wen X, Jin J. **Potential associated factors of functional disability in Chinese older inpatients: a multicenter cross-sectional study.**. *BMC Geriatrics* (2020) **20** 319. DOI: 10.1186/s12877-020-01738-x
23. Peltzer K, Pengpid S. **Physical inactivity among older adults with and without functional disabilities in South Africa.**. *African Journal for Physical Activity and Health Sciences (AJPHES).* (2020) **26** 252-60. DOI: 10.37597/ajphes.2020.26.3.2
24. Wu F, Guo Y, Chatterji S, Zheng Y, Naidoo N, Jiang Y. **Common risk factors for chronic non-communicable diseases among older adults in China, Ghana, Mexico, India, Russia and South Africa: the study on global AGEing and adult health (SAGE) wave 1.**. *BMC public health.* (2015) **15** 1-13. DOI: 10.1186/s12889-015-1407-0
25. Parahyba MI, Veras R, Melzer D. **Incapacidade funcional entre as mulheres idosas no Brasil.**. *Revista de Saúde Pública* (2005) **39** 383-91. DOI: 10.1590/s0034-89102005000300008
26. Curtis J, White P, McPherson B. **Age and physical activity among Canadian women and men: findings from longitudinal national survey data**. *Journal of Aging and Physical Activity* (2000) **8** 1-19
27. Sun F, Norman IJ, While AE. **Physical activity in older people: a systematic review.**. *BMC Public Health* (2013) **13** 449. DOI: 10.1186/1471-2458-13-449
28. Aro AA, Agbo S, Omole OB. **Factors influencing regular physical exercise among the elderly in residential care facilities in a South African health district**. *African Journal of Primary Health Care and Family Medicine* (2018) **10** 1-6
29. Millán-Calenti JC, Tubío J, Pita-Fernández S, González-Abraldes I, Lorenzo T, Fernández-Arruty T. **Prevalence of functional disability in activities of daily living (ADL), instrumental activities of daily living (IADL) and associated factors, as predictors of morbidity and mortality.**. *Archives of Gerontology and Geriatrics* (2010) **50** 306-10. DOI: 10.1016/j.archger.2009.04.017
30. Wandera SO, Ntozi J, Kwagala B. **Prevalence and correlates of disability among older Ugandans: evidence from the Uganda National Household Survey.**. *Global Health Action* (2014) **7** 25686. DOI: 10.3402/gha.v7.25686
31. Kowal P, Chatterji S, Naidoo N, Biritwum R, Fan W, Lopez Ridaura R. **Data resource profile: the World Health Organization Study on global AGEing and adult health (SAGE).**. *International Journal of Epidemiology* (2012) **41** 1639-49. DOI: 10.1093/ije/dys210
32. Charlton K, Ware LJ, Menyanu E, Biritwum RB, Naidoo N, Pieterse C. **Leveraging ongoing research to evaluate the health impacts of South Africa’s salt reduction strategy: a prospective nested cohort within the WHO-SAGE multicountry, longitudinal study**. *BMJ Open* (2016) **6** e013316. DOI: 10.1136/bmjopen-2016-013316
33. 33World Health Organisation. Measuring health and disability: manual for WHO Disability 326 Assessment Schedule (WHODAS 2.0).
Geneva: World Health Organisation;
2010
2012. https://www.who.int/publications/i/item/measuring-health-and-disability-manual-for-who-disability-assessment-schedule-(-whodas-2.0).. *Measuring health and disability: manual for WHO Disability 326 Assessment Schedule (WHODAS 2.0).* (2010) **2012**
34. Andrews G, Kemp A, Sunderland M, Von Korff M, Ustun TB. **Normative data for the 12 item WHO Disability Assessment Schedule 2.0.**. *PloS One* (2009) **4** e8343. DOI: 10.1371/journal.pone.0008343
35. Kirchberger I, Braitmayer K, Coenen M, Oberhauser C, Meisinger C. **Feasibility and psychometric properties of the German 12-item WHO Disability Assessment Schedule (WHODAS 2.0) in a population-based sample of patients with myocardial infarction from the MONICA/KORA myocardial infarction registry.**. *Population Health Metrics* (2014) **12** 27
36. Keysor JJ. **Does late-life physical activity or exercise prevent or minimize disablement?: a critical review of the scientific evidence**. *American Journal of Preventive Medicine* (2003) **25** 129-36. DOI: 10.1016/s0749-3797(03)00176-4
37. Netz Y, Wu M-J, Becker BJ, Tenenbaum G. **Physical activity and psychological well-being in advanced age: a meta-analysis of intervention studies**. *Psychology and Aging* (2005) **20** 272. PMID: 16029091
38. Ueshima K, Ishikawa-Takata K, Yorifuji T, Suzuki E, Kashima S, Takao S. **Physical activity and mortality risk in the Japanese elderly: a cohort study**. *American Journal of Preventive Medicine* (2010) **38** 410-8. DOI: 10.1016/j.amepre.2009.12.033
39. Motl RW, McAuley E. **Physical activity, disability, and quality of life in older adults.**. *Phys Med Rehabil Clin N Am* (2010) **21** 299-308. DOI: 10.1016/j.pmr.2009.12.006
40. Makino K, Lee S, Lee S, Bae S, Jung S, Shinkai Y. **Daily Physical Activity and Functional Disability Incidence in Community-Dwelling Older Adults with Chronic Pain: A Prospective Cohort Study.**. *Pain Med* (2019) **20** 1702-10. DOI: 10.1093/pm/pny263
41. Belza B, Walwick J, Schwartz S, LoGerfo J, Shiu-Thornton S, Taylor M. **pEER REvIEWED: older Adult perspectives on physical Activity and Exercise: voices From Multiple cultures.**. *Preventing Chronic Disease* (2004) **1**. PMID: 15670441
|
---
title: 'Linkage to HIV care and hypertension and diabetes control in rural South Africa:
Results from the population-based Vukuzazi Study'
authors:
- Itai M. Magodoro
- Stephen Olivier
- Dickman Gareta
- Olivier Koole
- Tshwaraganang H. Modise
- Resign Gunda
- Kobus Herbst
- Deenan Pillay
- Emily B. Wong
- Mark J. Siedner
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021540
doi: 10.1371/journal.pgph.0001221
license: CC BY 4.0
---
# Linkage to HIV care and hypertension and diabetes control in rural South Africa: Results from the population-based Vukuzazi Study
## Abstract
Non-communicable diseases (NCDs) account for half of all deaths in South Africa, partly reflecting unmet NCDs healthcare needs. Leveraging existing HIV infrastructure is touted as a strategy to alleviate this chronic care gap. We evaluated whether HIV care platforms are associated with improved NCDs care. We conducted a community-based screening of adults in rural KwaZulu-Natal, collecting BP, HbA1c, and health services utilization data. Care cascade indicators for hypertension and diabetes mellitus were defined as: 1) aware, if previously diagnosed, 2) in care, if seeing a provider within last 6 months; 3) treated, if reporting medication use within preceding 2 weeks; and 4) controlled, if BP<$\frac{140}{90}$mmHg or HbA1c<$6.5\%$. We fit multivariable adjusted logistic regression models to compare successful completion of each step of the care cascade for hypertension and diabetes between people with virally suppressed HIV and HIV-negative comparators. Inverse probability sampling weights were applied to derive population-level estimates. The analytic sample included 4,933 individuals [mean age 58.4 years; $77\%$ female]. Compared to being HIV-negative, having suppressed HIV was associated with lower adjusted prevalence of being aware (-$6.0\%$ [$95\%$ CI: -11.0, -$1.1\%$]), in care (-$5.7\%$ [-10.6, -$0.8\%$]), and in treatment (-$4.8\%$ [-9.7, $0.1\%$]) for diabetes; but higher adjusted prevalence of controlled diabetes ($3.2\%$ [0.2–$6.2\%$]). In contrast, having suppressed HIV was associated with higher adjusted prevalence of being aware ($7.4\%$ [5.3–$9.6\%$]), in care ($8.0\%$ [5.9–$10.2\%$]), in treatment ($8.4\%$ [6.1–$10.6\%$]) and controlled ($9.0\%$ [6.2–$11.8\%$]), for hypertension. Overall, disease control was achieved for $40.0\%$ (38.6–$40.8\%$) and $6.8\%$ (5.9–$7.8\%$) of individuals with hypertension and diabetes, respectively. Engagement in HIV care in rural KwaZulu-Natal was generally associated with worse diabetes care and improved hypertension care. While further work should explore how success of HIV programs can be translated to NCD care, strengthening of primary healthcare will also be needed to respond to the growing NCDs epidemic.
## Introduction
Noncommunicable diseases (NCDs) are exacting rising human and economic costs in sub-Saharan Africa (SSA) [1–3]. It is estimated that nearly half ($46\%$) of all deaths in the region will be attributable to NCDs by 2030, representing a two-fold increase from $25\%$ in 2004 [4]. However, national health systems in SSA appear largely ill-prepared to meet this challenge [5–8]. Continuity of care across the lifespan and care settings is one of the critical elements of chronic diseases programs [9–11]. Health care provision in much of SSA, in contrast, is typically geared towards acute and/or episodic ill-health [12, 13]. Given the region`s resources constraints, creative strategies will be required to close this growing chronic care delivery gap [14, 15].
By contrast, HIV primary care programs in SSA have been enormously successful [16], commonly achieving global targets of disease diagnosis, enrollment in care, and control, or the so-called “cascade of care” [12]. HIV care programs typically outperform existing NCDs care delivery. For example, systematic reviews have suggested that in SSA $81\%$ of persons living with HIV (PWH) have been diagnosed, versus $30\%$ of those with hypertension; $70\%$ of PWH are on antiretroviral therapy (ART) [17, 18] versus $18\%$ of hypertensives receiving treatment [19, 20]; and at least $80\%$ of PWH on ART are virally suppressed compared with only $7\%$ drug-treated hypertensive individuals achieving adequate blood pressure control [19, 20]. Based on the comparative success of HIV care programs, many have proposed adapting the healthcare infrastructure that developed in response to the HIV epidemic to meet the growing care needs occasioned by NCDs [12, 21].
In response, beginning in 2011, South Africa adopted a national integrated chronic diseases management policy [22]. This policy, among other goals, envisaged integrated HIV, hypertension and diabetes care at primary healthcare (PHC) level. It was further expected that integration would extend disease prevention and management from health facilities into households and communities driven mainly by community health workers. While challenges with implementation of this policy in South Africa have been highlighted [23–26], the broader evidence for the clinical benefits and cost effectiveness of adapting HIV care programs for NCD care in general is equivocal, contextual and sparse [27–30]. Studies to date in SSA have mostly focused on integrated HIV and NCD clinical care. These studies are largely facility-based, with relatively small and/or highly selected patient populations. Similarly, the economic justification for HIV/NCD integration in SSA remains unproven [31]. Overall, these knowledge gaps suggest that, while leveraging HIV platforms for NCDs care is intuitively appealing, it nonetheless lacks a robust evidence base. To help address this gap in the literature, we analyzed data from a large population-based cohort [32] in one of the world`s high HIV burden settings [33, 34], to examine whether engagement in HIV primary care delivers improved hypertension and diabetes mellitus care in rural South Africa using the care cascade framework. We hypothesized that PWH with a suppressed viral load would have evidence of improved indicators along the NCD cascade of care, supporting a role of expanding the HIV care model for management of chronic diseases more broadly.
## Study design, population and setting
We analyzed individual-level data from persons aged ≥15 years old participating in the Vukuzazi Population Health Study (Vukuzazi Study) and with either hypertension or diabetes mellitus. The Vukuzazi Study was a population-based clinical phenotyping survey in uMkhanyakude District, KwaZulu-Natal, South Africa, with goals of describing the burden and intersection of HIV, tuberculosis (TB), hypertension and diabetes mellitus (DM) in rural South Africa [35]. It is nested within a 20-year demographic and health surveillance site (DHSS) started in 2000 and presently covering a population of c.140,000 persons [32]. The DHSS combines an annual household-based census with individual level socio-demographic data collection and HIV testing, among others. HIV clinical care data are also available through centralized electronic patient records (TIER.net) with district-wide coverage, and the HIV care continuum for the district has been reported elsewhere [35]. All 36,314 resident adults over the age of 15 years in the southern DHSS study area were eligible for recruitment into the Vukuzazi Study. Individual households were visited by research assistants who explained study objectives and invited all consenting adults to attend a single mobile clinic visit for data collection. All study activities were conducted in keeping with the principles of the Declaration of Helsinki, and had prior approval of the Africa Health Research Institute (AHRI) review board. Study participants gave written informed consent or informed parental/guardian written consent and participant assent if <16 years old as per standard South African practice.
## Data collection
A questionnaire was administered by research nurses to collect data on socio-demographics, smoking, medical history of HIV, hypertension, tuberculosis (TB), diabetes and their respective drug treatment, and to establish current TB symptomatology. Measurements of brachial blood pressure (BP), weight, height and waist circumference were obtained according to the WHO STEPS (STEPwise Approach to Surveillance) protocol [36]. The last two of three BP readings were averaged to estimate the final BP reading. Non-fasting venous blood was collected to measure glycated hemoglobin (HbA1c) using the VARIANT II TURBO Hemoglobin testing system [Bio-Rad, Marnes-la-Coquette, Paris, France] and to test for HIV (Genscreen Ultra HIV Ag-Ab enzyme immunoassay [Bio-Rad]). Participants with a positive HIV immunoassay had reflex measurement of HIV-1 RNA viral load (Abbott RealTime HIV-1 Viral Load [Abbott, IL, USA]) and CD4+ cell count (BD FACS Calibur flow cytometer, BD Bioscience [San Jose, CA, USA]).
## Definitions of hypertension, diabetes mellitus, comorbidity and NCD care indicators
We defined hypertension as systolic BP ≥140 mmHg and/or diastolic BP ≥90 mmHg or self-reported use of antihypertensive medication in the preceding 2 weeks; and diabetes mellitus (diabetes) as HbA1c ≥$6.5\%$ or self-reported use of hypoglycemic medication in the past 2 weeks. Cascade of care indicators were defined as follows: 1) aware: a self-reported prior hypertension or diabetes diagnosis, 2) in care: seeing a health provider within the past 6 months for the relevant NCD; 3) treated: reported use of appropriate medication in the preceding 2 weeks; and 4) controlled: BP<$\frac{140}{90}$mmHg or HbA1c <$6.5\%$, for hypertension and diabetes, respectively.
## Definitions of linkage to HIV care
Linkage to HIV care was defined as enrolment in ongoing ART for at least six preceding months based on centralized electronic patient records (TIER.net), and was further categorized as virologically suppressed (HIV/suppressed) if the current HIV-1 RNA load ≤40 copies/mL.
## Data analysis
We limited our analyses to Vukuzazi Study participants with either hypertension or diabetes (as defined above). Our primary analysis compared HIV-negative with HIV/suppressed individuals overall with respect to hypertension and diabetes care indices. To construct a sample representative of the local population, we constructed sampling weights from the stabilized inverse probability of participation in the Vukuzazi Study. Weights were based on the predicted probability of participation in the Vukuzazi Study and calculated by fitting logistic regression models with study participation as the outcome of interest and predictors comprised of age and sex as derived from the 2018 DHSS census.
We summarized participant characteristics and compared differences by HIV serostatus using t-test and chi-squared tests. We categorized age as <25 years old, 25–44 years old, and 45–64, and ≥65 years old; body mass index (BMI) as underweight (<18.5), normal (18.5–24.9), overweight (25.0–29.9), and obese (≥ 30 kg/m2); waist circumference as increased if >102cm (male) or 88cm (female); smoking status as current, former, and never; and HbA1c as ≥$6.5\%$ (diabetic), 5.7–$6.4\%$ (pre-diabetic) and <$6.4\%$ (normoglycemic) [37]. Blood pressure was further classified as normal (SBP <120mmHg and DBP <80mmHg), prehypertension (SBP 120–139 mmHg or DBP 80–89 mmHg), and stage 1 (SBP 140-159mmHg or DBP 90-99mmHg) and stage 2 (SBP ≥160mmHg or DBP ≥100mmHg) hypertension [38]. We assessed socioeconomic status using household-owned assets and housing characteristics aggregated into a Filmer-Pritchett asset wealth index and divided into tertiles [39].
For each of hypertension and diabetes, we compared successful completion in each step of the “cascade of care”, between HIV-negative and HIV/suppressed persons fitting multivariable logistic regression models to estimate the prevalence of [1] NCD disease awareness, [2] NCD engagement in care, [3] NCD treatment, [4] and NCD disease control. Models were adjusted for age, sex, BMI, education, smoking status and wealth tertiles. We also fit linear regression models, adjusted for the same covariates, to estimate differences in mean HbA1c% and systolic BP by stage in the NCD care cascade and HIV care status. We conducted a sensitivity analysis to assess the impact of HIV clinical care, irrespective of virologic suppression, by including all people with HIV who had been on ART for at least 6 months, without regard to their viral load. All statistical analyses were performed using R statistical software [2021] (R Foundation for Statistical Computing, Vienna, Austria) with a 2-sided p value of < 0.05 considered statistically significant.
## Derivation of analytic sample
Between May 2018 –March 2020, 18,027 [out of 36,314 ($50\%$)] adults had participated in the Vukuzazi study, and were thus eligible for inclusion in this analysis. We excluded 78 with missing BP or HbA1c data, and 20 with implausible BP readings (systolic BP <60 or >240 mmHg, and/or diastolic BP <45 or >160mmHg). We excluded an additional 4 with missing HIV test results, and if sero-positive, with missing viral load testing results, leaving 17,924 participants (HIV positive = 6,090; HIV negative = 11,854) with complete data. Among the 6,090 people with HIV, 1,221 ($20.0\%$) had hypertension and 380 ($6.2\%$) had diabetes, and among the 11,854 without HIV, 3,382 ($28.5\%$) had hypertension and 1,352 ($11.4\%$) diabetes. We subsequently excluded 12,744 participants who had neither hypertension nor diabetes mellitus, and a further 187 with an unsuppressed HIV viral load (who were included in sensitivity analyses only). The final analytical sample for our primary analysis included 4,993 participants (S1 Fig), and when weighted, it was comparable to the true DHSS population indicating the external validity of IPTW adjustment (S1 Table).
## General characteristics
The unweighted analytic sample including adults ≥15 years old with either hypertension and/or diabetes, who had a median (IQR) age of 60 years (interquartile range 50–69) years, and of whom $77\%$ (CI 76.2–$78.5\%$) were females. HIV/suppressed persons were younger (<45 years old: $29.0\%$ [CI 26.4–$31.6\%$] vs. $14.3\%$ [CI 13.3–$15.5\%$]; $P \leq 0.001$), more frequently female ($80\%$ [CI 78.2–$82.8\%$] vs. $76\%$ [CI 74.9–$77.7\%$]; $$p \leq 0.007$$) and they had higher rates of formal educational attainment (post-secondary: $5.7\%$ [CI 4.4–$7.3\%$] vs. $2.8\%$ [CI 2.3–$3.4\%$]; $p \leq 0.001$) and employment (full-time: $22\%$ [CI 19.2–$24\%$] vs. $12\%$ [CI 11.4–13.6]; $p \leq 0.001$) than HIV negative comparators (Table 1 and S2 Table). After inverse probability weighting to account for non-participation in the Vukuzazi study, these relationships were largely consistent (S1 Table).
**Table 1**
| Characteristica | HIV Negative (n = 3,798) | HIV Positive with viral suppressionb (n = 1,195) | P value | Overall (n = 4,993) |
| --- | --- | --- | --- | --- |
| Proportion (%) | | | | |
| Female | 2,899 (76%) | 963 (81%) | 0.002 | 3,862 (77%) |
| Marital Status | | | | |
| Single (never married) | 845 (35%) | 461 (65%) | | 1,306 (42%) |
| Married/Informal union | 791 (33%) | 108 (15%) | | 899 (29%) |
| Widowed/divorced/separated | 782 (32%) | 136 (19%) | <0.001 | 918 (29%) |
| Age (years) | 62.0 (53.0, 71.0) | 52.0 (43.0, 59.0) | <0.001 | 60.0 (50.0, 69.0) |
| <25 | 172 (4.5%) | 16 (1.3%) | | 188 (3.8%) |
| 25–44 | 374 (9.8%) | 330 (28%) | | 704 (14%) |
| 45–64 | 1,607 (42%) | 704 (59%) | | 2,311 (46%) |
| ≥65 | 1,645 (43%) | 145 (12%) | <0.001 | 1,790 (36%) |
| Highest Attained Formal Education | | | | |
| Primary or less | 2,445 (68%) | 547 (49%) | | 2,992 (63%) |
| Secondary | 1,046 (29%) | 513 (46%) | | 1,559 (33%) |
| Post-secondary | 101 (2.8%) | 64 (5.7%) | <0.001 | 165 (3.5%) |
| Household Wealth Tertiles | | | | |
| Low | 1,246 (34%) | 421 (36%) | | 1,667 (34%) |
| Middle | 1,262 (34%) | 397 (34%) | | 1,659 (34%) |
| High | 1,202 (32%) | 338 (29%) | 0.087 | 1,540 (32%) |
| Employment Statusc | | | | |
| Unemployed | 3,076 (84%) | 859 (74%) | | 3,935 (82%) |
| Employed part-time | 112 (3.1%) | 55 (4.7%) | | 167 (3.5%) |
| Employed full-time | 453 (12%) | 251 (22%) | <0.001 | 704 (15%) |
## Cardiometabolic profile
Among those with either diabetes or hypertension, the HIV/suppressed group tended to have a more favorable cardiometabolic profile (Table 2). They were less likely to be obese (BMI ≥30.0 kg/m2: $51\%$ [CI 48.1–$53.9\%$] vs. $56\%$ [CI 54.0–$57.3\%$]; $p \leq 0.001$) and more likely to be normoglycemic ($31\%$ [CI 28.2–$33.5\%$] vs. $24\%$ [CI 22.4–$25.1\%$]; $p \leq 0.001$) than HIV negative peers. Similarly, they had lower mean systolic BP (130.0 vs. 134.5 mmHg; $p \leq 0.001$) and less severe hypertension (stage 2: $12\%$ [CI 10.7–$14.5\%$] vs. $17\%$ [CI 15.5–$17.9\%$%]; $p \leq 0.001$). Notable, HIV-negative persons were almost twice as likely as HIV/suppressed persons to have comorbid hypertension and diabetes ($25\%$ [CI 23.3–$26.1\%$] vs. $15\%$ [CI 13.3–$17.5\%$]; $p \leq 0.001$). These observed differences were maintained in sensitivity analyses comparing HIV negative versus HIV positive in care persons (S3 Table). When analyses were restricted to HIV/ART persons only, successful versus failing ART was associated with central obesity (increase waist circumference: $75\%$ vs. $65\%$; $$p \leq 0.028$$) and more severe hypertension (stage 2: $12\%$ vs. $5.5\%$; $$p \leq 0.017$$) (S4 Table).
**Table 2**
| Characteristica | HIV Negative (n = 3,798) | HIV Positive with suppressed viral load (n = 1,195) | P value | Overall (n = 4,993) |
| --- | --- | --- | --- | --- |
| Obesity | | | | |
| Mean BMI (kg/m2) | 31.2 (26.0–36.9) | 30.1 (25.1–35.6) | <0.001 | 30.8 (25.8–36.5) |
| Underweight | 74 (2.0%) | 31 (2.6%) | | 105 (2.2%) |
| Normal | 686 (19%) | 257 (22%) | | 943 (19%) |
| Overweight | 873 (24%) | 291 (25%) | | 1,164 (24%) |
| Obese | 2,050 (56%) | 602 (51%) | | 2,652 (55%) |
| Mean Waist Circumference (cm) | 98.0 (86.0–109.0) | 31 (2.6%) | 0.018 | 105 (2.2%) |
| Increased | 2,461 (65%) | 751 (63%) | | 3,212 (65%) |
| Diabetes Mellitus | | | | |
| Mean HbA1c (%) | 6.0 (5.7–6.7) | 5.9 (5.6–6.4) | <0.001 | 6.0 (5.6–6.6) |
| Normal (<5.7%) | 901 (24%) | 368 (31%) | | 1,269 (25%) |
| Pre-diabetic (5.7–6.4%) | 1,633 (43%) | 529 (44%) | | 2,162 (43%) |
| Raised (≥6.5%) | 1,263 (33%) | 298 (25%) | <0.001 | 1,561 (31%) |
| Current diabetes mellitusc | 1,352 (36%) | 326 (27%) | <0.001 | 1,678 (34%) |
| Hypertension | | | | |
| Mean Systolic BP (mmHg) | 134.5 (122.0–148.5) | 130.0 (119.0–143.0) | <0.001 | 133.5 (121.0–147.0) |
| Mean Diastolic BP (mmHg) | 81.5 (73.5–91.0) | 82.5 (74.5–91.5) | 0.028 | 82.0 (73.5–91.0) |
| Normal | 647 (17%) | 232 (19%) | | 879 (18%) |
| Pre-hypertension | 1,156 (30%) | 395 (33%) | | 1,551 (31%) |
| Stage 1 hypertension | 1,360 (36%) | 419 (35%) | | 1,779 (36%) |
| Stage 2 hypertension | 633 (17%) | 149 (12%) | 0.002 | 782 (16%) |
| Current hypertensiond | 3,382 (89%) | 1,052 (88%) | 0.300 | 4,434 (89%) |
| Smoking | | | | |
| Never | 3,586 (94%) | 1,128 (94%) | | 4,714 (94%) |
| Former | 43 (1.1%) | 11 (0.9%) | | 54 (1.1%) |
| Current | 169 (4.4%) | 56 (4.7%) | | 225 (4.5%) |
| Previous CVDe | 229 (6.0%) | 1,128 (94%) | 0.600 | 4,714 (94%) |
| Comorbidity (hypertension AND diabetes) | | | | |
| Comorbidity | 936 (25%) | 183 (15%) | <0.001 | 1,119 (22%) |
| HIV Disease | | | | |
| Current CD4+ count (cells/mL)b | - | 763.0 (562.0–979.0) | - | |
## Diabetes mellitus care cascade
After adjustment for cardiovascular disease risk factors, being HIV/suppressed, relative to being HIV negative, was associated with a lower adjusted prevalence difference of being aware (diff = -$6.0\%$ [$95\%$ CI -11.0, -$1.1\%$]), in care (diff = -$5.7\%$ [$95\%$ CI -10.6, -$0.8\%$]), and in treatment (diff = -$4.8\%$ [$95\%$ CI -9.7, $0.1\%$]), for diabetes (Fig 1). Diabetes control prevalence, however, was greater in HIV/suppressed ($9.1\%$[CI 6.3–$12.0\%$]) than HIV negative individuals ($5.9\%$ [CI 4.9–$7.0\%$; adjusted; $$p \leq 0.04$$]) (Fig 1B). Despite the comparatively poorer diabetes care indices, HIV/suppressed persons attained lower mean HbA1c% than their HIV negative peers within each aspect of the care cascade (aware: 8.2 HbA1c% [CI 7.6–$8.7\%$] vs. 9.5 HbA1c%[CI 9.3–$9.8\%$]; in care: 8.2 HbA1c%[CI 7.6–$8.7\%$] vs. 9.6 HbA1c%[CI 9.3–$9.8\%$] and treated: 8.2 HbA1c%[CI 7.7–$8.8\%$] vs. 9.5 HbA1c%[CI 9.2–$9.7\%$]; all $p \leq 0.001$) (Fig 2). Overall, population coverage of diabetes care was very low with, approximately $6.8\%$ [$95\%$CI 5.9–$7.8\%$] persons with diabetes mellitus achieving disease control. Socio-demographic and behavioral factors, like age, household wealth or smoking were not correlated with diabetes control in multivariable models (Table 3).
**Fig 1:** *The cascade of diabetes care among HIV negative versus HIV positive/successful ART adults in uMkhanyakude, KwaZulu-Natal, South Africa, according to HIV/ART status.A. Minimally adjusted diabetes care cascade. * Estimates adjusted for age and sex only and include inverse probability of sampling weights. B. Fully adjusted diabetes care cascade. *Estimates adjusted for age, sex, BMI, wealth tertile, education and smoking status, and include inverse probability of sampling weights.* **Fig 2:** *Mean glycated hemoglobin (HbA1c) among adults with hypertension in uMkhanyakude, KwaZulu-Natal, South Africa, according to HIV/ART status.A. Minimally adjusted predicted mean HbA1c. * Estimates adjusted for age and sex only and include inverse probability of sampling weights. B. Fully adjusted predicted mean HbA1c. * Estimates adjusted for age, sex, BMI, wealth tertile, education, and smoking status, and include inverse probability of sampling weights.* TABLE_PLACEHOLDER:Table 3
## Hypertension care cascade
In contrast to trends seen with diabetes care, population coverage of hypertension care was relatively higher (Fig 3). Having HIV with virologic suppression was associated with greater adjusted prevalence of being aware (diff = $7.4\%$ [CI 5.3–$9.6\%$]), in care (diff = $8.0\%$ [5.9–$10.2\%$]), and in treatment (diff = $8.4\%$ [6.1–$10.6\%$]) for hypertension relative to being HIV negative. Hypertension control prevalence was also higher for HIV/suppressed persons ($49.1\%$[46.6–$51.6\%$]) than HIV negative individuals ($40.1\%$ [38.8–$41.5\%$; adjusted; $p \leq 0.001$]) (Fig 3B). Mean systolic BP attained across the care cascade was comparable between the two groups (aware: 132.5 [130.9–134.2 mmHg] vs. 134.0 [133.1–134.9 mmHg]; in care: 132.4 [130.8–134.1 mmHg] vs. 133.6 [132.7–134.5 mmHg] and treated: 131.9[130.2–133.5 mmHg] vs. 132.9 [132.0–133.8 mmHg]; all $p \leq 0.05$) (Fig 4). Overall, $40.0\%$ [38.6–$40.8\%$] of persons with hypertension attained blood pressure control across the population. Other correlates of hypertension control in multivariable models included female sex and increasing age (Table 3).
**Fig 3:** *The cascade of hypertension care among HIV negative versus HIV positive/successful ART adults in uMkhanyakude, KwaZulu-Natal, South Africa.A. Minimally adjusted hypertension care cascade. *Estimates adjusted for age and sex only and include inverse probability of sampling weights. B. Fully adjusted hypertension care cascade. *Estimates adjusted for age, sex, BMI, education, smoking status and wealth tertile, and include inverse probability of sampling weights.* **Fig 4:** *Mean predicted systolic blood pressure (SBP) among adults with hypertension in Umkhanyakude, Kwazulu-Natal, South Africa, according to HIV/ART status.A. Minimally adjusted predicted mean SBP. * Estimates adjusted for age and sex only and include inverse probability of sampling weights. B. Fully adjusted predicted mean SBP. * Estimates adjusted for age, sex, BMI, wealth tertile, education and smoking status and include inverse probability of sampling weights.*
## Sensitivity analysis
PWH with and without viral suppression had largely similar cardiometabolic profiles (S4 Table). The exceptions were central obesity ($65\%$ vs $75\%$; $$p \leq 0.028$$) and hypertension ($81\%$ vs. $88\%$; $$p \leq 0.061$$) which were less frequent among those failing ART compared to comparators with successful treatment. We did not find meaningful differences compared to our primary results in models that included those on ART for at least six months but with detectable viral loads in the category of people with HIV (S2–S5 Figs).
## Discussion
In this large population-based study set in rural South Africa, we found little evidence suggesting that engagement in HIV care is associated with substantial improvement in hypertension or diabetes care. Enrolment in HIV care and confirmation of virologic suppression was associated with modestly higher (7.4–$9.0\%$) prevalence of hypertension awareness, diagnosis, and treatment compared to HIV-uninfected comparators. With the exception of disease control, which was improved by less than $5\%$, indicators for diabetes care, however, were poorer among people with HIV and virologic suppression. Even for those indicators with relative improvements, these indices did not appear to translate into a substantial clinical disadvantage for PWH, based on the observation of similar mean systolic blood pressure–or lower HbA1c in the case of diabetes–between the two groups. Overall, our findings call into question the successful track record of expanding HIV care platforms to address NCDs in this setting. This is particularly evident when contrasting the overall prevalence of disease control between hypertension ($40\%$), diabetes ($6.8\%$) and HIV ($78\%$) control in this population [35], and further highlights the expansive divide between the strength of the HIV and NCD healthcare systems in rural South Africa [40, 41].
Two prior quasi-experimental studies have also examined the impact of integration of NCD services into HIV care in South Africa using interrupted time series designs. Rawat et al., [ 2018] [42] assessed the impact of integrated PHC, including HIV/ART, on hypertension and diabetes outcomes over a 4-year period in the Free State Province. Their data, covering 131 PHC public sector clinics with a catchment population of 1.5 million people, representing $54\%$ of the province, suggested potential compromise in the quality of hypertension care but not diabetes care at two years post-integration. There were fewer new hypertension patients placed on treatment two years post-integration than prior. By contrast, a smaller study of 878 individuals at 12 PHC clinics in rural Mpumalanga province that also used interrupted time series analyses found small and clinically insignificant improvements in blood pressure control over 2 years post-integration [43].
Similar to reports elsewhere [29, 44, 45], these two studies noted a number of challenges to the integration of services for multiple chronic conditions, including increased workload for over-burdened healthcare staff, suboptimal motivation, drug stock-outs and increased patient wait times, among others. While optimal BP and HbA1c control are difficult to achieve even in ideal conditions, these challenges may also explain in part why the anticipated NCD care benefits from HIV care were not observed in this setting. It is noteworthy that the bulk of available evidence supporting HIV and NCD care linkage in South Africa and SSA is drawn from cross-sectional studies whose limited generalizability, have been previously highlighted [28, 29, 46–48].
Our data reinforce the need for further research to better understand optimal strategies of chronic care service integration in the public sector. There remain key gaps in our understanding of best practices and in the optimal approaches to implementation. These additional data are essential to inform actionable recommendations towards improving NCD care outcomes in the region both for PWH and the general population.
## Strengths
The evidence base to date for leveraging HIV platforms for NCD care in SSA rests largely on facility-based studies with relatively small samples and/or highly selected participants. Our results, deriving from a large population-based study, represent an improvement on the generalizability of the evidence by adding a community-based focus on this prior work and capturing data from individuals who have not yet linked to care or dropped out of care. This is further enhanced by our incorporation of sampling probability weights (IPTW) to enable population-level estimates of the care cascade and adjust for potential selection bias or uneven odds of participation.
## Limitations
Our study also has important limitations. As a cross-sectional study, we cannot determine the timing of HIV care services in relationship to NCD care. Our data are also susceptible to disease misclassification, either masked or “white coat” hypertension, since BP readings and A1c testing for this analysis were derived from a single measurement [49, 50]. Masked hypertension, for example, may be common in this population as suggested by previous South African surveys reporting misclassification rates of up to $18\%$ [51, 52]. Similarly, care cascade indices were self-reported. The present analysis was undertaken at nearly $50\%$ enrolment of target eligible sample, potentially threatening the external validity of our results from selection bias. Lastly, our definition of NCDs was relatively narrow, limited to hypertension and diabetes, and thus overlooking increasingly important respiratory, non-AIDS cancer and mental health related morbidity [35].
## Conclusions
Engagement in HIV care with successful viral suppression was not associated with meaningful improvements in hypertension or diabetes care for PLWH in a low-income and rural district of South Africa where both HIV and NCDs are common. In fact, we found that engagement in HIV care was associated with lower prevalence of successful completion along the cascade of care for diabetes. The enormous successes of HIV care in the region demonstrate the capacity of the health system to effectively care for people with chronic disease. However, our findings add to data suggesting that much work is needed to understand the optimal design and implementation of integrating additional chronic disease services into HIV programs in rural South Africa; as well as to extend the lessons learned from the HIV care program to the general population.
## References
1. Dalal S, Beunza JJ, Volmink J, Adebamowo C, Bajunirwe F, Njelekela M. **Non-communicable diseases in sub-Saharan Africa: what we know now**. *Int J Epidemiol* (2011.0) **40** 885-901. DOI: 10.1093/ije/dyr050
2. Gouda HN, Charlson F, Sorsdahl K, Ahmadzada S, Ferrari AJ, Erskine H. **Burden of non-communicable diseases in sub-Saharan Africa, 1990–2017: results from the Global Burden of Disease Study 2017**. *Lancet Glob Health* (2019.0) **7** e1375-e87. DOI: 10.1016/S2214-109X(19)30374-2
3. 3WHO. Noncommunicable diseases. 2014 [20 January 2021]. https://www.afro.who.int/health-topics/noncommunicable-diseases.
4. Mathers CD, Loncar D. **Projections of global mortality and burden of disease from 2002 to 2030**. *PLoS Med* (2006.0) **3** e442. DOI: 10.1371/journal.pmed.0030442
5. Peck R, Mghamba J, Vanobberghen F, Kavishe B, Rugarabamu V, Smeeth L. **Preparedness of Tanzanian health facilities for outpatient primary care of hypertension and diabetes: a cross-sectional survey**. *Lancet Glob Health* (2014.0) **2** e285-92. DOI: 10.1016/S2214-109X(14)70033-6
6. Chikowe I, Mwapasa V, Kengne AP. **Analysis of rural health centres preparedness for the management of diabetic patients in Malawi**. *BMC Res Notes* (2018.0) **11** 267. DOI: 10.1186/s13104-018-3369-7
7. Bintabara D, Mpondo BCT. **Preparedness of lower-level health facilities and the associated factors for the outpatient primary care of hypertension: Evidence from Tanzanian national survey**. *PLoS One* (2018.0) **13** e0192942. DOI: 10.1371/journal.pone.0192942
8. Kibirige D, Sanya RE, Sekitoleko I, Katende D, Atuhe D, Milln J. *Availability and affordability of essential medicines and diagnostic tests for diabetes mellitus in sub-Saharan Africa: A systematic review* (2020.0). DOI: 10.21203/rs.3.rs-16103/v1
9. Haggerty JL, Reid RJ, Freeman GK, Starfield BH, Adair CE, McKendry R. **Continuity of care: a multidisciplinary review**. *BMJ* (2003.0) **327** 1219-21. DOI: 10.1136/bmj.327.7425.1219
10. Wagner EH, Davis C, Schaefer J, Von Korff M, Austin B. **A survey of leading chronic disease management programs: are they consistent with the literature?**. *Managed care quarterly* (1999.0) **7** 56-66. PMID: 10620960
11. Naithani S, Gulliford M, Morgan M. **Patients’ perceptions and experiences of ‘continuity of care’in diabetes**. *Health Expectations* (2006.0) **9** 118-29. DOI: 10.1111/j.1369-7625.2006.00379.x
12. Rabkin M, El-Sadr WM. **Why reinvent the wheel? Leveraging the lessons of HIV scale-up to confront non-communicable diseases**. *Glob Public Health* (2011.0) **6** 247-56. DOI: 10.1080/17441692.2011.552068
13. Zeltner T, Riahi F, Huber J, Groth H, May JF. *Africa’s Population: In Search of a Demographic Dividend* (2017.0) 283-97
14. Shakarishvili G, Atun R, Berman P, Hsiao W, Burgess C, Lansang MA. **Converging health systems frameworks: towards a concepts-to-actions roadmap for health systems strengthening in low and middle income countries**. *Global Health Governance* (2010.0) **3**
15. Travis P, Bennett S, Haines A, Pang T, Bhutta Z, Hyder AA. **Overcoming health-systems constraints to achieve the Millennium Development Goals**. *Lancet* (2004.0) **364** 900-6. DOI: 10.1016/S0140-6736(04)16987-0
16. De Cock KM, El-Sadr WM, Ghebreyesus TA. **Game changers: why did the scale-up of HIV treatment work despite weak health systems?**. *J Acquir Immune Defic Syndr* (2011.0) **57** S61-3. DOI: 10.1097/QAI.0b013e3182217f00
17. 17Global A. Update. Seizing the moment, tackling entrenched inequalities to end epidemics. 2020.
18. 18AVERT. HIV and AIDS in East and Southern Africa regional overview. 2020 [cited 20 January 2021]. https://www.avert.org/professionals/hiv-around-world/sub-saharan-africa/overview.
19. Ataklte F, Erqou S, Kaptoge S, Taye B, Echouffo-Tcheugui JB, Kengne AP. **Burden of undiagnosed hypertension in sub-saharan Africa: a systematic review and meta-analysis**. *Hypertension* (2015.0) **65** 291-8. DOI: 10.1161/HYPERTENSIONAHA.114.04394
20. Geldsetzer P, Manne-Goehler J, Marcus ME, Ebert C, Zhumadilov Z, Wesseh CS. **The state of hypertension care in 44 low-income and middle-income countries: a cross-sectional study of nationally representative individual-level data from 1.1 million adults**. *Lancet* (2019.0) **394** 652-62. DOI: 10.1016/S0140-6736(19)30955-9
21. El-Sadr WM, Goosby E. **Building on the HIV platform: tackling the challenge of noncommunicable diseases among persons living with HIV**. *AIDS* (2018.0) **32** S1-S3. DOI: 10.1097/QAD.0000000000001886
22. 22Health SDo. Integrated Chronic Disease Management Manual, A step-by-step guide for implementation. National Department of Health South Africa; 2012.
23. Naledi T, Barron P, Schneider H. **Primary health care in SA since 1994 and implications of the new vision for PHC re-engineering**. *South African health review* (2011.0) **2011** 17-28
24. Puoane TR, Tsolekile LP, Egbujie BA, Lewy M, Sanders D. **Advancing the agenda on noncommunicable diseases: prevention and management at community level**. *South African Health Review* (2017.0) **2017** 171-9
25. Nxumalo N, Choonara S. **A rapid assessment of ward-based PHC outreach teams in Gauteng Sedibeng district–emfuleni sub-district**. *Johannesburg: Centre for Health Policy, University of the Witwatersrand* (2014.0)
26. Schneider H, Schaay N, Dudley L, Goliath C, Qukula T. **The challenges of reshaping disease specific and care oriented community based services towards comprehensive goals: a situation appraisal in the Western Cape Province, South Africa**. *BMC Health Serv Res* (2015.0) **15** 436. DOI: 10.1186/s12913-015-1109-4
27. Manne-Goehler J, Montana L, Gomez-Olive FX, Rohr J, Harling G, Wagner RG. **The ART Advantage: Health Care Utilization for Diabetes and Hypertension in Rural South Africa**. *J Acquir Immune Defic Syndr* (2017.0) **75** 561-7. DOI: 10.1097/QAI.0000000000001445
28. Muddu M, Tusubira AK, Sharma SK, Akiteng AR, Ssinabulya I, Schwartz JI. **Integrated Hypertension and HIV Care Cascades in an HIV Treatment Program in Eastern Uganda: A Retrospective Cohort Study**. *J Acquir Immune Defic Syndr* (2019.0) **81** 552-61. DOI: 10.1097/QAI.0000000000002067
29. Pfaff C, Singano V, Akello H, Amberbir A, Berman J, Kwekwesa A. **Early experiences integrating hypertension and diabetes screening and treatment in a human immunodeficiency virus clinic in Malawi**. *Int Health* (2018.0) **10** 495-501. DOI: 10.1093/inthealth/ihy049
30. Kemp CG, Weiner BJ, Sherr KH, Kupfer LE, Cherutich PK, Wilson D. **Implementation science for integration of HIV and non-communicable disease services in sub-Saharan Africa: a systematic review**. *AIDS* (2018.0) **32** S93-S105. DOI: 10.1097/QAD.0000000000001897
31. Nugent R, Barnabas RV, Golovaty I, Osetinsky B, Roberts DA, Bisson C. **Costs and cost-effectiveness of HIV/noncommunicable disease integration in Africa: from theory to practice**. *AIDS* (2018.0) **32** S83-S92. DOI: 10.1097/QAD.0000000000001884
32. Gareta D, Baisley K, Mngomezulu T, Smit T, Khoza T, Nxumalo S. **Cohort Profile Update: Africa Centre Demographic Information System (ACDIS) and population-based HIV survey**. *Int J Epidemiol* (2021.0) **50** 33-4. DOI: 10.1093/ije/dyaa264
33. Chimbindi N, Mthiyane N, Birdthistle I, Floyd S, McGrath N, Pillay D. **Persistently high incidence of HIV and poor service uptake in adolescent girls and young women in rural KwaZulu-Natal, South Africa prior to DREAMS**. *PLoS One* (2018.0) **13** e0203193. DOI: 10.1371/journal.pone.0203193
34. Kharsany ABM, Cawood C, Khanyile D, Lewis L, Grobler A, Puren A. **Community-based HIV prevalence in KwaZulu-Natal, South Africa: results of a cross-sectional household survey**. *Lancet HIV* (2018.0) **5** e427-e37. DOI: 10.1016/S2352-3018(18)30104-8
35. Wong EB, Olivier S, Gunda R, Koole O, Surujdeen A, Gareta D. **Convergence of infectious and non-communicable disease epidemics in rural South Africa: a cross-sectional, population-based multimorbidity study**. *Lancet Glob Health* (2021.0) **9** e967-e76. DOI: 10.1016/S2214-109X(21)00176-5
36. 36The STEPS Manual. Geneva: World Health Organization, 2014.
37. American Diabetes A.. **Standards of medical care in diabetes—2012**. *Diabetes Care* (2012.0) **35** S11-63. DOI: 10.2337/dc12-s011
38. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL. **The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report**. *JAMA* (2003.0) **289** 2560-72. DOI: 10.1001/jama.289.19.2560
39. Filmer D, Pritchett LH. **Estimating wealth effects without expenditure data—or tears: an application to educational enrollments in states of India**. *Demography* (2001.0) **38** 115-32. DOI: 10.1353/dem.2001.0003
40. 40Organization WH. WHO package of essential noncommunicable (PEN) disease interventions for primary health care. 2020.
41. Demaio AR, Kragelund Nielsen K, Pinkowski Tersbol B, Kallestrup P, Meyrowitsch DW. **Primary Health Care: a strategic framework for the prevention and control of chronic non-communicable disease**. *Glob Health Action* (2014.0) **7** 24504. DOI: 10.3402/gha.v7.24504
42. Rawat A, Uebel K, Moore D, Yassi A. **Integrated HIV-Care into primary health care clinics and the influence on diabetes and hypertension care: an interrupted time series analysis in free state, South Africa over 4 years**. *JAIDS Journal of Acquired Immune Deficiency Syndromes* (2018.0) **77** 476-83. DOI: 10.1097/QAI.0000000000001633
43. Ameh S, Klipstein-Grobusch K, Musenge E, Kahn K, Tollman S, Gomez-Olive FX. **Effectiveness of an Integrated Approach to HIV and Hypertension Care in Rural South Africa: Controlled Interrupted Time-Series Analysis**. *J Acquir Immune Defic Syndr* (2017.0) **75** 472-9. DOI: 10.1097/QAI.0000000000001437
44. Topp SM, Chipukuma JM, Giganti M, Mwango LK, Chiko LM, Tambatamba-Chapula B. **Strengthening health systems at facility-level: feasibility of integrating antiretroviral therapy into primary health care services in Lusaka, Zambia**. *PloS one* (2010.0) **5** e11522. DOI: 10.1371/journal.pone.0011522
45. Yu D, Souteyrand Y, Banda MA, Kaufman J, Perriëns JH. **Investment in HIV/AIDS programs: does it help strengthen health systems in developing countries?**. *Globalization and health* (2008.0) **4** 1-10. DOI: 10.1186/1744-8603-4-8
46. Manne-Goehler J, Montana L, Gómez-Olivé FX, Rohr J, Harling G, Wagner RG. **The ART advantage: Healthcare utilization for diabetes and hypertension in rural South Africa**. *Journal of acquired immune deficiency syndromes (1999)* (2017.0) **75** 561. DOI: 10.1097/QAI.0000000000001445
47. Manne-Goehler J, Siedner MJ, Montana L, Harling G, Geldsetzer P, Rohr J. **Hypertension and diabetes control along the HIV care cascade in rural South Africa**. *J Int AIDS Soc* (2019.0) **22** e25213. DOI: 10.1002/jia2.25213
48. Edwards JK, Bygrave H, Van den Bergh R, Kizito W, Cheti E, Kosgei RJ. **HIV with non-communicable diseases in primary care in Kibera, Nairobi, Kenya: characteristics and outcomes 2010–2013**. *Trans R Soc Trop Med Hyg* (2015.0) **109** 440-6. DOI: 10.1093/trstmh/trv038
49. Seedat YK, Rayner BL, Veriava Y. **South African hypertension practice guideline 2014**. *Cardiovasc J Afr* (2014.0) **25** 288-94. DOI: 10.5830/CVJA-2014-062
50. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines**. *Journal of the American College of Cardiology* (2018.0) **71** e127-e248. DOI: 10.1016/j.jacc.2017.11.006
51. Thompson JE, Smith W, Ware LJ, C MCM, van Rooyen JM, Huisman HW. **Masked hypertension and its associated cardiovascular risk in young individuals: the African-PREDICT study**. *Hypertens Res* (2016.0) **39** 158-65. DOI: 10.1038/hr.2015.123
52. Ware LJ, Rennie KL, Gafane LF, Nell TM, Thompson JE, Van Rooyen JM. **Masked Hypertension in Low-Income South African Adults**. *J Clin Hypertens (Greenwich)* (2016.0) **18** 396-404. DOI: 10.1111/jch.12768
|
---
title: 'The impact of IOM recommendations on gestational weight gain among US women:
An analysis of birth records during 2011–2019'
authors:
- Vidhura S. Tennekoon
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021552
doi: 10.1371/journal.pgph.0000815
license: CC BY 4.0
---
# The impact of IOM recommendations on gestational weight gain among US women: An analysis of birth records during 2011–2019
## Abstract
The prevailing guidelines of the Institute of Medicine (IOM) of United States on gestational weight gain (GWG) are based on women’s prepregnancy body mass index (BMI) categories. Previous research has shown that the guidelines issued in 1990 and revised in 2009 had no effect. We investigate the effectiveness of new guidelines issued in 2009 analyzing the records of all singleton births in the U.S. during 2011–2019 (34.0 million observations). We use the discontinuity in recommended guidelines at the threshold values of BMI categories in a regression discontinuity (RD) research design to investigate the effect of IOM guidelines on GWG. We also use an RD analysis in a difference in difference (DID) framework where we compare the effect on women who had any prenatal care to others who did not receive prenatal care. The naïve RD estimator predicts an effect in the expected direction at the threshold BMI values of 18.5 and 25.0 but not at 30.0. After the DID based correction, the RD analyses show that the GWG, measured in kg, drop at the BMI values of 18.5, 25.0 and 30.0 by 0.189 [CI: 0.341, 0.037], 0.085 [CI: 0.179, -0.009] and 0.200 [CI: 0.328, 0.072] respectively when the midpoint of the recommended range in kg drops by 1.5, 4.5 and 2.25. This implies a responsiveness of $12.6\%$, $1.9\%$ and $8.9\%$ respectively to changes in guidelines at these BMI values. The findings show that the national guidelines have induced some behavioral changes among US women during their pregnancy resulting in a change in GWG in the expected direction. However, the magnitude of the change has not been large compared to the expectations, implying that the existing mechanisms to implement these guidelines have not been sufficiently strong.
## Introduction
Improving the well-being of mothers, infants, and children is an important public health goal. Over the past half century, the guidance provided to women before, during, and after pregnancy to achieve this goal has changed dynamically when public health trends change. A positive relationship between gestational weight gain (GWG) and birth weight was identified in a report Maternal Nutrition and the Course of Pregnancy published by the Committee on Maternal Nutrition of National Research Council (NRC) in 1970 which recommended an average GWG of 20–25 pounds (9.1–11.3 kg) against the previous recommendation of 10–14 pounds (4.5–6.4 kg) [1]. During the 20 years since publishing this report, several additional studies enhanced the existing knowledge on the effect of nutrition during pregnancy and the perinatal period [2, 3]. During the same period, the Supplemental Food Program for Women, Infants, and Children (WIC) was established in the U.S. Department of Agriculture as a permanent program with the aim of improving the health of pregnant mothers, infants, and children. During the period from 1971 to 1980, mean birth weight increased by approximately 60g for whites and 30g for blacks, prevalence of low birth weight was reduced by about $20\%$ for whites and $7\%$ for blacks, and the prevalence of high birth weight increased by $30\%$ for whites and by $15\%$ for blacks, probably due to increased prepregnancy weight, increased height, decreased smoking during pregnancy, increased participation in the WIC program, and earlier prenatal care [4]. These changes in birth outcomes motivated the revised IOM guidelines introduced in 1990, which were based on BMI categories (underweight, normal weight, overweight, and obese) unlike the then existed uniform guidelines (see Table 1).
**Table 1**
| Prepregnancy BMI (kg/m2) | Recommended GWG in lbs. | Recommended GWG in kg |
| --- | --- | --- |
| During 1970–2009 | | |
| All BMI categories | 20–25 | 9.1–11.3 |
| During 1990–2009 | | |
| ≤ 19.7 (underweight) | 28–40 | 12.5–18.0 |
| 19.8–26.0 (normal weight) | 25–35 | 11.5–16.0 |
| 26.1–30.0 (overweight) | 15–25 | 7.0–11.5 |
| ≥ 29.1 (obese) | ≥ 15.0 | ≥ 6.8 |
| From 2009 | | |
| ≤ 18.4 (underweight) | 28–40 | 12.5–18.0 |
| 18.5–24.9 (normal weight) | 25–35 | 11.5–16.0 |
| 25.0–29.9 (overweight) | 15–25 | 7.0–11.5 |
| ≥ 30.0 (obese) | 11–20 | 5.0–9.0 |
Trends in pregnancy and birth related outcomes continued to change during the 19 years after introducing the 1990 guidelines. The population of U.S. women of childbearing age became more diverse while the average age of a pregnant woman increased. The percentage of women who were overweight or obese when entering pregnancy increased gradually. These trends and newer research prompted to revise the 1990 guidelines and issue new guidelines in 2009. The 2009 IOM guidelines for singleton births which are still in use are presented in Table 1, together with 1970 and 1990 guidelines [5].
The new guidelines also were based on BMI categories, like the previous guidelines. The main difference between the guidelines issued in 1990 and 2009 was on how BMI categories were defined. The cutoff points used to separate the four BMI categories in 1990 corresponded to $90\%$, $120\%$, and $135\%$ of the weight-for-height standards prepared by the Metropolitan Life Insurance Company and most widely used in the US at that time [6]. A few years after issuing the 1990 IOM recommendations, the World Health Organization (WHO) introduced a different set of cutoff points to define BMI categories which were subsequently endorsed by the National Institutes of Health [7, 8]. The BMI categories of WHO were widely used in the US and internationally by the time when 2009 IOM recommendations were prepared, and the new guidelines were based on those categories. The recommendation for underweight, normal weight and overweight categories in 2009 were the same as those prevailed since 1990 but the cutoff point between the first two categories was shifted from 19.8 to 18.5 and the cutoff point separating the next two categories was shifted from 26.0 to 25.0. The lower limit of the obese category was moved from 29.0 to 30.0. In addition, the recommended range of GWG for that category was changed to 11–20 pounds from the previous recommendation of over 15 pounds with no upper limit.
When the 2009 IOM guidelines were prepared, the effect of GWG on a broad spectrum of outcomes including pregnancy and birth outcomes, neonatal outcomes, postpartum outcomes as well as the long-term health consequences affecting both mother and the child were considered. Increases in body weight during pregnancy involve both maternal components including RBC mass, body water, fat, and uterine and breast tissue; and the products of conception, the fetus, placenta, and amniotic fluid [9]. Maternal components contribute to approximately $65\%$ of GWG; the rest is products of conception [10]. Changes in body weight, fat mass, and fat-free mass during pregnancy reflect changes in maternal nutritional status throughout this nutritionally challenging period [10, 11]. Maternal prepregnancy weight and GWG are strong predictors of intrauterine growth and birth weight which influences the growth and survival of an infant [12]. The multifaceted process of transmitting a woman’s dietary intake during pregnancy towards the birth weight of an infant is moderated through GWG patterns [11].
The effectiveness of IOM guidelines depends crucially on the adherence to these guidelines by pregnant women. Recognizing this, the IOM/NRC committee recommended ways to promote adoption of these GWG guidelines through consumer education, strategies to assist practitioners, and public health strategies [5]. In 2013, IOM and NRC published implementation guidelines for their 2009 GWG recommendations which prescribed practitioners to record prepregnancy BMI, chart weight gain throughout pregnancy, share the results with the patient and provide counselling, among other things [13].
The IOM guidelines introduced in 2009 have been adopted by many other countries even though the report explicitly mentioned that they are intended for use among women in the United States. A comparative study of national GWG guidelines published in 2012 found that the guidelines of about half of the countries they studied were similar to 2009 IOM guidelines [14]. Benefits of following these guidelines are documented for US women and infants as well as for populations outside the US [15–19]. Despite the wider acceptance of these guidelines among researchers and medical practitioners, its acceptance within the targeted audience, the pregnant women, has not been satisfactory as indicated by the low rate of compliance with these guidelines. A meta-analysis based on more than 1.3 million women in 23 studies found that only $30\%$ of women in the respective samples had a GWG within the ranges of IOM recommendations [16]. Data from US birth records confirm this number. A part of the problem could be due to weaknesses in implementing these guidelines. Some researchers have pointed out that the existing system of healthcare delivery does not always adhere to evidence based best practice guidelines [20–22]. According to one study, healthcare providers typically provide GWG information early in pregnancy, but not again unless there is a concern [23]. In addition, a review of recent literature on GWG highlights many inconsistencies in the interpretation and application of 2009 IOM guidelines [24]. Even if the existing healthcare delivery system is efficient, inducing behavioral changes that would impact GWG through guidelines alone is not an easy task. At the first place, behaviors of a women influencing her GWG such as calory intake and physical exercise is a utility maximizing combination of choices from a behavioral economic perspective and may not necessarily reflect a knowledge gap. A recommendation, unlike a financial incentive or a penalty, can alter behaviors only when there is a knowledge gap due to an information asymmetry. Therefore, the adherence to a guideline is not guaranteed even when a woman is fully aware of the evidence-based benefits.
In this study, we use US birth records during 2001–2019 to investigate the effectiveness of 2009 IOM guidelines. The only existing work with a comprehensive evaluation of IOM guidelines on GWG is Hamad, Cohen and Rehkopf [2016], HCR hereafter [25]. Using a quasi-experimental study HCR show that 1990 IOM guidelines on GWG had no effect on pregnant women. No study has evaluated the effectiveness of IOM guidelines after the 2009 revision in a broader way as HCR analyzed the 1990 guidelines. That study, however, is based on a sample of 4,173 women (7,442 pregnancies) from the National Longitudinal Survey of Youth who self-reported their GWG for pregnancies during 1979–2000. Their main identification strategy is based on the discontinuity in recommended GWG resulted from the introduction of new guidelines in July 1991. In other words, they investigate whether there is any discrete change in average GWG before and after implementing 1990 guidelines in a regression discontinuity (RD) framework. The authors also use a difference-in-difference (DID) strategy which also relies on the difference before and after implementing the guidelines. A limitation of this approach is that the identification strategy fails if the policy change diffused to practitioners and patients slowly and the new recommendation was not implemented across US at the same time in July 1991.
One of the key observations when the IOM guidelines were introduced in 2009 was that the data sources then available were inadequate for studying national trends in GWG. The IOM/NRC committee recommended all states to adopt the revised version of the birth certificate, which includes fields for maternal prepregnancy weight, height, delivery weight and weight at delivery, and gestational age. This recommendation was followed subsequently, and the details are available for all births registered in the US since 2011. We use records of all singleton births in the US during 2011–2019 for our analysis. This amounts to 34.0 million observations representing $10.3\%$ of the current US population. We utilize this data to see a more accurate and detailed picture of the effect of 2009 IOM guidelines on GWG. This is the first comprehensive study to investigate the effectiveness of prevailing IOM guidelines on GWG of US women. Unlike an analysis based on a representative sample survey, our analysis is free from sampling bias for the population it represents, and the large sample size allows us to accurately identify even a smaller effect. Ours also is a RD based identification strategy as HCR. However, we use discontinuity in recommended GWG at the cutoffs separating the BMI categories, rather than the discontinuity at the date of implementing these guidelines.
## Dataset
All analyses presented here are based on 2011–2019 public use US birth data files available at the Vital Statistics Online Data Portal of the National Center for Health Statistics. The combined full dataset includes the details contained in 35,229,670 birth records representing $10.7\%$ of 2019 US population. Our analyses are based on 34,021,301 singleton births which represent $96.6\%$ of these records.
## Measures
The two measures of interest in our study are prepregnancy BMI and GWG. Both these measures have been derived using the information about maternal prepregnancy weight, delivery weight, and height. US national standard birth certificate did not include information on maternal prepregnancy weight and height prior to the 2003 revision [26]. It was recommended during the 2003 revision of US national standard birth certificate that prepregnancy weight (in pounds) and height (in feet and inches) be collected at the time of delivery. The exact question asked about prepregnancy weight is “What was your prepregnancy weight, that is, your weight immediately before you became pregnant with this child?”. Information on prepregnancy weight and height are collected from the self-reported Mother’s Worksheet [27]. Weight at delivery is collected directly from the medical record using the Facility Worksheet [28]. GWG was derived from mother’s prepregnancy weight and mother’s weight at delivery and converted to grams. Prepregnancy BMI was derived from mother’s prepregnancy weight and height as, mother’s pre-pregnancy weight in pounds x 703/ (mother’s height in inches)2 rounded to one decimal place [29]. Other confounding variables we use in our study also were based on self-reported information in Mother’s Worksheet.
## The RD approach
The 2009 IOM guidelines have three thresholds at the BMI values of 18.5, 25.0 and 30.0 when BMI categories change (Table 1). At each of these threshold values the upper bound of the recommended range in kg drops by 2.0, 4.5 and 2.5; lower bound drops by 1.0, 4.5 and 2.0; and the midpoint drops by 1.5, 4.5 and 2.25 as shown in Fig 1. The idea behind the RD approach is that the relationship between GWG (outcome variable) and prepregnancy BMI value (rating variable) could potentially change at the BMI values equal to 18.5, 25.0 and 30.0 (the threshold values) because of IOM guidelines but for no other reason. Therefore, any changes in GWG at one of these threshold values can be interpreted as the causal effect of IOM guidelines in GWG at that value.
**Fig 1:** *Average GWG at each value of prepregnancy BMI and IOM guidelines.*
The first step of any RD analysis is the visual inspection of data to identify the nature of the relationship between the outcome variable and the rating variable including any clearly visible shift in the value of the outcome at the threshold values. We noticed a clear and consistent negative relationship between the two variables; average GWG drops with prepregnancy BMI in its entire range. Moreover, this relationship was non-linear. However, we did not observe any clearly visible shift in outcome at any of the three threshold values. Therefore, we moved to test whether the mean of the outcome is statistically different before and after the cutoff value in its local neighborhood. For that purpose, the mean GWG within a unit (in kg/m2) below each BMI cutoff was t-tested against the mean GWG within a unit above each BMI cutoff. The results presented in Table 2 indicate that the means are statistically different, implying a possible change at the cutoffs in response to the guidelines. The mean differences were -0.146, -0.094 and -0.350kg, respectively, around the cutoff values of 18.5, 25.0 and 30.0. However, these differences may represent the general trend in the relationship between the rating variable and the outcome variable, not necessarily a response to IOM guidelines. It also may represent differences between the two groups due to confounders.
**Table 2**
| Cutoff value (kg/m2) | Mean GWG below cutoff (kg)a | Mean GWG above cutoff (kg)b | Mean difference (kg)c |
| --- | --- | --- | --- |
| 18.5 | 14.972 | 14.826 | -0.146*** (16.650) |
| 25.0 | 14.61 | 14.515 | -0.094*** (14.831) |
| 30.0 | 12.817 | 12.497 | -0.320*** (33.643) |
The effect of IOM recommendations at the BMI cutoff values can be identified precisely using a regression model which identifies the nature of the relationship between the two variables as well as any change in that relationship at the threshold values. This approach also allows to measure the statistical accuracy of the estimated effect. In order to identify the effect of IOM recommendations on GWG at a given threshold value, we used data belonged to each couple of two adjacent categories. For example, the data belonged to underweight and normal weight categories were used to measure the effect at the BMI value of 18.5. This allows to model the relationship between GWG and BMI more precisely which we assumed to be quadratic in each of the three local regions based on the non-linear relationship between the rating variable (BMI value) and the outcome variable (GWG) we noticed in Fig 1. In the regression estimates restricted to a specific range of BMI values, the assumed quadratic relationship potentially breaks only at one value. Each regression model we estimated using OLS takes the following form where *Di is* a dummy variable which takes the value 0 at the points below the threshold and 1 at the values above the threshold.
Our parameter of interest is δ, the discontinuity at the cutoff point in response to each discrete jump in IOM guidelines. Xi is a vector of additional control variables.
## The RD approach in a DID setting
In above quasi-experimental research design, we assumed there is no reason for the relationship between GWG and prepregnancy BMI to break at the threshold values except for the change in IOM guidelines. These exact thresholds determine the BMI categories of a woman. If the identity with a specific BMI category of a woman induces any psychological effect and thereby a behavioral change which affects GWG, our identifying assumption fails. A modified RD research design can circumvent this problem. In this research design, we make the additional assumption that a woman is unaware of IOM guidelines if she never receives prenatal care during her pregnancy and therefore any estimated effect using our naïve RD design on this group represents a placebo effect. The true effect of IOM guidelines on GWG net of this placebo effect can be found by the difference in RD estimates on the group of women who received any prenatal care and others who did not. An identified weakness in HCR, which we also need to address, is the measurement error in self-reported anthropometric measurements [30, 31]. If any measurement error in self-reported anthropometric measurements were due to social desirability bias or recall bias and that bias is uncorrelated with prenatal care this DID-based RD approach also eliminates that bias by differencing [32]. The regression equation we estimated was the following and the parameter of interest was δ.
For identification, we assumed that each of the two groups have quadratic relationship between GWG and prepregnancy BMI, an assumption justified by Fig 2. We did not assume similarity of the two groups or a parallel trend.
**Fig 2:** *GWG and prepregnancy BMI with and without prenatal care.*
## Control variables
We have included age, age squared, education, marital status, race categories, smoking during pregnancy (as a proxy for risky health behaviors) and WIC participation (as a proxy for low household income) as controls. Sensitivity of results to each of these covariates was tested.
## Sample characteristics
The mean prepregnancy BMI of women was 26.7 in the full sample, a value falls within the overweight category. Only $3.3\%$ of women were in underweight category while $41.3\%$ were in normal weight category, $24.0\%$ were overweight, and $31.4\%$ were obese. Overall, $29.8\%$ of women had a GWG within the recommended range, $51.0\%$ were above and $20.1\%$ were below the recommended range. Rate of compliance with IOM guidelines decreased when prepregnancy BMI increased. Other differences and similarities between BMI categories are summarized in Table 3.
**Table 3**
| Variable (mean) | Under weight | Normal weight | Overweight | Obese |
| --- | --- | --- | --- | --- |
| GWG (kg) | 15.13 | 14.83 | 13.64 | 10.55 |
| Age (years) | 26.28 | 28.33 | 28.69 | 28.6 |
| Education | | | | |
| 8th grade or less | 0.029 | 0.03 | 0.045 | 0.039 |
| 9th through 12th grade | 0.158 | 0.103 | 0.11 | 0.116 |
| High school graduate or GED | 0.29 | 0.224 | 0.251 | 0.289 |
| Some college credit | 0.191 | 0.182 | 0.21 | 0.24 |
| Associate degree | 0.059 | 0.074 | 0.084 | 0.088 |
| Bachelor’s degree | 0.164 | 0.231 | 0.188 | 0.143 |
| Master’s degree | 0.072 | 0.109 | 0.082 | 0.057 |
| Doctorate or professional degree | 0.025 | 0.036 | 0.02 | 0.012 |
| Married | 0.495 | 0.606 | 0.57 | 0.534 |
| Race | | | | |
| White | 0.687 | 0.761 | 0.751 | 0.716 |
| Black | 0.14 | 0.118 | 0.159 | 0.207 |
| American Indian and Alaska Native | 0.007 | 0.007 | 0.01 | 0.013 |
| Asian | 0.14 | 0.089 | 0.053 | 0.031 |
| Native Hawaiian and other Pacific Islanders | 0.002 | 0.002 | 0.003 | 0.005 |
| Smoked during pregnancy | 0.129 | 0.071 | 0.071 | 0.222 |
| WIC participation | 0.451 | 0.35 | 0.416 | 0.407 |
| Number of observations | 1126929.0 | 14059319.0 | 8156377.0 | 10678676.0 |
## Regression discontinuity analyses
The naïve RD estimator predicts an effect in the expected direction at the BMI values of 18.5 and 25.0 but not at 30.0. After the DID based correction, a drop in GWG was noticed at each of the three threshold values (Table 4). The midpoint of the recommended range in kg drops by 1.5, 4.5 and 2.25 respectively at the threshold values of 18.5, 25.0 and 30. Therefore, our estimates imply a responsiveness of $12.6\%$, $1.9\%$ and $8.9\%$, respectively, at these BMI values in response to changes in guidelines. The results were robust (qualitatively similar) to the exclusion of any or all the covariates.
**Table 4**
| Threshold | RD Estimates [CI] | RD/DID Estimates [CI] |
| --- | --- | --- |
| 18.5 | -0.244*** [-0.265, -0.224] | -0.189** [-0.341, -0.037] |
| 25.0 | -0.020*** [-0.031, -0.009] | -0.085* [-0.179, 0.009] |
| 30.0 | 0.218*** [0.203, 0.233] | -0.200***[-0.328, -0.072] |
## Discussion
Maternal prepregnancy BMI and GWG independently increase various health risks [33]. During the prenatal period, pregnant women are willing to make lifestyle changes for the benefit of their offspring and they are in close contact with their healthcare providers through routine prenatal care visits [34]. Lifestyle interventions initiated during this period may yield lasting positive benefits for both mother and the child [35, 36]. The IOM guidelines are useful for monitoring GWG and intervene so that the risks of several adverse outcomes such as cesarean delivery, macrosomia, preterm birth and LBW could be minimized [37]. In this study, we examined the effect of 2009 IOM recommendations on GWG using 2011–2019 US birth records. The dataset used has over 34 million observations and includes $10.3\%$ of US population in 2019. Our estimates show that the overall responsiveness to the guidelines is not very high. At the prepregnancy BMI value of 25.0, the threshold value which separates the overweight and normal weight categories, the responsiveness is only $1.9\%$.
The only previous study (HCR) which exclusively investigates the effect of IOM recommendations on GWG is based on a survey data sample and covers a period before the 2009 revision of guidelines but uses quasi experimental research designs closely related to ours [25]. They found no statistically significant change in GWG both when the guidelines were revised in 1990 and at the BMI value which separates overweight and normal weight categories. We found statistically significant effects at each of the threshold values separating the BMI categories. While the results can be different due to the time gap between the datasets used in two studies, our results can easily be reconciled with HCR using the differences in research designs and the number of observations in each dataset.
Authors of HCR explain their null findings arguing that the recommendations, probably, were not adequately disseminated to patients and providers, citing prior research in support of their argument [38]. This argument is also supported by the facts that implementation guidelines for 2009 recommendations were not published until 2013 and the American College of Obstetricians and Gynecologists delayed endorsing these guidelines until the same year [13, 39]. Therefore, the observation that the GWG did not change immediately following the 1990 revision is no surprise. The second null finding in HCR is based on the discontinuity at the threshold value between normal and overweight categories. While we found a significant effect at that value, the magnitude of the effect was only $1.9\%$ of the change in recommended range of guidelines (0.085kg). This effect was too small to be detected using a smaller sample size. While responsiveness to the guidelines based on our RD estimates was only $1.9\%$ at the BMI value of 25.0, it was $12.6\%$ at 30.0 and $8.9\%$ at 18.5. This implies that pregnant women have been less likely to respond to sharp changes in recommended guidelines as at the BMI value of 25.0. Therefore, a gradual decline in recommended range of GWG with prepregnancy BMI could be more effective in improving compliance.
While our results show a statistically significant effect of guidelines, that effect is far less than the intended effect. Addressing this gap requires clearly identifying why the gap exists. It has been suggested that this gap could be a result of limitations in the implementation procedure and if that is the case the appropriate solution is motivating healthcare providers through some action. However, knowledge about a guideline alone does not guarantee compliance, a best example is vaccine hesitancy. A woman’s prepregnancy BMI as well as the GWG is a result of her lifestyle choices though other factors such as genetics also play a role. The only tool available for physicians and other healthcare provides to implement IOM guidelines is patient education but lifestyle choices are hard to change through education alone and may require other types of interventions. A survey of pregnant women may help to understand the role of various factors causing the gap we identify here and respond as appropriate. Only through such a well-designed survey we can identify whether most women are unaware of the guidelines, they are aware but rely more on misinformation, or they simply ignore the guidelines. If most women are aware of the guidelines and they ignore those, a survey will also help to identify the reasons underlying such behavior.
Accuracy of our estimates are limited by measurement error in data. While the delivery weight was based on clinical records, prepregnancy weight was based on self-reports which were subject to various sources of bias. Measuring the BMI of a women accurately at the time of conception is challenged by the fact that over a half of pregnancies in the US are unplanned [40]. Self-reported weight is significantly underreported and the extent of underreporting increases with BMI, even though some research finds a higher level of agreement between self-reported and clinical measurements [41–43]. Another clearly observable tendency was reporting ‘round’ numbers, multiples of five, for example [44]. Since GWG was derived using the delivery weight and prepregnancy weight, this ‘heaping issue’ may have added systemic measurement error on GWG. The measurement error in prepregnancy weight could also propagate to prepregnancy BMI and thus distort the relationship between observed prepregnancy BMI and GWG. If any bias due to this distortion has been similar among the two groups who received prenatal care and who did not, our DID strategy may have eliminated this bias but not if this assumption was violated.
While we found an effect of IOM guidelines on GWG, the effect was not very large compared to the expectations. One of the potential reasons behind this observation is limited compliance; less than $30\%$ of women in our dataset has followed the recommendations. Future research should focus on investigating the cost of noncompliance so that the cost of any intervention to improve compliance can be justified.
## References
1. 1National Research Council. Maternal Nutrition and the Course of Pregnancy. Washington, DC: National Academies Press; 1970.. *Maternal Nutrition and the Course of Pregnancy* (1970.0)
2. 2Institute of Medicine. Preventing Low Birth weight. Report of the Committee to Study the Prevention of Low Birth weight, Division of Health Promotion and Disease Prevention. National Academies Press, Washington DC; 1985.. *Report of the Committee to Study the Prevention of Low Birth weight, Division of Health Promotion and Disease Prevention* (1985.0)
3. 3Prenatal Care: Reaching Mothers, Reaching Infants. Report of the Committee to Study Outreach for Prenatal Care, Division of Health Promotion and Disease Prevention. National Academies Press, Washington DC; 1988. https://nap.nationalacademies.org/catalog/731/prenatal-care-reaching-mothers-reaching-infants. *Reaching Infants. Report of the Committee to Study Outreach for Prenatal Care, Division of Health Promotion and Disease Prevention* (1988.0)
4. 4Institute of Medicine. Nutrition during pregnancy: part I, weight gain: part II, nutrient supplements. Institute of Medicine: Washington, DC, USA; 1990, Report no. 0309041384.. *Nutrition during pregnancy: part I, weight gain: part II, nutrient supplements* (1990.0)
5. 5IOM (Institute of Medicine). Weight Gain During Pregnancy: Reexamining the Guidelines. The National Academies Press, Washington DC; 2009.. *Weight Gain During Pregnancy: Reexamining the Guidelines* (2009.0)
6. **New weight standards for men and women. Stat**. *Bull Metropol. Life Insur. Co* (1959.0) **40** 1-4
7. 7WHO (World Health Organization). Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee. World Health Organization Technical Report Series. 1995; 854: 1–452.. *Report of a WHO Expert Committee* (1995.0) **854** 1-452
8. Heart National. *National Institutes of Health Publication 98–4083* (1998.0)
9. Pipe NG, Smith T, Halliday D, Edmonds CJ, Williams C, Coltart TM. **Changes in fat, fat-free mass and body water in human normal pregnancy**. *Br J Obstet Gynaecol* (1979.0) **86** 929-940. DOI: 10.1111/j.1471-0528.1979.tb11240.x
10. Gernand AD, Christian P, Paul RR, Shaikh S, Labrique AB, Schulze KJ. **Maternal weight and body composition during pregnancy are associated with placental and birth weight in rural Bangladesh**. *J Nutr* (2012.0) **142** 2010-2016. DOI: 10.3945/jn.112.163634
11. Devaki G, Shobha R. **Maternal anthropometry and low birth weight: a review**. *Biomed Phamacol J* (2018.0) **11** 815-820
12. Kelly A, Kevany J, de Onis M, Shah PMA. **WHO Collaborative Study of Maternal Anthropometry and Pregnancy Outcomes**. *Int J Gynaecol Obstet* (1996.0) **53** 219-233. DOI: 10.1016/0020-7292(96)02652-5
13. 13National Research Council. Implementing Guidelines on Weight Gain and Pregnancy. Washington DC: The National Academies Press; 2013. 10.17226/18292.. *Implementing Guidelines on Weight Gain and Pregnancy* (2013.0). DOI: 10.17226/18292
14. Alavi N, Haley S, Chow K, McDonald SD. **Comparison of national gestational weight gain guidelines and energy intake recommendations**. *Obes Rev* (2013.0) **14** 68-85. DOI: 10.1111/j.1467-789X.2012.01059.x
15. Wang L, Wen L, Zheng Y, Zhou W, Mei L, Li H. **Association between gestational weight gain and pregnancy complications or adverse delivery outcomes in Chinese Han Dichorionic Twin Pregnancies: Validation of the Institute of Medicine (IOM) 2009 Guidelines**. *Med Sci Monit* (2018.0) **24** 8342. DOI: 10.12659/MSM.911784
16. Goldstein RF, Abell SK, Ranasinha S, Misso M, Boyle JA, Black MH. **Association of gestational weight gain with maternal and infant outcomes: a systematic review and meta-analysis**. *JAMA* (2017.0) **317** 2207-2225. DOI: 10.1001/jama.2017.3635
17. Goldstein RF, Abell SK, Ranasinha S, Misso ML, Boyle JA, Harrison CL. **Gestational weight gain across continents and ethnicity: systematic review and meta-analysis of maternal and infant outcomes in more than one million women**. *BMC Med* (2018.0) **16** 1-4. DOI: 10.1186/s12916-018-1128-1
18. Kominiarek MA, Saade G, Mele L, Bailit J, Reddy UM, Wapner RJ. **Association between gestational weight gain and perinatal outcomes**. *Obstet Gynecol* (2018.0) **132** 875. DOI: 10.1097/AOG.0000000000002854
19. Haugen M, Brantsæter AL, Winkvist A, Lissner L, Alexander J, Oftedal B. **Associations of pre-pregnancy body mass index and gestational weight gain with pregnancy outcome and postpartum weight retention: a prospective observational cohort study**. *BMC Pregnancy Childbirth* (2014.0) **14** 1-1. DOI: 10.1186/1471-2393-14-201
20. Grol R, Wensing M. **What drives change? Barriers to and incentives for achieving evidence‐based practice**. *Med J Aust* (2004.0) **180** S57-S60. DOI: 10.5694/j.1326-5377.2004.tb05948.x
21. Wilkinson S, Beckmann M, Donaldson E, McCray S. **Implementation of gestational weight gain guidelines-what’s more effective for ensuring weight recording in pregnancy?**. *BMC Pregnancy Childbirth* (2019.0) **19** 19. DOI: 10.1186/s12884-018-2162-x
22. Wilkinson SA, Stapleton H. **Overweight and obesity in pregnancy: The evidence–practice gap in staff knowledge, attitudes and practices**. *Aust N Z J Obstet Gynaecol* (2012.0) **52** 588-592. DOI: 10.1111/ajo.12011
23. Morris J, Nikolopoulos H, Berry T, Jain V, Vallis M, Piccinini-Vallis H. **Healthcare providers’ gestational weight gain counselling practises and the influence of knowledge and attitudes: a cross-sectional mixed methods study**. *BMJ Open* (2017.0) **7** e018527. DOI: 10.1136/bmjopen-2017-018527
24. Gilmore LA, Redman LM. **Weight gain in pregnancy and application of the 2009 IOM guidelines: toward a uniform approach**. *Obesity (Silver Spring)* (2015.0) **23** 507-511. PMID: 25521748
25. Hamad R, Cohen AK, Rehkopf DH. **Changing national guidelines is not enough: the impact of 1990 IOM recommendations on gestational weight gain among US women**. *Int J Obes (London)* (2016.0) **49** 1529-1534
26. Branum A, Kirmeyer SE, Gregory ECW. *Natl Health Stat Report; 65 (6)* (2016.0)
27. 27Mother’s Worksheet for Child’s Birth Certificate. Available from: https://www.cdc.gov/nchs/data/dvs/moms-worksheet-2016.pdf
28. 28Guide to Completing the Facility Worksheet for the Certificate of Live Birth and Report of Fetal Death. Available from: https://www.cdc.gov/nchs/data/dvs/GuidetoCompleteFacilityWks.pdf
29. 29National Center for Health Statistics. User guide to the 2019 natality public use file. Available from: https://wonder.cdc.gov/wonder/help/natality/NatalityPublicUseUserGuide2019-508.pdf.
30. Mandujano A, Huston-Presley L, Waters TP, Catalano PM. **Women’s reported weight: is there a discrepancy?**. *J Matern Fetal Neonatal Med* (2012.0) **25** 1395-1398. DOI: 10.3109/14767058.2011.636099
31. McClure CK, Bodnar LM, Ness R, Catov JM. **Accuracy of maternal recall of gestational weight gain 4 to 12 years after delivery**. *Obesity (Silver Spring)* (2011.0) **19** 1047-1053. DOI: 10.1038/oby.2010.300
32. Rosenman R, Tennekoon V, Hill LG. **Measuring bias in self-reported data**. *Int J Behav Healthc Res* (2011.0) **2** 320-332. DOI: 10.1504/IJBHR.2011.043414
33. O’Reilly JR, Reynolds RM. **The risk of maternal obesity to the long‐term health of the offspring**. *Clin Endocrinol* (2013.0) **78** 9-16
34. Smith GN, Pudwell J, Roddy M. **The Maternal Health Clinic: a new window of opportunity for early heart disease risk screening and intervention for women with pregnancy complications**. *J Obstet Gynaecol Canada* (2013.0) **35** 831-839. DOI: 10.1016/S1701-2163(15)30841-0
35. Rodgers GP, Collins FS. **The next generation of obesity research: no time to waste**. *JAMA* (2012.0) **308** 1095-1096. DOI: 10.1001/2012.jama.11853
36. Gillman MW, Ludwig DS. **How early should obesity prevention start?**. *New Eng J Med* (2013.0) **369** 2173-2175. DOI: 10.1056/NEJMp1310577
37. Asvanarunat E.. **Outcomes of gestational weight gain outside the Institute of Medicine Guidelines**. *J Med Assoc Thai* (2014.0) **97** 1119-1125. PMID: 25675675
38. Herring SJ, Platek DN, Elliott P, Riley LE, Stuebe AM, Oken E. **Addressing obesity in pregnancy: what do obstetric providers recommend?**. *J Womens Health* (2010.0) **19** 65-70. DOI: 10.1089/jwh.2008.1343
39. **Weight gain during pregnancy: committee opinion no. 548**. *Obstet Gynecol* (2013.0) **121** 210-212. PMID: 23262962
40. Finer LB, Zolna MR. **Shifts in intended and unintended pregnancies in the United States, 2001‐2008**. *Am J Public Health* (2014.0) **104** S43-S48. DOI: 10.2105/AJPH.2013.301416
41. Stewart AL. **The reliability and validity of self‐reported weight and height**. *J Chronic Dis* (1982.0) **35** 295-309. DOI: 10.1016/0021-9681(82)90085-6
42. Thomas D, Halawani M, Phelan S, Butte NF, Redman LM. **Prediction of pre‐pregnancy weight from first trimester visit**. *J Fed Am Soc Exp Biol* (2014.0) **28** 1031-1032
43. Phelan S, Phipps MG, Abrams B, Darroch F, Schaffner A, Wing RR. **Randomized trial of a behavioral intervention to prevent excessive gestational weight gain: the Fit for Delivery Study**. *Am J Clin Nutr* (2011.0) **93** 772-779. DOI: 10.3945/ajcn.110.005306
44. Dharmalingam A, Navaneetham K, Krishnakumar CS. **Nutritional status of mothers and low birth weight in India**. *Matern Child Health J* (2010.0) **14** 290-298. DOI: 10.1007/s10995-009-0451-8
|
---
title: 'Maternal and paternal employment in agriculture and early childhood development:
A cross-sectional analysis of Demographic and Health Survey data'
authors:
- Lilia Bliznashka
- Joshua Jeong
- Lindsay M. Jaacks
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021554
doi: 10.1371/journal.pgph.0001116
license: CC BY 4.0
---
# Maternal and paternal employment in agriculture and early childhood development: A cross-sectional analysis of Demographic and Health Survey data
## Abstract
Considerable literature from low- and lower-middle-income countries (LLMICs) links maternal employment to child nutritional status. However, less is known about the role of parental employment and occupation type in shaping child development outcomes. Additionally, little empirical work has examined the mechanisms through which parental occupation influences child outcomes. Our objective was to investigate the associations between maternal and paternal employment (comparing agricultural and non-agricultural employment) and child development and to examine childcare practices and women’s empowerment as potential mechanisms. We pooled nine Demographic and Health Surveys (Benin, Burundi, Cambodia, Congo, Haiti, Rwanda, Senegal, Togo, and Uganda) with data on 8,516 children aged 36–59 months. We used generalised linear models to estimate associations between parental employment and child development, child stimulation (number of activities provided by the mother, father, and other household members), child supervision (not left alone or with older child for >1 hour), early childhood care and education programme (ECCE) attendance, and women’s empowerment. In our sample, all fathers and $85\%$ of mothers were employed. In $40\%$ of families, both parents were employed in agriculture. After adjusting for child, parental and household confounders, we found that parental agricultural employment, relative to non-agricultural employment, was associated with poorer child development (relative risk (RR) 0.86 ($95\%$ CI 0.80, 0.92), more child stimulation provided by other household members (mean difference (MD) 0.26 ($95\%$ CI 0.09, 0.42)), less adequate child supervision (RR, 0.83 ($95\%$ 0.78, 0.80)), less ECCE attendance (RR 0.46 ($95\%$ CI 0.39, 0.54)), and lower women’s empowerment (MD -1.01 ($95\%$ CI -1.18, -0.84)). Parental agricultural employment may be an important risk factor for early childhood development. More research using more comprehensive exposure and outcome measures is needed to unpack these complex relationships and to inform interventions and policies to support working parents in the agricultural sector with young children.
## 1. Introduction
In low- and lower-middle income countries (LLMICs), up to $40\%$ of children are developmentally off-track [1]. Improving child development in early life can improve adult educational, labour, and health outcomes [2, 3]. Prior literature has examined many biopsychosocial (e.g., nutritional deficiencies, suboptimal childcare practices) and contextual (e.g., societal violence) risk factors for poor child development [4]. However, parental employment, which affects caregivers’ capacities and children’s ecological environments, remains understudied in relation to child development even though $61\%$ of adults in LLMICs participate in the labour force [5].
## 1.1. Parental employment and child nutritional outcomes
To the best of our knowledge, no studies to date have examined the associations between parental occupation and child development in LLMICs. Instead, numerous studies have examined the association between maternal employment and child nutritional status and yielded mixed results. In some cases, maternal employment is associated with improved child nutritional outcomes [6–8], whereas in other cases maternal employment is associated with poor child nutritional outcomes [9–14]. Still other studies have found no significant associations between maternal employment and child nutritional outcomes [15–17]. Similarly, findings with respect to maternal occupation type are also mixed. In some contexts, maternal agricultural and manual employment are associated with poor child nutritional outcomes [9], whereas in others, maternal non-agricultural employment is associated with poor child nutritional outcomes [18].
Importantly, none of these studies in LLMICs considered the associations between paternal employment or occupation type and child outcomes. This is despite prior work highlighting that fathers are often the primary breadwinner of the family, providing substantial financial resources that can support positive outcomes for all household members [19, 20]. Only a few known studies have investigated the role of both maternal and paternal employment in child outcomes [21, 22]. The emerging evidence is largely inconclusive and none of the studies considered child development as an outcome. For example, one study from the United Kingdom found that employment of either parent was associated with improved child nutritional outcomes, compared to both parents being unemployed [21]. Another study from China found that associations differed depending on which parent was employed: paternal unemployment was negatively associated with child health, while maternal unemployment was positively associated with child nutrition and health [22]. However, this nascent literature from upper-middle and high-income countries may not be generalizable to LLMICs, where employment opportunities are more informal and of lower wages [23] and where agricultural employment is the predominant type of labour ($49\%$ in LLMICs vs. $21\%$ in upper-middle income countries vs. $3\%$ in high-income countries [5]).
These mixed findings are not surprising given the complexity of the relationships between parental occupation and child nutritional outcomes, varying study methodologies, and the competing mechanisms through which parental employment and occupation type can influence child outcomes (see Section 1.2). Given that family and ecological caregiving environments are similar for child nutrition and development [24], associations between parental occupation and child development are plausible and equally important to unpack.
## 1.2. Hypothesised mechanisms
Parental occupation can influence child development through five potential mechanisms: [1] household income, [2] women’s empowerment, [3] childcare, [4] parental physical and mental health, and [5] pesticide exposure. The relative importance of these mechanisms likely varies by occupation type. In LLMICs, agricultural employment is common with $60\%$ of working adults ($63\%$ of women, $57\%$ of men) in low-income countries employed in agriculture and $38\%$ ($42\%$ of women, $36\%$ of men) in lower-middle income countries [5].
First, parental employment increases household income [11, 25], and thus the availability of additional resources to allocate towards child health, nutrition, and development [26, 27]. Families with higher income may have more and better access to child healthcare services in times of child illness [28] and financially provide more enriching opportunities for early learning like preschool fees and toys [1].
Second, parental employment can improve child development by increasing women’s empowerment. Extensive empirical literature from LLMICs has demonstrated that women’s employment is associated with greater women’s empowerment [29–31]. In fact, women’s employment, seasonality of employment, and type of remuneration (cash vs. in-kind) are often included as indicators in composite measures of women’s empowerment [27, 32, 33]. Further, evidence suggests that women’s non-agricultural employment is associated with greater women’s empowerment relative to agricultural employment [34–36], likely because non-agricultural employment allows women to learn non-farm skills and exposes them to knowledge and information which can improve their household decision-making [34]. More empowered women allocate more resources towards their children, which in turn support better child development [27, 37]. Of note, prior literature on women’s empowerment has largely been in the context of women’s employment, despite scholars highlighting the need for research on employment and empowerment to include both men and women [38].
Third, parental employment can influence child development by impacting childcare arrangements, such as reducing the amount of time parents spend with their children and/or increasing reliance on alternative caregivers within (e.g., other adults or children) or outside (e.g., preschool) the household [7, 11, 39]. Given the morning and seasonal nature of farming, parents may have limited time with their children and require alternative non-parental childcare (e.g., supervision by older siblings). Evidence indicates different time patterns in childcare between working and non-working parents [6, 40–43]. However, the relationship between parental childcare time patterns and child development is complex and can vary depending on how parents manage time trade-offs and negotiate caregiving activities [43]. Studies of women’s time use in agriculture show that women in agricultural settings face more severe time constraints and stricter trade-offs than women in non-agricultural settings [44, 45], which could adversely affect child development. Finally, considering that fathers are the primary breadwinners in many LLMICs contexts and can positively influence child development [46], there may also be interactions between maternal and paternal occupation with respect to childcare. However, evidence considering both maternal and paternal work and childcare is lacking from LLMICs.
Fourth, employed women have to balance paid work and unpaid childcare and household work, which can compromise their physical and mental health and, in turn, affect their ability to care for themselves and their children [40]. The potential adverse effects of employment on parental health are likely not unique to women, though they are probably exacerbated relative to men given employed women’s dual burden of occupational and household work [47]. Stark gender inequities in the distribution of other household responsibilities remain, with women largely responsibility for all household chores, including physically intensive chores like collecting water and firewood [40]. Moreover, parental health is likely worse among parents employed in agriculture, which is generally more physically taxing than other occupations.
Lastly, pesticide exposure is a unique mechanism through which agricultural work can influence child development. Pesticides are widely used in agricultural contexts in LLMICs, with higher exposure due to continued use of harmful pesticides banned in high-income countries and unsafe handling and application practices [48]. Children of parents engaged in agriculture can be exposed to pesticides by: [1] spending time on the farm where they may inhale pesticides directly during spraying or ingest them by touching objects and putting them or their fingers in their mouths; or [2] consuming contaminated foods or water [49, 50]. Extensive evidence has linked pesticide exposure to suboptimal child development, largely through inhibition of acetylcholinesterase activity [51, 52]. Pesticide exposure can also adversely affect parental health [53] and compromise parents’ ability to care for their child.
## 1.3. Current study
Given the high proportion of developmentally off-track children in LLMICs and employed parents, unpacking the relationships between parental occupation and child development is crucial. Understanding the strength and direction of these relationships can help inform the design and targeting of interventions to support working parents with young children. Therefore, in this paper, we used nationally representative data from nine Demographic and Health Surveys (DHS) to investigate the associations between parental employment and child development, and the role of childcare and women’s empowerment as potential mechanisms. Given the high proportion of adults employed in agriculture in LLMICs, we compare agricultural to non-agricultural employment. Based on our review of the literature and mechanisms (Sections 1.1 and 1.2), in this analysis, we tested five exploratory hypotheses. Due to the limited literature considering both maternal and paternal employment, we had no a priori hypotheses about parental sex and occupation type. Compared to non-agricultural employment, we hypothesised that parental agricultural employment is associated with: We build on existing literature by examining child development as our primary outcome and considering both maternal and paternal occupation as determinants.
## 2.1. Data and study population
We pooled data from all DHS that collected information on parental employment and child development and were publicly available as of February 2022. All DHS collect data on women’s employment in the last 12 months in the woman’s interview. For a random sub-sample of households, an adult man is also interviewed, and the same employment questions are asked during the man’s interview. A couples’ file is then generated by the DHS Program pairing women and men who are married or cohabitating. With respect to child development, this optional module is applied to the youngest child aged 36–59 months, and collects data on children’s attainment of developmental milestones, child stimulation, child supervision, and ECCE attendance. Since the child development module is optional, we could only include surveys which collected this module. However, within a DHS survey, the sub-samples of households completing the child development module and the man’s interview are not always overlapping. Therefore, not all DHS surveys collecting data on child development contributed data to our analysis. In total, we included nine DHS spanning 2011–2020: Benin, Burundi, Cambodia, Congo, Haiti, Rwanda, Senegal, Togo, and Uganda (S1 Table). These were all the countries with an overlapping sub-sample of households with data on child development and paternal employment.
## 2.2. Measures
We created two variables for parental employment and occupation in the last 12 months. First, we created a binary variable for whether each parent was employed in the last 12 months (employed vs. unemployed). Second, for those employed in the last 12 months, we created a categorical variable for whether one or both parents were employed in agriculture (occupation): [1] mother employed in agriculture, father employed in non-agriculture, [2] mother employed in non-agriculture, father employed in agriculture, [3] both parents employed in agriculture, and [4] both parents employed in non-agriculture. Non-agricultural occupations included: clerical, sales, services, professional/technical/managerial, household and domestic, skilled manual, and unskilled manual occupations. These groupings are pre-specified by the DHS Program [54].
Child development was assessed using the Early Childhood Development Index (ECDI), a population-based measure designed to assess four domains of development in children aged 36–59 months: cognitive, socio-emotional, literacy-numeracy, and physical [55]. The child’s mother reports on whether the child can perform each of ten developmental milestones. Per the ECDI guidelines, we first created binary indicators for whether children attained each one of the ten developmental milestones. We then constructed binary indicators for whether children were developmentally on-track in each domain and for whether children were overall developmentally on-track (on-track in at least three out of the four domains). We also constructed the count ECDI score as the number of milestones achieved by each child, range 0–10 [55].
Child stimulation was assessed using three indicators for the number of stimulation activities (range 0–6) provided by the mother, father, or other household members in the last three days (all based on maternal report) [56]. The six activities were: [1] reading books/looking at pictures, [2] telling stories, [3] naming/counting/drawing, [4] singing, [5] taking the child outside, and [6] playing with the child. The adequacy of child supervision over the last week was assessed using three indicators: [1] child was not left alone for >1 hour, [2] child was not left under the supervision of another child for >1 hour, and [3] child was not left alone or under the supervision of another child (referred to as “adequate supervision” for brevity) [56]. ECCE attendance was assessed using a single indicator for whether the child attended an organised learning or ECCE programme.
We assessed three dimensions of women’s empowerment: [1] access to and control over resources, [2] decision-making, and [3] attitudes towards wife beating. Factor scores for each dimension were derived from a form-invariant model using confirmatory factor analysis. We also calculated a total women’s empowerment score as the sum of the three dimensions’ factor scores. The only indicator related to women’s employment included in the derivation of the “access to and control over resources” dimension was an indicator for seasonality of employment (throughout the year vs. seasonal/occasional). Full details on the indicators comprising the domains and the derivation of the factor scores have been previously published [27].
## 2.3. Statistical analysis
We restricted the analytic sample to children aged 36–59 months with available data on child development. We merged the couples and child files to create a mother-father-child triad, and we therefore refer to women and men as mothers and fathers. Since only 97 ($1.11\%$) fathers were not employed in the last 12 months, we restricted the sample to households where the father was employed. We tested for differences between included and excluded households using a Wald test, considered significant at $p \leq 0.05.$ Excluded households with unemployed fathers were generally similar to included households with employed fathers, except for the former provided less stimulation (S2 Table). We further excluded households with missing data on parental occupation ($$n = 142$$, $1.64\%$). Women in households without data on parental occupation were less empowered and more likely to have no education compared to women in households with data on parental occupation (S2 Table). In addition, fathers in households without data on parental occupation were older and more likely to have no education than fathers in households with data on parental occupation. Excluded households were also more likely to live in an urban area and were somewhat wealthier than included households (S2 Table). The final analytic sample included 8,516 children aged 36–59 months with data on child development and maternal and paternal occupation.
We used generalised linear models to assess the associations of interest. First, given that all fathers in our sample were employed, we examined the associations between maternal employment and child development (the outcome) and maternal employment and child stimulation, child supervision, ECCE attendance, and women’s empowerment (the potential mechanisms). Then, among employed parents, we examined the associations between parental occupation and child development, and parental occupation and child stimulation, child supervision, ECCE attendance, and women’s empowerment. We used log-Poisson models for the binary variables (overall development on-track, child supervision, and ECCE attendance) and calculated unadjusted and adjusted relative risk (RR) and $95\%$ confidence intervals (CIs). For child stimulation and women’s empowerment, we used linear models and calculated unadjusted and adjusted mean differences (MD) and $95\%$ CIs. Adjusted models controlled for child age and sex, maternal age and education, paternal age and education, household size, wealth, and location (urban vs. rural). Standard errors were clustered at the primary sampling unit level. Models were weighted for representativeness using country-specific weights. Missing data on mechanisms ($$n = 64$$; $0.75\%$) were imputed using country-specific mean imputation.
In models including the four-category variable for parental occupation, we considered the category “both parents employed in non-agriculture” as the reference category. We tested for equality across exposure categories using a Wald test. In all models, the Wald tests indicated differences across exposure categories (all p-values <0.001). Given this non-equivalence, we then re-estimated the models changing the reference category to “both parents employed in agriculture” to assess whether associations differed depending on which parent was employed in agriculture. Thus, for a given outcome, the models using the categorical exposure were the same, except for this change in the reference category to assess the different associations.
Given the role of child nutrition (safe and nutritious foods) in supporting child wellbeing [57] and evidence that severe childhood malnutrition can impair child development [58], we conducted a sensitivity analysis accounting for child wasting (defined as weight-for-height Z-score <-2 SD) in the analysis. We first examined biserial correlations between child wasting, parental employment, and child development in the pooled sample and separately by country. We then included child wasting as a control variable in the adjusted models for the associations between maternal employment, parental occupation, and child development. Of note, child wasting was not available for the sample of children with data on child development and maternal and paternal occupation in Cambodia and Congo. Therefore, these sensitivity analyses were conducted on a dataset pooling the remaining seven countries.
To explore whether adjusted associations between parental occupation and child development, child stimulation, child supervision, ECCE attendance, and women’s empowerment differed across household location (urban vs. rural), household wealth, maternal and paternal education, and country income level (lower income vs. lower-middle income), we included an interaction term between the categorical exposure variable and each of these modifiers. Given the exploratory nature of these analyses, interactions were considered significant at $p \leq 0.10$ based on a Wald test. All analyses were conducted in Stata 17 [59].
## 2.4. Ethical considerations
Ethical clearance for each DHS is granted by the relative institutions in the respective country. DHS data are publicly available de-identified secondary data and thus exempt from further ethical review. Access and permission to use the data for the present analyses was granted by the DHS Program (http://www.dhsprogram.com).
## 3.1. Sample descriptives
Children in our sample were 46.6 months of age on average and $49\%$ were girls (Table 1). All fathers and $85\%$ of mothers in our sample were employed in the last 12 months. Among employed parents, $44\%$ of mothers and $53\%$ of fathers were employed in agriculture. Child development was generally poor with $40\%$ of children developmentally off-track. Childcare practices and women’s empowerment were also suboptimal. Child, maternal, paternal, and household characteristics by maternal employment and parental occupation are shown in S3 Table. Both parents were engaged in agriculture in $35\%$ of the poorest households and $3\%$ of the wealthiest households. Both parents were engaged in non-agriculture in $8\%$ of the poorest households and $38\%$ of the wealthiest households (S3 Table). Overall, households where both parents were employed in agriculture appeared poorer and less educated than households where only one or neither parent was employed in agriculture. Further, both agricultural and non-agricultural occupations differed by parental education, household wealth, and household location (S4 Table). The proportion of mothers and fathers employed in agriculture was lower among more educated individuals, wealthier households, and urban households. Respectively, the proportion of parents employed in professional/technical/managerial, sales, or skilled manual occupations was higher among more educated individuals, wealthier households, and urban households (S4 Table).
**Table 1**
| Unnamed: 0 | Mean±SD or proportion |
| --- | --- |
| Household characteristics | |
| Size | 6.81±3.71 |
| Number of children <5 y | 2.03±1.1 |
| Lives in rural area | 75.73 |
| Is in poorest quintile | 24.24 |
| Maternal characteristics | |
| Age, years | 31.49±6.37 |
| Highest level of education | |
| | 37.87 |
| Primary | 40.81 |
| Secondary or higher | 21.32 |
| Employment status | |
| Unemployed | 15.27 |
| Employed in agriculture | 44.07 |
| Employed in non-agriculture | 40.66 |
| Paternal characteristics | |
| Age, years | 37.16±7.84 |
| Highest level of education | |
| | 28.00 |
| Primary | 43.54 |
| Secondary or higher | 28.46 |
| Employment status | |
| Unemployed | 0.00 |
| Employed in agriculture | 53.39 |
| Employed in non-agriculture | 46.61 |
| Parental occupation | |
| Mother employed in agriculture; father employed in non-agriculture | 11.96 |
| Mother employed in non-agriculture; father employed in agriculture | 15.71 |
| Both parents employed in agriculture | 40.06 |
| Both parents employed in non-agriculture | 32.27 |
| Child characteristics | |
| Male | 50.83 |
| Age, months | 46.61±7.11 |
| Overall development on-track | 59.98 |
| Early Childhood Development Index Score (range 0–10) | 5.3±1.76 |
| Childcare practices | |
| Number of stimulation activities provided by | |
| Mother | 1.77±1.77 |
| Father | 0.83±1.44 |
| Other household members | 1.69±1.96 |
| Supervision in the past week | |
| Child not left alone for >1 hour | 81.64 |
| Child not left under the supervision of another child for >1 hour | 71.2 |
| Child provided adequate supervision | 64.56 |
| Child attended an early childhood education programme | 24.39 |
| Women’s empowerment | |
| Access to and control over resources | -0.04±0.92 |
| Decision-making | 0.38±0.69 |
| Attitudes towards wife-beating | -1.21±1.64 |
| Total empowerment | -0.88±2.21 |
Overall, $2.4\%$ of children in our sample were wasted, ranging from $0.6\%$ in Rwanda to $10\%$ in Senegal. In the pooled sample, child wasting was significantly correlated with maternal employment ($p \leq 0.01$); however, the magnitude of this correlation was small (-0.03). Child wasting was not correlated with child development (overall or by domain) in our sample (all p-values >0.3 and all correlations were between -0.01 and 0.01).
## 3.2. Associations between maternal employment and child outcomes
Maternal employment was not associated with child development in adjusted models (Table 2). Results were consistent in sensitivity analyses controlling for child wasting (RR 1.01 ($95\%$ CI 0.95, 1.08)). However, maternal employment was associated with more stimulation provided by fathers and other household members, but with less adequate child supervision (Table 2). Maternal employment was not associated with stimulation provided by the mother or ECCE attendance. Maternal employment was positively associated with women’s total empowerment, access to and control over resources, and decision-making (Table 2).
**Table 2**
| Unnamed: 0 | Employed vs. unemployed mothers | Employed vs. unemployed mothers.1 |
| --- | --- | --- |
| | Unadjusted | Adjusted |
| Child development | | |
| Overall development on-track† | 0.96 (0.91, 1.02) | 0.98 (0.92, 1.04) |
| Childcare practices | | |
| Number of stimulation activities provided by | | |
| Mother | -0.05 (-0.19, 0.10) | 0.04 (-0.11, 0.19) |
| Father | 0.07 (-0.04, 0.17) | 0.12 (0.02, 0.22) |
| Other household members | 0.16 (0.02, 0.31) | 0.16 (0.01, 0.30) |
| Supervision in the past week | | |
| Child not left alone for >1 hour† | 0.97 (0.93, 1.00) | 0.96 (0.92, 0.99) |
| Child not left under the supervision of another child for >1 hour† | 0.87 (0.83, 0.90) | 0.90 (0.86, 0.93) |
| Child provided adequate supervision† | 0.84 (0.80, 0.88) | 0.87 (0.83, 0.91) |
| Child attended an early childhood education programme† | 0.82 (0.73, 0.93) | 0.96 (0.86, 1.08) |
| Women’s empowerment | | |
| Access to and control over resources | 1.75 (1.71, 1.79) | 1.77 (1.73, 1.81) |
| Decision-making | 0.33 (0.27, 0.40) | 0.26 (0.21, 0.32) |
| Attitudes towards wife beating | -0.18 (-0.31, -0.05) | -0.07 (-0.20, 0.06) |
| Total empowerment | 1.90 (1.74, 2.06) | 1.96 (1.80, 2.12) |
## 3.3. Associations between parental agricultural employment and child development
Among employed parents, children whose parents were employed in agriculture relative to non-agriculture were less likely to be developmentally on-track (Table 3). Results were consistent in sensitivity analyses controlling for child wasting (RR 0.85 ($95\%$ CI 0.79, 0.92)). More specifically, overall ECDI score, cognitive, socio-emotional, and literacy-numeracy development were poorer among children whose parents were both employed in agriculture relative to non-agriculture (S5 Table). When assessing whether associations differed by which parent was employed in agriculture, we observed that children were more likely to be developmentally on-track if only one parent was employed in agriculture as compared to both, regardless of whether it was the mother or father (S6 Table). The one exception was that children were more likely to be cognitively on-track if only the father was employed in agriculture relative to both parents (S6 Table).
**Table 3**
| Unnamed: 0 | Overall development on-track | Overall development on-track.1 |
| --- | --- | --- |
| | Unadjusted RR | Adjusted RR |
| Both parents employed in non-agriculture | Ref | Ref |
| Mother employed in agriculture; father employed in non-agriculture | 0.88 (0.81, 0.95) | 0.92 (0.85, 1.00) |
| Mother employed in non-agriculture; father employed in agriculture | 0.95 (0.89, 1.01) | 1.04 (0.97, 1.11) |
| Both parents employed in agriculture | 0.77 (0.73, 0.82) | 0.86 (0.80, 0.92) |
## 3.4. Associations between parental agricultural employment and hypothesised mechanisms
With respect to the potential mechanisms we examined, parental agricultural employment relative to non-agricultural employment was not associated with stimulation by mothers or fathers (Table 4). However, stimulation by other household members was higher ($15\%$ additional activities) when both parents were employed in agriculture compared to when both parents were employed in non-agriculture (Table 4). Stimulation by other household members was significantly lower when only one parent was employed in agriculture relative to both parents (S7 Table). In addition, stimulation by mothers was higher if only the mother was employed in agriculture relative to both parents (S7 Table). Further, parental agricultural employment relative to non-agricultural employment was associated with inadequate child supervision, particularly leaving the child with an older sibling (Table 4). Parental agricultural employment was also associated with less ECCE attendance than parental non-agricultural employment. These associations with child supervision and ECCE attendance were largely driven by paternal employment in agriculture (S7 Table).
**Table 4**
| Panel A Childcare practices | Panel A Childcare practices.1 | Panel A Childcare practices.2 | Panel A Childcare practices.3 | Panel A Childcare practices.4 | Panel A Childcare practices.5 | Panel A Childcare practices.6 | Panel A Childcare practices.7 | Panel A Childcare practices.8 | Panel A Childcare practices.9 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Number of stimulation activities provided by the mother | Number of stimulation activities provided by the mother | Number of stimulation activities provided by the father | Number of stimulation activities provided by the father | Number of stimulation activities provided by other household members | Number of stimulation activities provided by other household members | Number of stimulation activities provided by other household members | | |
| | Unadjusted MD | Adjusted MD | Unadjusted MD | Adjusted MD | Unadjusted MD | Unadjusted MD | Adjusted MD | | |
| Both parents employed in non-agriculture | Ref | Ref | Ref | Ref | Ref | Ref | Ref | | |
| Mother employed in agriculture; father employed in non-agriculture | -0.16 (-0.33, 0.01) | 0.11 (-0.07, 0.286) | -0.34 (-0.47, -0.21) | 0.02 (-0.13, 0.16) | -0.05 (-0.24, 0.13) | -0.05 (-0.24, 0.13) | 0.08 (-0.11, 0.27) | | |
| Mother employed in non-agriculture; father employed in agriculture | -0.45 (-0.63, -0.28) | -0.04 (-0.22, 0.14) | -0.34 (-0.47, -0.21) | 0.00 (-0.13, 0.14) | -0.13 (-0.3, 0.04) | -0.13 (-0.3, 0.04) | -0.01 (-0.19, 0.17) | | |
| Both parents employed in agriculture | -0.50 (-0.64, -0.36) | -0.06 (-0.21, 0.09) | -0.40 (-0.51, -0.28) | -0.04 (-0.16, 0.09) | 0.11 (-0.04, 0.26) | 0.11 (-0.04, 0.26) | 0.26 (0.09, 0.42) | | |
| | Child not left alone for >1 hour in the past week | Child not left alone for >1 hour in the past week | Child not left under the supervision of another child for >1 hour in the past week | Child not left under the supervision of another child for >1 hour in the past week | Child provided adequate stimulation | Child provided adequate stimulation | Child provided adequate stimulation | Child attended an early childhood education programme | Child attended an early childhood education programme |
| | Unadjusted RR | Adjusted RR | Unadjusted RR | Adjusted RR | Unadjusted RR | Unadjusted RR | Adjusted RR | Unadjusted RR | Adjusted RR |
| Both parents employed in non-agriculture | Ref | Ref | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| Mother employed in agriculture; father employed in non-agriculture | 1.02 (0.98, 1.07) | 1.01 (0.96, 1.05) | 0.81 (0.98, 0.86) | 0.84 (0.79, 0.9) | 0.84 (0.78, 0.90) | 0.84 (0.78, 0.90) | 0.86 (0.80, 0.93) | 0.49 (0.40, 0.58) | 0.66 (0.55, 0.79) |
| Mother employed in non-agriculture; father employed in agriculture | 0.95 (0.90, 0.99) | 0.96 (0.91, 1.00) | 0.86 (0.82, 0.91) | 0.91 (0.85, 0.96) | 0.85 (0.80, 0.91) | 0.85 (0.80, 0.91) | 0.90 (0.83, 0.96) | 0.62 (0.54, 0.72) | 0.95 (0.82, 1.09) |
| Both parents employed in agriculture | 1.01 (0.97, 1.05) | 1.01 (0.97, 1.05) | 0.76 (0.73, 0.8) | 0.81 (0.76, 0.85) | 0.79 (0.75, 0.83) | 0.79 (0.75, 0.83) | 0.83 (0.78, 0.88) | 0.27 (0.24, 0.32) | 0.46 (0.39, 0.54) |
| Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment | Panel B Women’s empowerment |
| | Access to and control over resources | Access to and control over resources | Decision-making | Decision-making | Attitudes towards wife beating | Attitudes towards wife beating | Attitudes towards wife beating | Total empowerment | Total empowerment |
| | Unadjusted MD | Adjusted MD | Unadjusted MD | Adjusted MD | Unadjusted MD | Adjusted MD | Adjusted MD | Unadjusted MD | Adjusted MD |
| Both parents employed in non-agriculture | Ref | Ref | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| Mother employed in agriculture; father employed in non-agriculture | -0.48 (-0.55, -0.42) | -0.42 (-0.49, -0.35) | 0.03 (-0.04, 0.09) | -0.02 (-0.08, 0.04) | -0.96 (-1.12, -0.80) | -0.67 (-0.84, -0.51) | -0.67 (-0.84, -0.51) | -1.41 (-1.60, -1.23) | -1.11 (-1.30, -0.91) |
| Mother employed in non-agriculture; father employed in agriculture | -0.05 (-0.13, 0.02) | 0.03 (-0.05, 0.10) | -0.04 (-0.10, 0.02) | 0.01 (-0.05, 0.06) | -0.22 (-0.37, -0.07) | 0.06 (-0.10, 0.23) | 0.06 (-0.10, 0.23) | -0.31 (-0.52, -0.11) | 0.10 (-0.11, 0.31) |
| Both parents employed in agriculture | -0.48 (-0.54, -0.43) | -0.39 (-0.45, -0.33) | -0.04 (-0.09, 0.01) | -0.01 (-0.06, 0.04) | -0.95 (-1.06, -0.84) | -0.61 (-0.75, -0.47) | -0.61 (-0.75, -0.47) | -1.47 (-1.62, -1.32) | -1.01 (-1.18, -0.84) |
In addition, we found that parental agricultural employment was negatively associated with women’s empowerment relative to parental non-agricultural employment: in households where both parents were employed in agriculture, women had lower scores on total empowerment, access to and control over resources, and attitudes towards wife beating (Table 4). These associations appeared to be largely driven by maternal agricultural employment (S9 Table).
## 3.5. Heterogeneity of associations
We found that the magnitude of the associations between parental occupation (both in agriculture vs. both in non-agriculture) and child development was larger among less educated parents and in low income countries (p-values for interaction <0.10) (Fig 1, S9 Table). There was also evidence that household location (urban vs. rural), parental education, and country income level modified the associations between occupation and stimulation (p-values for interaction <0.10). We observed less beneficial associations for stimulation by the mother among rural households and more educated parents, a less beneficial association for stimulation by the father in lower-middle income countries, and a more beneficial association for stimulation by other household members among rural households and in low income countries (Fig 2, S9 Table). Further, household wealth and country income level modified the association between occupation and adequate supervision, whereas maternal education and country income level modified the association between occupation and ECCE attendance (p-values for interaction <0.10). However, no clear patterns appeared by household wealth quintile or maternal education level. With respect to country income level, associations generally appeared less beneficial among lower-middle income countries. Lastly, household location, household wealth, and country income level modified the associations between parental occupation and women’s empowerment (p-value for interaction <0.10) with larger, more negative associations among rural household, wealthier households, and lower-middle income countries (Fig 2, S9 Table).
**Fig 1:** *Heterogeneity of the associations between parental occupation and child development by household and parental characteristics, comparing households where both parents were employed in agriculture and households where both parents were employed in non-agriculture.Models adjusted for child age and sex, maternal age and education, paternal age and education, household size, wealth, and location (urban vs. rural), and accounted for representativeness. SEs were clustered at the primary sapling unit level.* **Fig 2:** *Heterogeneity of the associations between parental occupation, childcare practices, and women’s empowerment by household and parental characteristics, comparing households where both parents were employed in agriculture and households where both parents were employed in non-agriculture.Models adjusted for child age and sex, maternal age and education, paternal age and education, household size, wealth, and location (urban vs. rural), and accounted for representativeness. SEs were clustered at the primary sapling unit level.*
## 4. Discussion
In this study, we used nationally representative data from nine DHS surveys to investigate the relationship between parental employment (comparing agricultural and non-agricultural employment) and child development, and the role of childcare practices and women’s empowerment as potential mechanisms. In support of four out of our five hypotheses, we found that parental agricultural employment, relative to non-agricultural employment, was associated with poorer child development, less adequate child supervision, less ECCE attendance, and lower women’s empowerment. Contrary to our third hypothesis, we found that parental agricultural employment, relative to non-agricultural employment, was associated with more child stimulation provided by other household members. In exploratory analyses, we found that parental education, household wealth, and household location (urban vs. rural) modified the associations between parental occupation, child development, childcare practices, and women’s empowerment.
Building on prior literature on the associations between maternal employment and child nutritional outcomes, we showed that any maternal employment was not associated with child development. However, given the mixed associations between maternal employment and child nutritional outcomes [6–11, 15, 16], our results should be taken with caution until replicated using more comprehensive maternal employment measures (e.g., considering duration and hours worked) and better child development tools (e.g., based on direct assessment). With respect to the mechanisms we examined, consistent with existing literature we found that employed women had greater women’s empowerment relative to unemployed women [29–31]. We further extend the evidence base by demonstrating that maternal employment was associated with certain caregiver practices, specifically with more child stimulation provided by fathers and other household members and less adequate supervision. These findings corroborate the hypothesis that employed women have to rely on alternative caregivers within the household [7, 11, 39]. It is worth noting that all fathers in our sample were employed. These relationships between maternal employment, child development, childcare practices, and women’s empowerment may be different in families where fathers are unemployed, which may limit the generalisability of our findings to such families.
In a major extension of existing literature, our analysis demonstrated that parental agricultural employment was negatively associated with child development. Specifically, we found that children whose parents were employed in agriculture relative to non-agriculture had poorer overall, cognitive, socio-emotional, and literacy-numeracy development. As hypothesised, we observed the poorest child development outcomes when both parents were employed in agriculture vs. both employed in non-agriculture. Our analysis of potential mechanisms showed that both childcare practices and women’s empowerment can help explain these differences in child development outcomes. First, with respect to caregiver practices, parental agricultural employment was associated with more stimulation provided by other household members and less adequate child supervision. These findings indicate that families where both parents are employed in agriculture are more reliant on alternative caregivers than families where both parents are employed in non-agriculture. However, given the cross-sectional nature of our analysis, it is also possible that other family members provide more childcare so that parents can farm, or that parents who farm delegate more childcare responsibilities to other family members. The presence and availability of potential alternative caregivers is likely to influence women’s employment [39] and occupation type. Further, prior work on intergenerational transfer of childcare suggests that employed women transfer childcare to older female family members [40]. However, in our case, we were unable to determine which other family members provided more stimulation because of lack of data. It is plausible that older children provided more stimulation, a hypothesis substantiated by the increase in inadequate supervision we observed (specifically, leaving the child with an older sibling). However, given that families in our sample were relatively large (6.8 members on average), it is also possible that other family members such as mothers-in-law stepped in to provide more stimulation. Lastly, parental agricultural employment was associated with less ECCE attendance relative to non-agricultural employment. This difference may be due to differences in household wealth (with wealthier households better able to afford ECCE) or location (with better access in urban settings) [1]. However, our analysis of heterogeneity did not confirm this hypothesis.
Second, with respect to women’s empowerment, we found lower women’s empowerment scores in families where both parents were employed in agriculture relative to non-agriculture. Further, women employed in agriculture had lower women’s empowerment scores than those employed in non-agriculture, results consistent with the literature [34–36], regardless of whether their partner was employed in agriculture or not. The limited role of paternal occupation in women’s empowerment scores may be due to paternal employment status (employed vs. unemployed) being a more important determinant of women’s empowerment than paternal occupation, or to the crude exposure indicator we used. The concurrent nature of the measures we used should also be noted. Women’s empowerment is a process [60] and the cumulative and/or lagged role of paternal occupation in shaping women’s empowerment is theoretically and empirically unclear. More research is needed to better understand the role of men’s employment in shaping women’s empowerment trajectories both concurrently and over time [38].
We did not investigate household income, parental physical and mental health, and pesticide exposure as mechanisms in our analysis due to lack of data. First, DHS surveys do not collect data on household income. Instead, the DHS Program calculates a household wealth index representing fixed assets [54] that cannot easily be converted into resources for child development. Second, with respect to parental physical and mental health, DHS surveys collect data on maternal perinatal health with respect to the most recent pregnancy but do not collect data on current/general physical health, which is likely a more viable mechanism between parental employment and child development. A mental health module was introduced in DHS Phase 8 in 2020 and was thus not available for the surveys in our sample. Data on paternal physical and mental health is also lacking. Third, with respect to pesticide exposure, DHS surveys do not collect data on agricultural or occupational pesticide exposure. However, given the wide use of pesticides in LLMICs [48] and pesticides’ direct adverse effects on child development through acetylcholinesterase inhibition [51, 52, 61], we cannot rule out that pesticide exposure was a major contributor towards the poorer child development outcomes we observed among agricultural families relative to non-agricultural families. Future work should collect data on these mechanisms to better understand their contribution towards child development and to help design interventions that support children and their families. Other early life biological and nutritional mechanisms beyond our proposed framework could also be explored. For example, evidence suggests that better diet in early infancy predicts improved development later in childhood [62, 63]. It is likely that families where both parents are engaged in agriculture rely more on staple foods, resulting in poor micronutrient and protein intake for themselves and their children. However, we lacked data on dietary intake in early childhood and were therefore unable to explore this mechanism. We considered the role of child wasting in sensitivity analyses and found no evidence that child wasting influenced the associations between maternal employment, parental occupation, and child development. This may be due to the low prevalence of child wasting in our sample ($2.4\%$) or because undernutrition may be less harmful in children 36–59 months of age than in younger children [64]. Further, a growing body of evidence suggests that malnutrition in early childhood impairs cognitive, academic, and human capital outcomes later in childhood and adulthood [58, 65–67]. Due to the lack of data, we could only assess the role of concurrent child wasting. Although we found no evidence of a cross-sectional association between child wasting and development, future studies should consider previous and persistent episodes of child malnutrition, and if and how they may influence the associations between parental employment and child development.
Our analysis also uncovered some important potential modifiers of the associations between parental occupation and child development, childcare practices, and women’s empowerment. Parental education modified the associations between parental occupation and child development, with somewhat more beneficial associations among educated parents. Educated parents may provide higher quality stimulation and early learning opportunities [68, 69], which could help explain the more beneficial child development outcomes in this subgroup. With respect to childcare practices, household location (rural vs. urban), household wealth, and parental education modified associations, but no clear patterns emerged. Further, these three factors also modified the associations between parental agricultural employment and women’s empowerment. It appeared that women in wealthier households or with educated partners had lower empowerment scores than women in poorer households or with uneducated partners. However, occupation types for mothers and fathers differed across parental education, household wealth, and household location categories (with non-agricultural occupations more common among more educated individuals, wealthier households, and urban households), which made these results difficult to interpret. Finally, country income level modified all associations we examined except for the association between parental agricultural employment and number of stimulation activities provided by the mother. However, patterns were inconsistent here as well with less beneficial associations for child development and adequate supervision in low income countries and for number of stimulation activities provided by the father, ECCE attendance, and women’s empowerment in lower-middle income countries. Country income level likely reflects various country-level characteristics that influence these associations, such as the size of the agricultural sector, access to and quality of healthcare and early childhood care and education services, and gender equity. Importantly, these heterogeneity analyses were exploratory and hypothesis generating. Replication in samples adequately drawn and powered for subgroup analyses is needed before any definitive conclusions can be drawn. Future studies should examine other country-level characteristics that may moderate the associations we examined. Future work should also consider additional factors like parental use of alcohol, tobacco, and nicotine-containing products, which is prevalent in LMICs [70] and in farming populations [71], and associated with child development, particularly with poor emotional development and behavioural difficulties [72–74]. More evidence is needed on if, and how, parental substance use influences the relationship between parental occupation and child development in LLMICs. Due to data limitations on parental substance use and the fact that ECDI does not assess emotional development or behaviour, we did not include substance use in our analysis.
Among the strengths of our study was the use of nationally representative data from nine LLMICs and the large sample size. Nevertheless, several limitations are worth noting. First, the parental employment and occupation indicators we used were crude and captured limited information about parents’ work. We had no data on employment duration (e.g., how many months or seasons parents were employed for), work schedule (e.g., day, night, weekend), numbers of hours work (e.g., full-time or part-time), or ability to bring the child to work. All these factors may influence childcare practices and women’s empowerment, and in turn child development. For example, one study from Australia showed that parental nonstandard work schedules (e.g., evenings, nights, weekends) were associated with child overweight and obesity [75]. A study from Nigeria showed that children had worse nutritional outcomes when mothers did not bring them to work [11]. Studies using more comprehensive measures of parental employment encompassing all these aspects are needed to better understand the role parental employment plays in child outcomes. To help unpack additional mechanisms such as household income and parental physical and mental health, other occupation aspects should be considered including job security, regularity of payments, and working conditions.
Second, we only examined nuclear families, i.e., parents and their children. However, in many LLMICs, the concept of nuclear family extends to other household members and this is especially true for male caregivers [20, 76]. We lacked data on other family members and were unable to unpack how parental employment choices may influence or be influenced by the characteristics of other household members (e.g., age, caregiving roles, employment status). Relatedly, we largely examined mechanisms specific to mothers, i.e., maternal caregiving and empowerment. Child stimulation was the only mechanism (from the ones we examined) on which we had data for fathers, albeit the indicator was based on maternal report and thus subject to reporting bias. Given that both mothers and fathers can promote child development beyond stimulation practice alone [77], a wider consideration of other plausible mechanisms (e.g., sensitivity, positive disciplinary practices, emotional affect, perceived parenting stress) obtained directly from both parents can improve our understanding of the relationship underlying parental employment and child development [78].
Finally, given the cross-sectional nature of the data, we were unable to establish causality or the temporal order of the exposure, mechanisms, and outcomes we examined. Thus, we were unable to conduct mediation analysis, formally test the mechanisms we examined, and quantity indirect effects. More longitudinal research is needed to establish the temporal order of the variables examined here and to assess if and how childcare practices and women’s empowerment mediate the relationship between parental employment and child development. Relatedly, we could not resolve issues of endogeneity inherent in cross-sectional samples. Specifically, we could not determine the direction of the relationship between parental occupation type and parental education, household wealth, and household location. For example, we could not disentangle whether poorer parents chose to work in agriculture or whether those working in agriculture remained poorer. As a result, given that parental occupation type differed across parental education, household wealth, and household location, we were unable to conclusively demonstrate that the associations we observed were driven by parental occupation rather than by these other factors. We controlled for these factors in the adjusted models, which helped minimize confounding. However, we cannot rule out the presence of residual confounding. In addition, small sample sizes in sub-groups of households may have biased the results: both parents were engaged in agriculture in $3\%$ of the wealthiest households, whereas both parents were engaged in non-agriculture in $8\%$ of the poorest households. Replication in larger samples where more affluent and educated families are engaged in agriculture and less affluent and less educated families are engaged in non-agriculture is needed to provide more definitive support for our hypotheses.
Despite these limitations, we found suggestive evidence that maternal and paternal agricultural employment was associated with poorer child development, childcare practices, and women’s empowerment. Our paper helps improve our understanding of the role of parental agricultural employment in shaping child development outcomes, childcare practices, and women’s empowerment in LLMICs, thus filling important gaps in the literature. Nevertheless, to our knowledge, this is the first analysis of the relationship between parental agricultural employment and child development in LLMICs. Our analysis was largely exploratory and results should be taken with caution. Much research is still needed to fully unpack the complex relationships we examined and to help inform policies and interventions to support working parents with young children in LLMICs.
## Transfer Alert
This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.
## References
1. Lu C, Cuartas J, Fink G, McCoy D, Liu K, Li Z. **Inequalities in early childhood care and development in low/middle-income countries: 2010–2018.**. *BMJ Glob Heal.* (2020.0) **5** e002314. DOI: 10.1136/bmjgh-2020-002314
2. Campbell F, Conti G, Heckman JJ, Moon SH, Pinto R, Pungello E. **Early Childhood Investments Substantially Boost Adult Health**. *Science (80-)* (2014.0) **343** 1478-1485. DOI: 10.1126/science.1248429
3. Gertler P, Heckman J, Pinto R, Zanolini A, Vermeersch C, Walker S. **Labor market returns to an early childhood stimulation intervention in Jamaica**. *Science (80-)* (2014.0) **344** 998-1001. DOI: 10.1126/science.1251178
4. Black MM, Walker SP, Fernald LCH, Andersen CT, DiGirolamo AM, Lu C. **Early childhood development coming of age: science through the life course**. *Lancet* (2017.0) **389** 77-90. DOI: 10.1016/S0140-6736(16)31389-7
5. 5The World Bank. World Development Indicators. Washington D.C.: The World Bank; 2021. Available: https://datacatalog.worldbank.org/dataset/world-development-indicators. *World Development Indicators* (2021.0)
6. Lamontagne JF, Engle PL, Zeitlin MF. **Maternal employment, child care, and nutritional status of 12–18-month-old children in Managua, Nicaragua.**. *Soc Sci Med* (1998.0) **46** 403-414. DOI: 10.1016/s0277-9536(97)00184-6
7. Diiro GM, Sam AG, Kraybill D. **Heterogeneous Effects of Maternal Labor Market Participation on the Nutritional Status of Children: Empirical Evidence from Rural India.**. *Child Indic Res.* (2017.0) **10** 609-632. DOI: 10.1007/s12187-016-9378-y
8. Shuhaimi F, Muniandy ND. **The Association of Maternal Employment Status on Nutritional Status among Children in Selected Kindergartens in Selangor, Malaysia.**. *Asian J Clin Nutr.* (2012.0) **4** 53-66. DOI: 10.3923/ajcn.2012.53.66
9. Nankinga O, Kwagala B, Walakira EJ, Navaneetham K. **Maternal employment and child nutritional status in Uganda.**. *PLoS One* (2019.0) **14** e0226720. DOI: 10.1371/journal.pone.0226720
10. Toyama N, Wakai S, Nakamura Y, Arifin A. **Mother’s Working Status and Nutritional Status of Children Under the Age of 5 in Urban Low-income Community, Surabaya, Indonesia.**. *J Trop Pediatr.* (2001.0) **47** 179-181. DOI: 10.1093/tropej/47.3.179
11. Ukwuani FA, Suchindran CM. **Implications of women’s work for child nutritional status in sub-Saharan Africa: a case study of Nigeria.**. *Soc Sci Med* (2003.0) **56** 2109-2121. DOI: 10.1016/s0277-9536(02)00205-8
12. Abbi R, Christian P, Gujral S, Gopaldas T. **The Impact of Maternal Work Status on the Nutrition and Health Status of Children.**. *Food Nutr Bull.* (1991.0) **13** 1-6. DOI: 10.1177/156482659101300128
13. Glick P, Sahn DE. **Maternal Labour Supply and Child Nutrition in West Africa.**. *Oxf Bull Econ Stat.* (1998.0) **60** 325-355. DOI: 10.1111/1468-0084.00103
14. Rashad AS, Sharaf MF. **Does maternal employment affect child nutrition status? New evidence from Egypt**. *Oxford Dev Stud* (2019.0) **47** 48-62. DOI: 10.1080/13600818.2018.1497589
15. Frongillo EA, de Onis M, Hanson KMP. **Socioeconomic and Demographic Factors Are Associated with Worldwide Patterns of Stunting and Wasting of Children**. *J Nutr* (1997.0) **127** 2302-2309. DOI: 10.1093/jn/127.12.2302
16. Burroway R.. **Are all jobs created equal? A cross-national analysis of women’s employment and child malnutrition in developing countries**. *Soc Sci Res* (2017.0) **67** 1-13. DOI: 10.1016/j.ssresearch.2017.07.003
17. Mugo MG. **Impact of Parental Socioeconomic Status on Child Health Outcomes in Kenya.**. *African Dev Rev.* (2012.0) **24** 342-357. DOI: 10.1111/1467-8268.12003
18. Debela BL, Gehrke E, Qaim M. **Links between Maternal Employment and Child Nutrition in Rural Tanzania**. *Am J Agric Econ* (2021.0) **103** 812-830. DOI: 10.1111/ajae.12113
19. Tolhurst R, Amekudzi YP, Nyonator FK, Bertel Squire S, Theobald S. **“He will ask why the child gets sick so often”: The gendered dynamics of intra-household bargaining over healthcare for children with fever in the Volta Region of Ghana.**. *Soc Sci Med* (2008.0) **66** 1106-1117. DOI: 10.1016/j.socscimed.2007.11.032
20. 20Anonymous. Details omitted for double-blind reviewing. 2018.. *Details omitted for double-blind reviewing* (2018.0)
21. Hope S, Pearce A, Whitehead M, Law C. **Parental employment during early childhood and overweight at 7-years: findings from the UK Millennium Cohort Study.**. *BMC Obes* (2015.0) **2** 33. DOI: 10.1186/s40608-015-0065-1
22. Pieters J, Rawlings S. **Parental unemployment and child health in China.**. *Rev Econ Househ* (2020.0) **18** 207-237. DOI: 10.1007/s11150-019-09457-y
23. Merotto D, Weber M, Aterido R. **Pathways to Better Jobs in IDA Countries: Findings from Jobs Diagnostics.**. *Washington, DC;* (2018.0)
24. Black MM, Lutter CK, Trude ACB. **All children surviving and thriving: re-envisioning UNICEF’s conceptual framework of malnutrition**. *Lancet Glob Heal* (2020.0) **8** e766-e767. DOI: 10.1016/S2214-109X(20)30122-4
25. Brauner-Otto S, Baird S, Ghimire D. **Maternal employment and child health in Nepal: The importance of job type and timing across the child’s first five years.**. *Soc Sci Med* (2019.0) **224** 94-105. DOI: 10.1016/j.socscimed.2019.02.009
26. Santoso M V, Kerr RB, Hoddinott J, Garigipati P, Olmos S, Young SL. **Role of Women’s Empowerment in Child Nutrition Outcomes: A Systematic Review.**. *Adv Nutr.* (2019.0). DOI: 10.1093/advances/nmz056
27. 27Anonymous. Details omitted for double-blind reviewing. 2021.. *Details omitted for double-blind reviewing* (2021.0)
28. Peters DH, Garg A, Bloom G, Walker DG, Brieger WR, Hafizur Rahman M. **Poverty and Access to Health Care in Developing Countries**. *Ann N Y Acad Sci* (2008.0) **1136** 161-171. DOI: 10.1196/annals.1425.011
29. Malhotra A, Mather M. **Do Schooling and Work Empower Women in Developing Countries? Gender and Domestic Decisions in Sri Lanka.**. *Sociol Forum* (1997.0) **12** 599-630. DOI: 10.1023/A:1022126824127
30. Acharya DR, Bell JS, Simkhada P, van Teijlingen ER, Regmi PR. **Women’s autonomy in household decision-making: a demographic study in Nepal.**. *Reprod Health* (2010.0) **7** 15. DOI: 10.1186/1742-4755-7-15
31. Kabeer N, Mahmud S, Tasneem S. **The Contested Relationship Between Paid Work and Women’s Empowerment: Empirical Analysis from Bangladesh.**. *Eur J Dev Res.* (2018.0) **30** 235-251. DOI: 10.1057/s41287-017-0119-y
32. Miedema SS, Haardörfer R, Girard AW, Yount KM. **Women’s empowerment in East Africa: Development of a cross-country comparable measure**. *World Dev* (2018.0) **110** 453-464. DOI: 10.1016/j.worlddev.2018.05.031
33. Ewerling F, Lynch JW, Victora CG, van Eerdewijk A, Tyszler M, Barros AJD. **The SWPER index for women’s empowerment in Africa: development and validation of an index based on survey data**. *Lancet Glob Heal* (2017.0) **5** e916-e923. DOI: 10.1016/S2214-109X(17)30292-9
34. Maligalig R, Demont M, Umberger WJ, Peralta A. **Off-farm employment increases women’s empowerment: Evidence from rice farms in the Philippines.**. *J Rural Stud.* (2019.0) **71** 62-72. DOI: 10.1016/j.jrurstud.2019.09.002
35. Twyman J, Useche P, Deere CD. **Gendered Perceptions of Land Ownership and Agricultural Decision-making in Ecuador: Who Are the Farm Managers?**. *Land Econ.* (2015.0) **91** 479-500. DOI: 10.3368/le.91.3.479
36. Abrar ul Haq M, Akram F, Ashiq U, Raza S, Scott MW. **The employment paradox to improve women’s empowerment in Pakistan.**. *Cogent Soc Sci.* (2019.0) **5** 1707005. DOI: 10.1080/23311886.2019.1707005
37. Ewerling F, Lynch JW, Mittinty M, Raj A, Victora CG, Coll CV. **The impact of women’s empowerment on their children’s early development in 26 African countries.**. *J Glob Health.* (2020.0) 10. DOI: 10.7189/jogh.10.020406
38. Kabeer N.. **Contextualising the Economic Pathways of Women’s Empowerment: Findings from a Multi-Country Research Programme.**. *Brighton* (2011.0)
39. Connelly R, DeGraff DS, Levison D. **Women’s Employment and Child Care in Brazil.**. *Econ Dev Cult Change.* (1996.0) **44** 619-656
40. Chopra D, Zambelli E. **No Time to Rest: Women’s Lived Experiences of Balancing Paid Work and Unpaid Care Work.**. *Brighton* (2017.0)
41. Fox L, Han W-J, Ruhm C, Waldfogel J. **Time for Children: Trends in the Employment Patterns of Parents, 1967–2009.**. *Demography* (2013.0) **50** 25-49. DOI: 10.1007/s13524-012-0138-4
42. Sayer LC, Bianchi SM, Robinson JP. **Are Parents Investing Less in Children? Trends in Mothers’ and Fathers’ Time with Children**. *Am J Sociol* (2004.0) **110** 1-43. DOI: 10.1086/386270
43. Hsin A, Felfe C. **When Does Time Matter? Maternal Employment, Children’s Time With Parents, and Child Development**. *Demography* (2014.0) **51** 1867-1894. DOI: 10.1007/s13524-014-0334-5
44. Johnston D, Stevano S, Malapit HJ, Hull E, Kadiyala S. **Review: Time Use as an Explanation for the Agri-Nutrition Disconnect: Evidence from Rural Areas in Low and Middle-Income Countries.**. *Food Policy.* (2018.0) **76** 8-18. DOI: 10.1016/j.foodpol.2017.12.011
45. Komatsu H, Malapit HJL, Theis S. **How Does Women’s Time in Reproductive Work and Agriculture Affect Maternal and Child Nutrition? Evidence from Bangladesh, Cambodia, Ghana, Mozambique, and Nepal.**. *IFPRI Discuss Pap. Washington D.C* (2015.0) 1486. DOI: 10.2139/ssrn.2741272
46. 46Anonymous. Details omitted for double-blind reviewing. 2016.. *Details omitted for double-blind reviewing* (2016.0)
47. Jaffe S.. **Work Won’t Love You Back: How Devotion to Our Jobs Keeps Us Exploited, Exhausted and Alone**. *C Hurst & Co Publishers Ltd* (2021.0)
48. 48FAO. Pesticides use. Global, regional and country trends, 1990–2018. Rome; 2021. Available: https://www.fao.org/3/cb3411en/cb3411en.pdf
49. Mostafalou S, Abdollahi M. **Pesticides: an update of human exposure and toxicity**. *Arch Toxicol* (2017.0) **91** 549-599. DOI: 10.1007/s00204-016-1849-x
50. Weiss B.. **Vulnerability of children and the developing brain to neurotoxic hazards**. *Environ Health Perspect* (2000.0) **108** 375-381. DOI: 10.1289/ehp.00108s3375
51. Sapbamrer R, Hongsibsong S. **Effects of prenatal and postnatal exposure to organophosphate pesticides on child neurodevelopment in different age groups: a systematic review**. *Environ Sci Pollut Res* (2019.0) **26** 18267-18290. DOI: 10.1007/s11356-019-05126-w
52. González-Alzaga B, Lacasaña M, Aguilar-Garduño C, Rodríguez-Barranco M, Ballester F, Rebagliato M. **A systematic review of neurodevelopmental effects of prenatal and postnatal organophosphate pesticide exposure**. *Toxicol Lett* (2014.0) **230** 104-121. DOI: 10.1016/j.toxlet.2013.11.019
53. Koureas M, Tsakalof A, Tsatsakis A, Hadjichristodoulou C. **Systematic review of biomonitoring studies to determine the association between exposure to organophosphorus and pyrethroid insecticides and human health outcomes**. *Toxicol Lett* (2012.0) **210** 155-168. DOI: 10.1016/j.toxlet.2011.10.007
54. Croft TN, Marshall AMJ, Allen CK. *Rockville, Maryland, USA: ICF: Rockville* (2018.0)
55. Loizillon A, Petrowski N, Britto P, Cappa C. **Development of the Early Childhood Development Index in MICS surveys. MICS Methodological Papers, No**. *6, Data and Analytics Section, Division of Data, Research and Policy, UNICEF New York* (2017.0)
56. Kariger P, Frongillo EA, Engle P, Britto PMR, Sywulka SM, Menon P. **Indicators of Family Care for Development for Use in Multicountry Surveys.**. *J Heal Popul Nutr* (2012.0) 30. DOI: 10.3329/jhpn.v30i4.13417
57. Black MM, Trude ACB, Lutter CK. **All Children Thrive: Integration of Nutrition and Early Childhood Development**. *Annu Rev Nutr* (2020.0) **40** 375-406. DOI: 10.1146/annurev-nutr-120219-023757
58. Kirolos A, Goyheneix M, Kalmus Eliasz M, Chisala M, Lissauer S, Gladstone M. **Neurodevelopmental, cognitive, behavioural and mental health impairments following childhood malnutrition: a systematic review**. *BMJ Glob Heal* (2022.0) **7** e009330. DOI: 10.1136/bmjgh-2022-009330
59. 59StataCorp. Stata Statistical Software: Release 17. College Station, TX: StataCorp LLC; 2021.. *Stata Statistical Software: Release 17* (2021.0)
60. Resources Kabeer N.. **Agency, Achievements: Reflections on the Measurement of Women’s Empowerment**. *Dev Change* (1999.0) **30** 435-464. DOI: 10.1111/1467-7660.00125
61. Kwong TC. **Organophosphate Pesticides: Biochemistry and Clinical Toxicology.**. *Ther Drug Monit* (2002.0) **24** 144-149. DOI: 10.1097/00007691-200202000-00022
62. Pollitt E, Jahari A, Walka H. **A developmental view of the effects of an energy and micronutrient supplement in undernourished children in Indonesia.**. *Eur J Clin Nutr* (2000.0) **54** S107-13. DOI: 10.1038/sj.ejcn.1601012
63. Iannotti L, Jean Louis Dulience S, Wolff P, Cox K, Lesorogol C, Kohl P. **Nutrition factors predict earlier acquisition of motor and language milestones among young children in Haiti**. *Acta Paediatr* (2016.0) **105** e406-e411. DOI: 10.1111/apa.13483
64. Karlsson O, Kim R, Guerrero S, Hasman A, Subramanian SV. **Child wasting before and after age two years: A cross-sectional study of 94 countries.**. *eClinicalMedicine.* (2022.0) **46** 101353. DOI: 10.1016/j.eclinm.2022.101353
65. Mwene-Batu P, Bisimwa G, Ngaboyeka G, Dramaix M, Macq J, Hermans MP. **Severe acute malnutrition in childhood, chronic diseases, and human capital in adulthood in the Democratic Republic of Congo: the Lwiro Cohort Study**. *Am J Clin Nutr* (2021.0) **114** 70-79. DOI: 10.1093/ajcn/nqab034
66. Mwene-Batu P, Bisimwa G, Baguma M, Chabwine J, Bapolisi A, Chimanuka C, Denis F. **Long-term effects of severe acute malnutrition during childhood on adult cognitive, academic and behavioural development in African fragile countries: The Lwiro cohort study in Democratic Republic of the Congo**. *PLoS One* (2020.0) **15** e0244486. DOI: 10.1371/journal.pone.0244486
67. Ampaabeng SK, Tan CM. **The long-term cognitive consequences of early childhood malnutrition: The case of famine in Ghana.**. *J Health Econ* (2013.0) **32** 1013-1027. DOI: 10.1016/j.jhealeco.2013.08.001
68. 68Anonymous. Details omitted for doube-blind reviewing. 2017.. *Details omitted for doube-blind reviewing* (2017.0)
69. Sun J, Liu Y, Chen EE, Rao N, Liu H. **Factors related to parents’ engagement in cognitive and socio-emotional caregiving in developing countries: Results from Multiple Indicator Cluster Survey 3.**. *Early Child Res Q.* (2016.0) **36** 21-31. DOI: 10.1016/j.ecresq.2015.12.003
70. Xu Y, Geldsetzer P, Manne-Goehler J, Theilmann M, Marcus M-E, Zhumadilov Z. **The socioeconomic gradient of alcohol use: an analysis of nationally representative survey data from 55 low-income and middle-income countries.**. *Lancet Glob Heal.* (2022.0) **10** e1268-e1280. DOI: 10.1016/S2214-109X(22)00273-X
71. Watanabe‐Galloway S, Chasek C, Yoder AM, Bell JE. **Substance use disorders in the farming population: Scoping review.**. *J Rural Heal* (2022.0) **38** 129-150. DOI: 10.1111/jrh.12575
72. Huq T, Alexander EC, Manikam L, Jokinen T, Patil P, Benjumea D. **A Systematic Review of Household and Family Alcohol Use and Childhood Neurodevelopmental Outcomes in Low- and Middle-Income Countries**. *Child Psychiatry Hum Dev* (2021.0) **52** 1194-1217. DOI: 10.1007/s10578-020-01112-3
73. 73World Health Organization. Tobacco control to improve child health and development. 2021. Available: https://apps.who.int/iris/bitstream/handle/10665/340162/9789240022218-eng.pdf. *Tobacco control to improve child health and development* (2021.0)
74. McGrath-Morrow SA, Gorzkowski J, Groner JA, Rule AM, Wilson K, Tanski SE. **The Effects of Nicotine on Development**. *Pediatrics* (2020.0) 145. DOI: 10.1542/peds.2019-1346
75. Champion SL, Rumbold AR, Steele EJ, Giles LC, Davies MJ, Moore VM. **Parental work schedules and child overweight and obesity.**. *Int J Obes* (2012.0) **36** 573-580. DOI: 10.1038/ijo.2011.252
76. Madhavan S, Townsend NW, Garey AI. **‘Absent Breadwinners’: Father–Child Connections and Paternal Support in Rural South Africa.**. *J South Afr Stud.* (2008.0) **34** 647-663. DOI: 10.1080/03057070802259902
77. Boothby N, Mugumya F, Ritterbusch AE, Wanican J, Bangirana CA, Pizatella AD. **Ugandan households: A Study of parenting practices in three districts.**. *Child Abuse Negl.* (2017.0) **67** 157-173. DOI: 10.1016/j.chiabu.2017.02.010
78. Chatterji P, Markowitz S, Brooks-Gunn J. **Effects of early maternal employment on maternal health and well-being.**. *J Popul Econ* (2013.0) **26** 285-301. DOI: 10.1007/s00148-012-0437-5
|
---
title: 'Clinical progression and outcomes of patients hospitalized with COVID-19 in
humanitarian settings: A prospective cohort study in South Sudan and Eastern Democratic
Republic of the Congo'
authors:
- Shannon Doocy
- Iris Bollemeijer
- Eva Leidman
- Abdou Sebushishe
- Eta Ngole Mbong
- Kathleen Page
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021555
doi: 10.1371/journal.pgph.0000924
license: CC0 1.0
---
# Clinical progression and outcomes of patients hospitalized with COVID-19 in humanitarian settings: A prospective cohort study in South Sudan and Eastern Democratic Republic of the Congo
## Abstract
Little information is available on COVID-19 in Africa and virtually none is from humanitarian and more resource-constrained settings. This study characterizes hospitalized patients in the African humanitarian contexts of Juba, South Sudan and North and South Kivu in Eastern Democratic Republic of the Congo. This observational cohort was conducted between December 2020 and June 2021. Patients presenting for care at five facilities or referred from home-based care by mobile medical teams were eligible for enrollment and followed until death or recovery. Disease progression was characterized for hospitalized patients using survival analysis and mixed effects regression model to estimate survival odds for patient characteristics and treatments received. 144 COVID-19 cases enrolled as hospitalized patients were followed to recovery/death. The observed mortality proportion among hospitalized patients was $16.7\%$ (CI: 11.2–$23.3\%$); mortality was three times higher in South Sudan, where patients presented later after symptom onset and in worse conditions. Age and diabetes history were the only patient characteristics associated with decreased survival; clinical status indicators associated with decreased survival included fever, low oxygen level, elevated respiratory and pulse rates. The only therapy associated with survival was non-invasive oxygen; invasive oxygen therapies and other specialized treatments were rarely received. Improving availability of oxygen monitoring and proven COVID-19 therapies in humanitarian and resource-poor settings is critical for health equity. Customizing training to reflect availability of specific medications, therapies and operational constraints is particularly important given the range of challenges faced by providers in these settings.
## Introduction
COVID-19 has caused 4.7 million deaths globally and available data suggests a lesser impact in Africa, where 5.9 million cases and ~143,000 deaths were reported as of September, 2021 [1] Africa accounts for $14\%$ of the global population and <$3\%$ of reported COVID-19 cases and deaths [1, 2]. While younger population structure and limited testing capacity likely contribute to lower case counts and mortality, the impact of COVID-19 is almost certainly underreported, in particular considering evidence from seroprevalence surveys which suggest that $38\%$ of the population in Juba, South Sudan and $41\%$ of the population in Bukavu, a city in eastern Democratic Republic of the Congo, had been infected with COVID-19 by late 2020 [3, 4] Given slow vaccine rollout and limited health infrastructure, the continent’s 1.2 billion residents continue to face tremendous risk of infection as the COVID-19 pandemic continues.
Most African countries have low vaccination coverage, inadequate diagnostic and laboratory capacity, and at hospitals limited staffing and availability of evidence-based COVID-19 treatments such as ventilators, antivirals and monoclonal antibodies hamper quality of care. Furthermore, tertiary facilities are often overcrowded and difficult to access, particularly for rural populations. Sub-Saharan Africa has 1.2 hospital beds per 1000 population compared to $\frac{2.3}{1000}$ in all low and middle-income countries; similarly, sub-Saharan Africa has 0.2 physicians/1000 compared to $\frac{1.3}{1000}$ in low and middle-income countries [5, 6]. While there is great variation in capacity across the continent, service availability is particularly limited in the lowest income countries, many of which have a history of protracted conflict. In South Sudan (SSD), there are 0.15 physicians/1000, and a recent health system assessment indicated poor quality of care (estimates of hospital bed per capita were not available) [7, 8]. Similarly, Democratic Republic of the Congo (DRC) has only 0.1 physicians/1000 and 0.8 hospital beds/1000, which are among the lowest in the world. In both countries, health system capacity and access are inadequate [5].
Relatively little is known about the profile and outcomes of hospitalized COVID-19 patients in Africa compared to other settings. The recent African COVID-19 Critical Care Outcomes Study (ACCCOS) study enrolled >3,750 inpatients in ten African countries and estimated the hospital mortality proportion at $48.2\%$, much higher than rates observed in European, Asian, and American hospitals; elevated mortality was attributed to insufficient critical care resources [9]. This paper characterizes clinical progression and outcomes of hospitalized COVID-19 patients in Juba, South Sudan and North and South Kivu, Eastern DRC and aims to inform the COVID-19 response in resource-poor and conflict-affected settings in Africa.
## Methods
A prospective observational cohort of COVID-19 cases was conducted between December 2020 and June 2021 in five health facilities operated or supported by International Medical Corps (IMC) in Eastern DRC ($$n = 4$$) and Juba, South Sudan ($$n = 1$$), including four that provided inpatient care. Hospital characteristics and COVID-19 related care capacity among the four facilities providing inpatient care are summarized in S1 Table; laboratory and diagnostic testing, oxygen therapies and available medications varied across facilities. Both COVID-19 cases receiving inpatient and outpatient care were enrolled in the study with the primary aim of identifying risk factors for poor outcomes, including hospitalization and death as described in the companion paper [10] a secondary aim was to document clinical progression and characterize COVID-19 clinical management. Sample size calculations were conducted based on the primary aim and are presented in the risk factor paper; due to the descriptive nature of this paper, where detecting differences between countries or hypothesis testing was not an aim, additional sample size calculations were not conducted. Individuals presenting for care at a study facility or referred by mobile medical teams providing home-based care in the facility catchment area with a positive RT-PCR or antigen test and inpatients not tested meeting the national suspect case definitions were eligible for enrollment. In DRC, a case met the syndromic case definition if they had one or more of the following sign(s) or symptom(s): fever, dry cough, headache, severe fatigue, sore throat, shortness of breath, dyspnea (difficulty breathing), muscle or joint pain, or coryza (common cold). In South Sudan, suspect cases presented with acute onset of fever ≥38°C and cough, or an acute onset of any three or more signs or symptoms, including those in the DRC case definition as well as anorexia, nausea, vomiting, diarrhea, and altered mental status. Cases were subsequently excluded from analysis if they tested negative following enrollment or were transferred to another facility for care. Cases treated as inpatient were considered recovered if they were discharged alive from inpatient care. All eligible cases ($$n = 751$$) were recruited of which 592 consented to participate and were enrolled and 519 were followed to recovery or death, including 144 patients hospitalized at four health facilities which are the focus of this paper (Fig 1).
**Fig 1:** *Study participant flow diagram.*
Oral consent was obtained from adults and parental consent for subjects <18 years by trained research nurses. A questionnaire-based interview was conducted by IMC research nurses or Ministry of Health facility staff at each study facility trained by the lead investigator in each country. The interview, including demographic, symptoms and health history was conducted along with direct observation of mid-upper arm circumference (MUAC), weight, height, pulse rate, oxygen saturation and hemoglobin levels using Masimo RAD 57 (Masimo, Irvine CA), the Multi-Parameter Patient Monitor YK8000 (Yonker, Jiangsu, China) and HemoCue 301 devices (HemoCue, Ängelholm, Sweden); devices varied due to availability but are comparable [11, 12]. Malaria rapid diagnostic tests (Malaria Ag, SD) were available if providers suspected malaria and Hemoglobin A1c (HbA1c) was measured with A1C Now+ Professional Test Kits (PTS Diagnostics, Whitestown, IN) for patients with self-reported diabetes history. Inpatients were intended to have daily follow-ups with clinical course data collected by either health facility staff or research nurses. Participants were followed until discharge (considered as recovery), death, or transfer to a different facility.
In SSD, data entry was direct into the CommCare Platform (Dimagi, Cambridge MA); in DRC data was recorded on paper and subsequently entered electronically. Data was verified by research managers prior to further real-time review for quality and completeness. Data analysis was conducted in R version 4.0.4 (RStudio, PBC, Boston MA). Nutrition status was classified using WHO body mass index (BMI) cut-offs for ages ≥19 years and BMI-for-age for 5–19 years [13, 14] anemia was defined using WHO age/sex specific thresholds for hemoglobin concentration [15]. Definitions and cutoffs for other clinical parameters, were developed based on clinician consensus and clinical practice guidelines. Descriptive analysis included a comparison of continuous variables using Kruskal-Wallis test and categorical variables using Fisher’s exact test. Mixed effects logistic regression models were used to evaluate odds of mortality using parameters significant at $p \leq 0.1$ in unadjusted models; models were adjusted for age, sex, country, and nationality as fixed effects and facility as a random effect. Survival analysis evaluated time from self-reported symptom onset by country and oxygen levels at enrollment using Kaplan-Meier survival functions; survival functions are unadjusted for severity and were evaluated with a log rank test.
The study was reviewed and approved by the Johns Hopkins University Institutional Review Board, the South Sudan Ministry of Health Ethics Committee, the University of Kinshasa School of Public Health, and the United States Centers for Disease Control and Prevention (US CDC). The study is registered with ClinicalTrials.gov (NCT04568499) and was funded by USAID (award 72OFDA20GR0221).
## Results
The study included 519 COVID-19 cases followed to discharge, including 144 ($27.7\%$) hospitalized patients who are the focus of this analysis. The hospital mortality proportion was $16.7\%$ (CI: 11.2–$23.3\%$) and was significantly greater in SSD than DRC ($29.1\%$ vs. $9.0\%$; $$p \leq 0.003$$). Patient characteristics differed by country, where SSD had a significantly larger proportion than DRC of inpatients who were non-national ($26.4\%$ vs. $3.4\%$, $p \leq 0.001$) and male ($72.7\%$ vs. $56.2\%$, $$p \leq 0.046$$); there was no significant difference in age or medical history. The most frequent self-reported symptoms at admission were cough ($71.5\%$), fatigue ($61.1\%$), headache ($53.8\%$), shortness of breath ($52.1\%$) and chest pain ($43.4\%$). Of note, fever was a less common symptom than anticipated (observed in $19\%$ and self-reported by $46\%$ of patients at enrollment). Patients in SSD presented in worse condition by various clinical parameters, likely a function of greater duration from symptom onset to study enrollment (10.2 vs. 6.1 days, $p \leq 0.001$) (Table 1).
**Table 1**
| Unnamed: 0 | All Inpatients | By Enrollment Country | By Enrollment Country.1 | By Enrollment Country.2 | By Outcome | By Outcome.1 | By Outcome.2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | All Inpatients | DR Congo | S Sudan | p-value | Deceased | Recovered | p-value |
| | N = 144 | N = 89 | N = 55 | p-value | N = 24 | N = 120 | p-value |
| Days from Symptom Onset to Enrollment (mean, SD) | 7.8 (7.4) | 6.1 (5.7) | 10.2 (8.8) | <0.001 | 8.3 (5.4) | 7.7 (7.8) | 0.27 |
| Time in Study (mean, SD) | 9.2 (6.0) | 7.9 (4.3) | 11.3 (7.6) | 0.004 | 4. 5 (4.6) | 10.2 (5.8) | <0.001 |
| Age in years (mean, SD) | 48.0 (18.8) | 45.7 (20.5) | 51.6 (15.2) | 0.086 | 61.2 (11.7) | 45.3 (18.9) | <0.001 |
| Age categories < 18 | 5 (3.5%) | 5 (5.6%) | 0 (0.0%) | 0.29 | 0 (0.0%) | 5 (4.2%) | 0.003 |
| 18–44 | 57 (39.6%) | 37 (41.6%) | 20 (36.4%) | 0.29 | 3 (12.5%) | 54 (45.0%) | 0.003 |
| 45–64 | 50 (34.7%) | 29 (32.6%) | 21 (38.2%) | 0.29 | 10 (41.7%) | 40 (33.3%) | 0.003 |
| 65+ | 32 (22.2%) | 18 (20.2%) | 14 (25.5%) | 0.29 | 11 (45.8%) | 21 (17.5%) | 0.003 |
| Sex Female | 54 (37.5%) | 39 (43.8%) | 15 (27.3%) | 0.046 | 6 (25.0%) | 48 (40.0%) | 0.170 |
| Male | 90 (62.5%) | 50 (56.2%) | 40 (72.7%) | 0.046 | 18 (75.0%) | 72 (60.0%) | 0.170 |
| Nationality National | 125 (88.0%) | 86 (96.6%) | 39 (73.6%) | <0.001 | 22 (91.7%) | 103 (87.3%) | 0.74 |
| Non-National | 17 (12.0%) | 3 (3.4%) | 14 (26.4%) | <0.001 | 2 (8.3%) | 15 (12.7%) | 0.74 |
| Symptoms (self-report) | | | | | | | |
| Any symptom(s) | 129 (89.6%) | 76 (85.4%) | 53 (96.4%) | 0.036 | 24 (100.0%) | 105 (87.5%) | 0.076 |
| Cough | 103 (71.5%) | 54 (60.7%) | 49 (89.1%) | <0.001 | 20 (83.3%) | 83 (69.2%) | 0.16 |
| Fatigue/ malaise | 88 (61.1%) | 50 (56.2%) | 38 (69.1%) | 0.12 | 20 (83.3%) | 68 (56.7%) | 0.014 |
| Headache | 77 (53.8%) | 41 (46.6%) | 36 (65.5%) | 0.028 | 11 (45.8%) | 66 (55.5%) | 0.39 |
| Shortness of breath | 75 (52.1%) | 33 (37.1%) | 42 (76.4%) | <0.001 | 22 (91.7%) | 53 (44.2%) | <0.001 |
| Chest pain | 62 (43.4%) | 27 (30.7%) | 35 (63.6%) | <0.001 | 17 (70.8%) | 45 (37.8%) | 0.003 |
| Runny nose | 54 (37.8%) | 27 (30.7%) | 27 (49.1%) | 0.027 | 5 (21.7%) | 49 (40.8%) | 0.084 |
| Sore throat | 52 (36.4%) | 16 (18.2%) | 36 (65.5%) | <0.001 | 8 (33.3%) | 44 (37.0%) | 0.74 |
| Muscle/ joint pain | 50 (35.0%) | 24 (27.3%) | 26 (47.3%) | 0.015 | 9 (37.5%) | 41 (34.5%) | 0.78 |
| Loss of taste/ smell | 25 (17.5%) | 11 (12.5%) | 14 (25.5%) | 0.047 | 8 (33.3%) | 17 (14.3%) | 0.037 |
| Chills | 24 (16.7%) | 16 (18.0%) | 8 (14.5%) | 0.59 | 3 (12.5%) | 21 (17.5%) | 0.77 |
| Abdominal pain | 24 (16.7%) | 12 (13.5%) | 12 (21.8%) | 0.19 | 1 (4.2%) | 23 (19.2%) | 0.079 |
| Vomiting/nausea | 22 (15.3%) | 16 (18.0%) | 6 (10.9%) | 0.25 | 3 (12.5%) | 19 (15.8%) | >0.99 |
| Wheezing | 13 (9.0%) | 8 (9.0%) | 5 (9.1%) | >0.99 | 5 (20.8%) | 8 (6.7%) | 0.043 |
| Diarrhea | 10 (6.9%) | 4 (4.5%) | 6 (10.9%) | 0.18 | 1 (4.2%) | 9 (7.5%) | >0.99 |
| Loss of appetite | 8 (5.6%) | 3 (3.4%) | 5 (9.1%) | 0.26 | 4 (16.7%) | 4 (3.3%) | 0.027 |
| Medical History (self-report) | | | | | | | |
| BCG vaccine | 120 (83.3%) | 78 (87.6%) | 42 (76.4%) | 0.078 | 18 (75.0%) | 102 (85.0%) | 0.24 |
| Tuberculosis (prior) | 2 (1.4%) | 1 (1.1%) | 1 (1.8%) | >0.99 | 1 (4.2%) | 1 (0.8%) | 0.31 |
| HIV positive (n = 66) | 2 (3.0%) | 2 (6.5%) | 0 (0.0%) | 0.22 | 1 (7.7%) | 1 (1.9%) | 0.36 |
| Diabetes | 28 (19.4%) | 16 (18.0%) | 12 (21.8%) | 0.57 | 11 (45.8%) | 17 (14.2%) | 0.001 |
| Chronic Cardiac Disease | 11 (7.9%) | 9 (10.3%) | 2 (3.8%) | 0.21 | 0 (0.0%) | 11 (9.4%) | 0.21 |
| Chronic Pumonary Disease | 3 (2.1%) | 3 (3.4%) | 0 (0.0%) | 0.29 | 1 (4.3%) | 2 (1.7%) | 0.42 |
| Asthma | 2 (1.4%) | 1 (1.1%) | 1 (1.9%) | >0.99 | 0 (0.0%) | 2 (1.7%) | >0.99 |
| Current smoker | 3 (2.1%) | 3 (3.4%) | 0 (0.0%) | 0.30 | 0 (0.0%) | 3 (2.5%) | >0.99 |
When examined by survival status, there was no significant difference by sex, however, deceased patients were significantly older (61.2 vs. 45.3 years, $p \leq 0.001$) and more likely to report symptoms. The only significant difference in medical history between deceased and surviving patients was prior diabetes diagnosis ($45.8\%$ vs. $14.2\%$, $p \leq 0.001$) (Table 2). Differences in survival probability by day since symptom onset are presented in Fig 2. While the mortality proportion differed significantly between SSD and DRC overall, there was no difference in survival probability when adjusted for the date of symptom onset. Oxygen saturation at admission was related to survival, where oxygen saturation ≥$94\%$ at admission had a five-day survival probability of 0.96 compared to 0.71 among patients with oxygen saturation <$94\%$ ($p \leq 0.001$).
**Fig 2:** *Survival probability by country and oxygen level at admission.* TABLE_PLACEHOLDER:Table 2 Patients in SSD were significantly more likely to receive oxygen support ($56.4\%$ vs. $29.2\%$, $$p \leq 0.001$$), less likely to be placed in a prone or half-seated position ($60.0\%$ vs. $100.0\%$, $p \leq 0.001$) and more likely to receive medications within 24 hours of admission (Table 2). Laboratory and diagnostic tests were not assessed due to the small sample size, and are described for the entire study population. Blood samples were collected for $38.2\%$ ($$n = 55$$) patients; kidney and liver function tests, respectively, were available for only $16.0\%$ ($$n = 23$$) and $2.8\%$ ($$n = 4$$) of patients. Blood sugar control was assessed for 17 of 28 patients ($60.8\%$) reporting diabetes history; $35.3\%$ ($$n = 6$$) had HbA1C levels >8.0, indicating poor diabetes control. There were 61 suspected malaria cases (defined as taking antimalarials and having fever or chills) of which $75.4\%$ ($$n = 46$$) had rapid diagnostic tests (RDTs); testing was more common in DRC than SSD ($45\%$ vs. $11\%$) and overall, $15.2\%$ ($$n = 7$$) tested positive. When interpreting these findings, it is important to consider the observational nature of the study and the fact that co-morbidities are likely to be under-described, as a result of underdiagnosis, lack of confirmatory findings or inaccurate reporting of medical history.
Deceased patients were significantly more likely to have received non-invasive oxygen support ($91.7\%$ vs. $29.2\%$, $p \leq 0.001$) within the first 24 hours of admission; one individual in DRC later received bilevel positive airway pressure (BIPAP) or continuous positive airway pressure (CPAP) and subsequently died. Although BIPAP/CPAP machines and ventilators were available in facilities, use was limited by irregular electricity and human resource constraints, including absence of skilled staff and the time-intensive nature of monitoring needs. Deceased patients were more likely to receive non-steroid anti-inflammatories ($66.7\%$ vs. $33.3\%$, $$p \leq 0.002$$), corticosteroids ($62.5\%$ vs. 39.2, $$p \leq 0.035$$), and anticoagulants ($41.7\%$ vs. $10.0\%$, $p \leq 0.001$) at admission, likely a function of a worse clinical condition. Deceased patients were less likely to receive multivitamins or vitamin C ($54.2\%$ vs. $74.2\%$, $$p \leq 0.049$$). The proportions of patients receiving these therapies ever during the hospital stay were slightly greater than at admission, as would be expected. ( S1 Fig). Of note, only one patient in DRC received vasopressors and three patients in SSD received remdesivir while hospitalized; in DRC, remdesivir was available only in one facility and was a second line treatment.
In regression models estimating adjusted odds ratio (AOR) of survival (Table 3), age was the only demographic variable significantly associated with survival, with a $6\%$ survival decrease per additional year (AOR = 0.94, CI:0.91–0.98). With regard to symptoms and health history, survival AORs were significantly reduced among patients with shortness of breath (AOR = 0.16, CI:0.03–0.80), and wheezing (AOR = 0.20, CI:0.05–0.81) and prior diabetes diagnosis (AOR = 0.28, CI:0.09–0.90). Nearly all abnormal clinical observations were significantly associated with decreased survival; these included low oxygen levels (SaO2<$94\%$ AOR = 0.10, CI:0.02–0.39; SaO2 <$90\%$ AOR = 0.04, CI:0.01–0.16); elevated respiratory rate (AOR = 0.08, CI:0.02–0.38), elevated pulse rate (AOR = 0.11, CI:0.03–0.35), fever (AOR = 0.27, CI:0.08–0.93), systemic inflammatory response syndrome (SIRS) (AOR = 0.14, CI:0.04–0.46) and acute respiratory distress syndrome (ARDS) (AOR = 0.14, CI:0.04–0.46).
**Table 3**
| Unnamed: 0 | Unadjusted Odds | Unadjusted Odds.1 | Unadjusted Odds.2 | Adjusted Odds1 | Adjusted Odds1.1 | Adjusted Odds1.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | Point Estimate | 95% CI | p-value | Point Estimate | 95% CI | p-value |
| Demographic Characteristics | | | | | | |
| Age | 0.94 | (0.92–0.97) | <0.001 | 0.94 | (0.91–0.98) | 0.004 |
| Male sex (ref: female) | 0.5 | (0.19–1.35) | 0.171 | 0.49 | (0.15–1.57) | 0.229 |
| South Sudan (ref: DRC) | 0.24 | (0.09–0.61) | 0.003 | 0.19 | (0.03–1.43) | 0.107 |
| National (ref: non-nationals) | 1.6 | (0.34–7.51) | 0.55 | 3.08 | (0.57–15.53) | 0.192 |
| Symptoms and Health History at Admission (self-reported) | Symptoms and Health History at Admission (self-reported) | Symptoms and Health History at Admission (self-reported) | | | | |
| Symptom onset to final disposition (days) | 1.10 | (1.02–1.18) | 0.013 | 1.26 | (1.12–1.43) | <0.001 |
| Fatigue/malaise | 0.26 | (0.08–0.81) | 0.020 | 0.29 | (0.08–1.12) | 0.073 |
| Shortness of breath | 0.07 | (0.02–0.32) | <0.001 | 0.16 | (0.03–0.80) | 0.026 |
| Chest pain | 0.25 | (0.10–0.65) | 0.004 | 0.55 | (0.18–1.71) | 0.303 |
| Runny nose | 2.48 | (0.86–7.14) | 0.091 | 5.35 | (1.34–21.38) | 0.018 |
| Loss of taste/smell | 0.33 | (0.12–0.9) | 0.030 | 0.79 | (0.24–2.55) | 0.687 |
| Abdominal pain | 5.45 | (0.7–42.48) | 0.105 | 9.05 | (0.97–84.67) | 0.054 |
| Wheezing | 0.27 | (0.08–0.92) | 0.036 | 0.20 | (0.05–0.81) | 0.023 |
| Loss of appetite | 0.17 | (0.04–0.75) | 0.019 | 0.29 | (0.06–1.53) | 0.145 |
| Prior diabetes diagnosis | 0.20 | (0.08–0.51) | <0.001 | 0.28 | (0.09–0.90) | 0.032 |
| Clinical Observations at Admission | | | | | | |
| Fever (>37.5 C) | 0.38 | (0.14–1.00) | 0.051 | 0.27 | (0.08–0.93) | 0.037 |
| Mean pulse rate | 0.95 | (0.92–0.97) | <0.001 | 0.93 | (0.89–0.96) | <0.001 |
| High pulse rate (≥100/min) | 0.19 | (0.07–0.47) | <0.001 | 0.11 | (0.03–0.35) | <0.001 |
| Mean respiratory rate | 0.83 | (0.78–0.89) | <0.001 | 0.74 | (0.65–0.85) | <0.001 |
| High respiratory rate (>22/min) | 0.05 | (0.01–0.21) | <0.001 | 0.08 | (0.02–0.38) | 0.002 |
| Mean oxygen saturation | 1.14 | (1.07–1.21) | <0.001 | 1.16 | (1.08–1.23) | <0.001 |
| Low oxygen level (<94%) | 0.06 | (0.02–0.22) | <0.001 | 0.10 | (0.02–0.39) | 0.001 |
| Low oxygen level (<90%) | 0.05 | (0.02–0.14) | <0.001 | 0.04 | (0.01–0.16) | <0.001 |
| Systemic Inflammatory Response Syndrome (SIRS) | 0.19 | (0.07–0.49) | <0.001 | 0.14 | (0.04–0.46) | 0.001 |
| Acute Respiratory Distress (ARDS) | 0.11 | (0.04–0.30) | <0.001 | 0.14 | (0.04–0.46) | 0.001 |
| Sepsis | 0.08 | (0.01–0.49) | 0.006 | 0.18 | (0.03–1.25) | 0.083 |
| Therapies and Medications Provided (within 24 hours of admission) | Therapies and Medications Provided (within 24 hours of admission) | Therapies and Medications Provided (within 24 hours of admission) | Therapies and Medications Provided (within 24 hours of admission) | | | |
| Non-invasive oxygen support (ref: none) | 0.04 | 0.01–0.14 | <0.001 | 0.07 | (0.01–0.38) | 0.002 |
| Prone Position (ref: cardiac half-seated) | 5.65 | (2.15–16.81) | <0.001 | 2.08 | (0.54–7.96) | 0.285 |
| Anticoagulants | 0.15 | (0.06–0.37) | <0.001 | 0.56 | (0.18–1.77) | 0.321 |
| Non-steroid Anti-inflammatories | 0.30 | (0.12–0.67) | 0.001 | 0.86 | (0.19–3.83) | 0.839 |
| Corticosteroids | 0.42 | 0.18–0.94) | 0.035 | 2.11 | 0.53–8.36) | 0.287 |
| Vitamin C or Multivitamin | 2.43 | (0.97–6.00) | 0.054 | 3.54 | (0.99–12.67) | 0.052 |
| Vitamin C | 6.00 | (2.01–25.82) | 0.004 | 2.69 | (0.55–13.23) | 0.222 |
| Multivitamins | 0.37 | (0.16–0.86) | 0.019 | 1.44 | (0.34–6.18) | 0.621 |
| Health Facility Characteristics | | | | | | |
| Adequate oxygen concentrators/cylinders | 0.21 | (0.01–0.10) | 0.140 | 0.93 | (0.04–22.60) | 0.963 |
| Ventilators available | 0.25 | (0.09–0.63) | 0.004 | 0.20 | (0.02–1.56) | 0.124 |
| Chest x-ray available | 4.00 | (1.6–10.70) | 0.004 | 5.12 | (0.64–40.88) | 0.124 |
| Electrocardiogram available | 4.00 | (1.6–10.70) | 0.004 | 5.12 | (0.64–40.88) | 0.124 |
| Remdesivir available | 0.43 | (0.16–1.09) | 0.084 | 1.08 | (0.04–26.25) | 0.963 |
Non-invasive oxygen was the only treatment significantly associated with survival (AOR = 0.07, CI:0.01–0.38). Prone positioning was significantly associated with increased survival in unadjusted but not adjusted models. Receipt of non-steroidal anti-inflammatories, corticosteroids and anticoagulants, were significantly associated with decreased survival in unadjusted but not in adjusted models. Receipt of Vitamin C or a multivitamin was associated with increased survival with borderline significance in both unadjusted and adjusted models (AOR = 3.52, CI:0.99–12.67). Adjusted odds of survival were also calculated by health facility characteristics and no differences were observed in adjusted models; it is important to note that certain treatments were rationed and odds may not reflect actual access to a treatment/diagnostic test. Given the small number of health facilities and differences in survival rates between countries, findings should be interpreted with caution, in particular considering that access to these resources was not ubiquitous among individual participants.
## Discussion
This cohort study details clinical progression and outcomes of 144 hospitalized COVID-19 patients in North and South Kivu in eastern DRC and Juba, SSD. The observed hospital mortality proportion of $16.7\%$ is comparable to findings from global systematic reviews of COVID-19 inpatient survival [16, 17] but lower than prior studies in DRC which observed mortality rates of $22\%$ and $32\%$ and were conducted earlier in the pandemic in Kinshasa [18, 19]. However, actual mortality in study facilities was higher due to the inability to enroll patients presenting near death; this figure was not tracked in SSD, however in DRC 24 patients died before being admitted to the study ($2.3\%$ of DRC study participants). Higher mortality in SSD ($21.9\%$ vs. $9.0\%$) can be attributed to later presentation and worse conditions at admission. Time from symptom onset to admission has also been shown to be independently associated with survival prior studies in Kinshasa, DRC [18] Late presentation is potentially the result of variety of a factors, including denial, fear and stigma; minimal testing availability; isolation policies; transportation constraints and costs; use of informal providers; and perceptions of poor treatment availability which deter care seeking [20].
Characteristics associated with increased mortality risk included age and prior diabetes diagnoses, which is consistent with known risk factors and similar to ACCCOS study findings which examined outcomes of hospitalized COVID-19 patients in Africa [9, 21, 22] there was no difference in mortality risk by sex. Comparatively few studies have examined infectious co-morbidities, and findings from this study did not suggest an association between increased COVID-19 mortality and concurrent malaria or HIV infection or history of tuberculosis. This is potentially a function of inadequate power due to low prevalence and small sample size; other research suggests that HIV/AIDS increases risk for poor outcomes whereas malaria coinfection does not [23–25]. To the best of our knowledge this is the first study to assess underweight as a COVID-19 risk factor and no significant association with mortality was observed, however, only eight subjects had low BMI (seven survived). Anemia was not associated with increased mortality in this study but a meta-analysis has shown anemia to increase risk for poor COVID-19 [26].
Fever was observed in a smaller proportion of patients than anticipated, however, it is important to note that data presented reflects fever observed at enrollment which is distinct from ever having fever during the course of COVID-19. Fever is dynamic and may not have been elevated at presentation to hospital; for example, other studies of hospitalized COVID-19 patients have reported only $30\%$ of patients with fever [27] and in both SSD and DRC self-medication with antipyretics is common and could have suppressed fever at the time of the observation, thereby resulting in a lower proportion of patients presenting with fever. Patients with low oxygen levels, elevated respiratory rate, SIRS, ARDS and elevated pulse rate and fever at admission had significantly decreased odds of survival which aligns with existing evidence-[28, 29] Non-invasive oxygen was the only therapy or medication significantly associated with decreased survival in both unadjusted and adjusted models; as models were not adjusted for patient severity, lower odds of survival among patients receiving oxygen likely reflects preferential use of oxygen on patients with poorer clinical presentation on admission. Anticoagulants, non-steroid anti-inflammatory drugs and corticosteroids were associated with decreased survival in unadjusted but not adjusted models. Only several study patients received antivirals or vasopressors during their hospital stay, both of which limited ability to do analysis. In DRC, remdesivir was a second line treatment and not all patients were eligible; in both countries, clinicians reported limited supply resulted in rationing medications.
Most evidence on COVID-19 progression and outcomes is from upper-middle and upper-income countries where comorbidity prevalence and clinical management capacities differ greatly as compared to resource poor settings. Expanding the available evidence to understand if and how outcomes vary by context for known risk factors is critical to informing the health response in resource constrained settings. This study suggests that risk factors for inpatient mortality do not differ greatly from those observed in upper-middle and upper-income settings despite the differing population profile of the African humanitarian context and greater infectious disease prevalence. Vitamin C and multivitamins provision to inpatients however pointed to a protective effect (marginally significant), which is consistent with current evidence that indicates provision of Vitamin C is not associated with improved COVID-10 outcomes [30–33] Differences in the health system capacity were striking, where few patients received recommended medications and ventilation, even in facilities where these resources were available. Human resource factors including lack of training to manage patients receiving invasive oxygen and insufficient staff for oversight of resource-intensive therapies were notable limitations in both countries. Other factors that restricted ventilator use were irregular electricity and lack of supplies to support their use.
## Limitations
Weak surveillance systems, low COVID-19 testing capacity and inconsistent information flow from the laboratory coupled with testing hesitancy, a large proportion of travel-related tests and many unreachable cases were factors that influenced the population that was tested and subsequently identified as eligible for the study. Inability to rapidly hire additional study staff during the February/March and a June health worker strike in SSD, along with security issues and the May 22 Mt. Nyiragongo volcanic eruption in DRC also contributed to a smaller sample than planned ($$n = 1000$$) and one that is unlikely to be representative of the population in facility catchment areas. The small sample size coupled with low prevalence of selected patient characteristics/treatment use translated to inadequate power to detect significant differences for some variables. Self-reporting of symptoms and comorbidities may have been inaccurate, in particular for conditions with stigma (e.g., HIV/AIDS, TB); relatedly, co-infections may have been underdiagnosed because tests were not ordered, were unavailable (e.g. malaria) or could not be paid for. It was not feasible to record all clinical course details for inpatients, leading to some information being missed; notably quantity of oxygen used was not collected, and temporality of treatments received was not analyzed due to the fact that many deaths occurred soon after admission. Finally, challenges with timely delivery of equipment intended for the study, particularly in SSD, necessitated the use of alternative equipment and resulted in missing data, most notably on anemia among those enrolled early in the study.
## Conclusions
In this study of hospitalized COVID-19 inpatients in SSD and DRC, the observed mortality proportion ($16.7\%$) was comparable to other findings globally. Age and history of diabetes were the only characteristics measured at presentation associated with decreased survival. Clinical status measures associated with decreased survival included fever, low oxygen level, elevated respiratory and pulse rates, SIRS and ARDS. Patient positioning and receipt of various classes of medications were not associated with differences in survival; the only therapy significantly associated with survival was non-invasive oxygen. Antivirals, vasopressors and invasive oxygen therapies were rarely received despite demonstrated effectiveness in other settings.
To mitigate the challenges of COVID-19 hospital care in resource poor settings, recommendations include strengthening of training of health providers and customizing trainings to reflect availability of specific medications, therapies and operational constraints at the particular facility. Raising provider awareness of factors associated with poor outcomes and providing clear guidance on recommended treatment paths (e.g., flow charts) according to clinical status could reduce the burden on providers and facilitate treatment that is more closely aligned with current best practice. Beyond health facilities, expanded testing capacity and use of appropriate and context specific social behavior change campaigns to reduce the stigma of COVID-19 and improve the perception of available care could facilitate earlier presentation which also would contribute to improved outcomes.
## References
1. 1World Health Organization (WHO). COVID-19 Dashboard. URL: https://covid19.who.int/. Accessed Sep 21, 2021.. *COVID-19 Dashboard. URL* **21** 2021
2. 2World Health Organization (WHO). The Global Health Observatory Data Repository. Population Estimates by WHO Region. URL: https://apps.who.int/gho/data/view.main.POP2020. Accessed Aug 21, 2021.
3. Katchunga PB, Murhula A, Akilimali P, Zaluka JC, Karhikalembu R, Makombo M. **Seroprevalence of anti-SARS-CoV-2 antibodies among travelers and workers screened at the Saint Luc clinic in Bukavu, eastern Democratic Republic of the Congo, from May to August 2020**. *Pan African Medical Journal* (2021.0) **38** 93. DOI: 10.11604/pamj.2021.38.93.26663
4. Wiens KE, Mawien PN, Rumunu J, Slater D, Jones FK, Moheed S. **Seroprevalence of Severe Acute Respiratory Syndrome Coronavirus 2 IgG in Juba, South Sudan, 2020**. *Emerg Infect Dis* (2021.0) **27** 1598-1606. DOI: 10.3201/eid2706.210568
5. 5The World Bank. Open Data on physicians and hospital beds by country. Available at https://data.worldbank.org/indicator/SH.MED.PHYS.ZS and https://data.worldbank.org/indicator/ SH.MED.BEDS.ZS. Accessed Aug 22, 2021.
6. 6World Health Organization (WHO) Global Health Work force Alliance. South Sudan Country Profile. URL: https://www.who.int/workforcealliance/countries/ssd/en/. Accessed Aug 22, 2021.. **22** 2021
7. Berendes S, Lno Lako R, Whitson D, Gould S, Valadez J. **Assessing the quality of care in a new nation: South Sudan’s first national health facility assessment**. *Trop Med Intl Health* (2014.0) **19**. DOI: 10.1111/tmi.12363
8. **Patient care and clinical outcomes for patients with COVID-19 infection admitted to African high-care or intensive care units (ACCCOS): a multicentre, prospective, observational cohort study**. (2021.0) **397** 1885-94. DOI: 10.1016/S0140-6736(21)00441-4
9. Leidman E, Doocy S, Heymsfield G, Sebushishe A, Ngole ME, Bollemeijer I. **Risk Factors for Hospitalization and Death from COVID-19 in South Sudan and Eastern Democratic Republic of the Congo.**. *Manuscript under review.*
10. Whitehead RD, Mei Z, Mapango C, Jefferds MED. **Methods and analyzers for hemoglobin measurement in clinical laboratories and field settings**. *Annals of the New York Academy of Sciences* (2019.0)
11. Hiscock R, Kumar D, Simmons SW. **Systematic review and meta-analysis of method comparison studies of Masimo pulse co-oximeters (Radical-7 or Pronto-7) and HemoCue(R) absorption spectrometers (B-Hemoglobin or 201+) with laboratory haemoglobin estimation.**. *Anaesth Intensive Care.* (2015.0) **43** 341-50. DOI: 10.1177/0310057X1504300310
12. 12World Health Organization (WHO). Growth reference data for children 5–19 years. URL: https://www.who.int/toolkits/growth-reference-data-for-5to19-years/indicators/bmi-for-age.
13. 13World Health Organization (WHO). The Global Health Observatory Data Repository. Body Mass Index (BMI). URL: https://apps.who.int/gho/data/node.main.BMIANTHROPOMETRY?lang=en. Accessed July 14, 2021.
14. 14World Health Organization (WHO). Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity. WHO/NMH/NHD/MNM/11.1. 2011. Available at URL: https://www.who.int/vmnis/indicators/haemoglobin.pdf.
15. Dessie ZG, Zewotir T. **Mortality-related risk factors of COVID-19: a systematic review and meta-analysis of 42 studies and 423,117 patients.**. *Published 2021 Aug 21* (2021.0) **21** 855. DOI: 10.1186/s12879-021-06536-3
16. Macedo A, Gonçalves N, Febra C. **COVID-19 fatality rates in hospitalized patients: systematic review and meta-analysis.**. (2021.0) **57** 14-21. DOI: 10.1016/j.annepidem.2021.02.012
17. Majer J, Udoh K, Baleke A, Ahmed D, Kumar D, Summers A. **Operational challenges and considerations for COVID-19 research in humanitarian settings: A qualitative study of a project in South Sudan and Eastern Democratic Republic of the Congo.**. *Manuscript in press.*
18. Makulo JR, Mandina MN, Mbala PK, Wumba RD, Akilmali PZ, Nlandu YM. **SARS-CoV2 infection in symptomatic patients: interest of serological tests and predictors of mortality: experience of DR Congo.**. *BMC Infect Dis* (2022.0) **22** 21. DOI: 10.1186/s12879-021-07003-9
19. Nlandu Y, Mafuta D, Sakaji J, Brecknell M, Engole Y, Abatha J. **Predictors of mortality in COVID-19 patients at Kinshasa Medical Center and a survival analysis: a retrospective cohort study.**. *BMC Infect Dis* (2021.0) **21** 1272. DOI: 10.1186/s12879-021-06984-x
20. 20Centers for Disease Control and Prevention. People with certain medical conditions. Updated March 29, 2021. https://www.cdc.gov/coronavirus/2019-ncov/need-extra-precautions/people-with-medical-conditions.html. Accessed April 8, 2021.
21. Nandy K, Salunke A, Pathak SK, Pandey A, Doctor C, Puj K. **Coronavirus disease (COVID-19): A systematic review and meta-analysis to evaluate the impact of various comorbidities on serious events.**. *Diabetes Metab Syndr 2020* **14** 1017-1025. DOI: 10.1016/j.dsx.2020.06.064
22. Dong Y, Li Z, Ding S, Liu S, Tang Z, Jia L. **HIV infection and risk of COVID-19 mortality: A meta-analysis.**. *Medicine* (2021.0) **100** e26573. DOI: 10.1097/MD.0000000000026573
23. Chanda D, Minchella PA, Kampamba D, Itoh M, Hines JZ, Fwoloshi S. **COVID-19 Severity and COVID-19–Associated Deaths Among Hospitalized Patients with HIV Infection---Zambia, March–December 2020.**. *MMWR Morb Mortal Wkly Rep* (2021.0) **70** 807-810. DOI: 10.15585/mmwr.mm7022a2
24. Achan J, Serwanga A, Wanzira H, Kyagulanyi T, Nuwa A, Magumba G. **Impact of current malaria infection and previous malaria exposure on the clinical profiles and outcome of COVID-19 in a high malaria transmission setting: a prospective cohort study**. *Lancet prepreint.*
25. Hariyanto TI, Kurniawan A. **Anemia is associated with severe coronavirus disease 2019 (COVID-19) infection.**. (2020.0) **59** 102926. DOI: 10.1016/j.transci.2020.102926
26. Kakodkar P, Kaka N, Baig MN. **A Comprehensive Literature Review on the Clinical Presentation, and Management of the Pandemic Coronavirus Disease 2019 (COVID-19).**. (2020.0) **12** e7560. DOI: 10.7759/cureus.7560
27. Richardson S, Hirsch JS, Narasimhan M, Crawford J, McGinn T, Davidson K. **Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the New York City Area**. (2020.0) **323** 2052-2059. DOI: 10.1001/jama.2020.6775
28. Bahl A, Van Baalen MN, Ortiz L, Chen NW, Todd C, Milad M. **Early predictors of in-hospital mortality in patients with COVID-19 in a large American cohort.**. (2020.0) **15** 1485-1499. DOI: 10.1007/s11739-020-02509-7
29. Del Valle DM, Kim-Schulze S, Huang HH, Beckman ND, Nirenburg S, Wang B. **An inflammatory cytokine signature predicts COVID-19 severity and survival**. *Nature Medicine* (2020.0) **26** 1636-1643. DOI: 10.1038/s41591-020-1051-9
30. Zhang J, Rao X, Li Y, Yuan Z, Liu F, Guo G. **Pilot trial of high-dose vitamin C in critically ill COVID-19 patients.**. (2021.0) **11** 5. DOI: 10.1186/s13613-020-00792-3
31. Rawat D, Roy A, Maitra S, Gulati A, Khanna P, Baidya DK. **Vitamin C and COVID-19 treatment: A systematic review and meta-analysis of randomized controlled trials**. *Diabetes & Metabolic Syndrome: Clinical Research & Reviews* (2021.0) **15** 102324. DOI: 10.1016/j.dsx.2021.102324
32. Schuetz P, Gregoriano C, Keller U. **Supplementation of the population during the COVID-19 pandemic with vitamins and micronutrients---how much evidence is needed?.**. (2021.0) **151** w20522. DOI: 10.4414/smw.2021.20522
33. Milani GP, Macchi M, Guz-Mark A. **Vitamin C in the Treatment of COVID-19.**. (2021.0) **13** 1172. DOI: 10.3390/nu13041172
|
---
title: Factors associated with the lethality of patients hospitalized with severe
acute respiratory syndrome due to COVID-19 in Brazil
authors:
- Ana Cristina Dias Custódio
- Fábio Vieira Ribas
- Luana Vieira Toledo
- Cristiane Junqueira de Carvalho
- Luciana Moreira Lima
- Brunnella Alcantara Chagas de Freitas
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021557
doi: 10.1371/journal.pgph.0000200
license: CC BY 4.0
---
# Factors associated with the lethality of patients hospitalized with severe acute respiratory syndrome due to COVID-19 in Brazil
## Abstract
Due to the high rates of transmission and deaths due to COVID-19, understanding the factors associated with its occurrence, as well as monitoring and implementing control measures should be priority actions in health surveillance, highlighting the use of epidemiological surveillance information systems as an important ally. Thus, the objectives of this study were to calculate the mortality rate of hospitalized patients with severe acute respiratory syndrome due to COVID-19 and to identify factors associated with death, in the period corresponding to epidemiological weeks 01 to 53 of the year 2020. This was a longitudinal study, using the national influenza epidemiological surveillance information system database, routinely collected by healthcare services. The sociodemographic and clinical characteristics of 563,051 hospitalized patients with severe acute respiratory syndrome due to COVID-19 in the five regions of Brazil were analyzed. Cox regression was performed to assess factors associated with patient death during hospitalization. The national lethality rate was $35.7\%$, and the highest rates of lethality occurred in the Northeast ($44.3\%$) and North ($41.2\%$) regions. During the hospital stay, death was associated with older age (Hazard Ratio—HR = 1.026; $p \leq 0.001$); male sex (HR = 1.052; $p \leq 0.001$); living in the North (HR = 1.429; $p \leq 0.001$), Northeast (HR = 1.271; $p \leq 0.001$) or Southeast regions of Brazil (HR = 1.040; $p \leq 0.001$), presenting any risk factor (HR = 1.129; $p \leq 0.001$), the use of invasive (HR = 2.865; $p \leq 0.001$) or noninvasive (HR = 1.401; $p \leq 0.001$) mechanical ventilation devices. A high case lethality rate was evidenced in patients with severe acute respiratory syndrome due to COVID-19, however, deaths were not evenly distributed across the country’s regions, being heavily concentrated in the Northeast and North regions. Older male patients living in the North, Northeast, or Southeast regions of Brazil, who presented any risk factor and were submitted to the use of invasive or noninvasive mechanical ventilation devices, presented a higher risk of evolving to death.
## Introduction
The COVID-19 pandemic triggered by a novel coronavirus (SARS-CoV-2) which was detected in the Hubei province, Wuhan city in China, in December 2019, has spread worldwide, requiring from countries and regions all over the world actions to improve health surveillance, especially to stop the spread of the transmitting agent [1]. By April 2020, the confirmed cases of SARS-CoV-2 infection and sickness due to COVID-19 had already exceeded one million people and registered more than fifty thousand deaths worldwide [2].
In Brazil, the Influenza Epidemiological Surveillance Information System, SIVEP-Gripe (Sistema de Informação de Vigilância Epidemiológica da Gripe), has been used as the information system for the surveillance of Severe Acute Respiratory Syndrome (SARS). This system has been implemented in response to the Influenza A (H1N1) pandemic since 2009, and it currently also includes the surveillance of hospitalized patients and/or deaths suspected of COVID-19 [3]. SIVEP-Gripe was established as the official channel for notifications of this hazard at the hospital level, reaching a peak of 22,497 hospitalized patients infected with SARS-CoV-2 in the epidemiological week (EW) 20 of 2020—Coronavirus Panel https://covid.saude.gov.br/ The management of the negative impacts caused by this pandemic, especially regarding the reduction and assistance to the high number of victims, has been the most challenging task for healthcare providers and managers. Mortality can be perceived as a multicausal event, influenced by factors inherent to the patients themselves, such as pre-existing clinical conditions, in addition to the structural and organizational issues faced by society and health services [4].
The occurrence of a higher number of deaths due to COVID-19 was verified in the economically disadvantaged population and among those who did not adhere to strategies and measures to control the spread of the virus during the pandemic [5]. Moreover, an association was found between higher mortality rates and the presence of comorbidities, such as smoking, diabetes mellitus, hypertension, and obesity [6]. The highest occurrence of deaths due to COVID-19 was associated with the population aged over 60 years in an ecological study involving 113 countries [7]. In Brazil, in the state of Espírito Santo, an association was identified between death and the presence of comorbidities and advanced age among patients receiving medical care in public institutions [8]. In another Brazilian state, Rondônia, the highest occurrence of death was associated with ages over 60 years, male gender, and brown and black skin colors [9].
In this context, due to the high rates of transmission and deaths due to COVID-19, understanding the factors associated with its occurrence, as well as monitoring and implementing control measures should be priority actions in health surveillance, highlighting the use of epidemiological surveillance information systems as an important ally.
Therefore, this study aimed to calculate the case lethality rate of patients hospitalized presenting SARS due to COVID-19 and to identify the risk factors associated with death, in the period corresponding to EW 01 to 53 of 2020.
## Methods
This was an observational, longitudinal study, based on data routinely collected by healthcare services, related to the epidemiological surveillance of patients hospitalized with SARS due to COVID-19 in the period corresponding to EW 01 to 53 of 2020. The study was carried out based on data available in SIVEP-Gripe, which is publicly accessible, non-nominal, without any identification of individuals, available on the following website: https://opendatasus.saude.gov.br/dataset/srag-2020, extracted in the update of April 26th, 2021.
The study population was composed of hospitalized patients presenting SARS due to COVID-19 in 2020, notified in SIVEP-Gripe ($$n = 691$$,985). All patients classified with SARS due to COVID-19, who required hospitalization and submitted complete information on the date of outcome (discharge or death) were included. Cases of hospitalized patients with SARS due to other etiologies and notification forms remaining in the system with no outcome of the case were excluded.
In this study, the following sociodemographic, clinical, and diagnostic investigation data available on the individual notification form of hospitalized SARS cases were assessed: age (years), gender (male/female); skin color (white/non-white), level of education (illiterate); elementary school (1st-5th grades); middle school (6th-9th grades); high school (10th-12th grades); higher education; (not applicable); region (North; Northeast; South; Southeast; Midwest); the presence of risk factors (no/yes); description of risk factors (puerperal; chronic cardiovascular disease; chronic hematologic disease; Down’s syndrome; chronic liver disease; asthma; diabetes mellitus; chronic neurological disease; chronic pneumopathy; immunodepression; chronic kidney disease; obesity; other morbidities); length of hospitalization (days); admission to Intensive Care Unit (ICU: no/yes); invasive mechanical ventilation (IMV: no/yes); noninvasive ventilation (NIV: no/yes).
Descriptive and inferential data analysis was performed using IBM SPSS Statistics 23 software, considering a type I error level of $5\%$. Simple and relative frequencies, measures of central tendency, and dispersion (mean, median, standard deviation, and interquartile values) were presented. The normality of the distribution of the numeric variables was assessed using the Kolmogorov-Smirnov test.
The case lethality rate of SARS due to COVID-19 was calculated based on the proportion of deaths in relation to the total number of patients. The survivor and non-survivor groups were compared for their characteristics. Categorical variables were compared using Pearson’s Chi-Squared test. Mann Whitney’s test was used to compare numeric variables. Values of $p \leq 0.05$ were considered statistically significant differences.
For the survival analysis, the dependent variable was the "observation time in days", considering the period between the "patients’ admission date", registered in the epidemiological surveillance system, and the "evolution date" of the case, which indicates the date of death (outcome of interest) or discharge, which indicates the end of the observation time (censored).
The Cox Regression analysis, estimating the Hazard Ratio (HR) and its $95\%$ Confidence Interval (CI), was used to evaluate the risk factors for death in patients with SARS due to COVID-19. Univariate and multivariate regression models were used. Variables presenting more than $20\%$ missing information (incompleteness) were not included in the Cox Regression, and only variables presenting more than $80\%$ completeness were included [10].
Since this was a survey that included only public domain data, without participant identification, no approval by the Research Ethics Committee was required.
## Results
During the period evaluated, 563,051 reported cases of patients with SARS due to COVID-19 that required hospital admission had their outcome reported in SIVEP-Gripe. The national case lethality rate was $35.7\%$. The greatest lethality rates were found in the Northeast ($44.3\%$) and North ($41.2\%$) regions according to the geographical analysis. The Southeast ($34.4\%$), Midwest ($31.7\%$) and South ($30.0\%$) regions presented a lethality rate lower than the nationally calculated value.
Table 1 presents the sociodemographic characteristics of the patients hospitalized with SARS due to COVID, comparing those who were discharged and those who passed away. Among the patients who passed away, elderly patients were prevalent, with a median age of 71 years ($Q = 60$; Q3 = 80), male ($57.4\%$), considered non-white ($52.6\%$), with a low level of education, and who had only attended elementary school ($54.5\%$). Regarding spatial distribution, it is noteworthy that in the Northeast ($44.3\%$) and North ($41.2\%$) regions, the percentage of patients who passed away was higher than in other regions (Southeast $34.4\%$, Midwest $31.7\%$, and South $30.0\%$ regions).
**Table 1**
| Variable | General (n = 563051) | Survivors (n = 362144) | Non-survivors (n = 200907) | p-value |
| --- | --- | --- | --- | --- |
| Age (n = 563051) – med (Q1-Q3) | 62 (48–74) | 56 (43–68) | 71 (60–80) | <0.001a |
| Sex (n = 562963) – n (%) | | | | <0.001b |
| Female | 247931 (44.0) | 162399 (44.8) | 85532 (42.6) | |
| Male | 315032 (56.0) | 119682 (55.1) | 115350 (57.4) | |
| Skin color (n = 442856) – n (%) | | | | |
| Non-white | 218531 (49.3) | 132670 (47.5) | 85861 (52.6) | <0.001b |
| White | 224325 (50.7) | 146913 (52.5) | 77412 (47.4) | |
| Level of education (n = 208911) – n (%) | | | | <0.001b |
| Illiterate | 14671 (7.0) | 6405 (4.8) | 8266 (10.9) | |
| Elementary school | 56229 (26.9) | 29906 (22.5) | 26323 (34.6) | |
| Middle school | 38172 (18.3) | 23071 (17.4) | 15101 (19.9) | |
| High school | 63531 (30.4) | 45178 (34.0) | 18353 (24.2) | |
| Higher education | 32495 (15.6) | 24928 (18.8) | 7567 (10.0) | |
| Not applicable | 3813 (18) | 3430 (2.6) | 383 (0.5) | |
| Region (n = 563007) – n (%) | | | | <0.001b |
| North | 45458 (8.1) | 26746 (7.4) | 18712 (9.3) | |
| Northeast | 95460 (17.0) | 53202 (14.7) | 42258 (21.0) | |
| South | 84881 (15.1) | 59431 (16.4) | 25450 (12.7) | |
| Southeast | 284767 (50.6) | 186931 (51.6) | 97836 (48.7) | |
| Midwest | 52441 (9.3) | 35811 (9.9) | 16630 (8.3) | |
From the comparison between the clinical characteristics of surviving and non-surviving patients, it was found that, among the non-survivors, there was a higher proportion of those who presented at least one risk factor ($77.3\%$), remained hospitalized for a longer period, with a median of 10 days (Q1 = 4; Q3 = 18), required ICU admission ($63.6\%$), and required IMV ($48.7\%$), as presented in Table 2.
**Table 2**
| Variable | General (n = 563051) | Survivors (n = 362144) | Non-survivors (n = 200907) | p-value |
| --- | --- | --- | --- | --- |
| Risk factors/ Comorbidities (n = 563051) – n (%) | | | | |
| No | 189380 (33.6) | 143814 (39.7) | 45566 (22.7) | <0.001a |
| Yes | 373671 (66.4) | 218330 (60.3) | 155341 (77.3) | |
| Puerperium (n = 231611) | 1596 (0.7) | 1322 (0.9) | 274 (0.3) | <0.001a |
| Chronic Cardiovascular Disease (n = 306950) | 200840 (65.4) | 112771 (62.9) | 88069 (68.9) | <0.001a |
| Chronic Hematologic Disease (n = 232800) | 4477 (1.9) | 2343 (1.7) | 2134 (2.3) | <0.001a |
| Down’s Syndrome (n = 232517) | 1497 (0.6) | 872 (0.6) | 625 (0.7) | 0.346a |
| Chronic Liver Disease (n = 232293) | 5266 (2.3) | 2348 (1.7) | 2918 (3.1) | <0.001a |
| Asthma (n = 235855) | 15461 (6.6) | 10986 (7.8) | 4475 (4.7) | <0.001a |
| Diabetes mellitus (n = 286688) | 147681 (51.5) | 81704 (48.8) | 65977 (55.2) | <0.001a |
| Chronic Neurological Disease (n = 238951) | 23298 (9.8) | 9839 (7.0) | 13459 (13.7) | <0.001a |
| Chronic Pneumopathy (n = 238677) | 22845 (9.6) | 10280 (7.3) | 12565 (12.8) | <0.001a |
| Immunodepression (n = 234988) | 15386 (6.5) | 7689 (5.5) | 7697 (8.0) | <0.001a |
| Chronic Kidney Disease (n = 238497) | 23962 (10.0) | 9912 (7.1) | 14050 (14.3) | <0.001a |
| Obesity (n = 237409) | 36254 (15.3) | 22785 (16.1) | 13469 (14.1) | <0.001a |
| Other morbidities (n = 279031) | 160081 (57.4) | 90087 (55.2) | 69994 (60.4) | <0.001a |
| Hospitalization Time (n = 563051) med (Q1-Q3) | 8 (4–14) | 7 (4–12) | 10 (4–18) | <0.001b |
| ICU Admission (n = 511274) – n (%) | | | | |
| No | 312729 (61.2) | 247299 (74.6) | 65430 (36.4) | <0.001a |
| Yes | 198545 (38.8) | 84012 (25.4) | 114533 (63.6) | |
| IMV (n = 489787) – n (%) | | | | |
| No | 384436 (78.5) | 296041 (93.2) | 88395 (51.3) | <0.001a |
| Yes | 105351 (21.5) | 21530 (6.8) | 83821(48.7) | |
| NIV (n = 489787) – n (%) | | | | |
| No | 229166 (46.8) | 127297 (40.1) | 101869 (59.2) | <0.001a |
| Yes | 260021 (53.2) | 190274 (59.9) | 70347 (40.8) | |
Table 3 presents the results of the univariate and multivariate Cox Regression, including all variables with completeness greater than $80\%$. In the univariate analysis, it was observed that the effect of all independent variables was significant to explain the risk of death in patients hospitalized with SARS due to COVID-19. Following multivariate analysis, higher age, increasing the risk with each passing year (HR = 1.026; $95\%$ CI = 1.025–1.026); male sex (HR = 1.052; $95\%$ CI = 1.042–1.062); living in the North (HR = 1.429; $95\%$ CI = 1.397–1.462), Northeast (HR = 1.271; $95\%$ CI = 1.247–1.297) or Southeast regions of Brazil (HR = 1.040; $95\%$ CI = 1.021–1.058), presenting some risk factor (HR = 1.129; $95\%$ CI = 1.115–1.143), using IMV (HR = 2.865; $95\%$ CI = 2.812–2.919) or NIV (HR = 1.401; $95\%$ CI = 1.378–1.425).
**Table 3**
| Unnamed: 0 | Univariate Analysis | Univariate Analysis.1 | Univariate Analysis.2 | Multivariate Analysis | Multivariate Analysis.1 | Multivariate Analysis.2 |
| --- | --- | --- | --- | --- | --- | --- |
| Variable | HR | CI 95% | p-value | HR | CI 95% | p-value |
| Age (years) | 1.027 | 1.027–1.027 | <0.001a | 1.026 | 1.025–1.026 | <0.001a |
| Sex | | | | | | |
| Female | 1.000 | - | - | 1.000 | - | - |
| Male | 1.012 | 1.003–1.021 | 0.007a | 1.052 | 1.042–1.062 | <0.001a |
| Region | | | | | | |
| Midwest | 1.000 | - | - | 1.000 | - | - |
| North | 1.341 | 1.313–1.369 | <0.001a | 1.429 | 1.397–1.462 | <0.001a |
| Northeast | 1.339 | 1.315–1.363 | <0.001a | 1.271 | 1.247–1.297 | <0.001a |
| South | 0.919 | 0.901–0.937 | <0.001a | 0.852 | 0.835–0.870 | <0.001a |
| Southeast | 1.075 | 1.057–1.093 | <0.001a | 1.040 | 1.021–1.058 | <0.001a |
| Risk Factor Comorbidity | 1.386 | 1.372–1.401 | <0.001a | 1.129 | 1.115–1.143 | <0.001a |
| ICU admission | 1.567 | 1.552–1.583 | <0.001a | 0.950 | 0.938–0.962 | <0.001a |
| IMV | 2.238 | 2.217–2.260 | <0.001a | 2.865 | 2.812–2.919 | <0.001a |
| NIV | 0.686 | 0.680–0.693 | <0.001a | 1.401 | 1.378–1.425 | <0.001a |
Fig 1 presents the accumulated risk of death of patients hospitalized with SARS due to COVID-19 according to the patients’ region of residence and time of hospitalization. Throughout their hospital stay, a higher accumulated risk was observed in the North and Northeast regions, while the South region presented the lowest risk.
**Fig 1:** *The accumulated risk of deaths of patients hospitalized with SARS due to COVID-19 in 2020, according to their region of residence and length of hospitalization.*
## Discussion
This study provided nationwide data on the epidemiology and clinical course during the hospitalization of patients with SARS due to COVID-19. In the epidemiological weeks analyzed, the case lethality rate among patients hospitalized with SARS due to COVID-19 was $35.7\%$. Patients who were older, male, living in the North, Northeast, or Southeast regions of Brazil, who presented any risk factor and used IMV or NIV were at a higher risk of dying during hospitalization.
In another nationwide study, whose data source was the registration of deaths by the Civil Registry Offices, an excess of mortality in Brazil was proven, as early as the onset of the pandemic, in the months of March to May 2020, totaling an excess of 39,146 deaths for the period studied [11]. It should be noted that the analysis of lethality by COVID-19 should consider a combination of factors, such as the intrinsic characteristics of infected individuals (age, previous diseases, lifestyle habits) and the availability of therapeutic resources (hospital beds, healthcare teams, mechanical ventilators, and drugs) [12].
In this study, it has been found that deaths prevailed among the elderly, with a median age of 71 years, and among male patients ($57.4\%$), and that the mortality risk increased with each passing year and for males. A study conducted in the city of Wuhan found that the median age of patients affected by COVID-19 was 60 years, with the greatest severity of disease occurring among those aged 65 years or older. In addition, slightly more than half of the affected patients ($50.9\%$) were male, and these patients developed the most severe form of the disease. Both factors were associated with mortality due to COVID-19 [13]. An Indian study also associated higher mortality rates with ages equal to or greater than 60 years [14]. In another study, also conducted in Wuhan, older age and low lymphocyte count were associated with higher mortality rates among hospitalized patients [15]. The worse outcomes for males were corroborated in a study conducted in Beijing, with a similar prevalence of COVID-19 between men and women, however with a higher risk of death (2.4 times higher) for men, regardless of age [16].
A major research project led by Global Health 5050 and its partners is working to build the world’s largest database on sex and gender and its interface with the health policies in place, especially regarding the COVID-19 pandemic. Among the countries already analyzed, it was observed that there are few or no public health policies related to gender, and those that exist are highly focused on maternal health. Among the confirmed cases of the disease, a higher lethality rate is observed among men when compared to women, even in countries with a higher number of females among the confirmed cases. The higher occurrence of deaths among men may be related to immunological and hormonal issues, such as higher levels of angiotensin-converting enzymes; lower access or use of health services by this population; as well as behavioral issues, such as lower adherence to preventive actions and greater exposure to the virus and other harmful agents such as smoking, which contributes to the development of comorbidities [17]. Sex and gender differences should be accounted for, including in therapeutic interventions for the treatment of COVID-19 [18].
According to this study’s results, the fact that these patients presented any comorbidity was associated with higher lethality rates, which is corroborated by several studies, which have identified the association between hypertension, diabetes, chronic obstructive pulmonary disease, heart disease, neoplasms, and HIV and morbidity and mortality due to COVID-19 [19]. The lethality rate is increased by $10.5\%$ for cardiovascular diseases, $7.3\%$ for diabetes, $6.3\%$ for chronic respiratory diseases, and $6\%$ for patients with hypertension [20]. Two studies conducted in Wuhan, an initial focus city for SARS-CoV-2, also identified the presence of comorbidities as one of the factors associated with mortality, with hypertension standing out [13, 21]. A North American study identified that the largest proportion of patients hospitalized due to COVID-19 was composed of males, with advanced age, a history of smoking, and coexisting medical conditions, such as asthma, chronic obstructive pulmonary disease, hypertension, obesity, diabetes mellitus, chronic kidney disease, and cancer [22]. Thus, special attention should be directed to the elderly population, which is also more vulnerable to the development of comorbidities.
Critically ill patients affected by SARS-CoV-2 may present distinct pathophysiological mechanisms, influencing the treatment and the definition of the most adequate ventilatory support strategies for each case [23]. In this study’s findings, lethality was associated with the use of IMV and noninvasive devices. A study conducted in Switzerland, when determining the predictors of hospital mortality related to COVID-19 in elderly patients aged 65 years or older, identified as one of the risk factors the higher requirement of a fraction of inspired oxygen (FiO2) in NIV, that is, each $2\%$ increase in FiO2 added $7\%$ to the risk of death [24]. A study conducted in Italy also identified the high requirement of FiO2 as one of the independent risk factors associated with mortality [25]. However, in Brazil, the negative influences of the absence of a national protocol for the treatment of critically ill patients and the shortage of a properly qualified and trained team for intensive care are highlighted [26].
Differences among the regions in Brazil have been found, highlighting the highest risk of death in the North and Northeast regions, and the lowest risk in the South region. A Brazilian observational, ecological and analytical study, with national coverage, related the mortality of the elderly due to COVID-19 to demographic aspects and income distribution and identified higher mortality in states of the North, Northeast, and Southeast regions. The North region was ranked first place since its large territorial extension and poor transportation routes possibly hinders access to healthcare services [27]. The North region and Northeast region’s first-place ranking in mortality can also be explained by the chronic state of social vulnerability in which these populations are found [12]. The scarcity of hospital resources, such as ICU and pulmonary ventilators contribute to the virus lethality rates, a concerning situation which is inherent to the response of healthcare services, with the Northern region having the lowest quantity of these hospital resources [28]. The local, social and demographic characteristics should be accounted for, since *Brazil is* composed of a large and non-homogeneously distributed population, with cultural and geographical differences, in addition to social inequalities and unequal access to healthcare services [29].
The analysis based on SIVEP-*Gripe data* enables monitoring the pandemic caused by COVID-19, the definition of strategies for the prevention and control of the disease, and the evaluation of its impact on morbidity and mortality at the national level. Thus, the importance of national health information systems is highlighted as sources of information that support the planning of health policies and programs, contribute to the decision-making process, and allow the evaluation of the impact of interventions. Although these systems present limitations, they are relevant tools for public health, especially for the epidemiological surveillance of diseases [30].
The low completeness in the entries of some variables of SIVEP-Gripe, along with notification errors or delays in feeding the system in some EWs could interfere with the number of cases or deaths. In order to minimize this limitation, the sample included only the notification forms with complete information regarding the outcome date (discharge or death), that is, cases considered closed in the system. Moreover, only variables presenting more than $80\%$ completeness were included in the Cox *Regression analysis* in order to avoid incorrect inferences. As a strength of this study, the sample size is highlighted, consisting of 563,051 patients hospitalized with SARS due to COVID-19, allowing the tracing of the sociodemographic and clinical epidemiological profile at a national level.
In conclusion, a high case lethality rate was evidenced in patients with SARS due to COVID-19, however, it was found that the fatalities were not equally distributed throughout all regions of Brazil, with the Northeast and North regions presenting the highest lethality rate. Among the factors associated with the occurrence of death, it was found that older patients, who were male, living in the North, Northeast, or Southeast regions of Brazil, who presented any comorbidity and were submitted to IMV or NIV were at a higher risk of death. The recognition of the epidemiological profile from the results obtained can foster decision-making by healthcare providers and managers regarding more effective and equitable interventions, considering the regional diversities found.
## References
1. 1World Health Organization. Folha informativa sobre COVID-19—OPAS/OMS | Organização Pan-Americana da Saúde [Internet]. Genebra. World Health Organization; 2020 [acessado em 25 Mai. 2021]. Disponível em: https://www.paho.org/pt/covid19. *Folha informativa sobre COVID-19—OPAS/OMS | Organização Pan-Americana da Saúde* (2020.0)
2. 2World Health Organization. Coronavirus disease 2019 (COVID-19) Situation Report– 75. [internet]. Genebra: World Health Organization, 2020 [cited 2021 May 23]. Available from: https://www.who.int/docs/default-source/coronaviruse/situation-reports/20200404-sitrep-75-covid-19.pdf?sfvrsn=99251b2b_4. *Coronavirus disease 2019 (COVID-19) Situation Report– 75* (2020.0)
3. 3Brasil. Ministério da Saúde. Secretaria de Vigilância em Saúde. Guia de Vigilância Epidemiológica: Emergência de Saúde Pública de importância nacional pela doença pelo Coronavírus 2019—COVID-19 [Internet]. 2021 [acessado em 12 maio 2021]. Disponível em: https://coronavirus.saude.mg.gov.br/images/1_2021/17-03-Guia_de_vigilancia_da_covid_16marc2021.pdf
4. Noronha KVMS, Guedes GR, Turra CM, Andrade MV, Botega L, Nogueira D. **Pandemia por COVID-19 no Brasil: análise da demanda e da oferta de leitos hospitalares e equipamentos de ventilação assistida segundo diferentes cenários. Cad**. *Saúde Pública* (2020.0) **36**. DOI: 10.1590/0102-311X00115320
5. Walker PGT, Whittaker C, Watson O, Baguelin M, Ainslie KEC, Bhatia S. **Report 12: The Global Impact of COVID-19 and Strategies for Mitigation and Suppression**. *Imperial College COVID-19 Response Team* (2020.0)
6. Alkundi A, Mahmoud I, Musa A, Naveed S, Alshawwaf M. **Clinical characteristics and outcomes of COVID-19 hospitalized patients with diabetes in the United Kingdom: a retrospective single centre study**. *Diabetes Res Clin Pract* **165** 108263. DOI: 10.1016/j.diabres.2020.108263
7. Alshogran OY, Altawalbeh SM, Al-Azzam SI, Karasneh R. **Predictors of Covid-19 case fatality rate: Na Ecological study**. *Annals of Medicine and Surgery* (2021.0) **65**. DOI: 10.1016/j.amsu.2021.102319
8. Maciel EL, Jabor P, Júnior EG, Tristão-Sá R, Lima RCD, Reis-Santos B. **Fatores associados ao óbito hospitalar por COVID-19 no Espírito Santo**. *Epidemiol. Serv. Saude* (2020.0) **29** e2020413. DOI: 10.1590/S1679-49742020000400022
9. Escobar AL, Rodriguez TDM, Monteiro JC. **Letalidade e características dos óbitos por COVID-19 em Rondônia: estudo observacional**. *Epidemiol. Serv. Saude* (2021.0) **30**. DOI: 10.1590/S1679-49742021000100019
10. Zhang X-B, Hu L, Ming Q, Wei X-J, Zhang Z-Y, Chen L-D. **Risk factors for mortality of coronavirus disease-2019 (COVID-19) patients in two centers of Hubei province, China: A retrospective analysis**. *PloS one* (2021.0) **16** e0246030. DOI: 10.1371/journal.pone.0246030
11. Silva GAe, Jardim BC, Santos CVBdos. **Excesso de mortalidade no Brasil em tempos de COVID-19**. *Ciênc. Saúde Coletiva* (2020.0) **25**. DOI: 10.1590/1413-81232020259.23642020
12. Souza CDFde, Paiva JPSde, Leal TC, Silva LFda, Santos LG. **Evolução espaçotemporal da letalidade por COVID-19 no Brasil, 2020**. *J Bras Pneumol* (2020.0) **46** e20200208. DOI: 10.36416/1806-3756/e20200208
13. Li X, Xu S, Yu M, Wang K, Tao Y, Zhou Y. **Risk Factors for Severity and Mortality in Adult COVID-19 inpatients in Wuhan**. *Journal of Allergy and Clinical Immunology* (2020.0) **146** 110-118. DOI: 10.1016/j.jaci.2020.04.006
14. Mishra V, Burma AD, Das SK, Parivallal MB, Amudhan S, Rao GN. **COVID-19-Hospitalized Patients in Karnataka: Survival and Stay Characteristics**. *Indian Journal of Public Health* (2020.0) **64** 221-224. DOI: 10.4103/ijph.IJPH_486_20
15. Sun H, Ning R, Tao Y, Yu C, Deng X, Zhao C. **Risk Factors for Mortality in 244 Older Adults With COVID-19 in Wuhan, China: A Retrospective Study**. *Journal of the American Geriatrics Society* (2020.0) **68** e19-e23. DOI: 10.1111/jgs.16533
16. Jin J-M, Bai P, He W, Wu F, Liu X-F, Han D-M. **Gender Differences in Patients With COVID-19: Focus on Severity and Mortality**. *Frontiers in Public Health* (2020.0) **8**. DOI: 10.3389/fpubh.2020.00152
17. 17Global Health 5050. The sex, gender and COVID-19 Project. [Internet]. Global Health 5050, 2020 [cited 2021 May 29]. Available from: https://globalhealth5050.org/the-sex-gender-and-covid-19-project/. *The sex, gender and COVID-19 Project* (2020.0)
18. Gebhard C, Regitz-Zagrosek V, Neuhauser HK, Morgan R, Klein SL. **Impact of sex and gender on COVID-19 outcomes in Europe**. *Biology of Sex Differences* (2020.0) **11**. DOI: 10.1186/s13293-020-00304-9
19. Ejaz H, Alsrhani A, Zafar A, Javed H, Junaid K, Abdalla AE. **COVID-19 and Comorbidities: Deleterious Impact on Infected Patients**. *Journal of Infection and Public Health* (2020.0) **13** 1833-1839. DOI: 10.1016/j.jiph.2020.07.014
20. Wu Z, McGoogan JM. **Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention.**. *JAMA* (2020.0) **323** 1239-1242. DOI: 10.1001/jama.2020.2648
21. Zhang J-J, Dong X, Cao Y-Y, Yuan Y-D, Yang Y-B, Yan Y-Q. **Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan**. *China. Allergy* (2020.0) **75** 1730-1741. DOI: 10.1111/all.14238
22. Mikami T, Miyashita H, Yamada T, Harrington M, Steinberg D, Dunn A. *Risk Factors for mortality in patients with COVID-19 in New York City* (2020.0). DOI: 10.1007/s11606-020-05983-z
23. Gattinoni L, Chiumello D, Caironi P, Busana M, Romitti F, Brazzi L. **COVID-19 pneumonia: different respiratory treatments for different phenotypes?**. *Intensive Care Med* (2020.0) **46**. DOI: 10.1007/s00134-020-06033-2
24. Mendes A, Serratrice C, Herrmann FR, Genton L, Périvier S, Scheffler M. **Predictors of In-Hospital Mortality in Older Patients With COVID-19: The COVIDAge Study**. *Journal of the American Medical Directors Association* (2020.0) **21** 1546-1554.e3. DOI: 10.1016/j.jamda.2020.09.014
25. Grasselli G, Greco M, Zanella A, Albano G, Antonelli M, Bellani G. **Risk Factors Associated With Mortality Among Patients With COVID-19 in Intensive Care Units in Lombardy, Italy**. *JAMA Internal Medicine* (2020.0) **180** 1345-1355. DOI: 10.1001/jamainternmed.2020.3539
26. Ranzani OT, Bastos LSL, Gelli JGM, Marchesi JF, Baião F, Hamacher S. **Characterisation of the first 250,000 hospital admissions for COVID-19 in Brazil: a retrospective analysis of nationwide data**. *Lancet Respir Med* (2021.0) **9** 407-18. DOI: 10.1016/S2213-2600(20)30560-9
27. Barbosa IR, Galvão MHR, Souza TA, Gomes SM, Medeiros AA, Lima KCde. **Incidence of and mortality from COVID-19 in the older Brazilian population and its relationship with contextual indicators: an ecological study**. *Rev. bras. geriatr. gerontol* (2020.0) **23**. DOI: 10.1590/1981-22562020023.200171
28. Moreira RS. **COVID-19: unidades de terapia intensiva, ventiladores mecânicos e perfis latentes de mortalidade associados à letalidade no Brasil**. *Cad. Saúde Pública* (2020.0) **36** e00080020. DOI: 10.1590/0102-311x00080020
29. Cavalcante JR, Cardoso-dos-Santos AC, Bremm JM, Lobo AP, Macário EM, Oliveira WKde. **COVID-19 no Brasil: evolução da epidemia até a semana epidemiológica 20 de 2020**. *Epidemiol. Serv. Saúde* (2020.0) **29**. DOI: 10.5123/S1679-49742020000400010
30. Araujo KLR, Aquino EC, Silva LLS, Ternes YMF. **Fatores associados à Síndrome Respiratória Aguda Grave em uma Região Central do Brasil**. *Ciênc. Saúde Coletiva* (2020.0) **25**. DOI: 10.1590/1413-812320202510.2.26802020
|
---
title: Perceived stress and hair cortisol concentration in a study of Mexican and
Icelandic women
authors:
- Rebekka Lynch
- Mario H. Flores-Torres
- Gabriela Hinojosa
- Thor Aspelund
- Arna Hauksdóttir
- Clemens Kirschbaum
- Andres Catzin-Kuhlmann
- Martín Lajous
- Unnur Valdimarsdottir
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021558
doi: 10.1371/journal.pgph.0000571
license: CC BY 4.0
---
# Perceived stress and hair cortisol concentration in a study of Mexican and Icelandic women
## Abstract
Hair cortisol concentration (HCC) represent a potential biomarker of chronic psychological stress. Previous studies exploring the association between perceived stress and HCC have been limited to relatively small and selected populations. We collected hair samples from 881 women from the Mexican Teachers’ Cohort (MTC) and 398 women from the Icelandic SAGA pilot-cohort following identical protocols. HCC was quantified using liquid chromatography coupled with tandem mass spectrometry. The self-reported Perceived Stress Scale (PSS, 10 and 4 item, range 0–40 and 0–16) was used to assess psychological stress. We conducted multivariable linear regression analyses to assess the association between perceived stress and log-transformed HCC in the combined sample and in each cohort separately. MTC participants had slightly higher HCC and PSS scores than SAGA participants (median HCC 6.0pg/mg vs. 4.7pg/mg and mean PSS-10 score 12.4 vs. 11.7, respectively). After adjusting for sociodemographic factors and health behaviors, we observed a $1.4\%$ ($95\%$ CI 0.6, 2.1) increase in HCC for each unit increase in the PSS-10 score in the combined sample. Furthermore, PSS-10 quintiles were associated with a $24.3\%$ ($95\%$ CI 8.4, 42.6, mean logHCC 1.8 vs 1.6) increase in HCC when comparing the highest to the lowest quintile, after multivariable adjustment. Similar results were obtained when we analyzed each cohort separately and when using the PSS-4. Despite relatively small absolute differences, an association between perceived stress and HCC was found in a sample of women from two diverse geographical and cultural backgrounds supporting the hypothesis that HCC is a viable biomarker in studies of chronic psychological stress.
## Introduction
Stress, despite its colloquial connotations, is a scientific term used to describe cognitive, emotional and physiologic reactions resulting from an imbalance between demands and resources as perceived by the individual [1]. This definition puts less weight on the nature of the event in question, rather it is the individual’s perception of the event that is paramount. Within this framework, the Perceived Stress Scale (PSS) was developed by Cohen et al [2,3] and has been widely used in research studies. Results from these studies provide evidence of the association of high PSS scores with psychiatric and somatic outcomes, including symptoms of depression and anxiety [4], increased susceptibility for infections [5] and cardiovascular disease [6]. However, the biological mechanisms of the stress-disease pathway in humans are poorly understood, partly due to the inherent complexities of measuring the stress response.
Cortisol is a hormone continuously secreted by the hypothalamic–pituitary–adrenal (HPA) axis in a diurnal rhythm but also in response to stress [7]. Cortisol levels can be measured in blood, urine, and saliva; however, these measurements may be more suited to measure acute stress responses, as their concentration changes rapidly, which may contribute to conflicting results in the stress literature [8]. Measuring cortisol concentration in hair has been proposed as an alternative way to consistently assess cumulative exposure to this stress-related hormone over a period of weeks to months [9], as well as a much simpler and non-invasive method of assessment. However, hair cortisol concentration differs both by gender and changes through the life course [10].
A 2017 meta–analysis of 2,441 individuals from 26 studies did not find an overall association between perceived stress and hair cortisol concentration (HCC) [10]. However, the study reported an increase in HCC in subpopulations with ongoing life conditions of severe stress, defined as exposure to a psychological condition that is threatening and/or exceeds coping resources that either persisted for at least one month or could be appraised as threatening for a month (e.g. trauma). The included studies were based on small samples (the majority with fewer than 100 participants) and very selected populations, e.g. fulltime dementia caregivers [11], soldiers [12]. Additionally, the majority of studies have been conducted in European populations, with only one study performed in Latin America [13], and used different assaying protocols to measure HCC. We aimed to assess the association between perceived stress and HCC in two culturally and geographically distinct samples of adult women from population-based studies in Mexico and Iceland. We hypothesized that HCC might be a viable biomarker for chronic psychological stress among women across these two different cultures.
## The Mexican Teachers’ Cohort (MTC)
The MTC is a prospective, closed cohort study of 115,314 women teachers, aged 25 years and older, that began in 2006–2008 in twelve geographically and economically diverse states in Mexico [14]. At baseline, and every three years, participants responded to questionnaires on their demographic and reproductive characteristics, lifestyle, and health. Between May 2016 and June 2017, a random sample of 2,003 study participants (aged ≥40 years and living within a 50 km radius from a clinical site) in two Mexican states were invited to take part in an ancillary study on stress and cardiovascular disease, among others. A clinical evaluation took place in two sites: the city of Monterrey, Nuevo Leon, and Mexico City. Close to $60\%$ ($$n = 1$$,145) of those invited participated in the study.
## The SAGA Cohort
The SAGA *Cohort is* a nationwide study on the impact of trauma on women’s health, that launched in 2018, with a recruitment period of one year [15]. The target population is all women, 18–69 years old, residing in Iceland. In the pilot study performed in 2014, a random sample of 689 adult women living in the greater *Reykjavik area* and attending the national Cancer Detection Clinic were invited to participate. All women had appointments for routine breast and cervical cancer screens from February through April of 2014. Nationally, $85\%$ and $92\%$ of eligible Icelandic women have attended the respective screens at least once [16]. Over $70\%$ ($$n = 509$$) of the invited women responded to the online questionnaire and then attended the clinical visit in conjunction with their screening.
## Inclusion and exclusion criteria
For the present study, we included all women, 70 years or younger, who provided a hair sample at a clinical visit in both studies (1,111 from the MTC, and 492 from SAGA). Every woman from the MTC that attended a clinical visit donated a hair sample, but ten women from SAGA did not. We excluded pregnant women (3 ($0.6\%$) from SAGA) (as cortisol levels rise during pregnancy [17]), women with an incomplete or incorrectly completed 10-item PSS-10 (79 ($7.1\%$) from the MTC, and 33 ($6.7\%$) from SAGA), and those with undetectable cortisol values (77 ($6.9\%$) from the MTC and 23 ($4.7\%$) from SAGA). We also excluded women with a cortisol: cortisone ratio > 3 (74 ($6.7\%$) from the MTC and 35 ($7.1\%$) from SAGA), as these values are implausible and probably due to exogenous cortisol [18,19]. Thus, the final study sample comprised 1,279 women, 881 from the MTC and 398 from SAGA.
All women provided written informed consent. In Mexico, the institutional review boards at the National Institute of Public Health and the School of Medicine and Health Sciences of the Monterrey Institute of Technology approved the project. In Iceland, the study was approved by the National Bioethics Committee (VSNb$\frac{2013010025}{03.07}$) and reported to the Icelandic Data Protection Authority. Data transfer was approved by the Mexican institutional review boards and combined analyses were performed at the University of Iceland.
## Assessment of hair cortisol concentration (HCC)
For participants in both countries (Mexico and Iceland), hair samples were collected during the clinical visit to measure cortisol and cortisone levels. Samples were taken from the posterior vertex by study personnel, if there was at least 3cm of hair growth, and stored in aluminum foil. As hair grows an estimated 1 cm per month, the 3 cm closest to the scalp represent the prior three months [20]. The hair samples were stored at room temperature in a dark and dry environment. The samples were centrally assayed to determine hair cortisol and cortisone levels at the Dresden University of Technology (TU Dresden) in Germany under the supervision of one of the coauthors (CK). Liquid chromatography coupled with tandem mass spectrometry (LCMS/MS) was used to measure cortisol and cortisone values in the hair samples (see Gao et al [21] for a more detailed description).
## Assessment of perceived stress
Stress was assessed with the 10-item PSS-10, developed by Cohen et al [3]. The PSS-10 focuses on “the degree to which respondents found their lives unpredictable, uncontrollable, and overloading” [3], rather than assessing a specific life-event. The scale covers the past month, using a 5-point response scale (0 = “never”, 1 = “almost never”, 2 = “sometimes”, 3 = “fairly often”, 4 = “very often”). As per standard practice PSS-10 scores were obtained by reversing the scores on the four positive items (4, 5, 7 and 8) and summing across all 10 items, with higher scores indicating higher levels of perceived stress (scores range from 0 to 40) [22]. We additionally calculated a shorter PSS-4 score based on questions 2, 4 (reversed), 5 (reversed) and 10 of the original PSS-10 (scores range from 0–16), as its brevity could make it valuable for larger epidemiological studies. Since the PSS is not a diagnostic instrument and does not have established cutoff values, we categorized participants’ scores using quintiles of the PSS-10 total scores. Participants responded to a paper questionnaire that included the PSS-10 during the clinical visit for the Mexican sample in Spanish and in an online questionnaire for the Icelandic sample in Icelandic. Internal consistency for the full scoring scale was previously found to be high (Cronbach’s alpha $r = 0.83$) in a validation study of 1,310 MTC participants [23] and is similar in a post-hoc analysis of the current Icelandic sample ($r = 0.82$).
## Covariate assessment
Other variables were selected and harmonized to compare characteristics of participants from both countries based on availability and proposed frameworks for studying stress in human populations [24]. Missing values were imputed to the most frequent value (categorical) or the median value (numerical) for each variable, with $89.1\%$ of MTC participants and $96.7\%$ of SAGA participants having complete data. S1 Table includes detailed information on variable selection, harmonization and missingness. The sociodemographic factors assessed in both samples were age, marital status, education level and employment status. Information on health behaviors including alcohol consumption, smoking status and body mass index (measured clinically) was also collected in both cohorts.
## Statistical analysis
We combined data from both cohort studies to maximize power and assess general patterns in the distribution of the effects. We explored the distribution of participant characteristics and the graphical distribution of PSS scores and HCC in the combined sample and in each cohort separately. HCC was log-transformed to normalize its distribution. We fit linear regression models adjusted by age and cohort (only for the overall sample) to estimate the percentage difference in mean HCC and $95\%$ confidence interval ($95\%$ CI). We repeated the analyses using the quintiles of PSS-10 and -4 instead of a continuous PSS score, with the lowest quintile as the reference. To test for trend, we assigned the median PSS score within each quintile and fit a model with the exposure as a continuous variable. Finally, we re-ran analyses as a multivariable model with age, cohort (only for the overall sample), marital status, occupation, educational level, BMI (continuous), and smoking status, as these present possible confounders in the association between perceived stress and HCC. Analyses were conducted using R version 3.4.2.
## Results
Characteristics of study participants in the combined sample and by cohort (881 from MTC and 398 from SAGA) are presented in Table 1. In the combined sample, the mean age was 51.2 (SD 7.7), with the majority married ($71.2\%$) and employed ($89.8\%$). The mean BMI for participants was 28.9 (5.8), with $12.9\%$ current smokers. MTC participants were on average younger and less likely to be ≥60 years than SAGA participants. MTC were also more likely to be obese, and less likely to smoke or consume alcohol compared with SAGA participants. When we explored the distribution of the combined sample’s participant characteristics by quintiles of PSS-10 scores (S2 Table), we observed that, compared with women in the lowest quintile, those in the highest quintiles were younger, less likely to have a graduate degree or to be married, as well as more likely to report other occupation status. Women in the highest quintiles were also more likely to be obese and to currently smoke.
**Table 1**
| Unnamed: 0 | OverallN = 1279 | MTCN = 881 | SAGAN = 398 |
| --- | --- | --- | --- |
| Hair Cortisol Concentration | | | |
| Median cortisol (pg/mg) | 5.5 | 6.0 | 4.7 |
| 25th -75th percentile | 3.4–8.9 | 3.7–9.5 | 3.0–7.6 |
| Mean log-cortisol (SD)a | 1.7 (0.8) | 1.8 (0.8) | 1.6 (0.7) |
| Mean age (SD)a | 51.2 (7.7) | 50.9 (6.0) | 52.0 (10.6) |
| Age in years (%) | | | |
| 20–39 | 72 (5.6) | 40 (4.5) | 32 (8.0) |
| 40–49 | 386 (30.2) | 268 (30.5) | 118 (29.6) |
| 50–59 | 664 (51.9) | 522 (59.3) | 142 (35.7) |
| 60–70 | 157 (12.3) | 51 (5.8) | 106 (26.6) |
| Graduate Degree (%) | 261 (20.4) | 190 (21.6) | 71 (17.8) |
| Occupation (%) | | | |
| Employed | 1148 (89.8) | 836 (94.9) | 312(78.4) |
| Retired | 62 (4.8) | 45 (5.1) | 17 (4.3) |
| Otherb | 69 (5.4) | 0 (0.0) | 69 (17.3) |
| Marriage/Cohabitation (%) | 911 (71.2) | 619 (70.3) | 292 (73.4) |
| BMIc (SD)a | 28.9 (5.8) | 29.5 (5.8) | 27.5 (5.6) |
| BMIc category (%) | | | |
| Normal weight | 328 (25.6) | 182 (20.7) | 146 (36.7) |
| Overweight | 510 (39.9) | 370 (42.0) | 140 (35.2) |
| Obese | 441 (34.5) | 329 (37.3) | 112 (28.1) |
| Smoking (%) | | | |
| Never | 763 (59.7) | 592 (67.2) | 171 (43.0) |
| Former | 351 (27.4) | 177 (20.1) | 174 (43.7) |
| Current | 165 (12.9) | 112 (12.7) | 53 (13.3) |
| Alcohol consumption (drinks/month) (SD) a | 3.2 (5.0) | 1.9 (3.9) | 6.0 (5.8) |
| Mean PSS d -10 score (SD) a | 12.2 (6.0) | 12.4 (6.0) | 11.7 (6.0) |
| Mean PSS d -4 score (SD) a | 4.0 (2.7) | 3.9 (2.7) | 4.4 (2.8) |
The 1,279 study participants had a mean PSS-10 score of 12.2 (SD 6.0) (Table 1). The mean HCC before log transformation was 8.1 pg/mg (SD 19.1) and the median was 5.5 pg/mg (IQR 3.4–8.9 pg/mg) (S1 Fig, panel A). After log transformation, mean HCC was 1.7 (0.8) and the median was 1.7 (IQR 1.2–2.2) (S1 Fig, panel B). MTC participants had, on average, both slightly higher PSS-10 scores and higher log-transformed HCC than SAGA participants (12.4 vs 11.7 and 1.8 vs 1.6, respectively) (Table 1 and S2 Fig). The same pattern was not observed for the shorter PSS-4, with a mean score of 4.0 (SD 2.7), and higher values of PSS-4 among SAGA participants (4.4 vs 3.9) (Table 1).
We observed a linear association between perceived stress (based on the PSS-10 score) and HCC (Fig 1, panel A). After multivariable adjustment, we found that a unit increase in the PSS-10 score was associated with $1.4\%$ ($95\%$ CI 0.6, 2.1) increase in mean HCC. Results by country demonstrated an increase of $1.3\%$ ($95\%$ CI -0.7, 3.3) in mean HCC by unit increase in the PSS-10 score in the MTC and $2.0\%$ ($95\%$ CI 0.7, 3.2) in the SAGA. We observed similar results in age-adjusted models (S3 Fig). Likewise, the shorter PSS-4 results indicate a $2.3\%$ ($95\%$ CI 0.7, 4.0) increase in HCC for each unit increase in the PSS-4 score in the combined sample after adjusting for age. Estimates varied somewhat when each country sample was analyzed independently, with a non-significant increase ($1.4\%$, $95\%$ CI -0.6, 3.4) in the MTC sample and a significant increase of $4.3\%$ ($95\%$ CI 1.6, 7.1) in the SAGA sample (S5 Fig, panels B and C).
**Fig 1:** *Multivariable adjusted linear regression for log-transformed cortisol with 95% confidence intervals, with data from the Mexican Teacher’s Cohort (MTC) (N = 881) and the Icelandic SAGA Cohort (N = 398), both as a combined sample and by cohort according to 10-item Perceived Stress Scale score.The figures are scatterplots with an overlying linear regression and 95% confidence intervals. The results are adjusted for marital status, occupation, educational level, BMI, smoking status as well as cohort for the combined sample. The overall sample (Panel A) had a 1.4% (95% CI 0.6, 2.11) increase in log-cortisol per unit increase of PSS. The Mexican sample (Panel B) had a 1.3% (95% CI –0.7, 3.3) increase in log-cortisol. The Icelandic sample (Panel C) had a 2.0% (95% CI 0.7, 3.2) increase in log-cortisol. See S5 Fig for age-adjusted results.*
Similar results were obtained when we explored perceived stress by quintiles of PSS-10 (Fig 2, panel A), with both raw and log-transformed cortisol levels increasing with each quintile of perceived stress (S2 Table). Women in the 5th quintile had $24.3\%$ ($95\%$ CI 8.4, 42.6) higher mean HCC after multivariable adjustment, compared with women in the lowest quintile (p-trend <0.001, mean logHCC 1.6 vs 1.8), and $24.1\%$ ($95\%$ CI 8.3, 42.2) higher mean HCC after age adjustment. The data also indicates a plateau between the 4th and 5th quintile. A similar pattern was observed in both cohorts after age adjustment (S4 Fig, panels B and C), though the difference in HCC levels between the highest and lowest quintile of perceived stress appeared to be larger in SAGA ($33.3\%$, $95\%$ CI 5.7, 68.0, multivariable adjusted) and there was a decrease in the magnitude of the estimated effect from the 4th to the 5th quintile for the MTC (4th quintile $26.0\%$, $95\%$ CI 6.6, 48.9 and 5th quintile $18.6\%$, $95\%$ CI 0.2, 40.3). With regard to the PSS-4, the results were non-significant, but were suggestive of a J curve (S6 Fig).
**Fig 2:** *Differences in mean log-transformed cortisol by quintiles of the 10-item Perceived Stress Scale with data from the Mexican Teacher’s Cohort (MTC) (N = 881) and the Icelandic SAGA Cohort (N = 398), both as a combined sample and by cohort, with multivariable adjustment.The panels depict the mean log-cortisol value with the black dot representing the mean value and the bands representing 95% confidence intervals (logHCC with 95%CI) for each quintile of the Perceived Stress Scale (PSS). The results are adjusted for marital status, occupation, educational level, BMI, smoking status as well as cohort for the combined sample. Panel A includes the overall sample, panel B, the Mexican sample, and panel C the Icelandic sample. See S6 Fig for age-adjusted results.*
## Discussion
In this study of two cohorts of women from the general population of Mexico and Iceland, we found a dose-response association between perceived stress and HCC. Women in the highest quintile of the Perceived Stress Scale had significantly higher HCC compared with the lowest quintile and this association remained unchanged after multivariable adjustment, though the absolute difference was small (mean logHCC 1.8 vs 1.6). This stepwise increase is also suggested when each cohort was examined separately, though a plateau was seen in the highest quintiles in the Mexican MTC sample.
To date, this is the largest study on perceived stress and HCC. One of its strengths is the harmonized assessments and centralized assaying of women from two considerably diverse ethnic and geographic backgrounds yielding largely similar results. Another strength of this study is that both cohorts were made up of women, as earlier research has shown significant differences in both hair cortisol concentration [10] and perceived stress [25] by gender. While the study is cross-sectional, the design fits nicely with the research question on the association between two dynamic measures of stress–self-reported and biological—at a single point in time. Disentangling causal direction is outside the scope of this study, however, our results are partly supported by previous studies examining hair cortisol repeatedly (two and four times) through a year at college [26] and a medical internship [27]. Both found that HCC increased in response to stressful events but did not find an association with the Perceived Stress Scale.
With two distinct cohorts, one an occupational cohort and the other a general cohort of cancer screening participants, and on two different continents, it is likely that the stressors faced are quite diverse. Nevertheless, although some differences were observed between cohorts regarding stress levels, both in terms of self-perceived stress and hair cortisol concentration, and absolute differences were small, this study demonstrates a positive association between self-perceived stress and hair cortisol levels across two cultures with diverse sources of stress.
A 2017 meta-analysis did not find an overall association between perceived stress and HCC [10]. The meta-analysis summarized the findings of mostly small-scale studies using varying measures of perceived stress which may contribute to the varying results. Two of the largest studies in this review did not use the Perceived Stress Scale but the 57-item Trier Inventory for the Assessment of Chronic Stress or its screening scale. Neither study found an association with hair cortisol concentration, one with 109 individuals from a convenience sample [28] and the other involving 654 older adults of both genders [29]. It is therefore possible that the Trier assessment does not adequately capture the facets of stress that are linked to cortisol dysregulation. Interestingly, a study of 37 couples ($$n = 74$$) that measured the Weekly Hassle Scale (WHS), the Perceived Stress Scale and the Triers Inventory for Assessment of Chronic Stress repeatedly over 12 weeks (12, 3, and 1 time respectively), and then assayed hair for cortisol found WHS alone predicted hair cortisol concentration. When using state-space modelling they found that all three explained an incremental variance portion of hair cortisol concentration–highlighting that no one scale perfectly captures hair cortisol changes [30].
Our results are more in line with a study of 324 Canadians, with oversampling of individuals with mental health problems or in abusive relationships, where a marginally positive association (Beta 0.107, unadjusted $$p \leq 0.057$$) was noted between the Perceived Stress Scale and hair cortisol concentration [31]. The curvilinear association reported in that study was similar to the results in the MTC cohort where the fourth quintile had the highest mean cortisol level.
The 2017 meta-analysis did find an association with on-going chronic stress, as defined by Miller et al [32], with a $43\%$ increase in hair cortisol concentration. Of eight studies, half included some measure of PSS, with no study showing an association between hair cortisol levels and perceived stress, but all showing an association with event-based chronic stress. The largest of these studies included 85 caregivers [13], underscoring the lack of power that plagues earlier studies.
The PSS-4 is a shorter version of the Perceived Stress Scale and may be better suited for large epidemiological studies. Few studies have used this scale, though they exist [26,33]. Our main result, a positive linear association, remained the same when this smaller scale was used, however when divided into quintiles the association was slightly attenuated compared with the full PSS-10. This may reflect the fact that a smaller scale, although a valid tool, needs greater power to ascertain an association.
In summary, we found a modest positive association between perceived stress and hair cortisol concentration in two distinct cohorts of women, from Mexico and Iceland. The findings from earlier studies–using varying measures of stress—have been quite conflicting, however our findings indicate that the Perceived Stress Scale may in part reflect the underlying cortisol concentration in the general population of women. Further research is needed to assess this association, both regarding possible gender differences as well as through longitudinal study designs for further understanding of the temporal pattern in the association between perceived stress and HCC.
## References
1. Folkman S, Lazarus RS, Dunkel-Schetter C, DeLongis A, Gruen RJ. **Dynamics of a stressful encounter: cognitive appraisal, coping, and encounter outcomes**. *Journal of personality and social psychology* (1986.0) **50** 992-1003. DOI: 10.1037//0022-3514.50.5.992
2. Cohen S, Kamarck T, Mermelstein R. **A global measure of perceived stress**. *J Health Soc Behav* (1983.0) **24** 385-96. PMID: 6668417
3. Cohen S, Williamson G, Spacapan S, Oskamp S. *Social Psychology of Health: Claremont Symposium on Applied Social Psychology Newbury Park* (1988.0) 31-67
4. Lee E-H. **Review of the Psychometric Evidence of the Perceived Stress Scale**. *Asian Nursing Research* **6** 121-7. DOI: 10.1016/j.anr.2012.08.004
5. Cohen S, Tyrrell DA, Smith AP. **Negative life events, perceived stress, negative affect, and susceptibility to the common cold**. *Journal of personality and social psychology* (1993.0) **64** 131-40. DOI: 10.1037//0022-3514.64.1.131
6. Richardson S, Shaffer JA, Falzon L, Krupka D, Davidson KW, Edmondson D. **Meta-analysis of perceived stress and its association with incident coronary heart disease**. *The American journal of cardiology* (2012.0) **110** 1711-6. DOI: 10.1016/j.amjcard.2012.08.004
7. Spencer RL, Deak T. **A users guide to HPA axis research**. *Physiology & behavior* (2017.0) **178** 43-65. DOI: 10.1016/j.physbeh.2016.11.014
8. Stalder T, Kirschbaum C, Kudielka BM, Adam EK, Pruessner JC, Wüst S. **Assessment of the cortisol awakening response: Expert consensus guidelines**. *Psychoneuroendocrinology* (2016.0) **63** 414-32. DOI: 10.1016/j.psyneuen.2015.10.010
9. Raul JS, Cirimele V, Ludes B, Kintz P. **Detection of physiological concentrations of cortisol and cortisone in human hair**. *Clin Biochem* (2004.0) **37** 1105-11. DOI: 10.1016/j.clinbiochem.2004.02.010
10. Stalder T, Steudte-Schmiedgen S, Alexander N, Klucken T, Vater A, Wichmann S. **Stress-related and basic determinants of hair cortisol in humans: A meta-analysis**. *Psychoneuroendocrinology* (2017.0) **77** 261-74. DOI: 10.1016/j.psyneuen.2016.12.017
11. Spijker AT. **Elevated hair cortisol levels in chronically stressed dementia caregivers**. *Stress* (2014.0) **47** 26-30
12. Boesch M, Sefidan S, Annen H, Ehlert U, Roos L, Van Uum S. **Hair cortisol concentration is unaffected by basic military training, but related to sociodemographic and environmental factors**. *Stress* (2015.0) **18** 35-41. DOI: 10.3109/10253890.2014.974028
13. Chen X, Gelaye B, Velez JC, Barbosa C, Pepper M, Andrade A. **Caregivers’ hair cortisol: a possible biomarker of chronic stress is associated with obesity measures among children with disabilities**. *BMC pediatrics* (2015.0) **15** 9. DOI: 10.1186/s12887-015-0322-y
14. Lajous M, Ortiz-Panozo E, Monge A, Santoyo-Vistrain R, Garcia-Anaya A, Yunes-Diaz E. **Cohort Profile: The Mexican Teachers’ Cohort (MTC)**. *International journal of epidemiology* (2017.0) **46** e10. DOI: 10.1093/ije/dyv123
15. Asgeirsdottir HG, Valdimarsdottir UA, Thornorsteinsdottir Thorn K, Lund SH, Tomasson G, Nyberg U. **The association between different traumatic life events and suicidality**. *European journal of psychotraumatology* (2018.0) **9** 1510279. DOI: 10.1080/20008198.2018.1510279
16. Society IC. **Yearly Report 2015–2016**. *Reykjavik: Icelandic Cancer Society* (2016.0)
17. D’Anna-Hernandez KL, Ross RG, Natvig CL, Laudenslager ML. **Hair cortisol levels as a retrospective marker of hypothalamic–pituitary axis activity throughout pregnancy: Comparison to salivary cortisol**. *Physiology & behavior* (2011.0) **104** 348-53. DOI: 10.1016/j.physbeh.2011.02.041
18. Wang X, Busch JR, Banner J, Linnet K, Johansen SS. **Hair testing for cortisol by UPLC–MS/MS in a family: External cross-contamination from use of cortisol cream**. *Forensic Science International* (2019.0) **305** 109968. DOI: 10.1016/j.forsciint.2019.109968
19. Wester VL, Noppe G, Savas M, van den Akker ELT, de Rijke YB, van Rossum EFC. **Hair analysis reveals subtle HPA axis suppression associated with use of local corticosteroids: The Lifelines cohort study**. *Psychoneuroendocrinology* (2017.0) **80** 1-6. DOI: 10.1016/j.psyneuen.2017.02.024
20. Wennig R.. **Potential problems with the interpretation of hair analysis results**. *Forensic Sci Int* (2000.0) **107** 5-12. DOI: 10.1016/s0379-0738(99)00146-2
21. Gao W, Stalder T, Foley P, Rauh M, Deng H, Kirschbaum C. **Quantitative analysis of steroid hormones in human hair using a column-switching LC-APCI-MS/MS assay**. *J Chromatogr B Analyt Technol Biomed Life Sci* (2013.0) **928** 1-8. DOI: 10.1016/j.jchromb.2013.03.008
22. Cohen S, Kessler RC, Gordon LU. *Measuring stress: A guide for health and social scientists* (1997.0) **xii** 236-xii
23. Flores-Torres MH, Tran A, Familiar I, López-Ridaura R, Ortiz-Panozo E. **Perceived Stress Scale, a tool to explore psychological stress in Mexican women**. *Salud Pública de México* (2021.0) 1-8. PMID: 34098592
24. Epel ES, Crosswell AD, Mayer SE, Prather AA, Slavich GM, Puterman E. **More than a feeling: A unified view of stress measurement for population science**. *Frontiers in neuroendocrinology* (2018.0) **49** 146-69. DOI: 10.1016/j.yfrne.2018.03.001
25. Lavoie JAA, Douglas KS. **The Perceived Stress Scale: Evaluating Configural, Metric and Scalar Invariance across Mental Health Status and Gender**. *Journal of Psychopathology and Behavioral Assessment* (2012.0) **34** 48-57
26. Stetler CA, Guinn V. **Cumulative cortisol exposure increases during the academic term: Links to performance-related and social-evaluative stressors**. *Psychoneuroendocrinology* (2020.0) **114** 104584. DOI: 10.1016/j.psyneuen.2020.104584
27. Mayer SE, Lopez-Duran NL, Sen S, Abelson JL. **Chronic stress, hair cortisol and depression: A prospective and longitudinal study of medical internship**. *Psychoneuroendocrinology* (2018.0) **92** 57-65. DOI: 10.1016/j.psyneuen.2018.03.020
28. Stalder T, Steudte S, Miller R, Skoluda N, Dettenborn L, Kirschbaum C. **Intraindividual stability of hair cortisol concentrations**. *Psychoneuroendocrinology* (2012.0) **37** 602-10. DOI: 10.1016/j.psyneuen.2011.08.007
29. Feller S, Vigl M, Bergmann MM, Boeing H, Kirschbaum C, Stalder T. **Predictors of hair cortisol concentrations in older adults**. *Psychoneuroendocrinology* (2014.0) **39** 132-40. DOI: 10.1016/j.psyneuen.2013.10.007
30. Weckesser LJ, Dietz F, Schmidt K, Grass J, Kirschbaum C, Miller R. **The psychometric properties and temporal dynamics of subjective stress, retrospectively assessed by different informants and questionnaires, and hair cortisol concentrations**. *Scientific reports* (2019.0) **9** 1098. DOI: 10.1038/s41598-018-37526-2
31. Wells S, Tremblay PF, Flynn A, Russell E, Kennedy J, Rehm J. **Associations of hair cortisol concentration with self-reported measures of stress and mental health-related factors in a pooled database of diverse community samples**. *Stress* (2014.0) **17** 334-42. DOI: 10.3109/10253890.2014.930432
32. Miller GE, Chen E, Zhou ES. **If it goes up, must it come down? Chronic stress and the hypothalamic-pituitary-adrenocortical axis in humans**. *Psychological bulletin* (2007.0) **133** 25-45. DOI: 10.1037/0033-2909.133.1.25
33. Lehrer HM, Steinhardt MA, Dubois SK, Laudenslager ML. **Perceived stress, psychological resilience, hair cortisol concentration, and metabolic syndrome severity: A moderated mediation model**. *Psychoneuroendocrinology* (2020.0) **113** 104510. DOI: 10.1016/j.psyneuen.2019.104510
|
---
title: 'Multi-purpose cash transfers and health among vulnerable Syrian refugees in
Jordan: A prospective cohort study'
authors:
- Emily Lyles
- Stephen Chua
- Yasmeen Barham
- Dina Jardenah
- Antonio Trujillo
- Paul Spiegel
- Ann Burton
- Shannon Doocy
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021566
doi: 10.1371/journal.pgph.0001227
license: CC BY 4.0
---
# Multi-purpose cash transfers and health among vulnerable Syrian refugees in Jordan: A prospective cohort study
## Abstract
Cash assistance has rapidly expanded in the Syrian refugee response in Jordan and global humanitarian programming, yet little is known about the effect of multipurpose cash transfers (MPC) on health in humanitarian contexts. A prospective cohort study was conducted from May 2018 through July 2019 to evaluate the effectiveness of MPC in improving access to healthcare and health expenditures by Syrian refugees in Jordan. Households receiving MPCs (US$113–219 monthly) were compared to control households not receiving MPCs using difference-in-difference analyses. Overall health care-seeking was consistently high (>$85\%$). Care-seeking for child illness improved among MPCs but declined among controls with a significant adjusted difference in change of $11.1\%$ ($P \leq 0.05$). In both groups, child outpatient visits significantly increased while emergency room visits decreased. Changes in care-seeking and medication access for adult acute illness were similar between groups; however, hospital admissions decreased among MPCs, yet increased among controls (-$8.3\%$ significant difference in change; $P \leq 0.05$). There were no significant differences in change in chronic illness care utilization. Health expenditures were higher among MPCs at baseline and endline; the only significant difference in health expenditure measures’ changes between groups was in borrowing money to pay for health costs, which decreased among MPCs and increased among controls with an adjusted difference in change of -$10.3\%$ ($P \leq 0.05$). The impacts of MPC on health were varied and significant differences were observed for few outcomes. MPC significantly improved care-seeking for child illness, reduced hospitalizations for adult acute illness, and lowered rates of borrowing to pay for health expenditures. No significant improvements in chronic health condition indicators or shifts in sector of care-seeking were associated with MPC. While MPC should not be considered as a stand-alone health intervention, findings may be positive for humanitarian response financing given the potential for investment in MPC to translate to health sector response savings.
## Introduction
Since 2011, more than 5.6 million people have fled conflict in Syria for neighboring countries throughout the region [1]. Jordan currently hosts over 650,000 Syrian refugees, more than $80\%$ of whom live outside of refugee camps [2]. The humanitarian response in Jordan largely provides assistance to refugees outside of camps through existing Jordanian systems. Health assistance is provided at public sector facilities throughout the country for reduced rates based upon policies decided by the Government of Jordan (GoJ). GoJ health policies have shifted numerous times since the initial policy providing free care for Syrian refugees registered with the United Nations High Commissioner for Refugees (UNHCR) at Ministry of Health (MoH)/public sector facilities. This policy was revised in November 2014, when refugees were required to pay the same rates as uninsured Jordanians. In February/March 2018, the policy was modified again, requiring that refugees pay $80\%$ of foreigner rates for health services directly to the MoH facility with no exemptions, resulting in out-of-pocket costs two to five times those incurred under the 2014–2018 policy [3]. Most recently, in March 2019 the health policy reverted back to provision of care for refugees at uninsured Jordanian rates with payment required directly at facilities and exemptions for services at MoH affiliated maternity and childhood centers [4]. Drivers of Syrian refugees not seeking or receiving necessary health services and medicines have fluctuated with many of these changes in GoJ policies, reflective of financial barriers consistently remaining the dominant barriers reported by Syrian refugees [5–7]. Far less often, those not receiving needed care/medicines have reported this stemming from limited availability of services/medicine, not knowing where to go, and long wait times, among other infrequently reported reasons not as directly impacted by the changing policies [5–7].
Aside from sector-specific responses, cash assistance has rapidly expanded in Jordan’s humanitarian response and in global humanitarian programming [8, 9]. In response to persistent funding shortfalls, the Grand Bargain, an outcome of the 2016 World Humanitarian Summit, generated a series of commitments for donors and implementing agencies, among which is increasing routine use of cash assistance as compared to in-kind assistance [10]. Approximately US $5.6 billion in humanitarian assistance was disbursed through cash and voucher assistance (CVA) in 2019, double that of 2015 [11]. Notably, from 2018 to 2019, two of the largest UN humanitarian agencies, the World Food Program (WFP) and UNHCR, both substantially increased use of CVA; WFP increased provision of CVA from US$1.7 billion in 2018 to US$2.1 billion in 2019 (a $23\%$ increase) and UNHCR scaled CVA from US$568 million to US$650 million ($14\%$ increase) [11, 12]. This increase is also reflected in funding Jordan’s humanitarian response. Of the total US$10.4 million allocated through the Jordan Humanitarian Fund in 2018, US$3.9 million ($38\%$) was apportioned to cash assistance [13].
In Jordan, UNHCR and WFP provide the largest refugee cash assistance programs. UNHCR’s cash assistance was created in 2008 to assist Iraqi refugees and expanded in 2012 to provide increasing numbers of Syrian refugees with additional means for meeting their basic needs; it is now UNHCR’s second largest cash assistance program [14]. Syrian refugee households’ eligibility for receipt of UNHCR cash assistance in *Jordan is* determined through the Vulnerability Assessment Framework (VAF), which uses data collected during home visits to create vulnerability ratings to identify those most vulnerable; using this vulnerability ranking, households are prioritized for assistance targeting in line with the level of funding available [15]. The range of multi-sectoral indicators and methods used to generate vulnerability rankings minimize incentives to manipulate information provided during home visits. This targeting method is comprehensively revised every two years (including during the period in which the present study was implemented) to ensure the most vulnerable households are targeted, leading to some existing beneficiaries no longer receiving MPC from UNHCR and others not previously receiving UNHCR MPC to begin receiving this assistance. UNHCR beneficiary lists may also change from one month to the next because of births/deaths and addition of new beneficiaries when previous beneficiaries are resettled to third countries [15]. Through UNHCR, eligible households receive unrestricted monthly transfers valued between 80–155 JOD (US$113–219) depending upon household size and specific vulnerabilities [15]. Barriers to accessing UNHCR MPCs in Jordan are tracked in UNHCR’s regular post distribution monitoring, which, at the time of this study, indicated primary access challenges to include insufficient number of ATMs in certain governorates, lingering issues with the iris scan mechanism for accessing MPCs through ATMs, and limited knowledge of how to operate ATM equipment, though many of these difficulties have decreased in frequency over time [15].
WFP similarly provides cash transfers to vulnerable refugee households outside of camps through their “Choice” program. WFP historically provided food assistance to Syrian refugees in Jordan through in-kind food, paper and electronic vouchers, and at a smaller scale, cash [16]. In August 2017, with the introduction and subsequent expansion of the “Choice” program, WFP shifted to providing food assistance through unrestricted/multipurpose cash (MPC). Unlike UNHCR, cash through WFP’s Choice program can be redeemed as cash or by purchasing food items at one of WFP’s 200 partner shops throughout the country [17]. In June 2018, WFP’s Choice program reached 186,732 beneficiaries across four governorates of Jordan and by August 2018 increased to a total of 292,226 beneficiaries across seven governorates with continued expansion during the period in which the present study was implemented [16]. Since April 2018, WFP transfers for Syrian refugees living outside of camps are valued from 23 JOD/US$32 to 15 JOD/US$21 per person per month [15].
Cash assistance, and particularly MPC, is generally more efficient than in-kind assistance and more supportive of beneficiary dignity and local economies [18]. The current evidence base comparing cash to in-kind assistance is largely concentrated in sectors outside of health, leaving a dearth of research comparing cash assistance to in-kind health assistance (e.g., in-kind assistance to health facilities) or other health financing approaches Cash assistance can, however, improve health through a range of channels including through both structural and intermediate determinants of health. Owusu-Addo et al. ’s 2019 conceptual framework of cash transfers and the social determinants of health posits that cash transfers impact structural determinants such as financial poverty, education, productive capacity, and gender/women’s empowerment, among several others; these impacts in turn positively influence intermediate determinants such as material (e.g., food security, housing) and psychosocial circumstances, which then impact utilization of health services and health outcomes [19]. By reducing financial barriers, cash assistance has been shown to improve health service utilization in several contexts with a clear bi-directional relationship between socioeconomic status and overall health, though Owusu-Addo note that the nature of health systems and contextual factors at national, local, and household levels can also directly influence utilization. Bailey and Hedlund’s 2012 review of the impact of cash transfers on nutrition in emergency and transitional contexts lays a conceptual framework for humanitarian settings, exploring potential ways that emergency cash transfers can impact the causes of malnutrition, including through cash’s impact on health and health behaviors [20]. Though not exclusive to MPC, this review indicated that cash transfers can impact health through increased household expenditures on healthcare (or indirectly through expenditure on hygiene products). While financial barriers are central to the cash—health pathway, cash may also improve health behaviors and outcomes by reducing employment demands (e.g., number of working hours) that may limit household members’ abilities to utilize healthcare [20, 21]. Even when not directly applied for health expenditures, unconditional cash assistance may be spent in other areas that facilitate beneficiaries’ inclusion in “a health-promoting social group” [21]. Combined, these mechanisms potentially underpinning the effects of MPC on health may be realized most clearly through increased health care-seeking, improved treatment adherence, and mitigation of stress, nutrition, and numerous other lifestyle practices known to impact overall health [20–22].
While there is ample evidence in support of the benefits of cash assistance in development contexts, little is known about the effects of cash on sector-specific outcomes such as health in humanitarian contexts [21, 23–25]. It is unclear whether cash transfers would have measurable impacts on health due to the diversity of health needs across households and their members, and transfer amounts relative to the extent of unmet needs and household priorities for other types of expenditures (e.g., food, rent). In light of this gap, this study examined the effects of MPC on health-seeking behavior, health service utilization, and health expenditures among Syrian refugees in Jordan to provide evidence to inform use of cash transfer programs in both the current and future humanitarian responses.
## Methods
A prospective cohort study was conducted from May 2018 through July 2019 to evaluate the effectiveness of MPC provided by UNHCR to vulnerable Syrian refugee households in increasing access to health. Systematically sampled households receiving MPC from UNHCR at the start of this study (intervention group) and similarly vulnerable households not receiving MPC (control group) were followed for one year to compare health expenditures, health-seeking behavior, and health service utilization between the two groups. In an effort to understand the effect of MPC on health in multiple contexts, a parallel study was also conducted in Lebanon in the same timeframe, utilizing the same study design to examine comparable intervention in a setting with important differences (e.g., health system structures, provision of humanitarian assistance, and refugee settlement patterns) [26]. For the purposes of this study, households were defined as people who share a living space and share both meals and financial resources.
## Sampling
Due to strategic targeting of MPC assistance to vulnerable Syrian refugee households in response to funding limitations, at the time of study initiation $30\%$ of registered refugee households ($$n = 27$$,932) received MPC from UNHCR while approximately $92\%$ of households were classified as highly or severely vulnerable with respect to basic needs in the 2017 Jordan VAF sector vulnerability review [27]. This distribution of households with similar levels of economic vulnerability (not) receiving MPC facilitated assessment of health outcomes associated with cash assistance in two reasonably comparable groups of households. To reduce variability between comparison groups and the risk of cross-over (i.e., existing beneficiaries stopping receipt of MPC and/or those not receiving MPC becoming MPC beneficiaries) due to changes in targeting methods based upon predicted household expenditures and anticipated scale up of MPC programs, the sample was restricted to households with predicted per capita monthly expenditures between 40–50 JOD/US$56–70.
Sample size calculations were based on the primary aim of comparing households receiving MPC to similar households not receiving MPC. Most outcome measures of interest (e.g., care-seeking or having out-of-pocket health expenditures) can be expressed as a proportion, thus, in the absence of comparable studies at the time of study design, calculations assumed the most conservative proportion of $50\%$, power = 0.80, a minimum detectable difference of ≥$10\%$, and were two-sided. Based on these assumptions, a minimum required sample size of 770 households (385 per group) was identified; this was increased to a minimum planned sample of 1,000 households to allow for loss to follow-up of ≤$30\%$. The sample ($$n = 1$$,000) was allocated by governorate proportionally to the location of UNHCR MPC beneficiaries with similar numbers in the intervention and control groups. Lists of MPC recipients and non-MPC recipients were ordered by estimated per capita expenditure and systematically sampled. Sampling lists included additional households to the projected sample to allow for replacement sampling when households were unreachable, declined to participate, or determined to be ineligible during screening questions ahead of enrollment interviews. Syrian refugees ($87\%$) and as a result, MPC beneficiaries and the study sample, are concentrated in four governorates of Jordan (Amman, Irbid, Mafraq, and Zarqa) with relatively few in the remaining eight governorates.
UNHCR revises targeting for MPC recipients every two years to reflect changes in households’ financial situations and ensure those currently considered most vulnerable are selected. Consequently, over the course of the study period, some existing beneficiaries stopped receiving MPC ($$n = 121$$; $24.2\%$) while others not previously receiving MPC were added to the MPC beneficiary list ($$n = 51$$; $10.2\%$). To maximize statistical power in light of sample changes, participants receiving MPC from UNHCR at endline (i.e., those who received MPC for the entire study period and those who began receiving MPC at the study mid-point) were analyzed as MPC beneficiary households (intervention group) while the control group included only those not receiving MPC through the entire study period. Additional detail on change in intervention receipt is presented in S1 File.
## Study implementation
UNHCR lists of registered refugee households, including household names, phone number, district of residence, vulnerability category, and receipt of cash assistance were used for recruitment. This information was used only by the study coordinator to identify prospective participants. Interviewers that conducted recruitment received training on privacy and confidentiality and were provided only with names and phone numbers to reduce the risk of sensitive information (such as vulnerability status) being shared. Records of all participant phone numbers and identifiable information were destroyed immediately after use for the study. Sampled households were contacted by phone and invited to participate; households were asked a series of questions to confirm eligibility before the enrollment interview. In order to participate, households were required to be Syrian refugees that are currently registered with UNCHR, classified as vulnerable or severely vulnerable per UNHCR, and registered with UNHCR as residing outside of a camp setting. All consenting eligible households were enrolled in the study until the target sample size ($$n = 1$$,000) was reached. Respondents were household heads or principal applicants on the sampled UNHCR registration case when possible; if that individual did not regularly reside with the household or could not be reached, another adult member of the sampled household (who thus also met eligibility criteria) was permitted to complete interviews. If a sampled household could not be reached by phone after three attempts, they were deemed ineligible to participate in the study.
Verbal informed consent was required prior to enrollment interviews and an abbreviated oral consent for continued participation was used prior to endline interviews. All enrolled households completed phone interviews at enrollment (May-July 2018) and endline one-year following enrollment (May-July 2019). Phone interviews have been used by UNHCR and academic institutions for health research in Jordan and throughout the region with acceptable response rates and are regarded as an appropriate medium for data collection [5, 28, 29]. In this study, phone interviews were preferable for logistical feasibility, cost efficiency, and to better protect participant privacy given the potential risks associated with others likely learning of their participation if interviews were in-person. Interviews lasted between 30–50 minutes and used a structured questionnaire (see S2 File) focused on household demographic and socioeconomic characteristics, receipt of humanitarian assistance, and household health seeking behavior, health service utilization, and health expenditures for child, adult acute, and chronic illnesses. Enumerators mostly had prior data collection experience and received two days of classroom training on data collection tools, mobile data collection platform, interview techniques, and basic principles of human subjects’ protections followed by additional supervised practice interviews. Data were collected on Android tablets using the Magpi mobile data platform by DataDyne LLC (Washington, DC). Data collection was supervised by a local study coordinator with daily checks of uploaded interviews for quality and completeness to promptly address concerns.
## Data analysis
Data analysis was performed using Stata 13 (College Station, TX). Differences between study groups (i.e., MPC vs. non-MPC beneficiaries) in descriptive analyses were examined using chi-square and t-test methods for binary/categorical and continuous variables, respectively. Regression models were used to evaluate the effects of MPC on health outcomes, both unadjusted and controlling for differences in household characteristics. Covariates for all adjusted models were selected a priori to include characteristics known or suspected to be associated with intervention receipt and outcomes of interest. Linear probability models were used to estimate differences in binary outcomes between study groups from baseline to endline with main terms for study group, time period, and the interaction between study group and time period. Log-linear models were similarly used to estimate differences in continuous outcomes; log transformation was required for health expenditure outcomes due to their skewed distribution. Coefficients for the interaction of study group and time period represent the estimated difference in change comparing MPC beneficiaries to non-beneficiaries (i.e., the difference-in-difference/“MPC effect”). Effect sizes associated with receipt of MPC were also calculated by dividing difference-in-difference (DiD) by the overall mean for each outcome. All models utilized cluster-robust standard errors with clustering defined at the household level, allowing for correlation between observations for each household.
Financial indicators are presented in U.S. Dollars (US$) using an exchange rate of 1.41 JOD/US$1 [30]. All monetary variables were assessed for outliers using visual inspection and individual consideration of points falling three or more standard deviations from the mean. Outliers believed to be the result of misreporting or entry errors were corrected or removed from the data set. Other outliers were checked with field teams for accuracy and corrected as needed. Preliminary analysis and findings were discussed by all the collaborating organizations prior to finalization of results to ensure their accuracy and the best possible interpretation of findings within the Jordanian context.
The research was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board and the Ministry of Planning and International Cooperation of Jordan prior to the start of the study. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the S1 Checklist.
## Study population characteristics
A total of 998 households were enrolled in the study, of which 885 ($88.7\%$) were followed for one year and completed endline interviews. During the study period, WFP scaled coverage of their Choice program, offering beneficiaries the option of MPC or e-vouchers and a small number of households changed beneficiary status for UNHCR MPC. Those receiving UNHCR MPC at enrollment who stopped receiving MPC during follow-up ($$n = 121$$) were excluded from the analysis. Final analyses compared participants receiving MPC from UNHCR at endline (i.e., “MPC households”/intervention group, $$n = 429$$) with control households comprising participants not receiving MPC for the entire study period ($$n = 448$$). Retention was $96\%$ among MPC households and $87\%$ among controls; reasons for loss to follow-up and baseline characteristics of participants lost to follow-up are provided in S1 File. The enrolled sample and the final analyzed sample are presented in Table 1.
**Table 1**
| Region | UNHCR MPC Recipients by Region at Enrollment | Enrolled Sample (n = 998) | Enrolled Sample (n = 998).1 | Enrolled Sample (n = 998).2 | Enrolled Sample (n = 998).3 | Analyzed Sample (n = 877) | Analyzed Sample (n = 877).1 | Analyzed Sample (n = 877).2 | Analyzed Sample (n = 877).3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Region | UNHCR MPC Recipients by Region at Enrollment | Intervention (n = 499) | Intervention (n = 499) | Control (n = 499) | Control (n = 499) | Intervention (n = 429) | Intervention (n = 429) | Control (n = 448) | Control (n = 448) |
| Region | UNHCR MPC Recipients by Region at Enrollment | N | % | N | % | N | % | N | % |
| Amman | 29.8% | 171 | 34.3% | 173 | 34.7% | 144 | 33.6% | 159 | 35.5% |
| Irbid | 27.4% | 126 | 25.3% | 126 | 25.3% | 101 | 23.5% | 112 | 25.0% |
| Mafraq | 18.8% | 86 | 17.2% | 88 | 17.6% | 86 | 20.0% | 72 | 16.1% |
| Zarqa | 10.6% | 51 | 10.2% | 49 | 9.8% | 41 | 9.6% | 45 | 10.0% |
| Rest of Jordan a | 13.3% | 65 | 13.0% | 63 | 12.6% | 57 | 13.3% | 60 | 13.4% |
Principal applicants on the sampled UNHCR case served as respondents in $80\%$ and $83\%$ of households at baseline and endline, respectively. Respondents were immediate family members of the principal applicant (e.g., husband/wife, son/daughter, mother/father) in $19\%$ of households at baseline and $16\%$ at endline.
Characteristics of the principal applicant, household composition, and living conditions are summarized in Table 2. At both baseline and endline, principal applicants in MPC households were more commonly female, significantly older, and comparably less educated than in control households. Control households had significantly larger average household size, though smaller dependency ratios. They were also more likely to have children, but relative to MPC households, significantly fewer control households had members 60 years or older, with a disability, or needing daily living support. At baseline, living conditions also significantly differed with control households being more likely to live in an entire apartment/house and having a higher ratio of household members to sleeping rooms in the residence.
**Table 2**
| Unnamed: 0 | Unnamed: 1 | BASELINE | BASELINE.1 | BASELINE.2 | BASELINE.3 | BASELINE.4 | ENDLINE | ENDLINE.1 | ENDLINE.2 | ENDLINE.3 | ENDLINE.4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | MPC HHs (N = 429) | MPC HHs (N = 429) | Control HHs (N = 448) | Control HHs (N = 448) | P value | MPC HHs (N = 411) | MPC HHs (N = 411) | Control HHs (N = 391) | Control HHs (N = 391) | P value |
| Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics | Principal Applicant/Household Head Characteristics |
| Female sex | Female sex | 257 | (59.9%) | 111 | (24.8%) | <0.001 | 245 | (59.6%) | 100 | (25.6%) | <0.001 |
| Age (mean years) | Age (mean years) | 57.1 | (15.4) | 38.0 | (13.5) | <0.001 | 57.4 | (15.7) | 38.9 | (13.8) | <0.001 |
| Highest level of education | | 152 | (35.6%) | 51 | (11.5%) | <0.001 | 140 | (34.1%) | 43 | (11.0%) | <0.001 |
| Primary school | Primary school | 91 | (37.7%) | 202 | (45.4%) | | 170 | (41.5%) | 210 | (53.7%) | |
| Preparatory school | Preparatory school | 22 | (12.9%) | 110 | (24.7%) | | 47 | (11.5%) | 71 | (18.2%) | |
| Secondary school | Secondary school | 7 | (13.8%) | 82 | (18.4%) | | 53 | (12.9%) | 67 | (17.1%) | |
| Marital status | Married | 193 | (45.0%) | 361 | (80.6%) | <0.001 | 188 | (45.7%) | 313 | (80.1%) | <0.001 |
| Widowed | Widowed | 133 | (31.0%) | 22 | (4.9%) | | 140 | (34.1%) | 22 | (5.6%) | |
| Never married / Divorced | Never married / Divorced | 43 | (10.0%) | 43 | (9.6%) | | 47 | (11.4%) | 32 | (8.2%) | |
| Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics | Household Demographic Characteristics |
| Household size (mean) | Household size (mean) | 4.2 | (2.3) | 4.5 | (2.1) | 0.028 | 4.1 | (2.3) | 4.6 | (1.8) | 0.001 |
| Dependency ratio a (mean) | Dependency ratio a (mean) | 1.4 | (1.0) | 1.0 | (0.8) | <0.001 | 1.3 | (1.0) | 1.1 | (0.9) | 0.005 |
| Multiple UNHCR registration cases (%) | Multiple UNHCR registration cases (%) | 192 | (44.8%) | 134 | (29.9%) | <0.001 | 142 | (34.5%) | 94 | (24.0%) | 0.001 |
| Vulnerable members (%) | Vulnerable members (%) | | | | | | | | | | |
| Child(ren) <5 yrs | Child(ren) <5 yrs | 132 | (30.8%) | 281 | (62.7%) | <0.001 | 107 | (26.0%) | 250 | (63.9%) | <0.001 |
| Child(ren) ≤ 17 yrs | Child(ren) ≤ 17 yrs | 260 | (60.6%) | 366 | (81.7%) | <0.001 | 240 | (58.4%) | 319 | (81.6%) | <0.001 |
| Older adult(s) (>60 yrs) | Older adult(s) (>60 yrs) | 267 | (62.2%) | 81 | (18.1%) | <0.001 | 259 | (63.0%) | 75 | (19.2%) | <0.001 |
| Member with a chronic health condition | Member with a chronic health condition | 352 | (82.1%) | 229 | (51.1%) | <0.001 | 338 | (82.2%) | 192 | (49.1%) | <0.001 |
| Member w/ disability or that needs daily support | Member w/ disability or that needs daily support | 151 | (35.2%) | 78 | (17.4%) | <0.001 | 116 | (28.2%) | 46 | (11.8%) | <0.001 |
| Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions | Living Conditions |
| Residence type | Apartment or house | 346 | (80.7%) | 383 | (85.5%) | 0.008 | 367 | (89.3%) | 352 | (90.0%) | 0.217 |
| Single room | Single room | 43 | (10.0%) | 30 | (6.7%) | | 14 | (3.4%) | 8 | (2.0%) | |
| Temporary shelter b | Temporary shelter b | 21 | (4.9%) | 29 | (6.5%) | | 22 | (5.4%) | 28 | (7.2%) | |
| Other c | Other c | 19 | (4.4%) | 6 | (1.3%) | | 8 | (1.9%) | 3 | (0.8%) | |
| Residence arrangement | Rented | 390 | (90.9%) | 411 | (91.7%) | 0.657 | 384 | (93.4%) | 374 | (95.7%) | 0.326 |
| Hosted for free / rent paid by NGO/charity | Hosted for free / rent paid by NGO/charity | 21 | (4.9%) | 15 | (3.3%) | | 14 | (3.4%) | 8 | (2.0%) | |
| Owned | Owned | 15 | (3.5%) | 19 | (4.2%) | | 13 | (3.2%) | 8 | (2.0%) | |
| Crowding (mean # people/sleeping room) | Crowding (mean # people/sleeping room) | 4.5 | (1.9) | 3.6 | (1.7) | <0.001 | 4.3 | (2.0) | 4.1 | (1.9) | 0.228 |
Household economic characteristics (Table 3) differed significantly at baseline in terms of both income and expenditure, with MPC recipients reporting significantly lower incomes and expenditures. Mean incomes and expenditures decreased from baseline to endline in both groups, though only income differed significantly between groups at endline. Significant differences in receipt of any regular cash assistance were observed at baseline and endline both in the proportion of households receiving assistance and total amounts received. Notably, the proportion of households receiving WFP assistance increased in both groups over the study period and WFP transitioned similarly large proportions of households in both groups from e-vouchers to “Choice”/MPC.
**Table 3**
| Unnamed: 0 | Unnamed: 1 | BASELINE | BASELINE.1 | BASELINE.2 | BASELINE.3 | BASELINE.4 | ENDLINE | ENDLINE.1 | ENDLINE.2 | ENDLINE.3 | ENDLINE.4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | MPC HHs (N = 429) | MPC HHs (N = 429) | Control HHs (N = 448) | Control HHs (N = 448) | P value | MPC HHs (N = 411) | MPC HHs (N = 411) | Control HHs (N = 391) | Control HHs (N = 391) | P value |
| Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) | Household Income and Expenditures (past month; mean US$ a) |
| Income (excluding humanitarian assistance) | Income (excluding humanitarian assistance) | 305 | (566.8) | 445 | (598.9) | <0.001 | 193 | (423.9) | 356 | (277.7) | <0.001 |
| Total expenditures | | 471 | (377.7) | 556 | (552.9) | 0.008 | 466 | (327.3) | 474 | (244.8) | 0.710 |
| Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b | Total Humanitarian Assistance (past month) b |
| Any regular cash transfer (% of HHs) | Any regular cash transfer (% of HHs) | 413 | (96.3%) | 387 | (86.4%) | <0.001 | 411 | (100.0%) | 346 | (88.5%) | <0.001 |
| Amount received (mean US$ a per HH) | Amount received (mean US$ a per HH) | 215 | (121.3) | 101 | (67.2) | <0.001 | 239 | (97.9) | 99 | (77.0) | <0.001 |
| Amount received (mean US$ a per HH member) | Amount received (mean US$ a per HH member) | 63 | (36.0) | 23 | (14.6) | <0.001 | 72 | (33.8) | 22 | (18.2) | <0.001 |
| In-kind assistance (past 3 months)b | In-kind assistance (past 3 months)b | 22 | (5.1%) | 27 | (6.0%) | 0.703 | 61 | (14.8%) | 52 | (13.3%) | 0.490 |
| WFP Food Assistance (past month) | WFP Food Assistance (past month) | | | | | | | | | | |
| Current WFP recipients | Current WFP recipients | 397 | (92.5%) | 387 | (86.4%) | 0.011 | 400 | (97.3%) | 343 | (87.7%) | <0.001 |
| Amount received (mean US$ a per HH) | Amount received (mean US$ a per HH) | 89 | (63.6) | 90 | (48.9) | 0.785 | 87 | (59.5) | 99 | (50.1) | 0.003 |
| Amount received (mean US$aper HH member) | Amount received (mean US$aper HH member) | 22 | (8.4) | 20 | (7.1) | 0.010 | 22 | (7.4) | 21 | (8.8) | 0.697 |
| Transfer modality | E-Voucher | 304 | (76.6%) | 300 | (77.5%) | 0.753 | 118 | (29.5%) | 99 | (28.9%) | 0.849 |
| Choice | Choice | 93 | (23.4%) | 87 | (22.5%) | | 282 | (70.5%) | 244 | (71.1%) | |
| Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing | Asset Sales and Borrowing |
| Sold assets in past 3 months (%) | Sold assets in past 3 months (%) | 76 | (17.7%) | 119 | (26.6%) | 0.002 | 90 | (21.9%) | 127 | (32.5%) | 0.003 |
| Borrowed money in past 3 months (%) | Borrowed money in past 3 months (%) | 271 | (63.2%) | 310 | (69.2%) | 0.018 | 302 | (73.5%) | 328 | (83.9%) | 0.002 |
| Current debt | Any Debt | 304 | (76.6%) | 345 | (82.7%) | 0.029 | 321 | (84.0%) | 346 | (93.5%) | <0.001 |
| Amount of debt (among those w/ debt; mean US$ a) | Amount of debt (among those w/ debt; mean US$ a) | 633 | (1201.8) | 701 | (1735.0) | 0.519 | 1340 | (7427.6) | 1182 | (1711.0) | 0.691 |
## Health service utilization
Care-seeking was assessed for the most recent household member illness (within the past six months) believed to be severe enough to warrant medical care; results are reported for children’s illnesses and for acute and chronic illnesses among adults. Baseline and endline descriptive comparisons of both groups are presented in S1 Table and in Fig 1; unadjusted and adjusted individual group change and differences in change between groups are provided in Table 4 and Fig 2.
**Fig 1:** *Reasons and timeframes for care-seeking.* **Fig 2:** *Change in health utilization measures by group and difference in magnitude of change between groups.* TABLE_PLACEHOLDER:Table 4 Reasons for needing medical care for a child’s illness were similar between MPC and control households and most commonly included respiratory infections, fever, diarrhea, and asthma (Fig 1). Care-seeking rates for childhood illness were significantly greater at baseline among controls than MPC recipients ($87.4\%$ vs. $77.0\%$, $$P \leq 0.014$$), but decreased, though not significantly, by endline with a significant $11.1\%$ adjusted difference in change between groups (CI: 0.8–$21.3\%$; $$P \leq 0.035$$; effect size: $13.3\%$). Small sample sizes for those not seeking care prohibited robust analysis of change; however, among households that did not seek needed care, cost was the most commonly reported reason. The proportion of households that sought care who reported not receiving all recommended care due to cost decreased from baseline to endline in both groups; however, as with nearly all child health utilization outcomes, the difference in change between groups was not statistically significant in adjusted analyses ($$P \leq 0.148$$).
While most children received care as outpatient visits, at baseline $26\%$ of MPC group children and $20\%$ of control children were treated at the emergency room (ER) and $3\%$ and $2\%$ of children, respectively, had a hospital admission the last time care was sought. At endline, both groups reported decreases in emergency room visits for child illness and increases in outpatient visits ($P \leq 0.001$ for all within-group changes); however, no significant difference in adjusted facility utilization change was observed between groups (outpatient $$P \leq 0.097$$; ER $$P \leq 0.091$$). Both MPC recipients and controls predominantly sought care for childhood illness in the private sector and, while private sector care-seeking decreased in both groups at endline, change was small, not statistically significant (adjusted MPC change $$P \leq 0.301$$; control change $$P \leq 0.508$$), and did not significantly differ between groups ($$P \leq 0.686$$). Access to medication for childhood illness was high, with more than $97\%$ of households in both MPC and control groups obtaining prescribed medications at both time periods. At endline, increases in obtaining medication were observed in both groups, but the increase was not significantly different among controls compared to MPC recipients ($$P \leq 0.333$$).
Compared to MPC recipients, control households reported needing care for an adult with acute illness more recently (Fig 1), though these differences were significant only at endline (baseline $$P \leq 0.867$$; endline $$P \leq 0.021$$). Reasons for adult care-seeking were similar between MPC and control households at baseline ($$P \leq 0.488$$) but differed significantly at endline ($$P \leq 0.001$$) when MPC households reported infection more commonly than controls, who more frequently cited needing care for dental and gynecological problems (Fig 1). Care-seeking rates were significantly higher in control households than MPC households at baseline ($73.5\%$ and $68.3\%$, respectively; $$P \leq 0.031$$) and small but statistically non-significant decreases in care-seeking were observed in both MPC ($$P \leq 0.415$$) and control ($$P \leq 0.857$$) households (adjusted DiD $$P \leq 0.618$$). Cost remained the primary reason for not seeking care; and while smaller proportions of households in both groups reported not receiving all needed care due to cost at endline compared to baseline, change was only significant among MPC recipients (-$10.3\%$, CI: -20.5,-0.1; $$P \leq 0.047$$) and did not significantly differ from controls (adjusted DiD $$P \leq 0.328$$).
Although most adults received care for acute illness through outpatient visits, significantly more MPC households than controls reported emergency room visits ($20.7\%$ and $11.0\%$, respectively) and inpatient admissions ($7.3\%$ and $0.0\%$, respectively) at baseline. Care-seeking location for acute adult illness was similar between the two groups at endline and changed over time; in adjusted models comparing change in the two groups, outpatient care-seeking increased by $15.8\%$ (CI: 2.5,29.2; $$P \leq 0.020$$) more in the MPC group compared to controls (effect size: $17.4\%$). Hospital admissions were reported by $7.3\%$ of MPC recipient adult acute illness care-seekers but no controls at baseline nor in either group at endline, leading to an adjusted difference in change of -$8.3\%$ (CI: -15.2,-$1.3\%$; $$P \leq 0.021$$). At baseline, a significantly larger proportion of control households sought care in the private sector, while more MPC households sought care at public sector or charity facilities; sector utilization was similar between the two groups at endline. In adjusted analyses, private sector utilization decreased among controls and MPC recipients; both groups increased public and charity sector utilization over the study period; however, none of these changes significantly differed between groups (private sector $$P \leq 0.431$$; public $$P \leq 0.633$$; charity $$P \leq 0.576$$). Like childhood illnesses, access to medication for adult acute illness was very high, with more than $94\%$ of households in both groups reporting they were able to obtain prescribed medications; this did not significantly change in either group (adjusted DiD $$P \leq 0.467$$).
The proportion of households with member(s) with chronic health conditions was significantly higher among MPC beneficiaries compared to controls at baseline ($82\%$ vs $51\%$) and endline ($82\%$ vs $49\%$). Among household members with chronic condition(s), the most frequently reported were hypertension, diabetes, arthritis, and cardiovascular disease (Fig 1). MPC and control households did not significantly differ at either baseline or endline in nearly all care utilization outcomes for adult chronic illnesses. Households receiving MPC reported seeking care for adult chronic illnesses more recently than controls (Fig 1), but this difference was significant only at endline ($$P \leq 0.010$$). Both study groups reported similarly high care-seeking rates at baseline and endline with more than $93\%$ of individuals with a chronic condition having received care for their condition in Jordan. Similar proportions of MPC and control participants reported visits to a general practitioner and/or specialist in the preceding six months. Hospital visits were also similarly reported by both groups at baseline and while both groups decreased during follow-up, a significantly ($$P \leq 0.035$$) greater proportion of MPC participants ($34.7\%$) reported hospital visits at endline compared to controls ($26.0\%$). No statistically significant differences in adjusted utilization change were observed between groups. As with child and adult acute illness, sector utilization for adult chronic illnesses were similar between MPC and control groups and, except for decreased public sector utilization among controls (-$9.6\%$, CI: -18.2,-1.1; $$P \leq 0.027$$), did not significantly change or differ between groups during follow-up. Significantly fewer MPC recipients reported not receiving needed care/services because of cost at endline (adjusted change -$9.6\%$, CI: -17.4,-1.9; $$P \leq 0.015$$); however, neither baseline to endline change among controls ($$P \leq 0.926$$) nor the difference in change between groups ($$P \leq 0.090$$) were statistically significant.
Access to medication for adult chronic illness improved in both groups. The proportion of individuals reporting ever facing difficulties obtaining medication for their chronic illness decreased significantly in MPC recipients (adjusted change: -$14.0\%$, CI: -21.9,-6.1; $$P \leq 0.001$$) and controls (adjusted change: -$10.7\%$; CI: -20.1,-1.2; $$P \leq 0.028$$), though the difference between these changes was not statistically significant ($$P \leq 0.573$$). Medication affordability followed a similar trend with significant decreases in the proportion of individuals unable to afford medication for their chronic condition in the MPC (-$20.7\%$; $P \leq 0.001$) and control groups (-$15.8\%$, $$P \leq 0.001$$) but with no significant difference in change between groups ($$P \leq 0.405$$).
## Health expenditures
Health expenditure for the most recent child, adult acute, and adult chronic illnesses (within the past six months), as well as overall household health expenditures were also evaluated. Baseline and endline descriptive analyses for each group are provided in S2 Table; individual change over time and differences in change between study groups are provided in Table 5; baseline and endline mean overall costs for child illness, adult acute illness, and overall household health expenditures are presented in Fig 3.
**Fig 3:** *Health expenditures by group (one-year period from baseline to endline).* TABLE_PLACEHOLDER:Table 5 After adjusting for sociodemographic characteristics, change in nearly all health expenditure outcomes were not significantly different ($P \leq 0.05$) between MPC recipients and controls. Despite this, change among MPC recipients and/or controls were significant for selected expenditure measures.
With regard to the most recent care visit for a child’s illness, the proportion of controls with an out-of-pocket payment to the health facility increased $11.5\%$ (CI: 4.6,18.3; $$P \leq 0.001$$) from baseline to endline; this also increased among MPC recipients, though not statistically significantly ($$P \leq 0.107$$). Consequently, facility payment amounts among controls were also 3.5 times higher (CI: 1.6,8.0; $$P \leq 0.001$$) at endline relative to baseline, though the analogous increase among MPC recipients was not statistically significant ($$P \leq 0.146$$). In adjusted models, the proportion of households with payment for medication obtained outside the facility significantly decreased $9.1\%$ among controls ($$P \leq 0.049$$), but change among MPC households ($$P \leq 0.866$$) and the difference in change between groups were not significant ($$P \leq 0.267$$); no significant change or group difference in change was observed for medication costs (MPC $$P \leq 0.909$$; control $$P \leq 0.148$$; DiD $$P \leq 0.406$$). Overall payments for child illness significantly increased $9.1\%$ (CI: 1.6,$16.7\%$; $$P \leq 0.018$$) among controls, as did overall payment amounts, which were 2.8 (CI: 1.1,7.0; $$P \leq 0.033$$) times higher at endline than baseline; however, changes in both the proportion of households with payments and mean payment amounts were not significant among MPC recipients ($$P \leq 0.969$$ and $$P \leq 0.987$$, respectively) and did not significantly differ between groups ($$P \leq 0.159$$ and $$P \leq 0.214$$).
While health facility and overall expenditures for adult acute care did not significantly change in either group, the proportion of households with out-of-pocket payments for medication obtained at a pharmacy or elsewhere significantly decreased for both the MPC (-$17.8\%$, CI: -30.9,-4.7; $$P \leq 0.008$$) and control (-$19.0\%$, CI: -30.9,-7.0; $$P \leq 0.002$$) groups. Accordingly, adjusted medication costs also decreased similarly (DiD $$P \leq 0.895$$) for both groups; MPC group medication costs at endline were 0.2 (CI: 0.0,0.7; $$P \leq 0.016$$) times those at baseline and control costs at endline were 0.1 (CI: 0.0,0.5; $$P \leq 0.003$$) times baseline amounts. MPC recipients saw more significant changes in medication expenditures for adult chronic illness relative to controls. Average monthly medication costs were reported by significantly fewer individuals in both groups at endline compared to baseline with an adjusted change of -$14.0\%$ (CI: -21.5,-$6.5\%$; $P \leq 0.001$) among MPC recipients and -$10.0\%$ (CI: -18.6,-1.3; $$P \leq 0.024$$) among controls (DiD $$P \leq 0.471$$). Monthly medication payment amounts also significantly decreased for MPC recipients (0.2, CI: 0.0,0.7; $$P \leq 0.016$$) and controls (0.3, CI: 0.1,0.8; $$P \leq 0.019$$). Neither group saw significant change in the proportion with facility payments (MPC $$P \leq 0.055$$; control $$P \leq 0.840$$; DiD $$P \leq 0.245$$) or payment amounts (MPC $$P \leq 0.101$$; control $$P \leq 0.831$$; DiD $$P \leq 0.336$$) for the most recent chronic illness care visit.
Total household health expenditures in the past month were significantly higher among MPC households (Fig 3) and increased from baseline to endline in both groups in adjusted analyses. While change was not significant among MPC recipients ($$P \leq 0.194$$), expenditures were 2.8 (CI: 1.2,6.4; $$P \leq 0.016$$) times higher among controls at endline than at baseline. Change in routine health spending did not significantly differ between groups (adjusted DiD $$P \leq 0.405$$). Asset sales to pay for health expenses in the past three months were uncommon among participants (<$11\%$ in both groups). Although asset sales increased in both groups during the study period, change was only statistically significant for controls, among whom asset sales increased $3.8\%$ (CI: 0.0,$7.5\%$; $$P \leq 0.049$$) (MPC adjusted $$P \leq 0.679$$); there was no significant difference in change between groups (adjusted $$P \leq 0.412$$). Borrowing money to pay for health expenses was more common in both groups, with marginally more MPC households borrowing money at baseline compared to controls ($39.9\%$ and $33.7\%$, respectively, $$P \leq 0.069$$), but similar proportions in both groups at endline ($35.0\%$ and $35.8\%$, respectively; $$P \leq 0.820$$). In adjusted models, borrowing decreased $7.7\%$ (CI: -14.8,-$0.6\%$; $$P \leq 0.034$$) among MPC households and increased $2.5\%$ (CI: -4.2,$9.3\%$; $$P \leq 0.461$$) among controls with a significant difference in change of -$10.3\%$ (CI:-19.9,-0.6; $$P \leq 0.037$$; effect size: -$28.3\%$) between groups.
## Discussion
Frequency of needing, though not necessarily seeking, care for child and adult acute illness was similar among MPC recipients and controls at baseline; however, rates for actualized care-seeking for child and adult acute illness were significantly higher among control households than MPC households at baseline. Changes over the study period for care-seeking for both child and adult illness within groups were not statistically significant; however, the between group difference in change for child care-seeking was statistically significant, which is reflective of increases in care-seeking in the MPC group and decreases in the control group. Across all types of care, cost was the predominate barrier for those not seeking care and among care recipients, many did not receive all recommended care due to cost. The proportion of care-seekers unable to afford all recommend care was statistically similar between the two comparison groups at baseline for child illness, adult acute illness, and chronic illness. At the end of the one year follow up period, statistically significant decreases in the proportion of households not receiving all needed care due to cost were reported for all three illness categories in the MPC group, suggesting that MPC may improve the affordability of care. Similar statistically significant declines were not observed among control households and there were also no significant differences in change over the study period between the two groups for two of the three categories of illness (a difference was observed for chronic illness), yielding a mixed result with respect to MPC’s impact on the ability to afford all needed care. While data from 2019 is not yet available, UNHCR’s annual Health Access and Utilization Survey (HAUS) demonstrates similarly mixed care-seeking trends over time. Although the proportion of households surveyed for the HAUS that sought care in the preceding month decreased substantially from $84\%$ in 2016 to $45\%$ in 2018, access to medical services for chronic conditions improved over the same period and perceived ability to afford necessary medical services for chronic conditions also greatly improved from 2017 to 2018 [5–7]. Mixed findings on the impact of MPC on care-seeking in the present study are aligned with increased child health care-seeking observed in a recent study of MPC in Lebanon, but in contrast to null findings of two MPC evaluations in Jordan [31–33].
With respect to access to medicines, more than $95\%$ of households in both groups were able to afford prescribed medications for child and adult acute illness at baseline possibly because many of these medications are included in the essential care package. MPC receipt was associated with marginally significant increases in access to medicines for both child illness and adult acute illness. Cost was a barrier to medication for a larger proportion of households that had a member with a chronic health condition at baseline than at endline with similar levels of change in both control and MPC households. Increased affordability of chronic disease medications was not related to MPC receipt; it is possible but unlikely that this is attributable to the March 2019 policy changes intended to reduce out-of-pocket expenditures because endline data collection was conducted soon after the change and many households would not yet have benefited from it. Improvements in access to medication for chronic conditions observed in our study contrast with a decline from 2016 through 2018 previously noted in UNHCR’s HAUS [5–7]. The HAUS did, however, note improvements in availability of chronic illness medications from 2017 to 2018, which, coupled with policy changes and care-seeking behaviors, may have implications for medication affordability.
Changes in the sector where care was received were assessed for child illness and adult acute and chronic illness, but were not significant for either group with the exception of a decline in public sector utilization for chronic illness among controls, suggesting that increased use of charity/non-governmental organization (NGO) facilities was not a factor in increasing affordability of chronic disease medication and that MPC was not associated with a shift in care-seeking away from the public and charity/NGO sectors to the private sector. This is a positive finding with respect to the objective of providing unconditional cash to meet basic needs and indicates that MPC receipt does not result in replacement of public sector care with more costly care in the private sector. UNHCR’s HAUS reported a contradictory shift from comparable utilization of government hospitals and private clinics in 2016 and 2017 to markedly higher refugee use of private pharmacies for care-seeking in 2018, likely due to GoJ policy changes increasing the cost of public sector care in early 2018 [5–7]. The absence of significant changes in our findings and, although not significant, decreased private sector utilization for most illness types may support the potential influence of 2019 policy changes decreasing public sector care costs.
Emergency room visits for child illness and adult acute illness declined over the study period for both MPC recipients and controls. For children, declines were statistically significant for both MPC and control households, but the magnitude of decline was marginally greater among MPC households; among adults a statistically significant decline was observed only in the MPC group and the difference in change between groups over the study period was not significant. The observed decreases in emergency room use are in contrast with what would be expected as a result of April 2019 GoJ policy changes, which should have translated to decreased out of pocket payments; however, it is likely that many refugees were not widely aware of policy changes (in particular if they had not sought care recently) or that changes were not fully implemented prior to data collection [4]. Hospitalizations remained constant for children but statistically significantly declined for adult acute illness among MPC recipients; the difference in change over the study period was statistically significant, suggesting that MPC receipt was associated with a decrease in hospital admissions for adult acute illness. Collectively, this suggests that MPC may facilitate better access to care and prevent hospitalizations, which are important impacts for individuals and for financing of the broader humanitarian health response where reduced hospitalizations translates to a significant cost savings. In the parallel study conducted in Lebanon, results were also mixed with MPC recipients having a significantly smaller increase in child hospitalization compared to controls (DiD -$6.1\%$; $$P \leq 0.043$$) yet no significant differences for adult hospitalizations [26].
Regarding health expenditures, the proportion of MPC recipients that reported out of pocket expenses and average [log transformed] expenditures were assessed for child illness and adult acute and chronic illness. Total child health expenditures at the most recent visit were constant between groups and over time; adult health expenditures were significantly higher among controls at both baseline and endline and in both groups. Total expenditures for the most recent adult acute illness were more varied; controls had consistently higher expenditures and both groups saw a decrease in expenditures over the study period. However, when analyzed as a proportion of households with payments and log transformed expenses, there were no significant differences in change in expenditures between the two groups for any health expenditure measure, demonstrating that MPC receipt was not associated with increased spending on health during the study period. Examination of total monthly household health expenditures reveals significantly higher expenditures among MPC recipients compared to controls at both baseline and endline. This aligns with evidence from Lebanon including our parallel study, which also observed significantly greater monthly household health expenditures among MPC recipients at both baseline and endline [26, 31]. There was a significant reduction in the proportion of households in the MPC group borrowing to pay for health expenses and when change over time was compared between the two groups, MPC receipt was associated with a reduction in borrowing to pay for health, suggesting that MPC was protective for household financial risks associated with health, though this same trend was not observed for asset sales to pay for health costs, which were less frequently reported. Overall reduced health expenditures in our findings align with temporal trends reported in the HAUS where household health expenditures decreased annually from 2016 through 2018 [5–7]. Our findings also echo those from two previous evaluations of cash assistance for Syrian refugees in Jordan, which both found no statistically significant association between MPC receipt and health expenditures [32, 33]. It is worth noting that while a 2017 study comparing CVA provided by UNHCR, UNICEF, and WFP in Jordan found no evidence of increased health spending, they did find that beneficiaries believed receiving cash assistance, and in turn having more fungible income, had improved their ability to seek needed health care [32]. Previous studies of Syrian refugee MPC assistance throughout the region, notably in Lebanon and Turkey, also found limited or no impact of MPC on health expenditures, suggesting the probable influence of policy changes and other factors on health spending [31, 34].
Average monthly household expenditures exceeded the Minimum Expenditure Basket (MEB), which reflects the monthly expenditure amount needed to meet basic needs and is typically calculated on a per capita or household size basis, in both the MPC and control groups at both baseline and endline [35]. MPC assistance amounts were considerable and equate to approximately 46–$51\%$ of average monthly expenditures. Monthly household health expenditures among MPC recipient households averaged US$20.6 (baseline) to US$26.3 (endline) more than control households, which translates to a difference in household health expenditures of US$247.2 to US$315.6 annually. Applying these figures to the approximately 28,000 Syrian refugee households that received UNHCR MPC from mid-2018 to mid-2019, it can be estimated that MPC expenditures contribute US$6.9–8.8 million annually to refugee health in Jordan, equating to approximately $10\%$ of health sector response funding, which was US$73 million in 2019 [36]. Although MPC assistance amounts were sizable in this context, post distribution monitoring during this time found that cash assistance was more often spent on rent, utilities, and food (though still frequently used on health), suggesting that higher transfer amounts may be needed before households increase spending on health [37]. While MPC increased spending on health and affordability of care in this study, they cannot be viewed as a replacement for direct support to the health sector and they should be considered as one of several options for increasing access to care. There are few comparative studies of MPC compared to conditional cash transfers, which either have qualifying criteria (e.g., for individuals with chronic health conditions) or use conditions (e.g., ability to demonstrate care-seeking or medication purchase); however, previous evidence from development settings has demonstrated the positive impact of conditional cash transfers on health behaviors and, to a lesser extent, health outcomes [38–40]. Much of this research is limited to Latin-American countries and does not capture the nuanced interplay of factors in displacement settings, hindering the ability to translate inferences to humanitarian contexts in other regions [38–40]. Moreover, the potential benefits of cash transfers on improved health outcomes are noted to rely on the provision of quality care rather than utilization alone, supporting the need to integrate supply-side interventions when implementing cash assistance to improve health [40].
While affordability is constantly a primary barrier to healthcare access and utilization in Jordan and other humanitarian contexts, competing demands on households, among other factors, often influence health-related behaviors, likely hindering the impact of intervention via MPC alone. Several reviews of cash transfers in development settings have emphasized the influence of non-financial factors on health in explaining areas of heterogeneity in the impacts of cash, furthering the idea that cash alone, whether conditional or unconditional, is likely most effective when combined with supply-size or health system strengthening interventions [38–40]. Moreover, evidence of the benefits of health education and community-based screening, monitoring, and counseling interventions for improving chronic disease care in humanitarian settings is increasing, but alone these interventions cannot address affordability barriers [41–44].
Previous evidence from Syrian refugees in Jordan suggests that conditional cash transfers and health education may be more effective in improving health indicators for chronic disease compared to MPC [45]. Health sector ‘top ups,’ where MPC transfer value is increased for households with specific characteristics (e.g., members with a chronic health condition or disability) necessitating higher routine health expenditures could be another strategy to deliver cash for health at scale, though this approach has not been tested in a humanitarian setting to our knowledge.
## Limitations
This study had several limitations. First, results may be limited by the quality of expenditure reporting given enumerator and respondent confusion about what to include in the various expenditure questions despite thorough training on interview techniques and probing for accurate responses. Self-report of expenditure data may also have introduced recall bias, particularly for households whose most recent care-seeking was months prior to interview. Second, expansion of WFP’s Choice program during the study period resulted in approximately half of participants in both the MPC and control groups switching to WFP Choice. Because the Choice program allows beneficiaries to receive unconditional cash assistance, although this was accounted for in adjusted analyses, the influence of WFP transfer modality changes on our outcomes cannot be precluded. Changes to the GoJ health policy directing healthcare costs for Syrian refugees near the beginning and end of the study may likely have influenced care utilization and are potential confounder. The analysis approach assumes that both comparison groups would have parallel trends during the study period if no intervention was received; however, it is difficult to determine the credibility of this assumption because pre-baseline data that could be used to assess trends prior to the study period are not available. In addition, the study assumes that both comparison groups are similarly impacted by GoJ policy changes and other temporal trends due to their similarity as vulnerable refugees residing in the same location, however, it is possible that they experienced differential impacts, which could have contributed to [lack of] differences in change over time that are attributed to the intervention. The timing of this change also adds difficulty to interpreting changes in health expenditures at public sector facilities. Small sample sizes should be considered in interpretation of findings. While sample sizes for those not seeking care across care categories (i.e., child, adult acute, and adult chronic illness) prohibited robust analysis of change, conditional denominators for many outcomes (e.g., those who needed care, sought care, etc.) restricted the analyzed sample well below the minimum sample sizes calculated in the initial study design. This likely affected the sufficiency of statistical power to detect significant differences between groups. Care should similarly be taken in interpreting DiD estimates in cases where comparison groups significantly change, though in different directions. Finally, ideally this study would be conducted with new rather than existing beneficiaries but, given the extended nature of the refugee crisis and the need to conduct research within the ongoing humanitarian response, such a design was not possible.
## Conclusions
The impacts of unrestricted cash transfers, which provided for approximately half monthly household expenditures, on Syrian refugee health outcomes in Jordan were varied. There were no significant changes in household health expenditures during the study period, though cash transfer recipients were significantly less likely to report borrowing to pay for health expenses. At both study time points significantly higher household health expenditures were found among cash recipients, which may indicate that benefits are realized closer to transfer initiation and then sustained. Improvements in care-seeking attributed to receipt of unrestricted cash transfers were observed for child illness but not acute or chronic illness among adults. The proportion of households unable to receive all needed care due to cost declined significantly in the MPC group for child illness and both acute and chronic illness among adults, suggesting that cash may have improved health access in at least some way, though differences in change compared to the control group were not significant. Medication was accessible for nearly all households for both child illness and adult acute illness, but not chronic disease; access and affordability of chronic disease medications improved over the study period for both groups and not as a result of cash transfers. Cash transfers were not associated with a shift in care-seeking away from the public and charity/NGO sectors to the private sector and appear to yield benefits in terms of reduced hospitalizations for adult acute illness. These are both positive findings with respect to humanitarian response financing, particularly because investment in unrestricted cash transfers may translate to savings in the health sector response.
## References
1. 1United Nations High Commissioner for Refugees (UNHCR). Syria Emergency [Internet]. c2001-2022 [cited 31 Jan 2020]. Available from: https://www.unhcr.org/en-us/syria-emergency.html. *Syria Emergency*
2. 2UNHCR. Situation Syria Regional Refugee Response: Jordan [Internet]. [cited 31 Jan 2020]. Available from: https://data2.unhcr.org/en/situations/syria/location/36. *Situation Syria Regional Refugee Response: Jordan*
3. 3UNHCR. New health policy impact and actions [Internet]. 2018 [cited 15 Jun 2020]. Available from: https://data2.unhcr.org/en/documents/download/62984. *New health policy impact and actions* (2018.0)
4. 4UNHCR. UNHCR Jordan Factsheet, April
2019 [Internet]. 2019. [cited 2 Feb 2020]. Available from: https://reliefweb.int/report/jordan/unhcr-jordan-factsheet-april-2019. *UNHCR Jordan Factsheet* (2019.0)
5. 5UNHCR. Health access and utilization survey [Internet]. 2018 [cited 24 Apr 2020]. Available from: https://data2.unhcr.org/en/documents/download/68539. *Health access and utilization survey* (2018.0)
6. 6UNHCR. Health access and utilization survey: Access to health services in Jordan among Syrian Refugees [Internet]. 2016 [cited 24 Apr 2020]. Available from: https://data2.unhcr.org/en/documents/download/55906. *Health access and utilization survey: Access to health services in Jordan among Syrian Refugees* (2016.0)
7. 7UNHCR. Health access and utilization survey: Access to health services in Jordan among Syrian Refugees [Internet]. 2017 [cited 24 Apr 2020]. Available from: https://data2.unhcr.org/en/documents/details/62498. *Health access and utilization survey: Access to health services in Jordan among Syrian Refugees* (2017.0)
8. Spender A, Parrish C, Lattimer C. **Counting cash: tracking humanitarian expenditure on cash-based programming**. *Working Paper* (2016.0) **505**
9. 9Cash and Learning Partnership (CaLP). The state of the world’s cash report: cash transfer programming in humanitarian aid [Internet]. 2018 [cited 31 Jan 2020]. Available from: http://www.cashlearning.org/downloads/calp-sowc-report-web.pdf. *The state of the world’s cash report: cash transfer programming in humanitarian aid* (2018.0)
10. Bargain Grand. *The Grand Bargain–a shared commitment to better serve people in need* (2016.0)
11. Metcalfe-Hough V, Fenton W, Willitts-King B, Spencer A. *Grand Bargain annual independent report* (2020.0)
12. Metcalfe-Hough V, Fenton W, Poole L. *Grand Bargain annual independent report* (2019.0)
13. 13Jordan Humanitarian Fund Annual Report
2018 [Internet]. [cited 22 Apr 2020]. Available from: https://www.unocha.org/sites/unocha/files/Jordan%20HF%20Annual%20Report%202018.pdf. *Jordan Humanitarian Fund Annual Report* (2018.0)
14. 14UNHCR. UNHCR Cash Assistance Global Factsheet
2018 [Internet]. 2018 [cited 22 Apr 2020]. Available from: https://www.unhcr.org/protection/operations/5c5c5acc4/unhcr-fact-sheet-cash-based-assistance-2018.html. *UNHCR Cash Assistance Global Factsheet* (2018.0)
15. 15UNHCR. Multi-purpose cash assistance 2018 post distribution monitoring report [Internet]. 2018 [cited 24 Apr 2020]. Available from: https://data2.unhcr.org/en/documents/download/68296. *Multi-purpose cash assistance 2018 post distribution monitoring report* (2018.0)
16. Majewski B, Lattimer C, Gil Baizan P, Shtayyeh S, Canteli C. *Decentralized evaluation: WFP’s general food assistance to Syrian refugees in Jordan 2015 to mid-2018* (2018.0)
17. 17World Food Program (WFP). Draft Jordan country strategic plan (2020–2022) [Internet]. 2019 [cited 24 Apr 2020]. Available from: https://docs.wfp.org/api/documents/WFP-0000106351/download. *Draft Jordan country strategic plan (2020–2022)* (2019.0)
18. Austin A, Frize J. *Ready or not? Emergency cash transfers at scale* (2011.0)
19. Owusu-Addo E, Renzaho AM, Smith BJ. **Cash transfers and the social determinants of health: a conceptual framework.**. *Health Promot Int* (2019.0) **34** e106-18. DOI: 10.1093/heapro/day079
20. Bailey S, Hedlund K. *The impact of cash transfers on nutrition in emergency and transitional contexts: A review of evidence* (2012.0)
21. Pega F, Liu SY, Walter S, Lhachimi SK. *Unconditional cash transfers for assistance in humanitarian disasters: effect on use of health services and health outcomes in low- and middle-income countries (A Review)* (2015.0)
22. Browne E.. *Theories of Change for Cash Transfers (GSDRC Helpdesk Research Report 913)* (2013.0)
23. 23UNHCR. Multi-purpose cash and sectoral outcomes: a review of evidence and learning [Internet]. 2018 [cited 31 Jan 2020]. Available from: https://www.unhcr.org/5b0ea3947.pdf. *Multi-purpose cash and sectoral outcomes: a review of evidence and learning* (2018.0)
24. Doocy S, Tappis H. *What is the evidence of the effectiveness and efficiency of cash based approaches in protracted and sudden onset emergencies: a systematic review* (2016.0)
25. 25UNHCR. Cash-based Interventions for health programmes in refugee settings: a review [Internet]. 2015 [cited 31 Jan 2020]. Available from: http://www.unhcr.org/research/evalreports/568bce619/cash-based-interventions-health-programmes-refugee-settings-review.html. *Cash-based Interventions for health programmes in refugee settings: a review* (2015.0)
26. Lyles E, Arhem J, Khoury G, Trujillo A, Spiegel P, Burton A. **Multi-purpose cash transfers and health among vulnerable Syrian refugees in Lebanon: A prospective cohort study.**. *BMC Public Health* (2021.0) **21** 1176. DOI: 10.1186/s12889-021-11196-8
27. 27UNHCR Cash Based Interventions (CBI) team, Jordan on behalf of the VAF Advisory Board. Jordan Vulnerability Framework. 2017 Population Survey Report [Internet]. 2018 [cited 14 Jun 2020]. Available from: https://data2.unhcr.org/en/documents/download/65404. *Jordan on behalf of the VAF Advisory Board. Jordan Vulnerability Framework. 2017 Population Survey Report* (2018.0)
28. Mahfoud Z, Ghandour L, Ghandour B, Mokdad AH, Sibai AM. **Cell phone and face-to-face interview responses in population-based surveys: how do they compare?**. *Field Methods.* (2015.0) **27** 39-54. DOI: 10.1177/1525822X14540084
29. Sibai AM, Ghandour LA, Chaaban R, Mokdad AH. **Potential use of telephone surveys for non-communicable disease surveillance in developing countries: evidence from a national household survey in Lebanon.**. *BMC Med Res Methodol* (2016.0) **16** 64. DOI: 10.1186/s12874-016-0160-0
30. 30XE. XE Currency Charts: JOD to USD [Internet]. [cited 20 Jun 2018]. Available from: https://www.xe.com/currencycharts/?from=JOD&to=USD&view=1Y
31. Chaaban J, Ghattas H, Salti N, Moussa W, Irani A, Jamaluddine Z. *Multi-purpose cash assistance in Lebanon–Impact evaluation on the well-being of Syrian refugees* (2020.0)
32. 32Action Against Hunger, UNHCR. Evaluation synthesis of UNHCR’S cash based interventions in Jordan [Internet]. 2017 [cited 23 Apr 2020]. Available from: https://www.unhcr.org/5a5e16607.pdf. *Evaluation synthesis of UNHCR’S cash based interventions in Jordan* (2017.0)
33. Abu Hamad B, Jones N, Samuels F, Gercama I, Presler-Marshall E, Plank G. *A promise of tomorrow: the effects of UNHCR and UNICEF cash assistance on Syrian refugees in Jordan* (2017.0)
34. Doocy S, Leidman E. *Multi-Purpose Cash Assistance and Health: Evaluating the Effect of the Emergency Social Safety Net (ESSN) Programme on Access to Health Care for Refugees in Turkey* (2019.0)
35. 35UNHCR. UNHCR Guidance Note: Minimum Expenditure Basket for Syrian [Internet]. 2019 [cited 14 Jun 2020]. Available from: https://reliefweb.int/sites/reliefweb.int/files/resources/74050.pdf. *UNHCR Guidance Note: Minimum Expenditure Basket for Syrian* (2019.0)
36. 36Regional Refugee and Resilience Plan in Response to the Syria Crisis (3RP). 2019
Annual Report [Internet]. 2020 [cited 24 Jun 2020]. Available from: http://www.3rpsyriacrisis.org/wp-content/uploads/2020/05/annual_report.pdf. *Annual Report* (2019.0)
37. 37UNHCR. Mid-Year Post Distribution Monitoring Report for Refugees and Asylum Seekers—Multi-Purpose Cash Assistance in Jordan
2018 [Internet]. 2018 [cited 19 Oct 2021]. Available from: https://data2.unhcr.org/en/documents/details/65143.. *Mid-Year Post Distribution Monitoring Report for Refugees and Asylum Seekers—Multi-Purpose Cash Assistance in Jordan* (2018.0)
38. Ranganathan M, Lagarde M. **Promoting healthy behaviours and improving health outcomes in low and middle income countries: a review of the impact of conditional cash transfer programmes.**. *Prev Med* (2012.0) **55** S95-105. DOI: 10.1016/j.ypmed.2011.11.015
39. de Souza Cruz RC, de Moura LB, Neto JJ. **Conditional cash transfers and the creation of equal opportunities of health for children in low and middle-income countries: a literature review.**. *Int J Equity Health* (2017.0) **16** 1-2. DOI: 10.1186/s12939-017-0647-2
40. Gaarder MM, Glassman A, Todd JE. **Conditional cash transfers and health: unpacking the causal chain.**. *J Dev Effect* (2010.0) **2** 6-50. DOI: 10.1080/19439341003646188
41. Glanz K, Rimer BK, Viswanath K. *Health behavior and health education: theory, research, and practice.* (2008.0)
42. Rabkin M, Fouad FM, El-Sadr WM. **Addressing chronic diseases in protracted emergencies: Lessons from HIV for a new health imperative.**. *Glob Public Health.* (2018.0) **13** 227-33. DOI: 10.1080/17441692.2016.1176226
43. Murphy A, Biringanine M, Roberts B, Stringer B, Perel P, Jobanputra K. **Diabetes care in a complex humanitarian emergency setting: a qualitative evaluation**. *BMC Health Serv Res* (2017.0) **17** 1-0. DOI: 10.1186/s12913-017-2362-5
44. Sethi S, Jonsson R, Skaff R, Tyler F. **Community-based noncommunicable disease care for Syrian refugees in Lebanon.**. *Glob Health Sci Pract* (2017.0) **5** 495-506. DOI: 10.9745/GHSP-D-17-00043
45. Lyles E, Chua S, Barham Y, Pfeiffer-Mundt K, Spiegel P, Burton A. **Improving diabetes control for Syrian refugees in Jordan: a longitudinal cohort study comparing the effects of cash transfers and health education interventions.**. *Confl Health* (2021.0) **15** 41. DOI: 10.1186/s13031-021-00380-7
|
---
title: 'Understanding older peoples’ chronic disease self-management practices and
challenges in the context of grandchildren caregiving: A qualitative study in rural
KwaZulu-Natal, South Africa'
authors:
- Dumile Gumede
- Anna Meyer-Weitz
- Anita Edwards
- Janet Seeley
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021571
doi: 10.1371/journal.pgph.0000895
license: CC BY 4.0
---
# Understanding older peoples’ chronic disease self-management practices and challenges in the context of grandchildren caregiving: A qualitative study in rural KwaZulu-Natal, South Africa
## Abstract
While chronic diseases are amongst the major health burdens of older South Africans, the responsibilities of caring for grandchildren, by mostly grandmothers, may further affect older people’s health and well-being. There is a paucity of information about chronic disease self-management for older people in the context of grandchildren caregiving in sub-Saharan Africa. Guided by the Self-Management Framework, the purpose of this qualitative methods study was to explore the chronic disease self-management practices and challenges of grandparent caregivers in rural KwaZulu-Natal, South Africa. Eighteen repeat in-depth interviews were carried out with six grandparent caregivers aged 56 to 80 years over 12 months. Thematic analysis was conducted based on the Self-Management Framework. Pathways into self-management of chronic illnesses were identified: living with a chronic illness, focusing on illness needs, and activating resources. Self-perceptions of caregiving dictated that grandmothers, as women, have the responsibility of caring for grandchildren when they themselves needed care, lived in poverty, and with chronic illnesses that require self-management. However, despite the hardship, the gendered role of caring for grandchildren brought meaning to the grandmothers’ lives and supported self-management due to the reciprocal relationship with grandchildren, although chronic illness self-management was complicated where relationships between grandmothers and grandchildren were estranged. The study findings demonstrate that grandchildren caregiving and self-management of chronic conditions are inextricably linked. Optimal self-management of chronic diseases must be seen within a larger context that simultaneously addresses chronic diseases, while paying attention to the intersection of socio-cultural factors with self-management.
## Introduction
While chronic diseases affect all age groups, ageing increases the risk of chronic conditions [1] such as hypertension, diabetes, and ischaemic heart disease [2] and is associated with physical functional decline [3]. The number of older people aged 50 and above living with chronic conditions continues to rise rapidly in South Africa [4, 5]. In a recent population cohort study of older, rural, black South Africans, approximately $71\%$ had two or more chronic illnesses [5], placing a heavy burden on the South African healthcare system [3, 6].
Chronic diseases are also combined with a high burden of HIV in South Africa [7, 8]. South Africa has 7.5 million people living with HIV, with nearly 5.2 million on antiretroviral therapy (ART) [9] and more individuals are ageing with HIV in South Africa [4]. Living with HIV in the context of ART is another chronic illness burden faced by older people as effective treatment enables people to live with HIV into older age [10].
While chronic conditions pose a health burden for older South Africans [5], at the same time many act as primary caregivers for their grandchildren. The grandparent caregivers assume responsibilities associated with caregiving for their grandchildren which include providing shelter, food, and clothing for themselves and for their grandchildren [11]. As parental figures, grandparents serve as role models, provide their grandchildren with love and support, discipline their grandchildren, and impacting values to their grandchildren [12]. Different situations exist which explains why grandparents are providing caregiving to their grandchildren. Historically, grandchildren caregiving has always been common for grandparents as they have been identified as a key family support system in ensuring a critical safety net for children [13]. However, the HIV epidemic further aggravated the responsibility of grandparents for their grandchildren due to the increased parental morbidity and mortality [14]. While HIV care and ART services have significantly reduced AIDS-related deaths, grandparents still assume caregiving responsibilities for the grandchildren [15]. In South Africa, labor migration and unemployment have also contributed to the caregiving of grandchildren by grandparents [11]. The responsibilities of caring for grandchildren combined with living with chronic conditions may further impact the health and well-being of grandparents [16] and this may be physically, economically, psychologically, and socially a burden for the grandparents. In Uganda, older grandparent caregivers reported chronic pains and stress as major challenges in their role as carers, thereby limiting their ability to effectively execute caring duties [17]. This caregiving has been linked to worse mental health issues and social isolation for the grandparent caregivers [18].
Grandchildren caregiving is often undertaken with minimal financial resources [15]. In response to supporting poor families, the South African government offers social security grants including old-age pensions, child support grants, and foster child grants to eligible grandparents, which they can claim for themselves and their grandchildren [12]. It has been reported that these government grants are important sources of income, especially for grandmothers raising grandchildren in South Africa [11, 19].
When discussing the caregiving provided by grandparents, it is worth noting that gender differences shape caregiving within families [20]. The literature pertaining to grandchildren caregiving shows that the caregiving role is performed predominantly by women [15, 21]. Given that it is normative in the South African context for grandmothers to provide care for grandchildren [11], this further contributes to the notion of gendered caregiving of grandchildren [12, 20]. While there is little available research globally on the number of grandparents raising children, a study in Uganda among grandparent caregivers reported that the majority of caregivers of children aged 13–17 years were grandmothers aged 50 years and older [17]. In 2019, it was estimated that nearly 4 million children in South Africa were living with a grandparent [22]. Older grandparent caregivers with chronic conditions are likely to experience less time for self-management because, as previous research has established, chronic disease affects the ability of older people to function [3, 10]. Self-management is defined as the individual’s ability, in conjunction with family, community, and healthcare professionals, to manage symptoms, treatments, lifestyle changes, and psychosocial, cultural, and spiritual consequences of chronic conditions [23].
Understanding how grandparent caregivers of adolescents navigate the self-management of chronic conditions while at the same time taking on the added responsibility of caring for their grandchildren is a neglected research area. The personal health and well-being challenges of grandparent caregivers are likely to take a toll not only at a personal level, but also impact the caring process for adolescents in different ways. In this study, we explored the chronic disease self-management practices and challenges of grandparent caregivers and their views on how this was shaped by caring for adolescent grandchildren. The Self-Management Framework was used to shape the analysis because it can be applied across a wide range of individual characteristics and chronic conditions [24]. Self-management activities cluster into three main processes: focusing on illness needs, activating resources, and living with a chronic illness, each of which includes tasks and skills that facilitate self-management [25]. Understanding factors that influence self-management may improve the assessment of self-management among grandparent caregivers with chronic illness and inform interventions to support older caregivers.
## Study context
The study was conducted in uMkhanyakude district (KwaZulu-Natal province, South Africa), one of the poorest and rural districts in the country [26]. People in the district live in predominantly multi-generation families consisting of grandparents, adult children, and grandchildren [27]. The area is among those with the highest HIV prevalence and incidence rates in South Africa [28]. Between 2009 and 2015, the prevalence of hypertension, HIV, and diabetes increased in the district [29]. While there is a high burden of chronic illness, there are only five district hospitals, 251 primary care clinics, and 17 mobile clinics servicing a population of 625,846 [30]. Approximately $10\%$ of households in the district can reach a primary healthcare facility within 15 minutes by vehicle [29]. A previous study, undertaken two decades ago, reported that livelihood strategies of most households depended on small-scale agriculture, government grants, and remittances from migrant workers [31]; the situation has remained the same to the present time.
This study was part of a larger study at Africa Health Research Institute (AHRI) that commenced in September 2017 investigating the caring of adolescents by grandparent caregivers in the context of DREAMS (Determined, Resilient, Empowered, AIDS-free, Mentored and Safe) partnership. DREAMS is a multi-component HIV prevention intervention designed to reduce HIV incidence among adolescent girls and young women (AGYW) [32, 33]. The DREAMS programme was implemented between April 2016 and September 2018 and delivered by different DREAMS implementing partners in the district.
## Study design, study population and sample
A qualitative research design located within an interpretive paradigm was adopted for this study because of the focus on exploring the experiences of grandparent caregivers and the meanings they attribute to self-management of chronic diseases whilst caring for grandchildren. This design has proven useful in previous research on chronic illness [34, 35] and we sought to generate depth accounts. Participants were selected using a purposive sampling method with the following criteria: grandparents aged 50 years and above who were primary caregivers of at least one adolescent child aged 13 to 19 years who participated in DREAMS programme.
## Data collection
The first author, a local social science researcher trained in qualitative methods, conducted eighteen repeat in-depth interviews with six grandparents from uMkhanyakude district. The grandparents were recruited through a community-based organisation that delivered DREAMS interventions and were primary grandparent caregivers of adolescents that had received DREAMS interventions. The repeat in-depth interviews were conducted on three occasions from September 2017 to October 2018, using a semi-structured interview guide. The first interview focused on two central, open-ended questions that were posed to each participant: “How do you care for yourself?” and “*How is* it for you to raise your grandchildren while caring for yourself?” As the participants reflected on their experiences, the interviewer began to identify the challenges that they were experiencing with chronic illnesses while raising their grandchildren and the manner in which these influenced their lives. The repeat interviews were conducted every four months to maintain contact with research participants over time. As opposed to single interviews, repeat interviews are useful for documenting participants’ lived experiences over time, allowing researchers to ask follow-up questions and can be tailored for each individual [36]. After the first interview, data were preliminary analysed to identify emerging themes and issues to follow-up with each individual in the subsequent interviews. Through these repeat in-depth interviews, we were able to capture the daily lives, the caregiving experiences of participants, the experience of ageing, and health challenges. Also, prolonged engagement strengthened rapport that was established with participants and increased trustworthiness in data collection. Interviews were carried out in the local language, isiZulu. The grandmothers opted that interviews be conducted at their homes. This ensured privacy and confidentiality. Each interview took between 30 and60 minutes.
## Data management and analysis
Interviews were audio-recorded, transcribed, and translated into English. Atlas.ti 8 software was used to manage, and code translated data. The Self-Management Framework [24] guided the thematic analysis of data, using deductive and inductive processes. The analysis followed a five-step process, as described by Babchuk [37], that involved assembling materials for analysis; structured reading and re-reading of transcripts; coding the data along the identified themes of the self-management framework; generating categories and assigning codes to them; generating themes from categories; and using verbatim quotes from the in-depth interviews to illustrate certain themes. Coded data, themes, and categories were repeatedly reviewed by DG and agreed with AMW and JS to increase the rigour of the analysis and to ensure a deeper understanding of the data.
## Ethical considerations
Ethics approval for the study was provided by the University of KwaZulu-Natal Humanities and Social Sciences Research Ethics Committee (HSS/$\frac{1109}{017}$D). Voluntary written informed consent was obtained from participants after information about the purpose of the study had been provided. Confidentiality and anonymity were maintained through the use of participant codes and pseudonyms when the participants referred to their grandchildren during the interviews. All identifying information were removed from quotes.
## Results
All six participants were grandmothers aged between 56 and 80 years old and primary caregivers of two to fifteen grandchildren, at least, one of whom was an adolescent child aged 13 to 19 years and a recipient of DREAMS. While we did not collect data about all biological parents of the grandchildren that were raised by grandmothers, the biological fathers of four adolescent grandchildren who were recipients of DREAMS were deceased and the biological mothers of all six grandchildren were still alive. None of the adolescent grandchildren receiving DREAMS had both biological parents deceased. Chronic conditions varied widely between participants with all of the participants reporting living with one or more chronic conditions. Most grandmothers relied on the government’s social security grants including old age pension (approximately $100 per month) and child support grant (approximately $24 per month) and survived by growing and selling produce, even though it was on a small scale. The demographic characteristics of the participants, including health conditions, are presented in Table 1.
**Table 1**
| Participant Code | Age | Marital status | Education | Number of children in care | Source of income | Chronic conditions |
| --- | --- | --- | --- | --- | --- | --- |
| P1 | 76 | Widow | | 3 | Old age pension, cash, or in-kind remittances | Arthritis, chronic pain |
| P2 | 64 | Widow | Secondary | 15 | Old age pension, child support grant | Arthritis, hypertension, chronic pain |
| P3 | 80 | Widow | | 6 | Old age pension, informal micro-enterprising | Arthritis, hypertension, stomach ulcers, chronic pain |
| P4 | 58 | Unmarried, living with partner | | 11 | Child support grant, informal micro-enterprising | Chronic pain |
| P5 | 56 | Unmarried | | 2 | Farm work, informal micro-enterprising | HIV, epilepsy, chronic pain |
| P6 | 64 | Widow | Primary | 9 | Old age pension, child support grant | HIV, hypertension, vision impairment, chronic pain |
## Chronic disease self-management practices and challenges
The findings are structured according to the processes drawn from the Self-Management Framework [24] (Table 2), namely living with a chronic illness, focusing on illness needs, and activating resources. The quotations included throughout the results section are all from different participants’ responses.
**Table 2**
| Themes | Sub-themes | Sub-themes.1 |
| --- | --- | --- |
| Living with a chronic illness | Processing emotions | • Feeling shocked, confused, sad, and angry• Blaming self for acquiring HIV |
| Living with a chronic illness | Adjusting to illness and to a new self | • Accepting and embracing the illness as part of ageing• Comparing self to others as a way of fostering self-motivation |
| Living with a chronic illness | Integrating illness into daily life | • Reorganising everyday life in order to adapt to chronic illness |
| Living with a chronic illness | Meaning making | • Illness as a pathway to end of life• Caring for grandchildren as sense of purpose |
| Focusing on illness needs | Following instructions from the healthcare workers | • Treatment adherence• Eating healthy diet• Regular clinic attendance |
| Focusing on illness needs | Completing health tasks | • Regular medical appointments• Collecting treatment from the facilities• Adhering to treatment |
| Focusing on illness needs | Performing health promotion activities | • Using home remedies, traditional medicine, and alternative therapy |
| Activating health resources | Healthcare resources | • Regular interactions with community healthcare workers• Seeking care from multiple healthcare facilities |
| Activating health resources | Spiritual resources | • Regular church attendance• Praying |
| Activating health resources | Family support | • Home remittances from adult children• Assistance with performing household tasks by grandchildren• Support with taking medication by grandchildren |
| Activating health resources | Community resources | • Participation in community saving clubs |
## Living with a chronic illness
The grandmothers explained how living with a chronic illness impacted on their lives and required them to come to terms with their conditions. They had to manage their emotions following their chronic illness diagnoses and then make any required adjustments in daily living.
Firstly, the older carers narrated stories about the diagnosis of their chronic illness(es) and how they processed the emotions of living with a chronic illness as the narratives of two grandmothers illustrate: Living with HIV and lacking financial resources to self-manage, generated feelings of self-blame for acquiring HIV infection, and helplessness among the grandmothers: However, one grandmother compared herself to other people living with HIV as a way of fostering self-motivation, as she stated: The grandmothers also shared their initial experiences of the adjustments they had to make because of the illness and to come to terms with the ‘new self.’ They identified two strategies that they used to adjust to the illness and their new self: accepting and embracing the illness as part of ageing and comparing themselves to others as a way of fostering self-motivation.
One grandmother talked about how she had expected that her life would deteriorate because of ageing and the illness: Seasonal weather conditions also affected the physical well-being of the grandmothers as they were unable to function optimally. In some instances, living with a chronic illness was very limiting: When describing life with chronic illness, the same participant, P2, recounted the strategies she employed to integrate illness into her daily life by reorganising her life to adapt to living with arthritis: *Our data* suggest that grandmothers’ focused on the challenges of chronic illnesses in their capacity to function and on the disruptions that the chronic illnesses brought to their lives.
Among the accounts of living with a chronic illness, meaning making about the chronic illnesses featured during the interviews. P6 described illness as a pathway to end of life: Some mentioned that they found a sense of purpose in life through caring for their grandchildren. Grandmothers perceived themselves as caregivers. Caring for their grandchildren was more important than receiving better care for themselves: Meeting needs such as food for the grandchildren was also performed by the grandmothers. Instead of focusing on managing their chronic illnesses, grandmothers opted to prioritise the needs of their grandchildren: However, while grandmothers valued meeting the needs of their grandchildren instead of focussing on themselves and their use of resources to self-manage their chronic conditions, regret was also felt because of the additional responsibilities and consequences of caring for their grandchildren.
Our data suggest that grandmothers struggled to find access to their desired proper healthcare, availability of food, and their ability to access nutritious food as they were living with chronic conditions.
## Focusing on illness needs
When grandmothers focused on managing their chronic illness needs, they indicated a range of activities that they perform.
The grandmothers mentioned that they were following instructions from the healthcare workers about the management of chronic conditions. These instructions were related to treatment adherence, a healthy diet, and regular clinic attendance for routine check-ups: Fear of being reprimanded by the nurses for not adhering to these instructions appeared to motivate the older people in ensuring that they followed dietary requirements: However, the ability of some grandmothers to take ownership of their health needs by following appropriate meal plans was limited by their adolescent grandchildren refusing to cook healthier meals or to provide needed support for healthy eating: While refusal to cook healthy food was mentioned as one of the main challenges that grandmothers faced with focusing on illness needs, another grandmother related how living with poor vision as a chronic condition limited her independence and her ability to eat healthily:
Rather than supporting their grandmothers’ efforts to eat well, the adolescent grandchildren were quite dismissive of the health concerns of their older carers. Dependence on adolescent grandchildren to prepare meals for the grandmothers presented a challenge to self-manage chronic conditions.
Grandmothers mentioned completing health tasks in their effort to self-manage chronic conditions. Attending regular medical appointments, collecting treatment from the facilities, and adhering to treatment were key in the management of their illnesses.
Some grandmothers indicated that they performed these health tasks independently while others relied on the support of their adolescent grandchildren. This was raised by two grandmothers living with HIV and on ART as they explained that their adolescent grandchildren supported them in taking medication and escorted them to the clinic: Lastly, grandmothers mentioned that they performed health promotion activities to minimise the impact of chronic conditions and engaged in self-initiated treatment like using home remedies, traditional medicine, and alternative therapy for their conditions. Alternative therapies and home remedies were mentioned to self-manage the treatment of chronic illnesses and minor ailments including fever, heartburn, body pains, and skin rashes. The use of alternative therapies was motivated by factors such as the lack of access to medication at healthcare facilities and standard healthcare treatments which the older carers perceived as ineffective: Salt was used as a common commodity for bathing, steaming, and soaking, for the grandmothers to cope with chronic pain, and also for fever: Grandmothers often reported anxiety and emotional distress related to living with chronic conditions, the strained relationships with their adolescent grandchildren, and their poverty. Burning incense to cope with emotional ‘pains’ was mentioned by some grandmothers:
Improving musculoskeletal function and mobility through performing physical activities was seen as a benefit by all the grandmothers. They referred to physical activity as ‘ukunyakazisa igazi’ meaning ‘an act of moving one’s body for blood circulation in the body system.’ Walking, dancing in church, doing household chores such as gardening, washing clothes and dishes by hand, and sweeping the yard and house were the main daily physical activities that the grandmothers performed to remain physically fit.
Interestingly, P5 mentioned how she had changed a negative experience into something positive by thinking differently about it i.e., reframing the “problem”: Directing negative emotions of living with chronic illnesses onto other activities in order to get rid of the negative feelings was a defence mechanism for the grandmothers.
## Activating resources
In exploring the resources that the grandmothers activated, they mentioned healthcare resources, spiritual resources, family resources, and community resources to manage various aspects of their chronic illnesses.
With regards to healthcare resources that were important for the grandmothers, community healthcare workers (CHWs), and healthcare facilities were mentioned as key resources. The grandmothers stated that they regularly interacted with CHWs who provided home-based care and health education: Grandmothers also mentioned that they found CHWs easily accessible in case of emergency as they lived in the community unlike going to the clinic to consult the nurses. Interestingly, some grandmothers indicated that they also made use of the services of the CHWs for healthcare needs related to their adolescent grandchildren: When grandmothers described how they navigated the healthcare system in order to ensure continuity of primary healthcare services, they talked about healthcare facilities that they opted to use. Grandmothers mentioned that they seek care from multiple healthcare facilities using both the provincial and the local government clinics interchangeably: It was clear that the quality of patient care, specifically the lack of a patient-centered approach, distance to the healthcare facility, and availability of medication influenced the grandmothers’ choices of the healthcare facilities to use for their chronic illnesses.
Moreover, spiritual resources were a key component of health and well-being that the grandmothers mentioned that sustained them. They all mentioned that they regularly attended church services to worship and receive spiritual counseling. In addition, praying was a source of strength: Churches provided valued support to manage chronic illnesses and difficult relationships with their adolescent grandchildren.
Some grandmothers mentioned that they obtained family support in a form of home remittances from their adult children to cover the costs of self-managing their chronic illnesses and to meet the basic needs of their grandchildren: With regards to household chores, all the grandmothers mentioned that they sought assistance with performing household tasks from their adolescent grandchildren. They assigned household chores to the grandchildren such as cooking, cleaning, fetching water and wood as they were often sick and physically unable to conduct the housework themselves.
Yet, while some grandmothers received support from their adolescent grandchildren, they were not always satisfied with that support: The grandmothers were distressed when their adolescent grandchildren did not meet their expectations.
Lastly, mobilising community resources through participation in community-saving clubs including burial schemes, grocery schemes, and loan schemes was also mentioned by the grandmothers. These clubs provided quick cash loans to the older carers to meet their needs to manage chronic illnesses and to care for their grandchildren, as seen in the interviews below: Community-saving clubs facilitated access to micro-finances for the grandmothers to support their chronic disease self-management practices.
## Discussion
This study has uncovered three processes that shape the lived experiences of chronic disease self-management by older grandparent caregivers in a rural community of KwaZulu-Natal. Self-management involves the tasks that people living with a chronic illness must do to gain control of their condition and to live successfully with the chronic disease [24]. Living with a chronic illness, focusing on illness needs, and activating resources are the processes employed by grandmothers for their chronic disease self-management practices. Self-perceptions of caregiving dictated that grandmothers are responsible to assume primary caregiving for grandchildren when they themselves needed care, lived in poverty and with chronic illnesses that require self-management. The gendered role of caregiving for grandchildren brought meaning to life and supported self-management due to the reciprocal relationship with grandchildren. However, chronic illness self-management was complicated where relationships between grandmothers and grandchildren were strained.
This study expands on previous studies by contributing to knowledge about caregiving experiences of grandparent caregivers living with HIV and the intersection of meaning making of illness and caring [4, 38, 39]. Having grandparent caregivers with chronic illnesses could have a significant impact on the adolescents and the care that the adolescents receive from their grandparent caregivers. HIV stigma by association is one of the challenges that may face adolescents who are being raised by HIV-positive grandparent caregivers. Studies in high-income settings reported that adolescents with HIV-positive parents perceived themselves as different from their peers or feared they would be discriminated against if their parents’ HIV status is disclosed [40–42]. In a South African study, adolescents with HIV-positive caregivers were reported to have increased risks of poor educational outcomes, mental health problems, stigma, and isolation from peers [43].
It is critical for older people to engage in physical activity to prevent diseases, maintain independence, and improve their quality of life [44]. Other authors have noted that physical activity programmes target younger people more than older people and it is also less accessible for older people due to smaller incomes [45]. Our findings show that older people can use minimal resources to meaningfully engage in physical activities without financial costs and within their home environments. Consistent with previous studies, working around the house as part of domestic responsibilities was viewed as a form of exercise [46]. Apart from engaging in physical activities to prevent immobility, it is possible that physical activity also improved their mental health as they were often distressed by living with chronic diseases and caring for grandchildren. Our findings show that the impact of living with chronic illness(es) and caring for adolescent grandchildren had a toll on the health and well-being of the grandparent caregivers. The grandparent caregivers indicated mental distress in relation to living with chronic illnesses, lack of finances, and strenuous relationships with their adolescent grandchildren. Consistent with previous studies, HIV-positive caregivers and caregivers of orphaned and vulnerable adolescents are vulnerable to mental health problems [47, 48].
Mental distress is likely to impact the self-management of chronic illness(es) among grandparent caregivers. Problems experienced in everyday living have been reported to negatively impact the self-care of chronic illness [49]. Consistent with our findings, in another study among older men and women living with HIV in Uganda [50], difficult relationships with their adolescent grandchildren, rather than chronic conditions, were the main stressors that often undermined the grandparent caregivers’ ability to self-manage. This finding differs from a study conducted in the United States, where older adults with chronic conditions often did not want to burden their children with the responsibilities of caring for them [51]. In this study, grandparent caregivers expected their adolescent grandchildren to care and support them. It is possible that the caregivers’ expectations determined the relationships between grandparent caregivers and their adolescent grandchildren.
Consistent with the literature, family support is critical in sustaining self-management behaviours and addressing the barriers among people living with chronic illnesses [25, 52, 53]. Adolescent grandchildren provided support to their grandparent caregivers in executing self-management tasks such as treatment adherence, regular healthcare attendance, and cooking food. In this study, the grandparent caregivers emphasised the key role that the emotional and physical support from their adolescent grandchildren played in self-managing their chronic conditions. Support from their adolescent grandchildren facilitated the grandparent caregivers in self-managing chronic conditions. This finding is supported by other studies [17, 54] that reported the grandparent caregivers counted on their grandchildren to perform household chores that were too physically demanding for the caregivers to perform. They emphasised the vital role of social support not only in helping them to take their medication but also in helping them to find a sense of purpose in caring for their grandchildren. A study in Uganda found that adolescents supported caregivers’ adherence to HIV treatment by reminding them to take ARVs and honour clinic appointments [55]. Consistent with a study in South Africa [54], grandparent caregivers also believed that focusing on their grandchildren contributed to their sense of resilience and living with chronic illnesses.
However, adolescent grandchildren can also pose barriers to self-management behaviours for the grandparent caregivers. Dietary changes are commonly used as essential strategies to improve the self-management of chronic illnesses [56]. While adolescent grandchildren played a vital role in cooking for their grandparent caregivers, the grandparent caregivers in this study reported that grandchildren usually provided the grandparent caregivers with unhealthy food. Rather than supporting their grandparent caregivers’ efforts to eat healthily, the adolescent grandchildren were dismissive of the health concerns of the older carers. The findings of this study fill an important knowledge gap about the influence of adolescent grandchildren on dietary modifications for older carers with chronic illnesses. Previous studies have focused on the caregiver influence in relation to their children’s eating behaviours [57, 58]. In this study, grandmothers had limited financial resources to manage chronic illnesses and to provide basic needs for the grandchildren. Lack of finances negatively impacted the grandparent caregivers in self-managing their chronic illnesses in this study. The specific resources that the grandparent caregivers chose to mobilise were influenced by their human agency and the nature of their relationship with their adolescent grandchildren. For instance, participating in micro-finance activities was influenced by the caregivers’ agency and the need to provide the adolescent grandchildren with food.
Our findings add depth to previous research relating to the role of CHWs in society and particularly in supporting grandparent caregivers with the self-management of chronic illnesses. Previous studies have shown that the role of CHWs includes health education, home-based care, and supporting adherence to treatment [59, 60]. This study also reveals their role in facilitating access to and utilisation of sexual and reproductive (SRH) services by adolescents in grandparent families. For instance, it was reported that the CHWs conducted pregnancy testing, referrals for antenatal care (ANC), and referrals for contraception, thus promoting healthcare service utilisation by adolescents in grandparent families. Home visits by CHWs seemed to be effective in identifying pregnant adolescents and those needing contraceptives in grandparent families. A study in South Africa showed that healthcare providers or nurses tended to impose their values upon adolescents regarding contraceptives and posed challenges for adolescents’ uptake of SRH services [30]. It is possible that home visits by CHWs and the relationships they have with the grandparent caregivers could positively influence adolescents’ uptake of SRH services. These are in keeping with findings from another study conducted in South Africa, where older adults valued the services that they received from the CHWs [61].
Our findings resonate with other work in Malawi, Uganda, and South Africa showing how chronic patients experienced difficulty obtaining medicines from public healthcare facilities, leading to non-adherence to healthcare services [34, 35, 62]. The shortages of medicines in public healthcare facilities can also be regarded as the health systems barrier to support chronic disease self-management [34]. Being dissatisfied with shortages of medicines was one of the reasons that grandparent caregivers practiced switching healthcare facilities and it compromised chronic disease self-management.
The findings provide a framework within which services and interventions can support grandparent caregivers. Caring for grandchildren while suffering from a chronic illness can be draining and the older carers may need a combination of support to self-manage these conditions and strengthen relationships with their grandchildren.
## Strengths and limitations
Our study has made a contribution to the limited literature about the intersection of chronic disease self-management and caring for grandchildren. It has highlighted the potential of the Self-Management Framework to illuminate the complex lived experiences of navigating chronic disease self-management and caring for grandchildren. The main limitation of our study is we interviewed only grandmothers and had a limited perspective on the experience of grandfathers. While this may be an imbalance in our participants, it does reflect the gender distribution of grandparent caregivers, most of whom are women, and the gendered nature of caregiving. In addition, while our study included a small sample of grandparent caregivers living with chronic conditions, a larger sample would have been desirable and would likely have given us insight into other chronic diseases that are a burden in South African older people e.g. diabetes. Nonetheless, repeat in-depth interviews with the small sample enabled participants to provide detailed and in-depth accounts. The repeat in-depth interviews ensured data saturation which gave us confidence that we achieved diverse themes that emerged from the repeat interviews.
## Conclusions
This study provides needed information for planning primary healthcare needs and chronic care health services that will increasingly have to support older people with chronic illnesses. This evidence points to a range of self-management practices used by the older grandparent caregivers which were often influenced by the nature of care relationships between the caregivers and their adolescent grandchildren. The findings show that self-management of chronic conditions and grandchildren caregiving are inextricably linked. Health promotion researchers and health providers cannot view self-management of chronic conditions by grandparent caregivers as a single issue and hope to attain optimal health and well-being for older populations. Optimal self-management of chronic diseases must be seen within a larger context that simultaneously addresses chronic diseases, while paying attention to the intersection of social and cultural factors on older caregivers’ self-management strategies. Understanding the role played by grandchildren in supporting their grandparent caregivers’ self-management practices can assist in developing an intervention in which young people participate in self-management education and support for grandparent caregivers living with chronic conditions.
## References
1. Maresova P, Javanmardi E, Barakovic S, Barakovic Husic J, Tomsone S, Krejcar O. **Consequences of chronic diseases and other limitations associated with old age—A scoping review.**. *BMC Public Health* (2019.0) **19** 1-17. DOI: 10.1186/s12889-019-7762-5
2. Gouda HN, Charlson F, Sorsdahl K, Ahmadzada S, Ferrari AJ, Erskine H. **Burden of non-communicable diseases in sub-Saharan Africa, 1990–2017: results from the Global Burden of Disease Study 2017**. *Lancet Glob Health* (2019.0) **7** e1375-87. DOI: 10.1016/S2214-109X(19)30374-2
3. Solanki G, Kelly G, Cornell J, Daviaud E, Geffen L, Moeti T, Padarath A. *Population ageing in South Africa: trends, impact, and challenge for the health sector.* (2019.0) 173-82
4. Chang AY, Gómez-Olivé FX, Payne C, Rohr JK, Manne-Goehler J, Wade AN. **Chronic multimorbidity among older adults in rural South Africa**. *BMJ Glob Health* (2019.0) **4** 1-10. DOI: 10.1136/bmjgh-2018-001386
5. Wade AN, Payne CF, Berkman L, Chang A, Gómez-Olivé FX, Kabudula C. **Multimorbidity and mortality in an older, rural black South African population cohort with high prevalence of HIV findings from the HAALSI Study**. *BMJ Open* (2021.0) **11** 1-9
6. Herbst K, Law M, Geldsetzer P, Tanser F, Harling G, Bärnighausen T. **Innovations in health and demographic surveillance systems to establish the causal impacts of HIV policies**. *Curr Opin HIV AIDS* (2015.0) **10** 483-94. DOI: 10.1097/COH.0000000000000203
7. Hirschhorn LR, Kaaya SF, Garrity PS, Chopyak E, Fawzi MCS. **Cancer and the ‘other’ noncommunicable chronic diseases in older people living with HIV/AIDS in resource-limited settings: A challenge to success.**. *AIDS* (2012.0) **26** S65-75. PMID: 22781178
8. Deeks SG, Lewin SR, Havlir D V. **The end of AIDS: HIV infection as a chronic disease**. *The Lancet* (2013.0) **382** 1525-33. DOI: 10.1016/S0140-6736(13)61809-7
9. 9UNAIDS. Global AIDS Update 2020: Tackling entrenched inequalities to end epidemics [Internet].
Geneva: Joint United Nations Programme on HIV/AIDS (UNAIDS); 2020 [cited 2020 July 22]. Available from: https://www.unaids.org/en/resources/documents/2020/global-aids-report.. *Global AIDS Update 2020: Tackling entrenched inequalities to end epidemics [Internet].*
10. Mugisha J, Schatz EJ, Randell M, Kuteesa M, Kowal P, Negin J. **Chronic disease, risk factors and disability in adults aged 50 and above living with and without HIV: findings from the Wellbeing of Older People Study in Uganda.**. *Glob Health Action* (2016.0) **9** 31098. DOI: 10.3402/gha.v9.31098
11. Nyasani E, Sterberg E, Smith H. **Fostering children affected by AIDS in Richards Bay, South Africa: A qualitative study of grandparents’ experiences.**. *African Journal of AIDS Research.* (2009.0) **8** 181-92. DOI: 10.2989/AJAR.2009.8.2.6.858
12. Dolbin-MacNab ML, Yancura LA. **International Perspectives on Grandparents Raising Grandchildren: Contextual Considerations for Advancing Global Discourse.**. *International Journal of Aging & Human Development.* (2018.0) **86** 3-33. DOI: 10.1177/0091415016689565
13. Schatz E, Madhavan S, Collinson M, Gomez-Olive F, Ralston M. **Dependent or Productive? A New Approach to Understanding the Social Positioning of Older South Africans Through Living Arrangements.**. *Res Aging [Internet].* (2015.0) **37** 581-605. PMID: 25651584
14. Jennings EA, Farrell MT, Kobayashi LC. **Grandchild Caregiving and Cognitive Health Among Grandparents in Rural South Africa.**. *J Aging Health.* (2021.0) **33** 661-73. DOI: 10.1177/08982643211006592
15. Kasedde S, Doyle AM, Seeley JA, Ross DA. **They are not always a burden: Older people and child fostering in Uganda during the HIV epidemic.**. *Soc Sci Med.* (2014.0) **113** 161-8. DOI: 10.1016/j.socscimed.2014.05.002
16. Munthree C, Maharaj P. **Growing old in the era of a high prevalence of HIV/AIDS: The impact of AIDS on older men and women in KwaZulu-Natal, South Africa.**. *Res Aging.* (2010.0) **32** 155-74
17. Rutakumwa R, Zalwango F, Richards E, Seeley J. **Exploring the care relationship between grandparents/older carers and children infected with HIV in South-Western Uganda: Implications for care for both the children and their older carers**. *Int J Environ Res Public Health* (2015.0) **12** 2120-34. DOI: 10.3390/ijerph120202120
18. Lachman J, Cluver L, Boyes M, Kuo C, Casale M. **Positive parenting for positive parents: HIV/AIDS, poverty, caregiver depression, child behavior and parenting in South Africa.**. *AIDS Care.* (2014.0) **26** 304-13. DOI: 10.1080/09540121.2013.825368
19. Muruthi JR, Dolbin-MacNab ML, Jarrott SE. **Social Capital Among Black South African Grandmothers Raising Grandchildren**. *Journal of Applied Gerontology* (2021.0) **40** 1280-7. DOI: 10.1177/0733464820966474
20. Mugisha J, Schatz E, Seeley J, Kowal P. **Gender perspectives in care provision and care receipt among older people infected and affected by HIV in Uganda.**. *African Journal of AIDS Research* (2015.0) **14** 159-67. DOI: 10.2989/16085906.2015.1040805
21. Matovu S, Wallhagen M. **Loss as Experienced by Ugandan Grandparent-Caregivers of Children Affected by HIV/AIDS**. *J Loss Trauma [Internet].* (2018.0) **23** 502-15. PMID: 31839755
22. Shung-King M, Lake L, Sanders D, Hendricks M, Shung-King M, Lake L, Sanders D, Hendricks M. *South African Child Gauge 2019* (2019.0)
23. Richard AA, Shea K. **Delineation of self-care and associated concepts**. *Journal of Nursing Scholarship* (2011.0) **43** 255-64. DOI: 10.1111/j.1547-5069.2011.01404.x
24. Schulman-Green D, Jaser S, Martin F, Alonzo A, Grey M, McCorkle R. **Processes of self-management in chronic illness**. *Journal of Nursing Scholarship* (2012.0) **44** 136-44. DOI: 10.1111/j.1547-5069.2012.01444.x
25. Schulman-Green D, Feder SL, Dionne-Odom JN, Batten J, En Long VJ, Harris Y. **Family Caregiver Support of Patient Self-Management During Chronic, Life-Limiting Illness: A Qualitative Metasynthesis.**. *J Fam Nurs.* (2020.0) **00** 1-18. DOI: 10.1177/1074840720977180
26. Gareta D, Baisley K, Mngomezulu T, Smit T, Khoza T, Nxumalo S. **Cohort Profile Update: Africa Centre Demographic Information System (ACDIS) and population-based HIV survey.**. *Int J Epidemiol.* (2021.0) **50** 33-34f. DOI: 10.1093/ije/dyaa264
27. Knight L, Hosegood V, Timæus IM. **The South African disability grant: Influence on HIV treatment outcomes and household well-being in KwaZulu-Natal.**. *Dev South Afr* (2013.0) **30** 135-47
28. Zaidi J, Grapsa E, Tanser F, Newell ML, Bärnighausen T. **Dramatic increase in HIV prevalence after scale-up of antiretroviral treatment: a longitudinal population-based HIV surveillance study in rural KwaZulu-Natal**. *AIDS* (2013.0) **27** 2301-5. PMID: 23669155
29. Sharman M, Bachmann M. **Prevalence and health effects of communicable and non-communicable disease comorbidity in rural KwaZulu-Natal, South Africa.**. *Tropical Medicine and International Health* (2019.0) **24** 1198-207. DOI: 10.1111/tmi.13297
30. Nkosi B, Seeley J, Ngwenya N, Mchunu SL, Gumede D, Ferguson J. **Exploring adolescents and young people’s candidacy for utilising health services in a rural district, South Africa.**. *BMC Health Serv Res* (2019.0) **19** 1-12. PMID: 30606168
31. Tanser F, Hosegood V, Benzler J, Solarsh G. **New approaches to spatially analyse primary health care usage patterns in rural South Africa**. *Tropical Medicine and International Health* (2001.0) **6** 826-38. DOI: 10.1046/j.1365-3156.2001.00794.x
32. Saul J, Bachman G, Allen S, Toiv NF, Cooney C, Beamon T. **The DREAMS core package of interventions: A comprehensive approach to preventing HIV among adolescent girls and young women.**. *PLoS One.* (2018.0) **13** e0208167. DOI: 10.1371/journal.pone.0208167
33. Chimbindi N, Birdthistle I, Floyd S, Harling G, Mthiyane N, Zuma T. **Directed and target focused multi‐sectoral adolescent HIV prevention: Insights from implementation of the ‘DREAMS Partnership’ in rural South Africa.**. *J Int AIDS Soc* (2020.0) **23** e25575. DOI: 10.1002/jia2.25575
34. Chikumbu EF, Bunn C, Kasenda S, Dube A, Phiri-Makwakwa E, Jani BD. **Experiences of multimorbidity in urban and rural Malawi: An interview study of burdens of treatment and lack of treatment.**. *PLOS Global Public Health* (2022.0) **2** e0000139
35. Bukenya D, Van Hout MC, Shayo EH, Kitabye I, Junior BM, Kasidi JR. **Integrated healthcare services for HIV, diabetes mellitus and hypertension in selected health facilities in Kampala and Wakiso districts, Uganda: A qualitative methods study.**. *PLOS Global Public Health.* (2022.0) **2** e000084
36. Vincent KA. **The advantages of repeat interviews in a study with pregnant schoolgirls and schoolgirl mothers: piecing together the jigsaw**. *International Journal of Research and Method in Education* (2013.0) **36** 341-54
37. Babchuk WA. **Fundamentals of qualitative analysis in family medicine.**. *Fam Med Community Health* (2019.0) **7** 1-10. DOI: 10.1136/fmch-2018-000040
38. Nyirenda M, Newell ML, Mugisha J, Mutevedzi PC, Seeley J, Scholten F. **Health, wellbeing, and disability among older people infected or affected by HIV in Uganda and South Africa.**. *Glob Health Action.* (2013.0) **6** 19201. DOI: 10.3402/gha.v6i0.19201
39. Scholten F, Mugisha J, Seeley J, Kinyanda E, Nakubukwa S, Kowal P. **Health and functional status among older people with HIV/AIDS in Uganda.**. *BMC Public Health* (2011.0) **11** 886. DOI: 10.1186/1471-2458-11-886
40. Cree VE, Kay H, Tisdall K, Wallace J. **Stigma and Parental HIV.**. *Qualitative Social Work.* (2004.0) **3** 7-25
41. Murphy DA, Roberts KJ, Hoffman D. **Stigma and ostracism associated with HIV/AIDS: Children carrying the secret of their mothers’ HIV+ serostatus**. *J Child Fam Stud* (2002.0) **11** 191-202
42. Reyland SA, Higgins-D’Alessandro A, McMahon TJ. **Tell them you love them because you never know when things could change: Voices of adolescents living with HIV-positive mothers**. *AIDS Care* (2002.0) **14** 285-94. DOI: 10.1080/09540120120076977
43. Sharer M, Cluver L, Shields J. **Mental health of youth orphaned due to AIDS in South Africa: biological and supportive links to caregivers**. *Vulnerable Child Youth Stud* (2015.0) **10** 141-52
44. Sun F, Norman IJ, While AE. **Physical activity in older people: a systematic review.**. *BMC Public Health* (2013.0) **13** 449. DOI: 10.1186/1471-2458-13-449
45. Burton E, Farrier K, Hill KD, Codde J, Airey P, Hill AM. **Effectiveness of peers in delivering programs or motivating older people to increase their participation in physical activity: Systematic review and meta-analysis.**. *J Sports Sci* (2018.0) **36** 666-78. DOI: 10.1080/02640414.2017.1329549
46. Phillips EA, Comeau DL, Pisa PT, Stein AD, Norris SA. **Perceptions of diet, physical activity, and obesity-related health among black daughter-mother pairs in Soweto, South Africa: A qualitative study.**. *BMC Public Health.* (2016.0) **16** 750. DOI: 10.1186/s12889-016-3436-8
47. Kuo C, Operario D. **Health of adults caring for orphaned children in an HIV endemic community in South Africa.**. *AIDS Care* (2011.0) **23** 1128-35. DOI: 10.1080/09540121.2011.554527
48. Rochat TJ, Bland R, Coovadia H, Stein A, Newell ML. **Towards a family-centered approach to HIV treatment and care for HIV-exposed children, their mothers and their families in poorly resourced settings.**. *Future Virol.* (2011.0) **6** 687-96. DOI: 10.2217/fvl.11.45
49. van Houtum L, Rijken M, Groenewegen P. **Do everyday problems of people with chronic illness interfere with their disease management?**. *BMC Public Health* (2015.0) **15** 1000. DOI: 10.1186/s12889-015-2303-3
50. Wright S, Zalwango F, Seeley J, Mugisha J, Scholten F. **Despondency Among HIV-Positive Older Men and Women in Uganda.**. *J Cross Cult Gerontol.* (2012.0) **27** 319-33. DOI: 10.1007/s10823-012-9178-x
51. Cahill E, Lewis LM, Barg FK, Bogner HR. **You Don’t Want to Burden Them: Older Adults’ Views on Family Involvement in Care.**. *J Fam Nurs.* (2009.0) **15** 295-317. DOI: 10.1177/1074840709337247
52. Pamungkas RA, Chamroonsawasdi K, Vatanasomboon P. **A systematic review: Family support integrated with diabetes self-management among uncontrolled type II diabetes mellitus patients**. *Behavioral Sciences* (2017.0) **7** 1-17. DOI: 10.3390/bs7030062
53. Rochat TJ, Mkwanazi N, Bland R. **Maternal HIV disclosure to HIV-uninfected children in rural South Africa: A pilot study of a family-based intervention.**. *BMC Public Health* (2013.0) **13** 147. DOI: 10.1186/1471-2458-13-147
54. Dolbin-MacNab ML, Jarrott SE, Moore LE, O’Hora KA, Vrugt MDC, Erasmus M. **Dumela Mma: an examination of resilience among South African grandmothers raising grandchildren.**. *Ageing Soc.* (2016.0) **36** 2182-212
55. Nalugya R, Russell S, Zalwango F, Seeley J. **The role of children in their HIV-positive parents’ management of antiretroviral therapy in Uganda.**. *African Journal of AIDS Research* (2018.0) **17** 37-46. DOI: 10.2989/16085906.2017.1394332
56. Guilleminault L, Williams EJ, Scott HA, Berthon BS, Jensen M, Wood LG. **Diet and asthma: Is it time to adapt our message?**. *Nutrients.* (2017.0) **9** 1227. DOI: 10.3390/nu9111227
57. Begen FM, Barnett J, Barber M, Payne R, Gowland MH, Lucas JS. **Parents’ and caregivers’ experiences and behaviours when eating out with children with a food hypersensitivity.**. *BMC Public Health* (2018.0) **18** 38
58. Scaglioni S, De Cosmi V, Ciappolino V, Parazzini F, Brambilla P, Agostoni C. **Factors influencing children’s eating behaviours.**. *Nutrients* (2018.0) **10** 706. DOI: 10.3390/nu10060706
59. Mottiar S, Lodge T. **The role of community health workers in supporting South Africa’s HIV/AIDS treatment programme**. *African Journal of AIDS Research* (2018.0) **17** 54-61. DOI: 10.2989/16085906.2017.1402793
60. Loeliger KB, Niccolai LM, Mtungwa LN, Moll A, Shenoi S V. **‘I have to push him with a wheelbarrow to the clinic’: Community health workers’ roles, needs, and strategies to improve HIV care in rural South Africa.**. *AIDS Patient Care STDS* (2016.0) **30** 385-94. DOI: 10.1089/apc.2016.0096
61. Chetty-Makkan CM, Desanto D, Lessells R, Charalambous S, Velen K, Makgopa S. **Exploring the promise and reality of ward-based primary healthcare outreach teams conducting TB household contact tracing in three districts of South Africa.**. *PLoS One.* (2021.0) **16** e0256033. DOI: 10.1371/journal.pone.0256033
62. Rampamba EM, Meyer JC, Helberg E, Godman B. **Medicines availability among hypertensive patients in primary health care facilities in a rural province in South Africa: Findings and implications.**. *J Res Pharm Pract* (2020.0) **9** 181-5. DOI: 10.4103/jrpp.JRPP_20_49
|
---
title: Socio-economic and environmental factors affecting breastfeeding and complementary
feeding practices among Batwa and Bakiga communities in south-western Uganda
authors:
- Giulia Scarpa
- Lea Berrang-Ford
- Sabastian Twesigomwe
- Paul Kakwangire
- Maria Galazoula
- Carol Zavaleta-Cortijo
- Kaitlin Patterson
- Didacus B. Namanya
- Shuaib Lwasa
- Ester Nowembabazi
- Charity Kesande
- Janet E. Cade
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021580
doi: 10.1371/journal.pgph.0000144
license: CC BY 4.0
---
# Socio-economic and environmental factors affecting breastfeeding and complementary feeding practices among Batwa and Bakiga communities in south-western Uganda
## Abstract
Improving breastfeeding and complementary feeding practices is needed to support good health, enhance child growth, and reduce child mortality. Limited evidence is available on child feeding among Indigenous communities and in the context of environmental changes. We investigate past and present breastfeeding and complementary feeding practices within Indigenous Batwa and neighbouring Bakiga populations in south-western Uganda. Specifically, we describe the demographic and socio-economic characteristics of breastfeeding mothers and their children, and individual experiences of breastfeeding and complementary feeding practices. We investigate the factors that have an impact on breastfeeding and complementary feeding at community and societal levels, and we analysed how environments, including weather variability, affect breastfeeding and complementary feeding practices. We applied a mixed-method design to the study, and we used a community-based research approach. We conducted 94 individual interviews ($$n = 47$$ Batwa mothers/caregivers & $$n = 47$$ Bakiga mothers/caregivers) and 12 focus group discussions ($$n = 6$$ among Batwa & $$n = 6$$ among Bakiga communities) from July to October 2019. Ninety-nine per cent of mothers reported that their youngest child was currently breastfed. All mothers noted that the child experienced at least one episode of illness that had an impact on breastfeeding. From the focus groups, we identified four key factors affecting breastfeeding and nutrition practices: marginalisation and poverty; environmental change; lack of information; and poor support. Our findings contribute to the field of global public health and nutrition among Indigenous communities, with a focus on women and children. We present recommendations to improve child feeding practices among the Batwa and Bakiga in south-western Uganda. Specifically, we highlight the need to engage with local and national authorities to improve breastfeeding and complementary feeding practices, and work on food security, distribution of lands, and the food environment. Also, we recommend addressing the drivers and consequences of alcoholism, and strengthening family planning programs.
## Introduction
In 2019, $21\%$ of children under 5 years globally were stunted and $2\%$ wasted [1]. Undernutrition in infants exacerbates the risk of mortality, morbidity, and chronic diseases, and causes delays in neuro-psychomotor development [2,3]. Poor children living in vulnerable settings are 20 times more at risk of undernutrition than others [1], and especially among Indigenous communities [2,4]. This is caused by economic disparities, socio-cultural discrimination, and colonial legacies that translate into health inequalities [5].
Research demonstrates that improvements in infant and young child feeding (IYCF), and traditional feeding practices are critical to supporting good health, enhancing child growth, and reducing child mortality [6]. Breastmilk is the only nutritional source recommended by the WHO for newborns and infants up to 6 months [7]. According to WHO statistics, however, only 1 out of every 3 children worldwide is exclusively breastfeed for their first six months of life, and only 2 of 5 are immediately breastfed in the first hour after birth [8]. In addition, the WHO and UNICEF recommend starting complementary feeding along with breastfeeding after the first 6 months of life to avoid stunting in childhood [9,10]. Breastmilk alone is insufficient to ensure adequate child growth after 6 months of age [2]. Low quality and quantity of foods, and late introduction of solid foods are found to be causes of undernutrition [11].
Maternal diets influence infant diets. Tiedje et al. [ 12] proposed an ecological approach to understand the influences of maternal nutrition on breastfeeding and infant feeding practices by analysing contextual factors such as family, community and healthcare system. Later, others extended the focus of this model to also include societal contexts and changing environments [13]. Indeed, child feeding has been observed in Indigenous populations [14] to be driven by social and environmental contexts, which were identified as critical in understanding breastfeeding and infant feeding practices. For example, Sellen documented changes in individual choices and food behaviour among Indigenous families due to changes in socio-environmental conditions, such as culture, work activities, natural environment, and traditional food supplies [15]. In some cases, mothers substituted traditional infant foods with packaged food during the complementary feeding period [16].
While there are well-established indicators to measure changes in child nutrition linked to health (breastfeeding duration, starting age of complementary feeding, minimal dietary diversity, minimum meal frequency, minimum acceptable diet), few studies have assessed IYCF among Indigenous groups or in the context of environmental change [14,17]. The aim of this study was to investigate past and present breastfeeding and complementary feeding practices within the Indigenous Batwa and Bakiga populations. The main objectives were: 1. To describe the demographic and socio-economic attributes of breastfeeding mothers and their children, and individual experiences of breastfeeding and complementary feeding practices; 2. To investigate the factors that have an impact on breastfeeding and complementary feeding at community and societal level; 3. To analyse if and how environments, including weather variability, affect breastfeeding and complementary feeding practices.
## Material and methods
Ethics approvals were obtained from the University of Leeds Research Ethics Board (AREA 18–156), the Ugandan National Council for Science and Technology (SS5164), and the Makerere University Research Ethics Committee (MAKS REC 07.19.313/PR1). For the minors included in the study, we obtained consent from their parents or guardians.
## Study population
The Batwa and Bakiga communities live in the District of Kanungu in south-western Uganda, located on the border with the Democratic Republic of Congo. The Bakiga are historically an agrarian society, who depend on agriculture and livestock, and represent the majority of the population [18]. Both populations suffer from high burden of illness, and especially the Batwa have high incidence of malaria, malnutrition and gastrointestinal diseases [19–21], while the Bakiga have higher levels of HIV [22]. From previous studies, there is evidence that malnutrition is very high among Batwa communities, especially among children under five: $8\%$ of male children classified as wasted [18], for example, compared to $4\%$ nationally [23]. The proportion of male undernourished cases among *Batwa is* higher than among females, with boys at greater odds of being severely malnourished. In every Batwa age-sex grouping, $15\%$ or more individuals are malnourished [20]. Also, according to Patterson et al. [ 24], more than $90\%$ of Batwa households are rated as “very highly food insecure”. In 2017, Batwa mothers reported being malnourished and having malnourished children due to a scarcity of food [25].
## Conceptual framework
We adapted the conceptual framework of Hector et al. [ 13] (Fig 1). Our framework is composed of four levels: individual, group, societal and environmental. The individual level analyses the characteristics of the mother, the child, and the dyad (mother-child pairing), including knowledge on breastfeeding, mother-child interactions, and health status. The group level describes factors that influence breastfeeding practices in their proximal social setting, for example accessibility to health facilities, and mothers’ work and community environments. The societal level considers contextual elements that have an impact on breastfeeding choices such as the role of women and men in society, and cultural norms [13]. Finally, we included an environmental level to the framework to further analyse the impact of environmental contexts and events (weather, land suitability for food production, and extreme events such as flooding or drought) on breastfeeding and complementary feeding practices.
**Fig 1:** *Conceptual framework used to analyse breastfeeding and complementary feeding among the Batwa and Bakiga communities.The group level refers to the level of household and community; the societal level refers to the level of region, ethnicities and country. There is interaction across all levels; for example, the environmental level interacts with the societal and group levels.*
Marginalisation of Indigenous Batwa communities is present at all levels of society. The Batwa represent $1\%$ of the Ugandan population, and were displaced in 1991 by the Ugandan Government from their ancestral forest lands to create the Bwindi Impenetrable National Park [26]. Traditionally hunters-gathers, since 1991 the Batwa have begun transitioning to cultivating crops [24]. They work mostly as farmers, hired by the Bakiga; some are craft makers or brick makers or work in tourism [18]. However, the absence of a traditional culture of farming and low socio-economic status exacerbate food insecurity [27] which have an impact on individual nutrition, and can affect negatively child health [28]. Although many Batwa and Bakiga families are poor, Batwa per capita income is substantially lower than the national average (0.36 US dollars/day compared to 0.99 US dollars/day) [24]. Inequities are also found in education. Less than $12\%$ of Batwa living in Kanungu District are able to write and read, and to access education [24]; the school drop-out rate is especially high following marriage at a young age [29].
Persistent poverty underpins poorer access to healthcare services among Batwa, despite similar health facilities for both populations [20]. For example, compared to the rest of Uganda where $57\%$ of mothers gave birth with health professionals, only $40\%$ of Batwa births occurred in health facilities in 2017 [22]. Also, in cases of child malnutrition, Batwa mothers tend to stay at home rather than go to the hospital for treatment [20]. One of the main reasons for scarce hospital attendance is the cost of health insurance premiums [30], which are not affordable for many Batwa families, as well as persistent social and ethnic discrimination [18]. Also, Batwa more frequently report not having soap or access to health facilities compared to non-Indigenous neighbours, with negative consequences on maternal and child health [31].
Poverty and discrimination are linked to poor mental health, alcoholism and domestic violence.. This link is documented in previous studies, although the research on mental health among the Batwa and Bakiga communities remains limited [24]. Previous research indicates that internal displacement, as occurred among the Batwa population, can have a negative effect on health, wellbeing and socioeconomic status [32]. Batwa and Bakiga report high levels of alcoholism, particularly among men and during periods of food insecurity. High consumption of alcohol is an emerging problem among these communities, and the local hospital has addressed this by offering alcohol rehabilitation services [33]. Alcoholism has been linked to domestic violence [24,34]. Reported implications of alcoholism among Batwa include compromised food security for adults and children, mental health concerns, and poorer health outcomes [24].
Poverty and marginalisation are projected to be exacerbated by the impacts of climate change with consequences on nutrition and health, especially for young children [35,36]. The District of *Kanungu is* affected by extreme climatic events, such as the floods that occurred in 2019. Climate change projections anticipate an increase in annual mean temperature and change in seasonal variation [37,38]. The dry season is usually between December and February, and between June and August, however rainfalls are now longer (from September to December, and then from March to June) and less predictable, with fewer sunny days [39]. These changes are impacting key local crops (groundnuts, beans and cassava) that are an important nutritional source for the Batwa and Bakiga communities [39].
## Study design
We undertook a mixed-method study to explore breastfeeding and complementary feeding practices among mothers with children under two years in Kanungu district.
The research was guided by a community-based participatory research approach which engage researchers and community participants as equal partners during the research process; the objective is to educate or promote social change [40,41]. This approach has been used among marginalised groups, especially Indigenous communities, as it helps to reinforce the respect and collaboration between stakeholders and researchers [42].
Batwa and non-Batwa participants assessed the barriers to breastfeeding and complementary feeding practices, evaluated and shared their level of knowledge on maternal and infant nutrition, and critically explored solutions to improve mother and child health.
## Settlement and individual sampling
Twelve communities, six Batwa and six geographically matched adjacent Bakiga settlements, were included in the study (Fig 2).
**Fig 2:** *In the map we represented the ten Batwa settlements that participated in the research; there is a correspondent Bakiga settlement for each Batwa settlement.In our study, we involved 6 Batwa and 6 Bakiga settlements: Bikuuto (Batwa) & Bikuuto cell (Bakiga), Kihembe (Batwa) & Kengoma cell (Bakiga), Kitariro (Batwa) & Kitariro cell (Bakiga), Mpungu (Batwa) & Kikome cell (Bakiga), Kebiremu (Batwa) &Kebiremu cell (Bakiga) and Byumba (Batwa) & Byumba cell (Bakiga). Map adapted from Patterson, 2017 [24]. Coordinates of settlements described in S1 Data.*
The selection of communities sought a sample representing variation in terms of geographic location and market access. Mothers with children aged under two years were sampled from 12 settlements (6 Batwa and 6 Bakiga): 1) two settlements located very close to the market and shops (Bikuuto/Bikuto cell and Kihembe/Kengoma cell), 2) two settlements close to the forest and located very far from the market and shops (Kitariro/Kitariro cell and Mpungu/Kikome cell area), and two settlements situated mid-way (Kebiremu/Kebiremu cell and Byumba/Byumba cell). The selection of samples in the different areas allowed for exploration of geographically linked variation in practices and availability of food for infants’ complementary feeding between Batwa and Bakiga communities and between settlements with different proximity to market centres.
Twelve focus groups were conducted, one in each of the communities ($$n = 6$$ Batwa and $$n = 6$$ Bakiga), with 7–11 mothers per group. First, the research assistant met the local leader of each community, sought the permission to conduct the study and asked for a list with the names of mothers with children under 2 years living in Kanungu District. We invited all eligible Batwa mothers to participate, as the number of mothers was limited (47 individuals). In the case of the Bakiga, we conducted a random sample of eligible mothers, as the number of women with children under 2 years was greater. A number was consecutively assigned to each eligible individual starting at one. A table of random numbers was used to sample the 47 individuals who were invited to participate [43].
## Data collection
Three Ugandan researchers collected the data from July to October 2019. The research team was composed of a Ugandan male researcher living in Kanungu district, and a Mutwa (Batwa singular) and a Mukiga (Bakiga singular) female researcher. Focus group discussions (FGDs) and the individual interviews were conducted and audio-recorded in the local language, Rukiga, and then the information was translated into English by the local researcher (ST). Before sampling individuals, the research team contacted the chairperson of each community to introduce the study and seek his/her approval. Written consent was obtained before all interviews.
We created an interview guide with open questions for the FGD (S1 and S1A Text). The questions followed the themes from the Optimal IYCF guidelines, specifically the complementary feeding section [9,10]; IYCF programmes aim to prioritize and improve breastfeeding and complementary feeding practices to reduce malnutrition worldwide [9]. FGDs helped to explore the complexity of lived experiences that were not capturable with standard questionnaires, and encouraged women to share knowledge and perceptions on IYCF. Each group discussion lasted on average 62 minutes.
The individual interview questionnaire included primarily closed questions, with some requiring brief explanation (S2 and S2A Text). Some were adapted from the Standardized Monitoring and Assessment of Relief and Transition Methodology [44] to evaluate some factors linked to the nutritional status such as hygiene, food and water accessibility level. Interviews lasted on average 12 minutes.
## Data analysis
To respect the multi-perspectivity and the multivocality of the data [45–47], we coded, analysed and organised the qualitative data using NVivo 12® software. Contextualized thematic analysis (latent and manifest) was used to analyse the data [48]. The analysis process involved data familiarization, generating initial codes, defining, reviewing, and naming themes. In this manuscript, we reported the findings of the themes that were discussed by the participants during the FGDs as key factors of successful or unsuccessful breastfeeding and complementary feeding practices. Additionally, we descriptively analysed demographic data on Batwa and Bakiga women and children using Stata® version 15.
## Inclusivity in global research
Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S3 Text).
## Characteristics of mothers/child caregivers and children
Ninety-four women (47 Batwa and 47 Bakiga) took part in the individual interviews and focus group discussions (FGDs). A description of the 94 mothers/child caregivers and 95 child participating in the individual interviews included data on attributes and interaction between mothers and children and can be found in Table 1. The response rate in both FGDs and individual interviews was $100\%$. In the case of four Batwa children, the participating primary carer was a grandmother rather than mother (e.g. the mother had died or left the settlement). The number of Bakiga children included in the study was 48 as there was one set of twins.
**Table 1**
| Sample: 94 women with 95 infants | Batwa women (N = 47) | Bakiga women (N = 47) |
| --- | --- | --- |
| | Counts (percentage) 1 | Counts (percentage) |
| Age of female primary caregivers | | |
| 15–19 | 5 (11) | 4 (9) |
| 20–24 | 10 (21) | 13 (28) |
| 25–29 | 7 (15) | 17 (36) |
| 30–34 | 14 (30) | 5 (11) |
| > = 35 | 11 (23) | 8 (17) |
| Community of residence | | |
| Kebiremu/Kebiremu cell | 8 (17) | 7 (15) |
| Kitarir/Kitariro cell | 7 (15) | 7 (15) |
| Mpungu/ Kikome cell | 6 (13) | 7 (15) |
| Bikuuto/Bikuuto cell | 10 (21) | 7 (15) |
| Byumba/Byumba cell | 7 (15) | 8 (17) |
| Kihembe/Kengoma cell | 9 (19) | 11 (23) |
| Number of children under 5 years in the household | | |
| One child | 9 (19) | 11 (23) |
| Two children | 8 (17) | 9 (19) |
| Three children | 8 (17) | 9 (19) |
| Four Children | 8 (17) | 8 (17) |
| More than four children | 14 (30) | 10 (21) |
| Mothers who lost at least 1 child during pregnancy or after birth | 20 (43) | 8 (17) |
| Birth interval (referred to last child) | | |
| < 24 months | 2 (4) | 3 (6) |
| > = 24 months | 45 (96) | 44 (94) |
| No illnesses/complications during child birth 2 | 41 (87) | 40 (85) |
| Received food from Government/NGOs | 43 (91) | 5 (11) |
| Own land (small size) | 44 (94) | 47(100) |
| Own animals 3 | 19 (40) | 39 (83) |
| Pit latrine access | 47 (100) | 47 (100) |
| Access to soap | | |
| Yes | 2 (4) | 30 (64) |
| Sometimes | 32 (68) | 16 (34) |
| No | 13 (28) | 1 (2) |
| Protected water facilities | 39 (83) | 47 (100) |
| Age at first pregnancy | Mean/(SE) 18.9 (0.4) | Mean/(SE) 19.5 (0.4) |
| Child age | Batwa children (N = 47) | Bakiga children (N = 48) |
| 0–6 months | 20 (43) | 9 (19) |
| 7–12 months | 13 (28) | 12 (25) |
| 13–18 months | 8 (17) | 9 (19) |
| 19–23 months | 6 (13) | 18 (38) |
| No illnesses/complications at birth 4 | 45 (96) | 44 (92) |
| Currently breastfed | 46 (98) | 48 (100) |
| Child looked after by the mother/caregivers only | 40 (85) | 38 (79) |
| Place of birth | | |
| Home | 19 (40) | 7 (15) |
| Health Centre | 4 (9) | 17 (35) |
| Hospital | 24 (51) | 24 (50) |
| Reported child illnesses/symptoms in the first six months of life by caregivers | | |
| Cough | 31 (66) | 32 (67) |
| Diarrhoea | 38 (81) | 35 (73) |
| Flu | 23 (49) | 20 (42) |
| Fever | 7 (15) | 8 (17) |
| Malaria | 40 (85) | 20 (42) |
| Pneumonia | 15 (32) | 11 (23) |
| Vomiting | 8 (17) | 4 (8) |
| Skin diseases | 4 (9) | 8 (17) |
| Other | 8 (17) | 5 (10) |
From the data collected through the individual interviews, there were some differences in demographic and health characteristics between Batwa and Bakiga, primarily related to wealth/assets, sanitation/hygiene, and place of birth. The majority of Batwa did not have regular access to soap (32 of 47 participants, $68\%$); although most of the participants reported washing their hands with water only before and after breastfeeding, after visiting the toilet and eating, none reported hand washing before cooking or preparing food for the children and family. Eighty-three percent ($$n = 39$$) of Bakiga mothers reported owning at least one animal compared to $40\%$ ($$n = 19$$) of Batwa mothers. During the interviews, most Batwa mothers ($91\%$, $$n = 43$$) reported receiving food aid from NGOs or government, while only a few Bakiga mothers ($11\%$, $$n = 5$$) living close to the forest reported receiving compensation in the form of food (when elephants destroyed their gardens).
Children of Batwa mothers in this study were generally younger ($$n = 20$$ or $43\%$ of children aged 0–6 months) compared to children of Bakiga mothers ($$n = 18$$ or $38\%$). Births among Batwa mothers occurred more frequently at home ($40\%$, $$n = 19$$) compared to Bakiga mothers ($15\%$, $$n = 7$$). Reporting of child loss during pregnancy and after birth was higher among Batwa ($43\%$, $$n = 20$$) compared to Bakiga ($17\%$, $$n = 8$$) mothers.
Mothers reported that nearly $99\%$ of children were currently breastfed, and were able to breastfeed at their place of work or livelihood activity, typically their family farm. Mothers of all children older than 7 days reported that their child had experienced one or more episodes of illness in the first 6 months of life that had had a negative impact on breastfeeding.
## Main findings from the focus group discussions
We identified the key factors constraining child feeding through the focus group discussions which are summarised in Table 2. The elements are interlinked and positioned across the four levels of the framework (Fig 1): marginalisation and poverty at group/societal level, crossing the environmental and individual level; environmental changes at environmental level, crossing the societal level; lack of breastfeeding and complementary feeding-related information at individual level, crossing the group/societal level; and poor support at group level, crossing the societal level.
**Table 2**
| Key factors | Level(s) | Description |
| --- | --- | --- |
| Marginalisation and Poverty | Group & Societal | Due to marginalisation and high poverty levels, Batwa and Bakiga participants cannot afford to buy enough and high-quality foods. This had a negative impact on child nutrition and growth. |
| Environmental changes | Environmental | Extreme climatic events, especially droughts and floods, negatively impacted food security, especially accessibility and availability of foods with consequences on child and maternal nutrition. |
| Lack of breastfeeding and complementary feeding information | Individual | Lack of information received from health workers on complementary feeding practices and cultural beliefs were found to be limiting factors to achieve good nutrition in children according to the Batwa and Bakiga mothers. |
| Poor support | Group | Support is fundamental during breastfeeding and complementary feeding practices. However, this was limited for the Batwa and Bakiga women, who were used to live in contexts of domestic violence and alcoholism. |
## Marginalisation and poverty
The Batwa and Bakiga participants identified poverty as the dominant driver limiting the success of breastfeeding and complementary feeding practices. They repeated during the discussion that “we don’t have enough money, we are poor, thus we cannot feed our children” (Batwa FGD), and drew a link between lack of money and resources, and poor nutrition outcomes: Batwa mothers added that the Bakiga are also impoverished: “not only the Batwa do not have food, but also the Bakiga, so the Batwa cannot really work for them for food” (Batwa FGD). Batwa and Bakiga participants talked about their eviction from the forest, their traditional land, where previously they looked for food, now forbidden by the Government; they linked their historic displacement—and persistent restriction of forest access—to their current poverty and food insecurity: During the discussion, mothers also gave advice on specific foods that pregnant and lactating women should eat to increase milk production: millet porridge, greens, beans, Amaranthus leaves, and cassava leaves (Batwa FGD), and some Bakiga participants added also the consumption of meat and fish.
Women explained that children have a greater risk of malnutrition due to maternal poor diet, and infants are more likely to get sick with vomiting and diarrhoea due to ‘low-quality’ food, poor in proteins. They agreed that the quality of nutrition depends on money availability: “only if we have money, we can buy eggs, which are very nutritious for the baby” (Bakiga FGD). The most common foods for complementary feeding listed by mothers, and given from 4–5 months of age are matoke (green bananas), sweet or Irish potatoes, cassava, groundnuts, soup from meat (if available), and beans.
A Mutwa woman highlighted how their situation has changed compared to the past in terms of food availability: infants now are fed with what is available at home, and they don’t always have ‘sauce’ (beans, groundnuts or greens) to accompany cereals. According to the participants, poor quality food causes sickness in both mothers and children: Participants added that appropriate complementary feeding practices get more complicated to follow in case of twins or multiple children under 5 years of age as it is difficult to provide sufficient foods to satisfy their nutritional needs. Also, they argued that is harder to produce enough milk for two or more children at the same time, therefore weaning usually starts before 4 months (Bakiga FGD).
Participants suggested that improving food availability and land accessibility through collaboration with the Government and district would have a positive effect on dietary diversity, “to have a balanced diet”, and increase the quantity of food consumed by household (Batwa and Bakiga FGD).
The participants explained that food insecurity is also a consequence of limited availability of lands, and overpopulation; therefore, food production is decreasing with negative effects on child nutrition: *For this* reason, the women addressed the need of more lands to distribute among the community, and they sought to assess and redress barriers related to insecure land tenure and poor land quality to increase food production and availability (Batwa and Bakiga FGD).
## Poor support
Batwa and Bakiga participants reported that they lack support from partners, the community, and health workers during the breastfeeding and complementary feeding period.
The women in the study identified the role of partners and older children as important in supporting successful breastfeeding and complementary feeding. They suggested that older children usually help mothers to collect food for the family and look after the younger siblings when the mother works or is out of the home (Batwa and Bakiga FGDs). According to participants, partners (generally a husband) provide support by purchasing food for the family, assisting during feeding the child when ready for complementary foods, but partners’ support varies from household to household (Batwa and Bakiga FGDs). They also added that if the father does not buy food for the family, there may be a delay in the start of complementary feeding even in the case of poor breastmilk supply, leading to a deterioration in child nutrition (Bakiga FGD). For this reason, women requested more economic and social support from men, including help from the local district authorities to reduce alcohol consumption (Bakiga FGD).
In fact, participants indicated that domestic violence (verbal and/or physical) is exacerbated by high rates of alcoholism in the communities, in particular among men but also, in lower percentage, among women; they suggested to ask for the help of the district and Government to as this is also disruptive to child feeding: The participants articulated their perspectives on the causes of alcoholism and domestic violence, suggesting that alcoholism and domestic violence are driven by extreme poverty, inequalities and general and persistent food insecurity (Fig 3):
**Fig 3:** *Word frequency analysis with the most recurrent terms used by the participants to articulate the key drivers at the nexus of alcoholism, domestic violence, and child nutrition (breastfeeding and complementary feeding).The bigger the word, the more often it was mentioned during the FGDs (Scale: 1–50).*
Women mentioned that other support is usually offered by the community, and especially by relatives, friends, parents and neighbours. They usually provide food or help a mother to look after the baby and share information and knowledge on breastfeeding and weaning practices: Participants also reported other conditions where mothers lack support for breastfeeding. Many mentioned that they are able to breastfeed while working, but they also argued that in some workplaces breastfeeding is not allowed, and they reported the case of sex workers: Additionally, Batwa women explained how breastfeeding was difficult among sex workers: The women also noted that they lack strong support from health professionals to support mothers to breastfeed longer:
Bakiga women highlighted the need to strengthen family planning services to help women not only to breastfeed longer, but also to decide when conceiving. According to the participants, young women and teenagers are in fact more likely to have their next child within 23 months due to lack of information and lower use of contraception. Specifically, they suggested to involve governmental institutions to address this need.
Additionally, Batwa and Bakiga participants argued that support is also needed to improve hospital and health centre accessibility as mothers face challenges in accessing medical services when they need assistance:
## Lack of information on breastfeeding and complementary feeding
Both Batwa and Bakiga participants indicated that they had a good understanding of breastfeeding practices, and noted that it is healthy for their child and should be started as soon as possible after birth (Table 4). However, mothers felt that information on complementary feeding was insufficient. For this reason, they asked for more advice during antenatal visits and immunisations as in these circumstances the health professionals talk about breastfeeding in particular (Batwa FGD). Information on child feeding is provided to mothers by the hospital or NGOs (Table 3); the main topics covered in these sessions includes breastfeeding practices, complementary foods and malnutrition, importance of giving birth at the hospital and vaccinating children, family planning, and HIV testing before delivery. The participants reported that some information on child nutrition was also available through radio programmes that could be easily accessed by Batwa and Bakiga families. Some mothers reported receiving information from peers, local religious leaders, and at school (Batwa and Bakiga FGDs).
**Table 3**
| Courses mentioned by the Batwa and Bakiga participants | Course organiser |
| --- | --- |
| i) the principles of breastfeeding practices from positioning the baby to the breast to hygiene measures to apply | Hospital |
| ii) some information on the introduction of complementary feeding, and solid foods | Hospital |
| iii) the importance of delivering at the hospital with qualified health workers, taking supplementation during pregnancy and breastfeeding, and vaccinate the child at appropriate age | Hospital |
| iv) the causes of malnutrition, and how to avoid it | Hospital/NGOs |
| v) the importance of using family planning measures to control births and breastfeeding longer | Hospital/NGOs |
| vi) the need of testing the mothers for HIV before delivery (factor highlighted in the Bakiga FGD only) to know how long to breastfeed | Hospital |
| vii) the importance of microfinance courses to teach how to save or earn money by selling animals, and use it for transport to the hospital at the time of delivery or to cover any expenses when the mother or the child need to be hospitalised (e.g. in case of malnutrition) | NGOs |
Knowledge on breastfeeding and HIV was limited, and participants mentioned that breastfeeding should be stopped at six months if the mother is HIV positive. Also, they reported that if the child has teeth, it is better to start food sooner to avoid any bites on the breast that could facilitate the spread of HIV from mother to child. They also added that this is a suggestion given by health professionals (Batwa and Bakiga FGD).
## Cultural beliefs
Mothers discussed how cultural beliefs and local knowledge influence choices during breastfeeding and the complementary feeding period. Specifically, they discussed the role of traditional birth attendants (TBAs), and the practice of ‘chewing’ food for the child to make the first solid food softer.
Mothers reported confidence in the expertise of TBAs, including their knowledge on breastfeeding and complementary feeding, and willingness to be supported by them for antenatal and postnatal care together with the health care workers: Some Batwa participants mentioned also that TBAs offer traditional herbs to help with milk production, especially when mothers cannot buy millet porridge or do not have enough food to eat; therefore, mothers approach TBAs to ask for help with breastfeeding. A Mutwa mother also added that according to traditional practice, mothers are expected to stay at home in the first four days after birth to produce more milk: “If you do not stay at home, you won’t produce milk”.
Mothers discussed other practices used in the past to help children eat first solid foods, such as ‘chewing’: “*This is* not in practice anymore to avoid the diseases that came here […]. Those changes (type of food and chewing practice) had an impact on child growth, children now grow better” (Batwa FGD). However, many Batwa and Bakiga women did not support this position, arguing that children were healthier in the past when mothers used to chew wild foods from the forest before feeding the chewed food to the infant.
## Environment changes and environmental change
Weather variability, shifting seasonality, and environmental factors were discussed by the mothers as playing a role in the success or failure of breastfeeding and complementary feeding practices. Participants perceived that due to climate change, the seasons ‘are not the same as before’ and production of food has decreased, with consequences on maternal and child health status and nutrition: Also, participants perceived that a rise in temperatures has spread new diseases, such as malaria that before was uncommon, and due to this, children eat less, and get malnourished easily: Further, women stated that ‘December, January’ and ‘May and June’ are the harvesting months, the preferred time to introduce solid food, but ‘these seasons may not be always good’ anymore; therefore, the introduction of complementary feeding may be started later than 6 months, and weaning may be postponed: Mothers also reported that the type of food consumed during the complementary feeding period is ‘not energetic’ as wild meat is not consumed or consumed rarely. Participants argued that some families depend on only one type of food for the whole week because of food unavailability (Bakiga FGD), and they mostly eat once or twice a day only. A Mukiga mother discussed the consequences of extreme weather events on child nutrition and growth, adding that environmental changes contributed to food insecurity and reduced breastmilk production:
## Discussion
This study explored the key factors affecting breastfeeding and complementary feeding among the Batwa and Bakiga communities in south-western Uganda. Other studies in Uganda have only quantitatively investigated factors affecting complementary feeding by using Uganda Demographic Health Survey (UDHS) data focussing on child health, vaccination status and maternal education [49,50], and data on Indigenous communities are scarce. Our mixed-methods design allowed to explore the socio-cultural and environmental barriers to child feeding practices among the Indigenous Batwa population for which there are no data on the UDHS. Also, we compared the information with the neighbouring Bakiga who shared a similar experience in terms of breastfeeding and complementary feeding pratices.
Additionally, in collaboration with the community we synthetisized recommendations to improve breastfeeding and complementary feeding practices among Batwa and Bakiga communities (Table 4). These included suggestions to enhance the food environment, such as food availability and accessibility, health services, and community and medical support as summarised in Table 4. The community, also, recognized barriers and opportunities across levels and themes identified in our framework (Fig 1).
**Table 4**
| Recommendations | Expected outcome | Themes |
| --- | --- | --- |
| 1. Improve food availability and accessibility | Expected positive effect on dietary diversity and quantity of food consumed by mothers and children. | Livelihood resources & Natural resources |
| 2. Targeted provision of ‘good foods’ to lactating and pregnant mothers, including millet porridge, greens, beans, dodo, cassava leaves, meat and fish | Expected increased milk production in lactating mothers with benefits for child nutrition. | Home/Family environment & Livelihood strategies |
| 3. Assess and redress barriers related to insecure land tenure and poor land quality in communities | Expected increased food production. | Livelihood resources |
| 4. Promote increased economic and social support from men/partners during breastfeeding and complementary feeding period | Expected decreased level of stress in mothers, increased household food availability, and increased milk production and the possibility to breastfeed longer. | Health services, Family environment, Community environment & Economy |
| 5. Support and treatment to address the drivers and consequences of alcoholism in Batwa and Bakiga families | Expected decreased level of stress in mothers, increased household food availability, and increased milk production and the possibility to breastfeed longer. | Family environment, Community environment & Public policy environment |
| 6. Targeted training and rural outreach to support complementary feeding practices, potentially in conjunction with immunisation/antenatal medical services and trained traditional birth attendants | Expected improvement in complementary feeding practices with better child nutrition outcome. | Health services, Family environment, Community environment |
| 7. Strengthening family planning services | Expected more awareness when making the decision of conceiving, taking into account the importance of breastfeeding longer before giving birth to another baby. | Health services |
| 8. Promote sustainable agriculture, including taking into account smallholder practices and local knowledge. | Expected better adaptation to extreme climatic events to improve food production and availability, which could benefit mothers and children. | Livelihood strategies |
The Batwa and Bakiga are aware of the consequences that unavailability of food can have on maternal and child health. A study conducted in 2019 showed that the diet of Batwa and Bakiga communities is low in proteins and fats, and the caloric content of the most commonly consumed foods is low or very low, with negative implications for child nutrition [28]. For these reasons, women highlighted the need for the District and Government to improve access to, and security and ownership of, more land to cultivate, and distribution of food to pregnant and lactating mothers and children.
The issue of lack of land, land ownership and access, and loss of Indigenous food is linked to the eviction in 1991 from the ancestral forest; this phenomenon is also common in many other Indigenous communities globally [24,51,52]. Assessing barriers to land access, insecure land tenure and being supported by the Government are priorities to increase food production and improve children’ nutritional status.
Lack of representation and socio-economic marginalisation of Indigenous communities persists in the region; the Batwa lack representation at political levels, have limited voice in political decisions, perceive disenfranchisement from local leaders, and need to compete with non-Batwa to be elected to leadership positions [29]. Batwa and Bakiga women are especially marginalised and are victims of domestic violence. Marginalisation is linked to poverty [53–55]; the Indigenous Batwa communities are among the poorest communities in the world [18]. Studies have shown how wealth index and education level influences nutritional status [56,57]; indeed, both Batwa and Bakiga populations have extremely low education levels and income, and suffer from undernutrition and stunting, especially in children under 5 years [20].
Similarly, studies have demonstrated that increasing female empowerement can be beneficial to improve nutritional and caloric intake, especially dietary diversity, in infants and children [58,59]. Batwa and Bakiga participants reported that domestic violence is common among the communities, suggesting that empowerement levels remain low, and support from—and for—the community is required. Alcoholism, linked to domestic violence, is a factor affecting breastfeeding at individual, group and societal level within our framework, and has been reported in other low income settings [53,60]. Alcohol is closely linked to poverty and food insecurity, with alcohol use perceived as partially driven by the desire to quell hunger pains. Women noted that being stressed by a non-supportive and alcohol-consuming partner coupled with food insecurity were some of the main factors associated with inadequate production of milk. For this reason, social and economic support from partners, and engagement from the district and Government to address the consequences of alcoholism in the community were identified as key needs. Indeed, research has shown that support from partners and family had, in fact, a positive effect on breastfeeding as mothers reported to be more confident and motivated [61]. Also, laws and regulations to support victims of abuse are needed at local level; the aim is to ensure an environment where women can safely nourish their children.
Additionally, women reported the importance of professional support at the hospital similarly to other research conducted in Kenya [62]. From the health care providers, mothers expected to get more information on feeding practices, especially complementary foods, and support to access family planning services to breastfeed longer. Previous research has shown that getting information and support at the health facility is not easy during antenatal visits due to the number of patients and amount of work of nurses and midwives [62]. Benefits from creation of breastfeeding and complementary feeding support groups led by community workers, volunteers or outreach nurses and midwives have been demonstrated in literature, and may be appropriate in this region and context [63].
However, participants also highlighted the important role of TBAs to provide essential newborn care in the communities, including advice on breastfeeding and complementary foods. The role of TBAs has long been discussed, although their potential contribution in helping mothers during the delivery and afterwards has been recognized in research [64]. In Uganda, there was a transition from the promotion of skilled birth attendants to the ban of their involvement during deliveries in according to the recommendation of WHO and the Safe Motherhood initiative [65]. However, TBAs continue to look after around $50\%$ of pregnant mothers, especially in remote areas [66]. Studies in some low-income countries have suggested that trained TBAs can help remote communities in accessing health services, and giving support to families by promoting effective neonatal care [67,68]. Therefore, there is the need for health workers to collaborate with TBAs in order to encourage culturally accepted care for the rural and Indigenous communities in the hospitals and health centres in south-western Uganda.
Our study also found that environmental variability and change can play a role in food insecurity by affecting breastfeeding and complementary feeding practices. For example, according to the participants, seasonality was taken into account when introducing the first solid foods; in fact, complementary feeding was found to be delayed or anticipated depending on the season, and availability of food in the household. The time of introduction of solid foods impacts child nutrition and nutritional status, especially if the food insecurity persists for long periods [69]. Studies conducted in the same area show linkage between environmental factors and maternal and neonatal health. For example, Bryson et al. [ 25] described the link between pregnancy outcomes and climate change in Kanungu District among Indigenous and non-Indigenous populations, highlighting the impact of environmental factors on food security and maternal diet. Also, MacVicar et al. [ 22] found a causal pathway between weather, seasonal variability and birth weight among the same communities. These findings suggest that changing climate may have an impact on breastfeeding and complementary feeding practices through pathways of environmental variability and food security in the region.
Research has demonstrated that promoting sustainable agriculture is key to maintain and/or increase food security in a changing climate [70]. Sustainable agriculture has been identified as an important climate response with the potential for double benefits: potential reduction in emissions and adaptation to the implications of climate change for food security [71]. Common principles of sustainable farming include intercropping, high plant diversity and seed recycling [72]. Efficient adaptation strategies to climate change stressors on agricultural productivity will be important in underpinning nutritional outcomes for Batwa and Bakiga mothers and children during the breastfeeding and complementary feeding period. However, transitioning to climate-smart or sustainable agriculture in many cases requires a strong knowledge base, financial resources, and adequate policies; for populations in extreme poverty such as the Batwa–where baseline development remains severely inadequate − transitions to sustainable pathways are unlikely to be feasible without substantial and targetted support [73].
## Strengths and limitations
This study followed Hector et al. ‘s breastfeeding model [13], which is used to plan health interventions, and was adapted to investigate not only breastfeeding, as in the original framework, but also complementary feeding practices. The use of a mixed methods design and a community-based approach enriched the quantitative data with qualitative narratives, describing in detail key factors affecting child nutrition; also, group discussions offered the opportunity to share experiences and compare situations of lactating mothers living in the community. In spite of a relatively low number of FGDs ($$n = 12$$) conducted among the communities, we reached data saturation.
Although in the FGDs may be difficult for all participants to raise their voice due to a few dominant personalities in the group and power imbalances [74], the local research team had worked for many years with the communities and knew how to engage with the participants, mitigating this issue by making sure that everyone had the possibility to talk. Despite the expertise of the local team, there is a possibility that the researchers’ positionality may have influenced the results [75]. To counteract this, the researchers ensured that interviews were conducted in a safe and confidential environment. Participants discussed sensitive topics such as domestic violence and alcoholism or mental health related-issues without any specific question on this matter because they were eager to highlight the problem, share their experience, and think of possible solutions.
Another strength of this study is the addition of the environmental level to the framework which allowed us to explore the relationship between weather and extreme climatic events with child feeding practices. Interesting recommendations have been suggested by the participants, and they may be useful for interventions in rural Uganda and other low-income regions where food insecurity and malnutrition are exacerbated by climatic changes [76–79].
The findings of this research, coming from a minority group, contribute to the research field of global public health and nutrition among Indigenous communities with a focus on women and children, and they will aid in tailoring nutrition interventions to specific local community’s needs. Future research exploring breastfeeding and complementary feeding practices in the area may involve fathers, TBAs and health workers to investigate different perspectives, and to create more awareness on child nutrition and feeding practices among men. Indeed, the focus on IYCF, and in particular on the first 1000 days of life, is needed to prevent malnutrition and ensure that the children grow healthy and fully develop their capacities into adulthood [7].
## References
1. 1UNICEF. Malnutrition 2021 [cited 2021 30th March].
2. Black RE, Allen LH, Bhutta ZA, Caulfield LE, de Onis M, Ezzati M. **Maternal and child undernutrition: global and regional exposures and health consequences**. *Lancet (London, England).* (2008.0) **371** 243-60. DOI: 10.1016/S0140-6736(07)61690-0
3. Victora CG, Adair L, Fall C, Hallal PC, Martorell R, Richter L. **Maternal and child undernutrition: consequences for adult health and human capital**. *The Lancet* (2008.0) **371** 340-57. DOI: 10.1016/S0140-6736(07)61692-4
4. Peña M, Bacallao J. **MALNUTRITION AND POVERTY**. *Annual Review of Nutrition* (2002.0) **22** 241-53. DOI: 10.1146/annurev.nutr.22.120701.141104
5. Hotez PJ. **Neglected infections of poverty among the indigenous peoples of the arctic.**. *PLoS neglected tropical diseases* (2010.0) **4** e606. DOI: 10.1371/journal.pntd.0000606
6. 6UNICEF. Global Strategy for Infant and Young Child Feeding. Geneva, Switzerland: 2003.. *Global Strategy for Infant and Young Child Feeding.* (2003.0)
7. 7WHO. Work programme of the United Nations Decade of Action on Nutrition (2016–2025)
2020 [4 September 2020].. *Work programme of the United Nations Decade of Action on Nutrition (2016–2025)* (2020.0)
8. 8WHO. Malnutrition 2018 [cited 2019 4th January]. Available from: https://www.who.int/news-room/fact-sheets/detail/malnutrition.
9. 9UNICEF. Infant and Young Child Feeding. 2011.. *Infant and Young Child Feeding* (2011.0)
10. 10WHO, UNICEF. Global Strategy for Infant and Young Child Feeding. Geneva, Switzerland: 2003.. *Global Strategy for Infant and Young Child Feeding* 2003
11. **The State of the World’s Children 2019. Children, Food and Nutrition: Growing well in a changing world**. (2019.0)
12. Tiedje LB, Schiffman R, Omar M, Wright J, Buzzitta C, McCann A. **An ecological approach to breastfeeding**. *MCN Am J Matern Child Nurs* (2002.0) **27** 154-61. DOI: 10.1097/00005721-200205000-00005
13. Hector D, King L, Webb K, Heywood P. **Factors affecting breastfeeding practices: applying a conceptual framework**. *N S W Public Health Bull* (2005.0) **16** 52-5. DOI: 10.1071/nb05013
14. Ruel MT, Brown KH, Caulfield LE. **Moving forward with complementary feeding: indicators and research priorities. International Food Policy Research Institute (IFPRI) discussion paper 146 (April 2003).**. *Food Nutr Bull* (2003.0) **24** 289-90. DOI: 10.1177/156482650302400309
15. Sellen DW. **Comparison of Infant Feeding Patterns Reported for Nonindustrial Populations with Current Recommendations**. *The Journal of Nutrition* (2001.0) **131** 2707-15. DOI: 10.1093/jn/131.10.2707
16. Kuhnlein HV, Erasmus B, Spigelski D. **burlingame b. Indigenous Peoples’ food systems & well-being**. *Interventions & policies for healthy communities* 2013
17. Kuperberg K, Evers S. **Feeding patterns and weight among First Nations children**. *Canadian journal of dietetic practice and research: a publication of Dietitians of Canada = Revue canadienne de la pratique et de la recherche en dietetique: une publication des Dietetistes du Canada* (2006.0) **67** 79-84. DOI: 10.3148/67.2.2006.79
18. Berrang-Ford L, Dingle K, Ford JD, Lee C, Lwasa S, Namanya DB. **Vulnerability of indigenous health to climate change: a case study of Uganda’s Batwa Pygmies.**. *Soc Sci Med* (2012.0) **75** 1067-77. DOI: 10.1016/j.socscimed.2012.04.016
19. Donnelly B, Berrang-Ford L, Labbé J, Twesigomwe S, Lwasa S, Namanya DB. **Plasmodium falciparum malaria parasitaemia among indigenous Batwa and non-indigenous communities of Kanungu district**. *Uganda. Malaria Journal* (2016.0) **15** 254. DOI: 10.1186/s12936-016-1299-1
20. Sauer J, Berrang-Ford L, Patterson K, Donnelly B, Lwasa S, Namanya D. **An analysis of the nutrition status of neighboring Indigenous and non-Indigenous populations in Kanungu District, southwestern Uganda: Close proximity, distant health realities.**. *Soc Sci Med* (2018.0) **217** 55-64. DOI: 10.1016/j.socscimed.2018.09.027
21. Berrang-Ford L, Harper SL, Eckhardt R. **Vector-borne diseases: Reconciling the debate between climatic and social determinants.**. *Can Commun Dis Rep* (2016.0) **42** 211-2. DOI: 10.14745/ccdr.v42i10a09
22. MacVicar S, Berrang-Ford L, Harper S, Steele V, Lwasa S, Bambaiha DN. **How seasonality and weather affect perinatal health: Comparing the experiences of indigenous and non-indigenous mothers in Kanungu District, Uganda. Social Science &**. *Medicine* (2017.0) **187** 39-48. DOI: 10.1016/j.socscimed.2017.06.021
23. **Uganda Demographic and Health survey 2016.**. (2018.0)
24. Patterson K, Berrang-Ford L, Lwasa S, Namanya DB, Ford J, Twebaze F. **Seasonal variation of food security among the Batwa of Kanungu, Uganda.**. *Public health nutrition* (2017.0) **20** 1-11. DOI: 10.1017/S1368980016002494
25. Bryson JM, Patterson K, Berrang-Ford L, Lwasa S, Namanya DB, Twesigomwe S. **Seasonality, climate change, and food security during pregnancy among indigenous and non-indigenous women in rural Uganda: Implications for maternal-infant health**. *PLOS ONE* (2021.0) **16** e0247198. DOI: 10.1371/journal.pone.0247198
26. Kidd C, Zaninka P, Programme FP. *Securing Indigenous people’s rights in conservation: a review of South-west Uganda* (2008.0)
27. Clark S, Berrang-Ford L, Lwasa S, Namanya DB, Edge VL, Harper S. **The burden and determinants of self-reported acute gastrointestinal illness in an Indigenous Batwa Pygmy population in southwestern Uganda**. *Epidemiol Infect* (2015.0) **143** 2287-98. DOI: 10.1017/S0950268814003124
28. Scarpa G, Berrang-Ford L, Bawajeeh AO, Twesigomwe S, Kakwangire P, Peters R. **Developing an online food composition database for an Indigenous population in south-western Uganda.**. *Public Health Nutrition* (2021.0) **2021** 1-10. DOI: 10.1017/S1368980021001397
29. 29UOBDU UOfBDiU. Loss of land is not the only challenge faced by Uganda’s Batwa women. 2020 [cited 2021 4th July]. Available from: https://www.forestpeoples.org/en/community-institutions-rights-land-natural-resources-gender-issues-economic-social-cultural-rights.
30. Brubacher LJ, Berrang-Ford L, Clark S, Patterson K, Lwasa S, Namanya DB. **We don’t use the same ways to treat the illness:’ A qualitative study of heterogeneity in health-seeking behaviour for acute gastrointestinal illness among the Ugandan Batwa.**. *Global Public Health* (2021.0) 1-16. DOI: 10.1080/17441692.2021.1937273
31. Busch J, Berrang-Ford L, Clark S, Patterson K, Windfeld E, Donnelly B. **Is the effect of precipitation on acute gastrointestinal illness in southwestern Uganda different between Indigenous and non-Indigenous communities?**. *PLOS ONE* (2019.0) **14** e0214116. DOI: 10.1371/journal.pone.0214116
32. Mooney E.. **The Concept of Internal Displacement and the Case for Internally Displaced Persons as a Category of Concern**. *Refugee Survey Quarterly* (2005.0) **24** 9-26. DOI: 10.1093/rsq/hdi049
33. BCH BCH. *BCH 15th Annual Report* (2018.0)
34. Graves K. **Resilience and Adaptation among Alaska Native Men**. *International Journal of Circumpolar Health* (2003.0) 63. DOI: 10.3402/ijch.v63i1.17652
35. 35FAO, USAI. The State of Food Security and Nutrition in the World 2017. Building Resilience for Peace and Food Security 2017 [cited 2019 5th March]. Available from: http://www.fao.org/3/a-I7695e.pdf.
36. Grace K, Davenport F, Funk C, Lerner AM. **Child malnutrition and climate in Sub-Saharan Africa: An analysis of recent trends in Kenya.**. *Applied Geography* (2012.0) **35** 405-13. DOI: 10.1016/j.apgeog.2012.06.017
37. Anyah RO, Qiu W. **Characteristic 20th and 21st century precipitation and temperature patterns and changes over the Greater Horn of Africa**. *International Journal of Climatology* (2012.0) **32** 347-63. DOI: 10.1002/joc.2270
38. Christensen JH, Kanikicharla KK, Aldrian E, An SI, Albuquerque Cavalcanti IF, de Castro M. **Climate phenomena and their relevance for future regional climate change**. *Climate Change 2013 the Physical Science Basis: Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.* 1217-308
39. Scarpa G, Berrang-Ford L, Twesigomwe S, Kakwangire P, Peters R, Zavaleta-Cortijo C. **A Community-Based Approach to Integrating Socio, Cultural and Environmental Contexts in the Development of a Food Database for Indigenous and Rural Populations: The Case of the Batwa and Bakiga in South-Western Uganda.**. *Nutrients* (2021.0) 13. DOI: 10.3390/nu14010013
40. Israel BA, Schulz AJ, Parker EA, Becker AB. **Review of community-based research: assessing partnership approaches to improve public health**. *Annual review of public health* (1998.0) **19** 173-202. DOI: 10.1146/annurev.publhealth.19.1.173
41. Baum F, MacDougall C, Smith D. **Participatory action research**. *Journal of epidemiology and community health* (2006.0) **60** 854-7. DOI: 10.1136/jech.2004.028662
42. Cargo M, Mercer SL. **The value and challenges of participatory research: strengthening its practice**. *Annual review of public health* (2008.0) **29** 325-50. DOI: 10.1146/annurev.publhealth.29.091307.083824
43. Ben-Shlomo Y, Brookes S, Hickman M. (2013.0)
44. 44SMART. Smart Methodology 2019 [cited 2019 19/04]. Available from: https://smartmethodology.org/about-smart/.
45. Bergold J, Thomas S. **Participatory Research Methods: A Methodological Approach in Motion.**. *Forum Qualitative Sozialforschung / Forum: Qualitative Social Research* (2012.0) **13**
46. Cook T.. **Where Participatory Approaches Meet Pragmatism in Funded (Health) Research: The Challenge of Finding Meaningful Spaces.**. *Forum Qualitative Sozialforschung / Forum: Qualitative Social Research* (2012.0) **13**. DOI: 10.17169/fqs-13.1.1783
47. Russo J.. **Survivor-Controlled Research: A New Foundation for Thinking about Psychiatry and Mental Health.**. *Forum Qualitative Sozialforschung / Forum: Qualitative Social Research* (2012.0) **13**
48. Baxter J, Eyles J. **Evaluating Qualitative Research in Social Geography: Establishing ‘Rigour’ in Interview Analysis.**. *Transactions of the Institute of British Geographers* (1997.0) **22** 505-25. DOI: 10.1111/j.0020-2754.1997.00505.x
49. Mokori A, Schonfeldt H, Hendriks SL. **Child factors associated with complementary feeding practices in Uganda.**. *South African Journal of Clinical Nutrition* (2017.0) **30** 7-14. DOI: 10.1080/16070658.2016.1225887
50. Kajjura RB, Veldman FJ, Kassier SM. **Maternal socio-demographic characteristics and associated complementary feeding practices of children aged 6–18 months with moderate acute malnutrition in Arua, Uganda.**. *Journal of Human Nutrition and Dietetics* (2019.0) **32** 303-10. DOI: 10.1111/jhn.12643
51. Ford JD, Smit B, Wandel J. **Vulnerability to climate change in the Arctic: A case study from Arctic Bay, Canada.**. *Global Environmental Change* (2006.0) **16** 145-60. DOI: 10.1016/j.gloenvcha.2005.11.007
52. Stephens C, Porter J, Nettleton C, Willis R. **Disappearing, displaced, and undervalued: a call to action for Indigenous health worldwide**. *The Lancet* (2006.0) **367** 2019-28. DOI: 10.1016/S0140-6736(06)68892-2
53. Oti SO, van de Vijver SJ, Agyemang C, Kyobutungi C. **The magnitude of diabetes and its association with obesity in the slums of Nairobi, Kenya: results from a cross-sectional survey.**. *Tropical medicine & international health: TM & IH.* (2013.0) **18** 1520-30. DOI: 10.1111/tmi.12200
54. Zulu EM, Dodoo FN, Chika-Ezee A. **Sexual risk-taking in the slums of Nairobi, Kenya, 1993–8**. *Population studies* (2002.0) **56** 311-23. DOI: 10.1080/00324720215933
55. Mberu BU, Ciera JM, Elungata P, Ezeh AC. **Patterns and Determinants of Poverty Transitions among Poor Urban Households in Nairobi**. *Kenya. African Development Review* (2014.0) **26** 172-85. DOI: 10.1111/1467-8268.12073
56. Gatica-Domínguez G, Neves PAR, Barros AJD, Victora CG. **Complementary Feeding Practices in 80 Low- and Middle-Income Countries: Prevalence of and Socioeconomic Inequalities in Dietary Diversity, Meal Frequency, and Dietary Adequacy**. *The Journal of Nutrition* (2021.0) **151** 1956-64. DOI: 10.1093/jn/nxab088
57. Imdad A, Yakoob MY, Bhutta ZA. **Impact of maternal education about complementary feeding and provision of complementary foods on child growth in developing countries.**. *BMC public health* (2011.0) **11** S25. DOI: 10.1186/1471-2458-11-S3-S25
58. Na M, Jennings L, Talegawkar S, Ahmed S. **Association between women’s empowerment and infant and child feeding practices in sub-Saharan Africa: An analysis of Demographic and Health Surveys**. *Public health nutrition* (2015.0) **18** 1-11. DOI: 10.1017/S1368980015002621
59. Yaya S, Odusina EK, Uthman OA, Bishwajit G. **What does women’s empowerment have to do with malnutrition in Sub-Saharan Africa? Evidence from demographic and health surveys from 30 countries**. *Global Health Research and Policy* (2020.0) **5** 1. DOI: 10.1186/s41256-019-0129-8
60. Ayah R, Joshi MD, Wanjiru R, Njau EK, Otieno CF, Njeru EK. **A population-based survey of prevalence of diabetes and correlates in an urban slum community in Nairobi, Kenya**. *BMC public health* (2013.0) **13** 371. DOI: 10.1186/1471-2458-13-371
61. Mannion CA, Hobbs AJ, McDonald SW, Tough SC. **Maternal perceptions of partner support during breastfeeding**. *International breastfeeding journal* (2013.0) **8** 4. DOI: 10.1186/1746-4358-8-4
62. Kimani-Murage EW, Wekesah F, Wanjohi M, Kyobutungi C, Ezeh AC, Musoke RN. **Factors affecting actualisation of the WHO breastfeeding recommendations in urban poor settings in Kenya.**. *Maternal* (2015.0) **11** 314-32. DOI: 10.1111/mcn.12161
63. Rayfield S, Oakley L, Quigley MA. **Association between breastfeeding support and breastfeeding rates in the UK: a comparison of late preterm and term infants**. *BMJ Open* (2015.0) **5** e009144. DOI: 10.1136/bmjopen-2015-009144
64. Campbell OM, Graham WJ. **Strategies for reducing maternal mortality: getting on with what works**. *Lancet (London, England).* (2006.0) **368** 1284-99. DOI: 10.1016/S0140-6736(06)69381-1
65. Turinawe EB, Rwemisisi JT, Musinguzi LK, de Groot M, Muhangi D, de Vries DH. **Traditional birth attendants (TBAs) as potential agents in promoting male involvement in maternity preparedness: insights from a rural community in Uganda.**. *Reproductive Health* (2016.0) **13** 24. DOI: 10.1186/s12978-016-0147-7
66. 66Statistics. UBo. Uganda demographic and health survey 2011–2012. Kampala: Government of Uganda, 2013.
67. Falle T, Mullany L, Thatte N, Khatry S, Leclerq S, Darmstadt G. **Potential Role of Traditional Birth Attendants in Neonatal Healthcare in Rural Southern Nepal**. *Journal of health, population, and nutrition* (2009.0) **27** 53-61. DOI: 10.3329/jhpn.v27i1.3317
68. Adatara P, Afaya A, Baku EA, Salia SM, Asempah A. **Perspective of Traditional Birth Attendants on Their Experiences and Roles in Maternal Health Care in Rural Areas of Northern Ghana**. *Int J Reprod Med* (2018.0) **2018** 2165627. DOI: 10.1155/2018/2165627
69. 69FAO, IFAD, UNICEF, WFP, WHO The state of food security and nutrition in the world. Transforming food systems for affordable healthy diets. FAO, Rome.: 2020.. (2020.0)
70. Altieri M, Toledo V. **Natural Resource Management among Small-scale Farmers in Semi-arid Lands: Building on Traditional Knowledge and Agroecology**. *Annals of Arid Zone* (2005.0) 44
71. Abubakar MS, Attanda ML. **The Concept of Sustainable Agriculture: Challenges and Prospects.**. *IOP Conference Series: Materials Science and Engineering* (2013.0) **53** 012001. DOI: 10.1088/1757-899x/53/1/012001
72. Altieri MA, Nicholls CI. **The adaptation and mitigation potential of traditional agriculture in a changing climate.**. *Climatic Change* (2017.0) **140** 33-45. DOI: 10.1007/s10584-013-0909-y
73. Antle JM, Ray S, Antle JM, Ray S. *Sustainable Agricultural Development: An Economic Perspective.* (2020.0) 95-138
74. Using Smithson J.. **analysing focus groups: Limitations and possibilities**. *International Journal of Social Research Methodology* (2000.0) **3** 103-19. DOI: 10.1080/136455700405172
75. Moore J.. **A personal insight into researcher positionality**. *Nurse researcher* (2012.0) **19** 11-4. DOI: 10.7748/nr2012.07.19.4.11.c9218
76. Phalkey RK, Aranda-Jan C, Marx S, Hofle B, Sauerborn R. **Systematic review of current efforts to quantify the impacts of climate change on undernutrition**. *Proc Natl Acad Sci U S A* (2015.0) **112** E4522-9. DOI: 10.1073/pnas.1409769112
77. Smith P, Bustamante M, Ahammad H, Clark H, Dong H, Elsiddig EA. (2014.0)
78. Lobell DB, Burke MB, Tebaldi C, Mastrandrea MD, Falcon WP, Naylor RL. **Prioritizing climate change adaptation needs for food security in 2030**. *Science* (2008.0) **319** 607-10. DOI: 10.1126/science.1152339
79. Challinor A, Wheeler T, Garforth C, Craufurd P, Kassam A. **Assessing the vulnerability of food crop systems in Africa to climate change**. *Climatic change* (2007.0) **83** 381-99
|
---
title: 'Factors associated with the availability and affordability of essential cardiovascular
disease medicines in low- and middle-income countries: A systematic review'
authors:
- Ali Lotfizadeh
- Benjamin Palafox
- Armin Takallou
- Dina Balabanova
- Martin McKee
- Adrianna Murphy
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021589
doi: 10.1371/journal.pgph.0000072
license: CC BY 4.0
---
# Factors associated with the availability and affordability of essential cardiovascular disease medicines in low- and middle-income countries: A systematic review
## Abstract
Despite their potential to prevent or delay the onset and progression of cardiovascular disease (CVD), medicines for CVD remain unavailable and unaffordable to many in low- and middle-income countries (LMICs). We systematically reviewed the literature to identify factors associated with availability and affordability of CVD medicines in LMICs. A protocol for this study was registered on the PROSPERO register of systematic reviews (CRD42019135393). We searched Medline, EMBASE, Global Health, Cumulative Index to Nursing and Allied Health Literature, EconLit, Social Policy and Practice, and Africa Wide Information for studies analyzing factors associated with the presence of medicines (availability) or the price of these medicines as it relates to ability to pay (affordability) in LMICs. We performed a narrative synthesis of the results using an access to medicines framework that examines influences at different levels of the health system. We did not conduct a meta-analysis because of the differences in analytic approaches and outcome measures in different studies. The search was conducted in accordance with PRISMA guidelines. Of 43 studies meeting inclusion criteria, 41 were cross-sectional. Availability and affordability were defined and measured in different ways. A range of factors such as sociodemographic characteristics, facility tier, presence of medicines on national essential medicine lists, and international subsidy programs were examined. The studies had variable quality and findings were often inconsistent. We find gaps in the literature on factors associated with availability and affordability of CVD medicines, particularly at the health program level. We conclude that there is a need for experimental and quasi-experimental studies that could identify causal factors and effective responses. Such studies would help further our understanding of how complex multifactorial influences impact these outcomes, which could inform policy decisions. Along with this, greater standardization of definitions and measurement approaches of availability and affordability are needed to allow for more effective comparisons.
## Introduction
Control of blood pressure and lipid levels by pharmacotherapy is a core element of primary and secondary prevention of cardiovascular disease (CVD). Antihypertensive drugs are known to reduce the incidence of cardiac and cerebrovascular diseases [1], as are statins [2] and antiplatelet agents [3]. Despite the known efficacy and cost-effectiveness of these medicines [4], their use remains far from optimal, especially in low- and middle-income countries (LMICs), where $80\%$ percent of CVD-related mortality occurs [5]. The World Health Organization (WHO) has advocated for a target goal of $50\%$ of eligible people receiving medicines for CVD [6]. Yet, in LMICs, fewer than $30\%$ with hypertension receive treatment and less than $8\%$ achieve blood pressure control [7]. In one study of only low-income countries, fewer than $10\%$ of those with a history of coronary heart disease or ischemic stroke received treatment with antiplatelet drugs or lipid-lowering agents, contrary to international guidelines [8].
One reason for this treatment gap is poor access to medicines. Access has multiple dimensions but two that are commonly explored are availability and affordability. A recommended definition of availability compares the quantity of a medicine required in relation to its presence at health facilities. Affordability is captured by a medicine’s price relative to an individual or household’s ability to pay [9]. Research in LMICs shows that the availability and affordability of CVD drugs are associated with greater odds of patients using them and with a lower risk of adverse cardiovascular outcomes [10].
Although the WHO recommends at least $80\%$ medicine availability for CVD [6], this target is not being met in LMICs. Many patients with hypertension require combination therapy but only $13\%$ of communities in low-income countries live in areas where all four main classes of antihypertensives are available and only $30\%$ of households can afford them [11]. While many studies describe the scale of the problem, fewer seek to explain the multitude of factors involved and their interrelationships. Our objective is to systematically review the literature to identify such factors in LMIC to inform appropriate policy responses and future research.
## Materials and methods
We conducted a systematic review of the literature on factors associated with availability and affordability of medicines for CVD in LMICs. The protocol for this review was registered and published on the PROSPERO register of systematic reviews (CRD42019135393). Findings are reported according to the PRISMA guidelines.
## Inclusion and exclusion criteria
We included published studies in any language that reported original data and analyzed interventions or factors associated with availability and/or affordability of CVD medicines in countries defined as low- or middle-income according to the 2021 World Bank classification. We included quantitative, qualitative, or mixed-method comparative studies, using experimental, quasi-experimental, or observational designs. We included any measure of availability providing information on the physical presence of medicines at health facilities or home. For affordability, we included studies that considered the price of a medicine incurred by an individual or household in relation to their ability to pay [12]. Studies that only provided information on the price of medicines were excluded. Studies that examined an individual or household’s ability to obtain medicines for free (as opposed to having to pay for medicines) were included, because free medicines are, by definition, more affordable than medicines at any price [13]. We included studies that directly measured availability and/or affordability, those that used information from other surveys, and those that asked respondents to report on availability and/or affordability.
Studies with data from more than one country were included if the majority were LMICs. We focused on CVD medicines in the following three categories: antihypertensive agents, platelet aggregation inhibitors, and lipid-lowering agents. Studies that reported data on a basket of medicines were included if CVD medicines were a part of that basket.
## Search strategy
We searched MEDLINE, EMBASE, Global Health, Cumulative Index to Nursing and Allied Health Literature, EconLit, Social Policy and Practice, and Africa Wide Information for studies in any language published after the year 2000. We also conducted a grey literature search using the website of the Institute of Development Studies, the WHO Repository, the World Bank Repository, and Google. We restricted our search to publications from the year 2000 onward because of a paucity of health systems literature prior to this date. We performed our initial search in June 2020 and updated the search in July 2021. The references of all included records were manually reviewed. The search strategy was developed in collaboration with a librarian at the London School of Hygiene and Tropical Medicine with expertise in systematic review methodology. The search terms were subsequently peer-reviewed by another information specialist at the London School of Hygiene and Tropical Medicine not involved in developing the search strategy. More details on the search strategy are presented in Table 1 and the complete search terms can be found in Appendix 1 (S1 Table). Non-communicable disease (NCD) medicines were included in our search because an initial scoping review revealed papers that examined availability and/or affordability of a basket of NCD medicines, including CVD drugs.
**Table 1**
| Main Concept | Components of Main Concept | Sample Terms from Medline Search |
| --- | --- | --- |
| | [Terms for CVD Medicines] | [cardiovascular agents/ OR antihypertensive agents/ OR hypolipidemic agents/] |
| Medicines for NCDs and CVD | OR[(Terms for CVD) AND (Terms for Medicines)] | [(cardiovascular diseases/ OR exp hypertension/ OR exp hyperlipidemias/ OR cardiovascular disease*.ti,ab.)AND(exp pharmaceutical preparations/ OR medication*.ti,ab.)] |
| AND | OR[(Terms for NCD) AND (Terms for Medicines)] | [(exp chronic disease/ OR chronic disease*.ti,ab. OR chronic condition*.ti,ab. OR NCD.ti,ab.)AND(exp pharmaceutical preparations/ OR medication*.ti,ab.)] |
| Availability/Affordability AND | [Terms for Availability OR Affordability] | [availab*.ti,ab. OR supply.ti,ab. OR drug costs/ OR exp fees, pharmaceutical/ OR affordab*.ti,ab.] |
| LMIC | [Terms for LMIC OR List of LMIC] | [developing countries/ OR exp africa south of the sahara/ OR Armenia/ OR Armenia.ti,ab.] |
The results were screened independently by two reviewers at the title/abstract level and studies not meeting inclusion criteria were excluded. Both reviewers subsequently screened full texts of retained articles independently, excluding those that did not meet inclusion criteria. Discrepancies were discussed with a third reviewer and a consensus was reached.
## Data extraction
The following information was extracted from all studies: language, study implementation year, country, study design, study setting, sample size, survey method used, factors analyzed, methodology, medicines studied, outcome measure, definition of availability and/or affordability, how availability and/or affordability was quantified, and study findings including the statistical parameters used (e.g. odds ratios, proportions, chi-squared values, p-values, and confidence intervals). Studies that were not in English were translated using Google Translate. Where translations were not clear, we used a dictionary and consulted colleagues with knowledge of the language.
## Risk of bias assessment
We assessed quality for observational studies using a method previously published by Maimaris [14], which examines three domains: selection bias, information bias (differential and non-differential misclassification), and confounding, thereby providing more precise information on studies than is done by some other instruments whose primary purpose is to determine whether to include a study or not. To evaluate non-differential misclassification, we assessed the reliability of the measures used for availability and affordability. The risk of bias tool for observational studies is presented in Appendix 2 (S2 Table). For randomized controlled trials, the revised Cochrane tool was used [15]. This tool measures bias in the following domains: randomization, timing of identification/recruitment in relation to timing of randomization, deviations in intended interventions, missing outcome data, outcome measurements, and selection of the reported results. Two reviewers independently performed risk of bias assessment and resolved discrepancies with discussion.
## Conceptual framework and narrative synthesis
We organized our findings according to a framework of health systems constraints adapted to access to medicines in LMICs [16]. This framework proposes different levels of the health system as follows: 1) individuals, households, and communities, 2) health service delivery, 3) health sector (or program), 4) national context (public policies cutting across sectors), 5) international context. We considered sociodemographic factors and geographic location at level 1. At level 2, we included characteristics of a health facility or service delivery arrangements at that facility. At the health sector level, we examined country or region-wide health programs or policies. At the national context, policies extending beyond the health sector were examined. At level 5, international programs or arrangements that were related to the availability or affordability of medicines were examined. We coded factors associated with availability and affordability into the different levels and performed a narrative synthesis of the data.
## Results
The results of the screening process are reported in the PRISMA flowchart (Fig 1). The initial search yielded 32,352 results and an additional 4,077 were retrieved from the updated search, resulting in 36,429 titles. Of these, 9,839 were duplicates, leaving 26,590 citations. Title and abstract screening resulted in 334 studies, of which 327 were retrieved for full-text screening. The remaining seven studies could not be accessed because they were in journals to which our library and no affiliated libraries subscribe. Thirty-five studies met inclusion criteria and an additional eight were included through a manual search of the references.
**Fig 1:** *PRISMA flowchart.Flowchart of the selection and screening process of studies on availability and affordability of CVD medicines.*
## Study characteristics
Of the included studies, 40 were in English [17–56], two were in Spanish [57, 58], and one was in Portuguese [59]. Thirty-four studies focused on availability [17, 18, 20–24, 26–36, 38, 40–42, 44, 45, 47, 48, 50–54, 56, 58, 59], seven on affordability [19, 25, 37, 39, 43, 46, 55], and two on both [49, 57]. Risk of bias assessment revealed heterogeneity in the quality of the included studies. Fifteen of the cross-sectional studies had one or more domains with high risk of bias [23, 24, 27, 32, 35, 37, 43, 49–55, 57] while another seven had unclear risk of bias in at least one domain [20, 22, 30, 44, 53, 54, 59]. Detailed characteristics of included studies and the results of risk of bias assessments are described in Table 2. Seven of the included studies used the WHO Health Action International survey [20, 22, 23, 26, 28, 30, 31, 53, 56] and six used the Service Availability and Readiness Assessment tool to measure medicine availability [17, 18, 24, 27, 29, 38, 52]. Definitions and quantification approaches used to measure availability and affordability are presented in Tables 3 and 4, respectively.
## Individual, household, and community level
Sociodemographic characteristics associated with medicine availability were assessed in four studies, with inconsistent findings [41, 45, 47, 49]. Age was examined in two and was only associated with greater availability in three of five countries in one of the studies [45, 47]. Associations between socioeconomic status and availability varied in the same five-country study, but two other studies found positive associations [41, 45, 47]. Education was also positively associated with availability in one of two studies but not in the other [45, 47]. There was no association between sex and availability of medicines in the three studies looking at this variable [45, 47, 49]. At the community level, seven studies compared medicine availability in urban and rural settings [21, 23, 34, 36, 47, 52, 58], with three reporting no differences [36, 47, 58]. In two studies from Zambia and Kenya that looked at different CVD medicine classes separately, only calcium channel blockers and hydrochlorothiazide had significantly lower availability in rural facilities, respectively [23, 34].
Five studies examined sociodemographic factors associated with affordability with four examining socioeconomic status [39, 43, 46, 49, 55]. Two from Brazil revealed a negative association between socioeconomic status and affordability [43, 46], while a third study showed no association [55]. Conversely, a study from Russia revealed a positive association [39].
## Service delivery level
Twenty one studies examined service delivery level arrangements [17, 18, 21–23, 27–29, 31–33, 35, 36, 38, 42, 47, 51, 53, 56, 58, 59]. Thirteen of these compared availability at private and public facilities with mixed results [18, 21–23, 28, 31, 32, 35, 36, 47, 51, 53, 56]. Additionally, 11 studies looked at the relationship between facility tier and medicine availability [17, 18, 21, 27, 29, 33, 36, 38, 42, 58, 59]. In seven studies from Tanzania, Uganda, Nigeria, China, and Brazil, higher tier facilities staffed by medical doctors or providing a wider range of services were significantly more likely to have antihypertensive medicines or a basket of medicines available [18, 21, 29, 33, 36, 42, 59]. Among these, three also looked at correlations between availability and the level of amenities at a facility such as equipment, a pharmacy refrigerator, or solar panels. The presence of such amenities was associated with greater availability of medicines in all three studies [18, 36, 59]. The relationship between service integration and medicine availability was examined in two studies. In Uganda, facilities providing HIV counseling and testing had greater availability of NCD medicines while those offering HIV care had lower availability [18]. In Brazil, primary healthcare units dispensing psychotropic drugs had 3.16 ($95\%$ CI: 2.85–3.51) times greater odds of having essential medicines available [59].
Three studies examined correlations between affordability and service delivery level indicators [19, 55, 57]. In Kenya, public facilities were more likely to provide free NCD medicines than private facilities ($47\%$ v $9\%$, $p \leq 0.0001$) [19]. In Mexico, medicines were obtained for free by $62\%$ of those visiting pharmacies that were operated by the state and by $70\%$ of those visiting outsourced pharmacies ($p \leq 0.01$) [57]. In Brazil, medicines were more affordable for individuals who reported primarily using public health facilities as part of the Sistemo Unico de Saude to procure medicines than those who used private pharmacies. The Sistemo Unico de *Saude is* a national program that aims to increase access to medicines by providing certain medications free of charge [55].
## Health sector (program) level
Eight studies examined health sector level arrangements [20, 26, 28, 35, 42, 44, 54, 56]. Three studies from Brazil, China, and Nigeria all point to an association between facility revenues and medicine availability [26, 35, 44]. For example, in China, the introduction of a policy that eliminated facility mark-ups on drug prices (and thus decreased revenue) was associated with a decrease in mean availability of lowest-priced generics from $25.5\%$ to $20.5\%$ ($p \leq 0.0001$) [26]. In a fourth study, from Indonesia, implementation of a national policy that included a shift from fee-for-service to more limited reimbursements based on the type of diagnosis was associated with variable results for different types of antihypertensive agents. This policy also included a requirement that medicines be ordered from the national formulary, however, the study provided little detail on the methodology employed [54].
The relationship between the presence of drugs on essential medicine lists and availability was examined in four studies [20, 26, 28, 56]. In China, the introduction of a provincial essential medicine list was associated with decreases in medicine availability [26]. In Bangladesh, availability of medicines was $50.3\%$ for those on the national essential medicine list and $57\%$ for those not on the list ($p \leq 0.05$) [28]. But in a survey of 23 LMICs, median availability of generic medicines on national lists was significantly greater [20]. In Pakistan as well, health facility availability of CVD medicines on the essential medicine list was significantly greater than those not on the list [56].
Four studies examined the association between affordability and the presence of health insurance [39, 43, 46, 55]. In Brazil, not having a health insurance plan was associated with greater medicine affordability [43, 46, 55]. An interrupted time series analysis from Brazil examining a policy removing copayments for antihypertensives found significant reductions in mean out-of-pocket expenditures after the policy was implemented [25].
## National and international levels
We found no studies assessing the relationship between national policies from outside the health sector and availability. In terms of international factors, one cluster-randomized controlled trial that evaluated a Novartis-sponsored program providing medicines for NCDs in Kenya at $1 per treatment per month found increased availability of amlodipine–but not other CVD medicines–at health facilities [40].
We did not identify any studies that examined factors at the national level (non-health sector) or international level associated with medicine affordability.
## Discussion
To the best of our knowledge, this is the first systematic review of health system factors associated with availability and affordability of CVD medicines in LMICs. Our findings highlight significant gaps in evidence and heterogeneity in results as well as variable quality of existing studies. While certain factors seemed consistently associated with availability, findings regarding other factors were inconsistent.
There are a number of possible reasons for these inconsistencies. First, most studies were cross-sectional and subject to potential confounding, precluding establishment of causal relationships. Intervention studies are needed but the feasibility of implementing them will depend on context, so they should be accompanied by qualitative policy analyses.
Second, availability and affordability of medicines are unlikely to be affected by a single factor. Instead, a multitude of factors must be aligned in complex relationships at different health systems levels [16]. This was evident in several studies, including the sole cluster-randomized trial included, where subsidies from Novartis had little effect on medicine availability in Kenya [40]. This differs from a similar intervention involving subsidies for artemisinin-based combination therapy in sub-Saharan Africa, which achieved notable increases in medicine availability [60]. As the authors of the first study acknowledge, subsidization is only one element in a complex system which also has to account for factors such as physician, pharmacist, and patient awareness of specific medicines, procurement, delivery mechanisms, and inclusion in clinical practice guidelines. This may also explain why the presence of medicines on national essential medicine lists did not translate to greater availability in all included studies. Different countries may go about deciding what medicines to include on these lists differently [20]. It may be that in some contexts, drugs on essential medicine lists are not included in clinical guidelines or are less accepted by prescribers and patients. Other factors may also influence whether medicines on essential lists are more available. One included study demonstrated an inverse relationship between country-income level and availability of medicines on these lists [20]. The authors posit that in countries with fewer resources, essential medicine lists drive prioritization. Interestingly, in two studies we identified from Pakistan [56] and Bangladesh [28], availability of medicines positively correlated with their presence on essential lists only in Pakistan, which has the lower national income level of the two [61]. Similarly, medicines on essential lists tend to be more available at public facilities [20]. Yet, the studies we included analyzed a heterogeneous sample of public and private facilities as well as different tier facilities together. This may explain some of the inconsistencies observed. In a study from China [26], where implementation of an essential medicine list was associated with lower availability, the policy was accompanied by elimination of mark-ups on drug prices at health facilities, which may have reduced revenues for medicine procurement.
We also noted inconsistencies and counterintuitive findings for affordability. In Brazil, the absence of health insurance was associated with greater affordability of medicines [43, 46, 55]. Individuals from lower socioeconomic backgrounds who lack health insurance rely more heavily on Brazilian public health programs like the Sistema Unico de Saude, which provide medicines for free particularly for NCDs. Those with private insurance, on the other hand, are more likely to procure medicines through the private sector where they will be required to pay [62]. Future research must take a more holistic and dynamic approach, looking at how individual correlates of inadequate access are influenced by the characteristics of the health systems in which they operate.
A third possible reason for inconsistent findings is the heterogeneity in measures of availability and affordability. For example, some studies on availability assess the percentage of facilities where a drug is available while others report the proportion of drugs available from a list of medicines. Employing even slightly different measures can yield widely disparate estimates in the same country [63]. This problem is compounded by the fact that most current measures are binary. Many of the included studies considered a medicine available if it was present on the day of the survey regardless of quantity, expiry date, or reliability of supply.
A consistent and comprehensive definition of affordability is also needed. In all but one study, affordability was either ascertained subjectively from respondents or by measuring the ability to obtain medicines for free, which does not fully encapsulate the degree to which medicines are unaffordable for those who have to pay. Many measures of affordability, including those employed by the studies included here, rely on a single measure at one point in time. These measures may not provide a full picture of the long-term financial burden that purchasing medications for chronic conditions places on families [64]. Other metrics of affordability have been described in the literature, however, no study meeting our inclusion criteria used such definitions. For example, the WHO Health Action International Survey measures affordability as the number of days of wages the lowest-paid government worker has to spend on medicines [65]. Many of these approaches involve making arbitrary judgments on thresholds for affordability. Doing so without taking into account households’ disposable income risks overestimating true affordability [13]. The downside of more comprehensive definitions is challenges in measurement. Yet, without them, interventions may be deemed effective if they meet certain predetermined benchmarks, even if they do not truly reflect how available or affordable medicines are for those who need them.
An important finding from our study is the paucity of factors studied that could influence availability and affordability. For example, no study examined the role of corruption, although this has been shown to be negatively associated with access to HIV antiretroviral therapy [66]. Nor were supply chain factors investigated, despite several studies showing how relevant interventions affected availability of medicines for other diseases [67, 68]. Similarly, intellectual property provisions may have implications on availability and affordability of CVD medicines. Different studies have shown that the vast majority of medicines on the WHO Essential Medicine List are not under patent protection [69, 70]. These findings, coupled with extensive evidence that medicines on the essential medicine list remain unavailable and unaffordable in most LMICs, imply that patent provisions are unlikely to be the sole contributor to non-availability and affordability. Nevertheless, patents on newer therapeutics or fixed-dose combination therapies for CVD that are not on essential medicine lists may impact their availability and affordability and warrant further investigation [71].
Our study has some limitations. We restricted our search to factors associated with availability and affordability but did not examine other parameters of access. Though beyond the scope of this review, aspects such as acceptability are also important in ensuring that those who need CVD medicines can receive them [9]. Furthermore, while we did not restrict our search based on language, studies in languages other than English may not be captured in the databases used. Finally, availability and affordability are conceptualized in a multitude of ways. Some studies on this topic may have used terms that were missed by our search terms.
## Conclusion
Evidence concerning factors that influence the availability or affordability of CVD medicines is limited. The majority of studies are observational and while they have identified a number of potential associations, they cannot establish causality. Factors at different levels of the health system likely act together in a multifactorial way to influence availability and affordability of CVD medicines as part of complex health systems. Future research involving uniform definitions and measurement approaches is needed with a particular focus on experimental and quasi-experimental methods that provide insight into causal mechanisms.
## References
1. Messerli FH, Williams B, Ritz E. **Essential hypertension**. (2007.0) **370** 591-603. DOI: 10.1016/S0140-6736(07)61299-9
2. Shepherd J, Cobbe SM, Ford I, Isles CG, Lorimer AR, MacFarlane PW. **Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia. West of Scotland Coronary Prevention Study Group**. (1995.0) **333** 1301-7. DOI: 10.1056/NEJM199511163332001
3. **Aspirin in the primary and secondary prevention of vascular disease: collaborative meta-analysis of individual participant data from randomised trials**. (2009.0) **373** 1849-60. DOI: 10.1016/S0140-6736(09)60503-1
4. Aminde LN, Takah NF, Zapata-Diomedi B, Veerman JL. **Primary and secondary prevention interventions for cardiovascular disease in low-income and middle-income countries: a systematic review of economic evaluations.**. (2018.0) **16** 22. DOI: 10.1186/s12962-018-0108-9
5. 5Global status report on noncommunicable diseases
2010
Geneva: World Health Organization; 2011 [cited 2020 Sept 14]. Available from: https://www.who.int/nmh/publications/ncd_report2010/en/.. *Global status report on noncommunicable diseases* (2010.0)
6. 6Global Action Plan for the Prevention and Control of Noncommunicable Diseases 2013–2020. Geneva: World Health Organization; 2013 [cited 2020 Sept 14]. Available from: https://apps.who.int/iris/bitstream/handle/10665/94384/9789241506236_eng.pdf?sequence=1.. (2013.0)
7. Mills KT, Bundy JD, Kelly TN, Reed JE, Kearney PM, Reynolds K. **Global Disparities of Hypertension Prevalence and Control: A Systematic Analysis of Population-Based Studies From 90 Countries**. *Circulation* (2016.0) **134** 441-50. DOI: 10.1161/CIRCULATIONAHA.115.018912
8. Yusuf S, Islam S, Chow CK, Rangarajan S, Dagenais G, Diaz R. **Use of secondary prevention drugs for cardiovascular disease in the community in high-income, middle-income, and low-income countries (the PURE Study): a prospective epidemiological survey**. (2011.0) **378** 1231-43. DOI: 10.1016/S0140-6736(11)61215-4
9. Wirtz VJ, Kaplan WA, Kwan GF, Laing RO. **Access to Medications for Cardiovascular Diseases in Low- and Middle-Income Countries**. (2016.0) **133** 2076-85. DOI: 10.1161/CIRCULATIONAHA.115.008722
10. Chow CK, Nguyen TN, Marschner S, Diaz R, Rahman O, Avezum A. **Availability and affordability of medicines and cardiovascular outcomes in 21 high-income, middle-income and low-income countries**. (2020.0) **5**. DOI: 10.1136/bmjgh-2020-002640
11. Attaei MW, Khatib R, McKee M, Lear S, Dagenais G, Igumbor EU. **Availability and affordability of blood pressure-lowering medicines and the effect on blood pressure control in high-income, middle-income, and low-income countries: an analysis of the PURE study data**. (2017.0) **2** e411-e9. DOI: 10.1016/S2468-2667(17)30141-X
12. Embrey M, Embrey M, Ryan M. (2012.0)
13. Niens LM, Brouwer WB. **Measuring the affordability of medicines: importance and challenges**. (2013.0) **112** 45-52. DOI: 10.1016/j.healthpol.2013.05.018
14. Maimaris W, Paty J, Perel P, Legido-Quigley H, Balabanova D, Nieuwlaat R. **The influence of health systems on hypertension awareness, treatment, and control: a systematic literature review**. (2013.0) **10** e1001490. DOI: 10.1371/journal.pmed.1001490
15. Sterne JAC, Savovic J, Page MJ, Elbers RG, Blencowe NS, Boutron I. **RoB 2: a revised tool for assessing risk of bias in randomised trials**. (2019.0) **366** l4898. DOI: 10.1136/bmj.l4898
16. Bigdeli M, Jacobs B, Tomson G, Laing R, Ghaffar A, Dujardin B. **Access to medicines from a health system perspective.**. (2013.0) **28** 692-704. DOI: 10.1093/heapol/czs108
17. Adinan J, Manongi R, Temu GA, Kapologwe N, Marandu A, Wajanga B. **Preparedness of health facilities in managing hypertension & diabetes mellitus in Kilimanjaro, Tanzania: a cross sectional study.**. (2019.0) **19** 537. DOI: 10.1186/s12913-019-4316-6
18. Armstrong-Hough M, Kishore SP, Byakika S, Mutungi G, Nunez-Smith M, Schwartz JI. **Disparities in availability of essential medicines to treat non-communicable diseases in Uganda: A Poisson analysis using the Service Availability and Readiness Assessment**. (2018.0) **13** e0192332. DOI: 10.1371/journal.pone.0192332
19. Ashigbie PG, Rockers PC, Laing RO, Cabral HJ, Onyango MA, Buleti JPL. **Availability and prices of medicines for non-communicable diseases at health facilities and retail drug outlets in Kenya: a cross-sectional survey in eight counties**. (2020.0) **10** e035132. DOI: 10.1136/bmjopen-2019-035132
20. Bazargani YT, Ewen M, de Boer A, Leufkens HG, Mantel-Teeuwisse AK. **Essential medicines are more available than other medicines around the globe.**. (2014.0) **9** e87576. DOI: 10.1371/journal.pone.0087576
21. Bintabara D, Mpondo BCT. **Preparedness of lower-level health facilities and the associated factors for the outpatient primary care of hypertension: Evidence from Tanzanian national survey**. (2018.0) **13** e0192942. DOI: 10.1371/journal.pone.0192942
22. Cameron A, Roubos I, Ewen M, Mantel-Teeuwisse AK, Leufkens HG, Laing RO. **Differences in the availability of medicines for chronic and acute conditions in the public and private sectors of developing countries**. (2011.0) **89** 412-21. DOI: 10.2471/BLT.10.084327
23. Cepuch C.. (2012.0)
24. Duong DB, Minh HV, Ngo LH, Ellner AL. **Readiness, Availability and Utilization of Rural Vietnamese Health Facilities for Community Based Primary Care of Non-communicable Diseases: A CrossSectional Survey of 3 Provinces in Northern Vietnam**. (2019.0) **8** 150-7. DOI: 10.15171/ijhpm.2018.104
25. Emmerick ICM, Campos MR, da Silva RM, Chaves LA, Bertoldi AD, Ross-Degnan D. **Hypertension and diabetes treatment affordability and government expenditures following changes in patient cost sharing in the "Farmacia popular" program in Brazil: an interrupted time series study**. (2020.0) **20** 24. DOI: 10.1186/s12889-019-8095-0
26. Fang Y, Wagner AK, Yang S, Jiang M, Zhang F, Ross-Degnan D. **Access to affordable medicines after health reform: evidence from two cross-sectional surveys in Shaanxi Province, western China**. (2013.0) **1** e227-37. DOI: 10.1016/S2214-109X(13)70072-X
27. Jigjidsuren A, Byambaa T, Altangerel E, Batbaatar S, Saw YM, Kariya T. **Free and universal access to primary healthcare in Mongolia: the service availability and readiness assessment**. (2019.0) **19** 129. DOI: 10.1186/s12913-019-3932-5
28. Kasonde L, Tordrup D, Naheed A, Zeng W, Ahmed S, Babar ZU. **Evaluating medicine prices, availability and affordability in Bangladesh using World Health Organisation and Health Action International methodology**. (2019.0) **19** 383. DOI: 10.1186/s12913-019-4221-z
29. Katende D, Mutungi G, Baisley K, Biraro S, Ikoona E, Peck R. **Readiness of Ugandan health services for the management of outpatients with chronic diseases**. (2015.0) **20** 1385-95. DOI: 10.1111/tmi.12560
30. Khanal S, Veerman L, Ewen M, Nissen L, Hollingworth S. **Availability, Price, and Affordability of Essential Medicines to Manage Noncommunicable Diseases: A National Survey From Nepal**. (2019.0) **56** 1-8
31. Kibirige D, Atuhe D, Kampiire L, Kiggundu DS, Donggo P, Nabbaale J. **Access to medicines and diagnostic tests integral in the management of diabetes mellitus and cardiovascular diseases in Uganda: insights from the ACCODAD study**. (2017.0) **16** 154. DOI: 10.1186/s12939-017-0651-6
32. Minaei H, Peikanpour M, Yousefi N, Peymani P, Peiravian F, Shobeiri N. **Country Pharmaceutical Situation on Access, Quality, and Rational Use of Medicines: An Evidence from a middle-income country**. (2019.0) **18** 2191-203. DOI: 10.22037/ijpr.2019.111636.13273
33. Musinguzi G, Bastiaens H, Wanyenze RK, Mukose A, Van Geertruyden JP, Nuwaha F. **Capacity of Health Facilities to Manage Hypertension in Mukono and Buikwe Districts in Uganda: Challenges and Recommendations**. (2015.0) **10** e0142312. DOI: 10.1371/journal.pone.0142312
34. Mutale W, Bosomprah S, Shankalala P, Mweemba O, Chilengi R, Kapambwe S. **Assessing capacity and readiness to manage NCDs in primary care setting: Gaps and opportunities based on adapted WHO PEN tool in Zambia.**. (2018.0) **13** e0200994. DOI: 10.1371/journal.pone.0200994
35. Oliveira MA, Luiza VL, Tavares NU, Mengue SS, Arrais PS, Farias MR. **Access to medicines for chronic diseases in Brazil: a multidimensional approach**. (2016.0) **50** 6s. DOI: 10.1590/S1518-8787.2016050006161
36. Oyekale AS. **Assessment of primary health care facilities’ service readiness in Nigeria.**. (2017.0) **17** 172. DOI: 10.1186/s12913-017-2112-8
37. Paniz VM, Fassa AG, Facchini LA, Piccini RX, Tomasi E, Thume E. **Free access to hypertension and diabetes medicines among the elderly: a reality yet to be constructed**. (2010.0) **26** 1163-74. DOI: 10.1590/s0102-311x2010000600010
38. Peck R, Mghamba J, Vanobberghen F, Kavishe B, Rugarabamu V, Smeeth L. **Preparedness of Tanzanian health facilities for outpatient primary care of hypertension and diabetes: a cross-sectional survey**. (2014.0) **2** e285-92. DOI: 10.1016/S2214-109X(14)70033-6
39. Perlman F, Balabanova D. **Prescription for change: accessing medication in transitional Russia**. (2011.0) **26** 453-63. DOI: 10.1093/heapol/czq082
40. Rockers PC, Laing RO, Ashigbie PG, Onyango MA, Mukiira CK, Wirtz VJ. **Effect of Novartis Access on availability and price of non-communicable disease medicines in Kenya: a cluster-randomised controlled trial**. (2019.0) **7** e492-e502. DOI: 10.1016/S2214-109X(18)30563-1
41. Rockers PC, Laing RO, Wirtz VJ. **Equity in access to non-communicable disease medicines: a cross-sectional study in Kenya.**. (2018.0) **3** e000828. DOI: 10.1136/bmjgh-2018-000828
42. Su M, Zhang Q, Bai X, Wu C, Li Y, Mossialos E. **Availability, cost, and prescription patterns of antihypertensive medications in primary health care in China: a nationwide cross-sectional survey**. (2017.0) **390** 2559-68. DOI: 10.1016/S0140-6736(17)32476-5
43. Tavares NU, Luiza VL, Oliveira MA, Costa KS, Mengue SS, Arrais PS. **Free access to medicines for the treatment of chronic diseases in Brazil.**. (2016.0) **50** 1-10. DOI: 10.1590/S1518-8787.2016050006118
44. Uzochukwu BS, Onwujekwe OE, Akpala CO. **Effect of the Bamako-Initiative drug revolving fund on availability and rational use of essential drugs in primary health care facilities in south-east Nigeria.**. (2002.0) **17** 378-83. DOI: 10.1093/heapol/17.4.378
45. Vialle-Valentin CE, Serumaga B, Wagner AK, Ross-Degnan D. **Evidence on access to medicines for chronic diseases from household surveys in five low- and middle-income countries**. (2015.0) **30** 1044-52. DOI: 10.1093/heapol/czu107
46. Viana KP, Brito Ados S, Rodrigues CS, Luiz RR. **Access to continued-use medication among older adults, Brazil.**. (2015.0) **49** 14. DOI: 10.1590/s0034-8910.2015049005352
47. Wirtz VJ, Turpin K, Laing RO, Mukiira CK, Rockers PC. **Access to medicines for asthma, diabetes and hypertension in eight counties of Kenya**. (2018.0) **23** 879-85. DOI: 10.1111/tmi.13081
48. Yang L, Liu C, Ferrier JA, Zhang X. **Organizational barriers associated with the implementation of national essential medicines policy: A cross-sectional study of township hospitals in China.**. (2015.0) **145** 201-8. DOI: 10.1016/j.socscimed.2015.08.044
49. Fernandopulle BMR, Gunawardena N, de Silva SHP, Abayawardana C, Hirimuthugoda LK. **Patient experiences of access to NCD medicines in Sri Lanka: Evidence of the success story towards universal coverage**. (2019.0) **3** 1-10
50. Albelbeisi AH, Albelbeisi A, El-Bilbeisi AH, Takian A, Akbari-Sari A. **Capacity of palestinian primary health care system to prevent and control of non-communicable diseases in Gaza Strip, Palestine: a capacity assessment analysis based on adapted WHO-PEN tool**. (2020.0) **35** 1412-25. DOI: 10.1002/hpm.3022
51. Bintabara D, Ngajilo D. **Readiness of health facilities for the outpatient management of non-communicable diseases in a low-resource setting: an example from a facility-based cross-sectional survey in Tanzania.**. (2020.0) **10** e040908. DOI: 10.1136/bmjopen-2020-040908
52. Ekenna A, Itanyi IU, Nwokoro U, Hirschhorn LR, Uzochukwu B. **ready is the system to deliver primary healthcare? Results of a primary health facility assessment in Enugu State, Nigeria**. (2020.0) **35** i97-i106. DOI: 10.1093/heapol/czaa108
53. Mohamed Ibrahim MI, Alshakka M, Al-Abd N, Bahattab A, Badulla W. **Availability of essential medicines in a country in conflict: A quantitative insight from Yemen**. (2021.0) **18** 1-13
54. Restinia M, Laksmitawati DR, Anggriani Y, Sekowati Y. **The availability of antihypertensive drug in era of nhi: A study in the primary health care centre jakarta-indonesia.**. (2021.0) **13** 93-8. PMID: 33451053
55. Restrepo SF, Vieira MRdS, Barros CRdS, Bousquat A. **Medicines’ private costs among elderly and the impairment of family income in a medium-sized municipality in the state of Sao Paulo.**. (2020.0) **23**
56. Saeed A, Saeed F, Saeed H, Saleem Z, Yang C, Chang J. **Access to Essential Cardiovascular Medicines in Pakistan: A National Survey on the Availability, Price, and Affordability, Using WHO/HAI Methodology**. (2020.0) **11** 595008. DOI: 10.3389/fphar.2020.595008
57. Contreras-Loya D, Reding-Bernal A, Gomez-Dantes O, Puentes-Rosas E, Pineda-Perez D, Castro-Tinoco M. **Abasto y surtimiento de medicamentos en unidades especializadas en la atención de enfermedades crónicas en México en 2012.**. (2013.0) **55** 618-26. PMID: 24715014
58. Resendez C, Garrido F, Gomez-Dantes O. **Disponibilidad de medicamentos esenciales en unidades de primer nivel de la Secretaría de Salud de Tamaulipas, México**. (2000.0) **42** 298-308. PMID: 11026071
59. Mendes LV, Campos MR, Chaves GC, Mendes da Silva R, da Silva Freitas P, Costa KS. **Disponibilidade de medicamentos nas unidades básicas de saúde e fatores relacionados: uma abordagem transversal.**. (2014.0) **38** 109-23
60. Sabot OJ, Mwita A, Cohen JM, Ipuge Y, Gordon M, Bishop D. **Piloting the global subsidy: the impact of subsidized artemisinin-based combination therapies distributed through private drug shops in rural Tanzania**. (2009.0) **4** e6857. DOI: 10.1371/journal.pone.0006857
61. 61World Bank Country and Lending Groups [Internet]
Washington, D.C.: World Bank; 2020 [cited 2020 Sept 14]. Available from: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519.. (2020.0)
62. Luiza VL, Chaves LA, Campos MR, Bertoldi AD, Silva RM, Bigdeli M. **Applying a health system perspective to the evolving Farmacia Popular medicines access programme in Brazil.**. (2017.0) **2** e000547. DOI: 10.1136/bmjgh-2017-000547
63. Robertson J, Mace C, Forte G, de Joncheere K, Beran D. **Medicines availability for non-communicable diseases: the case for standardized monitoring.**. (2015.0) **11** 18. DOI: 10.1186/s12992-015-0105-0
64. Murphy A, McGowan C, McKee M, Suhrcke M, Hanson K. **Coping with healthcare costs for chronic illness in low-income and middle-income countries: a systematic literature review**. (2019.0) **4** e001475. DOI: 10.1136/bmjgh-2019-001475
65. 65Measuring medicine prices, availability, affordability and price components Geneva: World Health Organization; 2008 [cited 2020 Sept 14]. Available from: https://www.who.int/medicines/areas/access/OMS_Medicine_prices.pdf.
66. Levi J, Pozniak A, Heath K, Hill A. **The impact of HIV prevalence, conflict, corruption, and GDP/capita on treatment cascades: data from 137 countries**. (2018.0) **4** 80-90. PMID: 29682299
67. Shieshia M, Noel M, Andersson S, Felling B, Alva S, Agarwal S. **Strengthening community health supply chain performance through an integrated approach: Using mHealth technology and multilevel teams in Malawi.**. (2014.0) **4** 020406. DOI: 10.7189/jogh.04.020406
68. Vledder M, Friedman J, Sjoblom M, Brown T, Yadav P. **Improving Supply Chain for Essential Drugs in Low-Income Countries: Results from a Large Scale Randomized Experiment in Zambia**. (2019.0) **5** 158-77. DOI: 10.1080/23288604.2019.1596050
69. Attaran A.. **How do patents and economic policies affect access to essential medicines in developing countries?**. (2004.0) **23** 155-66. DOI: 10.1377/hlthaff.23.3.155
70. Mackey TK, Liang BA. **Patent and exclusivity status of essential medicines for non-communicable disease**. (2012.0) **7** e51022. DOI: 10.1371/journal.pone.0051022
71. Beall RF, Schwalm JD, Huffman MD, McCready T, Yusuf S, Attaran A. **Could patents interfere with the development of a cardiovascular polypill?**. (2016.0) **14** 242. PMID: 27538505
|
---
title: 'High impact health service interventions for attainment of UHC in Africa:
A systematic review'
authors:
- Humphrey Cyprian Karamagi
- Araia Berhane
- Solyana Ngusbrhan Kidane
- Lizah Nyawira
- Mary Ani-Amponsah
- Loise Nyanjau
- Koulthoume Maoulana
- Aminata Binetou Wahebine Seydi
- Jacinta Nzinga
- Jean-marie Dangou
- Triphonie Nkurunziza
- Geoffrey K. Bisoborwa
- Jackson Sophianu Sillah
- Assumpta W. Muriithi
- Harilala Nirina Razakasoa
- Francoise Bigirimana
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021619
doi: 10.1371/journal.pgph.0000945
license: CC BY 4.0
---
# High impact health service interventions for attainment of UHC in Africa: A systematic review
## Abstract
African countries have prioritized the attainment of targets relating to Universal Health Coverage (UHC), Health Security (HSE) and Coverage of Health Determinants (CHD)to attain their health goals. Given resource constraints, it is important to prioritize implementation of health service interventions with the highest impact. This is important to be identified across age cohorts and public health functions of health promotion, disease prevention, diagnostics, curative, rehabilitative and palliative interventions. We therefore explored the published evidence on the effectiveness of existing health service interventions addressing the diseases and conditions of concern in the Africa Region, for each age cohort and the public health functions. Six public health and economic evaluation databases, reports and grey literature were searched. A total of 151 studies and 357 interventions were identified across different health program areas, public health functions and age cohorts. Of the studies, most were carried out in the African region ($43.5\%$), on communicable diseases ($50.6\%$), and non-communicable diseases ($36.4\%$). Majority of interventions are domiciled in the health promotion, disease prevention and curative functions, covering all age cohorts though the elderly cohort was least represented. Neonatal and communicable conditions dominated disease burden in the early years of life and non-communicable conditions in the later years. A menu of health interventions that are most effective at averting disease and conditions of concern across life course in the African region is therefore consolidated. These represent a comprehensive evidence-based set of interventions for prioritization by decision makers to attain desired health goals. At a country level, we also identify principles for identifying priority interventions, being the targeting of higher implementation coverage of existing interventions, combining interventions across all the public health functions–not focusing on a few functions, provision of subsidies or free interventions and prioritizing early identification of high-risk populations and communities represent these principles.
## Introduction
As countries in Africa pursue the way for attainment of the sustainable development goals and universal health coverage, there is a need to re-pivot health systems, identify the diseases and conditions that cause the primary country-specific Disability-Adjusted Life Years (DALYs) and divert their limited resources to interventions that result in most lives saved and disabilities averted. These high impact interventions categorised as activities or actions undertaken at an individual, community or programmatic levels are expected to improve the health condition of individuals or communities by preventing infections or diseases, cure diseases or suppressing disease-causing germs/viruses or reducing the severity or duration of an existing disease, or by restoring function lost due to diseases or injuries [1]. Effective or high impact interventions, implemented strategically at the right time for the right population with high coverage, can eliminate conditions that cause the majority of morbidities, mortalities and disabilities. Cost-effectiveness has been one of the criteria used by WHO CHOICE [2] and Disease Control Priorities 3rd edition (DCP-3) [3] in prioritization of interventions for the recommended essential health care packages. The proof of effectiveness in averting DALYs for major conditions in the region could be an additional impetus for countries in prioritizing interventions when compiling their packages.
With the global advancement of science, technology and human innovations, several high-impact interventions of low cost but proven effectiveness have been implemented across regions. These interventions have reduced the burden of diseases and conditions globally and in the African Region. Of particular interest are interventions that contributed to the reduction in communicable diseases. A good example is malaria interventions that resulted in marked reduction in communicable diseases within the region. As a result of the global concerted effort and implementation of high impact interventions of insecticide treated bed nets, indoor residual spraying and artesunate based case management, an estimated 1.7 billion malaria cases and 10.6 million malaria deaths with $85\%$ and $95\%$, respectively in the WHO Africa region, were averted in the last two decades (2000–2020) [4]. Equally, low-cost and effective interventions also exist for reducing the risks and diseases related to non-communicable diseases (NCDs) [5] and Reproductive Maternal Neonatal, Child and Adolescent Health (RMNCAH) conditions [6]. If implemented across all public health functions and age cohorts, these interventions could help countries achieve a significant reduction in morbidities and mortalities across populations. Even more critical is how the evidence on the effectiveness of these interventions can provide countries with guidance on how they can re-pivot their health system development efforts towards attaining the results expected of them. The question then is centred on which diseases constitute the current burden of disease and what interventions are most effective in addressing them? Among the myriad of diseases that afflict the population in the region, some are more common in some specific age groups than others and contribute to DALYs more than others [7]; similarly, some interventions are also more effective than others.
Therefore, it is essential to have a common platform against which the effectiveness of different interventions can be compared to ensure countries can make the best decisions about what to focus on to reduce cost and maximize health outcomes. DALY estimates allow comparability between the impact of diseases and potential benefit of proposed measures set against similar and comparable data of other health problems as it aggregates the total health loss at population level into a single index by summarising years of life lost due to premature death (YLLs) and years lived with disability (YLDs) [7, 8]. Therefore, this comprehensive review is aimed at consolidating the effectiveness of various interventions across public health functions in averting the DALYs contributed by the leading communicable & NCDs, maternal and neonatal conditions and injuries in the different age cohorts in the African Region. This review is intended to influence and shape policies, practices and service delivery, while advancing the road to UHC in the continent.
## Measures of disease burden
Before compiling the interventions, the leading causes of DALYs in the African region were identified and ranked for each age cohort (new born/neonate, under five, 5–9, 10–24, 25–59, 60+. The primary source for this task was the data base of the Institute for Health Metrics and Evaluation (IHME) 2019 [9], which provided information on incidence, DALYs, and deaths for different age groups over extended period of time. Based on this analysis five leading causes of DALYs for each age cohort were prioritized for identifying appropriate interventions. All data on disease burden extracted from the IHME database and are attached in the supplementary appendix S1 Data.
## Measures of effectiveness
This review employed the DALY as the measure of effectiveness for all identified interventions for the preventive, promotive, curative and rehabilitative/palliative public health functions. In the rehabilitation and palliative care setting, Quality Adjusted Life Years (QALYs) gained was used as additional measure of effectiveness.
## Inclusion and exclusion criteria
Articles that were published i) in English or French ii) between 2000–2022 iii) included all types of interventions with health effects expressed in DALY or QALY (specifically for rehabilitative and palliative care interventions) iv) limited to the diseases and conditions identified as major contributors of DALY in the Africa region v) stratified across age cohorts and public health functions vi) all countries were included.
Articles i) languages was not English or French ii) publication date was before the year 2000 iii) different outcome measures from those in our inclusion criteria iv) interventions targeted conditions beyond the determined high burden contributors vi) study protocols, opinions and comments were excluded.
## Search strategy
A comprehensive search strategy adopting key search terms on the subject area based on the PICO framework was conducted. Search terms included “specific age cohorts”, “specific interventions”, “outcome measures of interest” and relevant synonyms. Medical subject headings (MeSH) and key text words were developed and combined with Boolean operators “AND” and/ “OR” across and within categories as necessary. A filter was applied to restrict the type of studies to systematic reviews, meta-analysis, randomized controlled trials, case-control studies or cohort studies. The entire search strategy for the PubMed database was tested and adapted to other databases.
The search followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines [10] [Fig 1]. A literature search in English and French in the following databases (PubMed, Cochrane, African Journals Online, CINAHL, African Index Medicus and Google Scholar) was conducted from December 2021 to February 2022 based on the predefined search strategy. Additional records were manually searched from reference lists of selected papers and grey literature. The search strategies used are attached in the supplementary appendix S1 Table.
**Fig 1:** *Prisma flow diagram of study selection.*
## Article screening
Selection of peer-reviewed articles involved initial screening of title and abstracts to include studies that met the inclusion criteria. Then, full-text articles that fulfilled the inclusion criteria were retrieved, and screening was done. Data screening and extraction was independently conducted by three members of the research team, focusing on their areas of expertise: communicable diseases (CDs), non-communicable diseases (NCDs) and reproductive health, maternal neonatal, child and adolescent health (RMNCAH). The other two members of the review team with health economics background counter-checked the accuracy of screened and extracted information to ensure fidelity. Any disagreement was resolved through discussion and consensus-building. The articles that met the inclusion criteria were imported into EndNote X9. Duplicate records were identified and eliminated. Data from the included articles were extracted onto an Excel worksheet and the evidence graded. The outputs are presented according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analysis) checklist in the supplementary appendix S1 Checklist.
## Data extraction
All relevant information required for analysis were collected using a data extraction template designed in Microsoft Excel 2013. Detailed data on the author, year of publication, the objective of the study, disease program/condition, population, country, study design, study year, intervention type, comparator, public health function, a measure of outcome (DALY), sample size and the summary findings from the study was conducted. Data extracted for included studies is in the supplementary appendix S1 File.
## Search results
The literature search yielded 29,853 articles, 29,820 titles and abstracts through database searching, and 33 records through bibliographic citation searches. A total of 19,722 articles were selected for the title and abstract evaluation. Full-text articles were then obtained for the 863 articles considered for inclusion, and then 151 full-text articles which met the inclusion criteria were included for the review. PRISMA flow [11] of selection and inclusion of the studies was used [Fig 1]. The list of the selected articles with their characteristics is provided as S1 File.
## Quality assessment
Quality assessment was conducted to appraise the quality of studies and establish the scientific processes involved in determining the value of high-quality research evidence to guide policy, inform decision making, direct resource allocation and leverage on strengths across similar contexts. The Quality of Health Economics Studies (QHES) 16-item appraisal tool, which is a validated quality-scoring instrument (with weight score range = 0–100; and >75 = high quality) [12] was used to garner the characteristics and quality of the included studies. Data extraction from the included studies was independently conducted by two research team members and compared for agreement; inter-rater agreement for the selection was $95\%$. Any disagreements were resolved through research team discussion, and agreement on scoring was determined at $79\%$. The quality indicators and their scoring are included in S2 File.
Based on the QHES appraisal, the minimum and maximum score of the included studies was 60 & 90, respectively, with an average quality score of 74 out of a possible 100. Almost $57\%$ of studies obtained a fair good quality score of between 50–$74\%$; and the remaining ($43\%$) score was of high-quality score between 75–$100\%$. None of the studies were extremely poor (0–24) or poor (25–49).
## Characteristics of included studies
Detailed data was extracted from included studies namely the author, year of publication, the objective of the study, disease program/condition, population, country, study design, study year, intervention type, comparator, public health function, measure of outcome, sample size, funding source declaration and the summary findings.
The majority of the effectiveness data was extracted from cost-effectiveness studies ($80.8\%$) that used different types of modelling such as decision- analysis model, Markov’s model and micro-simulation methods among others while the remaining ($19.2\%$) were built on other types of studies such as impact evaluations, economic evaluation among others. Different time horizons were used for modelling the data. A total of 35 ($23.2\%$) decision model studies adopted annual time, 29 ($19.2\%$) lifetime horizon and 27 ($17.2\%$) used 10-year time horizon. For $16.6\%$ of the studies, the period was not explicit or indicated. Most studies were carried out in the African region ($43.5\%$) followed by Asia ($20.1\%$) and multiple regions ($12.3\%$). They included 75 ($50.6\%$) articles for communicable diseases, 56 ($36.4\%$) for NCDs and the remaining 19 ($12.3\%$) for RMNCAH. Studies that covered interventions applicable to all age groups were the majority ($24.1\%$) followed by reproductive age ($15.1\%$) and early childhood ($13.9\%$). Several studies covered the elderly in combination with adults while elderly specific studies were the least represented. Most interventions in the included studies addressed the public health function of disease prevention ($41.2\%$), followed by treatment or curative care ($28.6\%$) and health promotion ($25.5\%$). A dearth of published literature on comparable effective interventions in the rehabilitation and palliative health functions existed. Table 1 below illustrates the characteristics of the included studies.
**Table 1**
| Characteristic | Percentage | Characteristic.1 | Percentage.1 |
| --- | --- | --- | --- |
| Region | Region | Publication years | |
| Africa | 43.5 | 2000–2009 | 18.5 |
| Asia | 20.1 | 2010–2019 | 64.3 |
| Multi-regions | 12.3 | 2020–2021 | 17.2 |
| Americas | 9.1 | Disease category | |
| Europe | 8.4 | Communicable diseases | 50.6 |
| Australia | 5.2 | Non-Communicable diseases | 36.4 |
| Non-specific | 1.3 | RMNCAH | 12.3 |
| | | CD/NCD | 0.6 |
| Income level | Income level | Intervention type | Intervention type |
| LMIC | 31.2 | Curative | 28.6 |
| HIC | 18.2 | Disease prevention | 41.2 |
| LIC | 13.0 | Health promotion | 25.5 |
| Regional | 11.0 | Rehab/palliative | 2.0 |
| Global | 9.1 | Combination | 2.8 |
| UMIC | 9.1 | Time horizon | Time horizon |
| Non-specific & combinations | 8.4 | Annual | 23.2 |
| Study design | | Lifetime | 19.2 |
| Cost-effectiveness studies | 80.8 | 10 years | 17.2 |
| Others* | 19.2 | Not indicated | 16.6 |
| | | 13–30 years | 8.6 |
| | | 2–7 years | 8.6 |
| | | Other time horizons | 6.6 |
## Characterization of interventions
A total of 357 interventions that measured their effectiveness in DALYs were identified for the selected significant conditions per age cohort. At least 162 ($45.74\%$) interventions for infectious diseases, 148 ($41.5\%$) interventions for NCDs and the remaining $11.2\%$ interventions for RMNCAH conditions across the age cohorts. The interventions were categorised first by the i) age cohort ii) disease area, and iii) public health function i.e., health promotion, disease prevention, curative, rehabilitative and palliative care. Almost half ($53.5\%$) of the interventions were for five conditions (HIV, malaria, cancer, mental disorders and CVDs. Studies assessed the effectiveness of single interventions ($58.2\%$), while others evaluated the effectiveness of multiple interventions ($41.7\%$). It should be noted that these 357 interventions were not unique and some of them were repeated in several studies with different settings, comparators and coverages. For example, intermittent malaria preventive treatment in pregnant women (IPTp) was counted seven times as it appeared in five studies. However, only single and unique intervention was selected for the final menu. A detailed menu of effective interventions for each age cohort is attached in the supplementary appendix S2 File.
## Maternal disease burden and interventions
Pregnant mothers and newborns are considered a standard cohort due to their intertwined biological relationships, however, distinct health afflictions occur in these two populations. In the region, four maternal conditions dominate the disease burden related to pregnancy and delivery that can affect negatively the outcome of the mother and newborn. Maternal haemorrhage, maternal abortion and miscarriage, hypertensive disorders and maternal sepsis contribute immensely to the global burden of diseases but typically in the African region. At the end of 2019, maternal disorders together contributed 11,533,751.36 DALYs in the African region. Forty five percent of the DALYs due to maternal conditions were contributed by the African region while almost $70\%$ of the DALYs caused by maternal abortions and $60\%$ of maternal sepsis were contributed by the region. These five maternal conditions were responsible for $32.3\%$ of the total maternal disorder DALYs in the region and $46\%$ of the global ones [Table 2]. This makes them important public health problems for countries to address them in their pursuit to reduce the maternal and neonatal mortalities.
**Table 2**
| Rank | Maternal disorders | DALYs, Africa region | % | DALYs, Global | Africa contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | Maternal haemorrhage | 1094474.07 | 9.5 | 3087856.85 | 35.4 |
| 2.0 | Maternal abortion and miscarriage | 783235.92 | 6.8 | 1130038.21 | 69.3 |
| 3.0 | Maternal hypertensive disorders | 712796.11 | 6.2 | 1823862.91 | 39.1 |
| 4.0 | Maternal sepsis and other maternal infections | 633631.1 | 5.5 | 1064670.02 | 59.5 |
| 5.0 | Maternal obstructed labour and uterine rupture | 504551.99 | 4.4 | 999540.67 | 50.5 |
| | Subtotal | 3728689.18 | 32.3 | 8105968.66 | 46.0 |
| | Total for all maternal disorders | 11533751.36 | 100.0 | 25352807.46 | 45.5 |
Overall, ten ($$n = 10$$) studies were identified with interventions that can reduce the leading causes of DALYs. Health promotion ($$n = 2$$); Disease Prevention ($$n = 4$$); Curative ($$n = 4$$); Rehabilitative/Palliative Care ($$n = 0$$)
## Health promotion
Two [2] health promotion studies were identified; these included for obstetric haemorrhage [13]; and hypertensive disorders of pregnancy [14]. A quality improvement programme covering personnel, system management and quality communication at facility levels was found to be effective at averting DALYs associated with obstetric haemorrhage [13]. Research evidence found that promoting the use of simple diagnostic tools in urine assessment was critical for identifying early signs of preeclampsia and was effective at averting DALYs associated with hypertensive disorders of pregnancy [14].
One study ($$n = 1$$) reported on health promotion campaigns for supplemental folate/folic acid consumption first, in women capable of or planning a pregnancy and second in population-wide campaigns to promote supplement use where mandatory fortification was found to be the most effective intervention at reducing neural tube defects [20]. The other study ($$n = 1$$) reported on screening all newborn using three possible screening options: pulse oximetry alone, clinical assessment alone, and pulse as an adjunct to clinical assessment in diagnosing critical congenital heart disease [21]. The most effective of the three neonatal screening options was pulse oximetry with clinical assessment. Newborn screening at the national level was recommended as an effective measure for early diagnosis and subsequent intervention to expand long term national health benefits and improve newborn health outcomes. These interventions were recommended across all levels of care, i.e., community, primary, secondary and tertiary. Ali et al., [ 22] reported on the use of the Augmented Infant Resuscitator (AIR) device as an effective device in a mixed microsimulation calibrated Markov model of the apnoea training programme when used as a supplement to existing resuscitation training programs. More DALYs were estimated to be averted on top of the benefits of the widely adopted Helping Babies Breathe (HBB) training for improving the quality of bag-valve-mask resuscitation among non-breathing newborn. The benefits of this AIR device have been identified; however, prioritisation of HBB implementation in birthing facilities was recommended until field test confirmations of the AIR device are conducted at scale. For intrapartum related complications, we identified that low-dose, high-frequency training in basic emergency obstetrics and newborn was an effective strategy in improving newborn health outcomes and averting DALYs. Integrating this intervention into existing in-service training programs and health systems is important for achieving quality health care outcomes [23].
We found several effective health promotion interventions for children aged 0–5. Exclusive breastfeeding promotion was highlighted in several studies as an effective intervention in averting DALYs for; diarrhoeal diseases [25, 26] and LRTIs [27]. No significant difference was found in community-based peer counselling to promote exclusive breastfeeding (in addition to HFP) relative to standard HFP (Health Facility Breastfeeding promotion) [26]. The promotion of population-wide oral rehydration therapy (ORT) was also found to be effective in averting DALYs caused by diarrhoeal diseases for this age cohort [25].
Community-based information activities was an effective strategy to avert DALYs caused by malaria [28]. Promotion of nutrient-rich diets through price subsidies, tax incentives and disincentives, and cash incentives effectively averted DALYs caused by malnutrition. Examples of effective price subsidies strategies included; price subsidies for fortified packaged infant cereals (F-PICs) for under 2’s within poorer households [29], price subsidies on fortified packaged complementary foods (FPCF) for children in different wealth quintiles [30]. Cash incentives such as standard cashback (SCB), double cash (DCB), and fresh food vouchers (FFV) also had a significant impact on nutrition outcomes in children under five years of age [31]. Other nutrition-based incentives included the mass distribution of daily micronutrient powders (MNPs) for infants between 6–18 months [32] and routine supplementation with EMMS [33]. Mass media campaigns addressing the main causes of post neonatal child mortality such as malnutrition were effective strategies in averting DALYs for this age cohort [34, 35]. All these interventions were implemented at the community level.
Malaria, dietary iron deficiency, diarrhoeal and meningitis health promotion interventions for this age cohort were similar to those in the early childhood group, with exceptions such as breastfeeding promotion for diarrhoeal diseases. Promotion of nutrient-rich diets through price subsidies, tax incentives and disincentives, and cash incentives effectively avert DALYs caused by malnutrition. These have been discussed within the early childhood section [29–35, 61]. Population-wide ORT interventions were important for diarrhoeal diseases [25].
We did not find any health promotion interventions for LRTIs for this age cohort. Meningitis health promotion interventions have also been addressed in the early childhood section [62]. HIV based interventions such as school-based sexual education on HIV and mass media campaigns were significant in averting DALYs related to HIV [63, 64]. Although at higher budget levels, school-based education yielded higher health gains than mass media, combining the two strategies averted the most DALYs caused by HIV. Information education and Communication (IEC) programmes also effectively averted DALYs in HIV disease [65]. Home-based HIV testing for partners of pregnant women was found to be effective [65].
HIV/AIDS health promotion interventions were similar to those of the late childhood age cohort and have been addressed in that section with the addition of condom promotion. Malaria disease health promotion interventions that were effective in averting DALYs have been addressed within the early childhood section.
Effective health promotion interventions that covered road traffic injuries and mental conditions included; increased taxation on alcohol, increased minimum legal drinking age, alcohol advertising bans and limited hours of alcohol sale [69, 70]. Random breath testing, selective breath testing, and mass media campaigns on road safety were also found to be effective against road traffic injuries [71]. Additionally, taxation of sugar-sweetened beverages to prevent obesity was found to be effective in averting DALYs caused by mental health conditions [35, 72].
In addition to earlier highlighted HIV based health promotion interventions, including school-based sexual education on HIV and mass media campaigns [62, 63], peer-led outreach and education [63, 67], peer education for sex workers [62], condom promotion for female sex workers (FSWs) and men who have sex with men [68] and condom distribution and promotion for sexually active populations were effective in averting HIV DALYs for this age cohort [63, 67]. Community-based information activities on malaria were found to be effective in averting DALYs caused by malaria [27]. Diet-related interventions had a significant role in averting DALYs for CVD/DM and cancers. These interventions included legislation over the following; population-wide reduction in daily salt consumption [54, 88, 89], tax reduction on fruits and vegetables, tax increase on fats and sugar, taxation of junk food and front of pack traffic light labelling of nutrition on processed foods [34]. Cancer awareness interventions averted cancer DALYs for this age cohort [90]. A combined approach to the prevention of behavioural risk factors was found to avert DALYs caused by NCDs significantly.
Diet based health promotion interventions that averted DALYs for NCDs, including CVD/DM and cancers, were; mass media campaigns on salt reduction [110], government legislation to reduce salt consumption [88], taxation of junk food, fats and sugars [111], front of pack-traffic light labelling of nutrition of processed foods [34], lower consumption of meat, dairy and fats [112] and reduction of tax for fruits and vegetables [111]. Cancer awareness-raising was effective in averting DALYs for cancers [90], and a combined approach in the prevention of behavioural risk factors was found to be very effective for NCDs generally. Effective interventions aimed at reducing the harmful effects of tobacco included tobacco price increase [25] and media campaigns on smoking [110].
Replacement of traditional stoves with solid fuel or liquid gas and attainment of air quality PM 2.5 concentration significantly was found to translate into an impressively large health gain [56, 113]. Among all these health promotion interventions, five interventions were based on LMIC settings in two countries (Vietnam and India). These included; mass media campaigns on salt reduction (Vietnam), front of pack-traffic light labelling of nutrition of processed foods (India), increased taxation on junk foods (India), legislations to replace traditional stoves with solid fuel or liquid gas (India) and attainment of air quality PM 2.5 concentration (India). The effect of legislation on salt reduction consumption was evaluated across multiple countries of all income classifications, while the rest of the interventions were based on high-income settings.
## Disease prevention
We identified four ($$n = 4$$) studies; one on obstetric haemorrhage [15]; two ($$n = 2$$) on unsafe abortions [13, 15] and one on maternal sepsis [15]. In the management of obstetric haemorrhage, early application of Non-pneumatic Anti-Shock Garment (NASG) at the primary health care level for women in hypovolemic shock can be -effective across diverse clinical settings rather than its application at the regional level [16]. The NASG is a low-cost first-aid compression device that limits the effects of PPH. Concerning unsafe abortions and its complications, sustained provision of Long-acting reversible contraceptives (LARCs) and depot-medroxyprogesterone acetate [DMPA] methods were effective in averting DALYs associated with unsafe abortions and its-related complications [17]. Women who were well counselled used these methods for a long time culminating into maternal and child health benefits. Vaginal misoprostol was found to be effective against the effects of unsafe abortions [15]. Tetanus toxoid for pregnant women significantly for averted DALYs associated with maternal and neonatal conditions; therefore, investing in universal coverage of this intervention and its scale-up was recommended [15].
Two ($$n = 2$$) studies were identified [15, 24], one related to neonatal sepsis and the other, on preterm birth complications. Research evidence from Ahmed et al., found that Group B streptococcal (GBS) hexavalent maternal vaccination program would avert more DALYs than the current standard of care for preventing neonatal sepsis [24]. The other study found that administration of- antibiotics at institutional levels for preterm pre-labour rupture of membrane (pPRoM), antenatal corticosteroids for preterm labour and Kangaroo mother care effectively averted DALYs associated with preterm birth complications in study [15].
Effective disease prevention interventions covering the major causes of DALYs for this age cohort included; vaccinations for various diseases and conditions such as Pneumococcal vaccine for pneumonia [27, 36–38], *Haemophilus influenzae* type B vaccination [27], cholera vaccination [39], RTS,S malaria vaccine [40, 41] and vaccinations for meningitis [42, 43]. Combinations of interventions such as pneumococcal conjugate vaccine and *Haemophilus influenza* type B vaccine averted more DALYs [27] than single entity vaccinations.
Exclusive breastfeeding for the first 6 months of life was an effective disease prevention strategy in averting DALYs for malnutrition [44]. Examples of nutrition modifications to prevent disease included; daily iron sulphate II supplementation [45], complementary feeding practices in infants six months to 3 years of age [44] and zinc supplementation for LTRIs [30] and diarrhoeal diseases [46].
Several interventions were effective in averting DALYs caused by malaria through disease prevention. These interventions included; intermittent preventive treatment of malaria in infants (IPTi) [47], intermittent malaria screening of school children [48], intermittent preventive treatment of malaria in pregnancy (IPTp) [49, 50], seasonal malaria chemoprophylaxis (SMC) [51], insecticide-treated nets (ITNs) [48, 49, 52, 53], indoor residual spraying (IRS) [25, 48, 49, 53], larviciding [54]. But WHO guidelines recommend larviciding as supplementary intervention in areas where optimal coverage with ITNs or IRS has been achieved., Long lasting ITNs averted more DALYs compared to conventional ITNs. Combinations of interventions averted more DALYs for example, IPTp + ITN [50], IRS with pirimiphos-methyl and pyrethroid ITNs [49, 53, 55], LIN plus IRS [48]. However, WHO guidelines recommend against combining ITNs and IRS and that priority be given to delivering either ITNs or IRS at optimal coverage and to a high standard.
Inhaled pollutants were considered risk factors for both infectious and non-infectious respiratory diseases; hence, the use of solid stoves and cleaner liquid fuels effectively averted DALYs caused by inhalation of polluted air [27, 56].
Malaria, dietary iron deficiency, diarrhoeal diseases, LRTIs and meningitis disease prevention interventions for this age cohort were similar to those in the early childhood group, with exceptions such as exclusive breastfeeding and complementary feeding practices unique to the younger age cohorts. These have been addressed within the early childhood section. Effective disease prevention interventions for HIV/AIDS included; blood safety [63,64], voluntary counselling and testing [62, 66], street children programs [64], treatment for STIs [62, 64, 67] and prevention of mother to child transmission [62, 64].
HIV/AIDS disease prevention interventions were similar to the late childhood age cohort. However, there were additional HIV/AIDS interventions for this age cohort and these included; Voluntary medical male circumcision [73], provision of oral PrEP ([74], intravaginal rings [74], microbicide gels, SILCS diaphragms used in concert with gel [74], injectable PrEP, provision of dual PrEP (both oral and injectable PrEP) and condom use [64, 65, 75, 76]. Malaria disease prevention interventions such ITNs, IRS, larviciding were effective in averting DALYs caused by malaria. These have been addressed within the early childhood cohort. Short course TB preventive therapy effectively averted TB DALYs [77].
Enforcement of road safety regulations through various means was found to be effective in averting DALYs. These included; speed limits via mobile speed cameras, breath testing campaigns, seatbelt use in cars, helmet use by motorcyclists and bicyclists, traffic codes, and building speed bumps at high-risk intersections [69, 70, 78]. School-based psychological interventions to prevent the onset of depression among students by a universal or targeted approach was an effective preventive method of averting DALYs caused by mental conditions [79].
Sustained free HIV voluntary counselling and testing [62, 64, 66], home-based HIV testing of pregnant womens’ [65], peer counselling for key populations [62, 63, 67], treatment of sexually transmitted infections (STIs) for the general population [67] and for key populations [62, 63], oral pre-exposure prophylaxis (Vogelzang et al. 2020, [28, 73, 74]), oral pre-exposure prophylaxis (PrEP) for MSMs [91], voluntary male medical circumcision [72], PMTCT [65] were found to be effective HIV disease prevention interventions and antenatal syphilis screening was effective in prevention syphilis in sub-Saharan Africa [92]. Dual-use of womens’ condoms for family planning and HIV prevention were shown to avert DALYs due to HIV [76]. The additional interventions for this age cohort mostly comprise interventions for key populations. TB preventive treatment for persons living with HIV [77] and tuberculin skin testing (TST) followed by IPT for TST-positive patients with no evidence of active TB was found to avert DALYs caused by TB [93] significantly. Malaria prevention for this age cohort also benefited significantly from interventions such as ITNs, IRS, IRS and ITN and larviciding, which have been addressed in earlier sections.
Cancer preventive interventions that were effective for this age cohort included; HPV cytology screening/triennial pap smear [94, 95] and HPV vaccination in pre-adolescent girls for cervical cancer [94, 96], biennial mammography in women above 40 years [90, 95, 97, 98] and biennial clinical breast examination for breast cancer [90, 97], colonoscopy every five years for colon cancer [95] and genomic screening of individuals at risk [53]. The effectiveness of annual mammography was found to be low due to the short interval of screening (Zehtab et al. [ 98]). Colonoscopy screening every seven years was the most effective strategy in terms of DALYs averted. However, it carried a high number of iatrogenic deaths associated with endoscopy-related perforations, while colonoscopy screening every five years was found to generate the most significant number of iatrogenic deaths [95]. Combined HPV vaccination in pre-adolescence, followed by screening (using HPV DNA testing) in adulthood, was also highly effective compared to one intervention [94].
Interventions for CVD/DM prevention that were effective included; community-based management of hypertension targeting everyone regardless of hypertension status as opposed to targeting only people with hypertension [99, 100], screening and addressing CVD risk factors [101–103], E-health interventions for CVD risk assessment and mitigation [103], integration of hypertension and diabetes screening with HIV programs [104] and use of polypill for secondary CVD prevention in high-risk candidates [25]. Reducing complications from diabetes in LMICs through scaling up blood pressure and statin medication treatment initiation and blood pressure medication titration was found to be a more effective strategy rather than focusing on increasing screening to increase diabetes diagnosis or a glycaemic treatment and control among people with diabetes [101].
Interventions to prevent CVD/DM that were effectively included; community-based management of hypertension [25, 99], screening and addressing CVD risk factors [100, 114], mobile technology/E-health enabled community-based CVD risk assessment mechanisms [103], higher consumption of grain-based foods, seafoods, fruits and vegetables [60, 102, 111, 112] and lower consumption of meat, dairy and fats [112]. Cancer disease prevention interventions that were effective were similar to that of the adult cohort addressed above.
LRTI interventions that were found to be effective included; COVID Vaccine) [115], H Influenzae type B vaccination [18], use of solid stoves [26, 56], use of cleaner liquid fuels [26, 56]. Replacement of indoor stoves with smokeless technology was effective in averting DALYs for COPD [56].
## Curative
We identified ($$n = 3$$) studies covering diverse interventions for obstetric haemorrhage ($$n = 3$$); and hypertensive disorders in pregnancy ($$n = 1$$). Concerning obstetric haemorrhage, Every Second Matters—Uterine Balloon Tamponade (ESM-UBT) was found to be effective and less expensive for managing atonic postpartum haemorrhage and averting DALYs compared to standard care—condom uterine balloon tamponade [18]. In another study, the use of NASG across three intervention scenarios: no women in shock receive the NASG, only women in severe shock receive the NASG, and all women in shock receive the NASG to be effective [19]. The application of NASG to women with haemorrhage significantly affected averting DALYs and decreasing mortality. Active management of the third stage of labour—prophylactic oxytocin, cord clamping and delivery of the placenta by controlled cord traction was effective for averting DALYs associated with obstetric haemorrhage [15]. In relation to preeclampsia and eclampsia management, packages of care involving anti-hypertensives and magnesium sulphates effectively averted DALYs associated with hypertensive disorders in pregnancy, hence, calling for universal coverage of the intervention and its scale-up [15]. In maternal sepsis case management, inpatient care, including treatment with antibiotics, were effective.
Injectable antibiotics were found to be effective for averting DALYs [15]. For infants with respiratory distress at birth identified as an intra-partum-related complication, immediate neonatal resuscitation was effective in averting both short and long sequelae of hypoxic-ischemic encephalopathy [15].
Effective curative interventions for LRTi included; treatment of non-severe clinical pneumonia at the community level and facility level [25, 27, 46, 57], zinc supplementation [27], standard of care (hospitalization, low flow oxygen and antibiotics) and bubble continuous positive airway (bCPAP) [58]. Diarrhoeal disease curative interventions included; oral rehydration therapy [46], therapeutic zinc supplementation [46], case management [46], and complementary feeding [44]. Malaria curative interventions included case management with artemisinin-based combination therapy [49], prereferral rectal artesunate for treatment of severe malaria [59] and parenteral artesunate for treating severe malaria [60].
Curative interventions for this age cohort were similar to those of early childhood for malaria, malnutrition, diarrhoeal diseases, LRTIs and meningitis except for exclusive breastfeeding and complementary feeding practices for malnutrition. These have been discussed within the early childhood interventions section. HIV based curative interventions that were found to be effective in averting more DALYs included; highly active antiretroviral therapy (HAART) and HAART with laboratory monitoring i.e., CD4 and viral load [68].
Effective curative interventions for HIV/AIDS were similar to those of late childhood and are addressed in that section. Effective curative interventions for diarrheal diseases were similar to those addressed within the early childhood section. TB effective curative interventions included; Diagnosis using Xpert MTB/RIF [61, 80], DOTS for management of TB [62, 81], DOTS-Plus treatment for multidrug-resistant tuberculosis [82], Active case finding (ACF) for TB in household contacts of index smear-positive TB patients [83]. Increased access to surgery significantly averted DALYs for road traffic injury victims. Mental conditions’ effective interventions included; community-based pharmacologic and psychosocial treatment for depression [75, 84, 85], roadside breath testing for alcohol use disorder [75], increased access to newer antipsychotic medications and psychosocial treatment [75, 84, 85], increased access to older mood stabilizers medications, newer antidepressant and psychosocial treatment for bipolar disorder [85], integration of family intervention with antipsychotic medications for schizophrenia, eHealth sessions [86] and face-to-face sessions [86, 87].
Highly active antiretroviral therapy with and without laboratory monitoring [62, 63, 68], injectable cabotegravir/rilpivirine in all people on ART [105] and co-morbidity and opportunistic infections treatment were effective curative interventions for HIV/AIDS. Lateral-flow urine lipoarabinomannan (LAM) with a standard for TB diagnosis in PLHIV [106], active case finding (ACF) for TB in household contacts of index smear-positive TB patients with Xpert MTB/RIF as a diagnostic tool [83], Xpert/LF-LAM diagnosis [107], DOTS for management of TB [25, 73, 81, 82, 108]. DOTS management of TB was family-based, community-based or facility-based. DOTS plus using Bedaquiline based regimen was also effective in MDR TB cure [109]. Case management with artemisinin-based combination therapy with high coverage [48] and parenteral artesunate for treating severe malaria [59] were effective in Malaria cure.
Improved access to cancer treatment averted cancer DALYs significantly [90]. Nicotine replacement and metformin for diabetes treatment were effective for CVD/DM [25].
Improved access to breast cancer treatment averted cancer DALYs [90]. Availability of metformin for diabetes treatment prevented CVD/DM DALYS [25]. Effective LRTI curative interventions included; treatment of non-severe clinical pneumonia at the community level [25, 26, 45], treatment of non-severe and severe clinical pneumonia at the facility level [26], the standard of care (hospitalization, low flow oxygen and antibiotics [26]. Improved access to inhaled corticosteroids and long-acting β agonists for asthma and COPD was found to be effective [116].
## Disease burden and interventions in newborn/neonate cohort
Newborns, also known as neonates, are considered individuals aged from birth up to 28 days of life. Complications from birth asphyxia and birth trauma followed by preterm birth and neonatal sepsis dominate the disease burden. These are followed by lower respiratory infections, congenital anomalies and syphilis. At the end of 2019, three of the leading conditions and diseases related to complicated birth contributed a total of 55,801, 796.50 DALYs, accounting for $62.2\%$ of the DALYs in the African region and $25.8\%$ of the Global DALYs in the neonate age group [Table 3].
**Table 3**
| Rank | Conditions | DALYs, Africa Region | % | DALYs, Global | African contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | Neonatal encephalopathy due to birth asphyxia and trauma | 24135392.07 | 26.9 | 48982146.29 | 49.3 |
| 2.0 | Neonatal preterm birth | 21544682.64 | 24.0 | 56492629.93 | 38.1 |
| 3.0 | Neonatal sepsis and other neonatal infections | 10121721.79 | 11.3 | 18688688.55 | 54.2 |
| 4.0 | Lower respiratory infections | 7762683.22 | 8.7 | 17442098.14 | 44.5 |
| 5.0 | Congenital abnormalities* | 5375147.82 | 6.0 | 16467535.72 | 32.6 |
| 6.0 | Syphilis | 3542416.85 | 3.9 | 5475407.22 | 64.7 |
| 7.0 | Diarrheal diseases | 2124956.32 | 2.4 | 3878981.52 | 54.8 |
| | Subtotal | 74607000.72 | 83.2 | 167427487.36 | 44.6 |
| | Total DALYs, Neonates | 89700472.1 | 100.0 | 215978571.79 | 41.5 |
Overall, seven ($$n = 7$$) studies with interventions that can reduce the leading causes of DALYs were identified. Health promotion ($$n = 4$$); Disease Prevention ($$n = 2$$); Curative ($$n = 1$$); Rehabilitative/Palliative Care ($$n = 0$$).
## Disease burden and interventions in early childhood (under 5 years)
The major contributors to DALYs for this age cohort were communicable diseases such as diarrhoeal diseases, lower respiratory infections, and malaria. Sequel of births such neonatal encephalopathy due to birth asphyxia and trauma, preterm birth and sepsis were also important contributors of DALY. This could be as a result of overlap with that of newborns. Three of the leading causes of DALYs in the under 5 age group in the African region (diarrheal diseases, lower respiratory tract infections and malaria) contributed a total of 91,744,812.03 DALYs which is $38.3\%$ for the African region and $19.3\%$ of the global DALYs for these three conditions in this age group. Almost $96\%$ of global malaria, $68.3\%$ of meningitis and $67.9\%$ of the diarrhoea DALYs in this age group were contributed by the African region [Table 4].
**Table 4**
| Rank | Conditions | DALYs, African Region | % | DALYs, Global | Africa Contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | Diarrheal diseases | 30925537.71 | 12.9 | 45544641.05 | 67.9 |
| 2.0 | Lower respiratory infections | 30503932.72 | 12.7 | 59162236.12 | 51.6 |
| 3.0 | Malaria | 30315341.6 | 12.6 | 31594030.51 | 96.0 |
| 4.0 | Neonatal encephalopathy due to birth asphyxia and trauma | 24653499.91 | 10.3 | 50775729.16 | 48.6 |
| 5.0 | Neonatal preterm birth | 22356728.45 | 9.3 | 60200062.02 | 37.1 |
| 6.0 | Neonatal sepsis and other neonatal infections | 10735404.53 | 4.5 | 20573473.61 | 52.2 |
| 7.0 | Meningitis | 6743317.42 | 2.8 | 9876050.87 | 68.3 |
| | Subtotal | 156233762.35 | 65.2 | 277726223.34 | 56.3 |
| | Total DALYs, Under-five | 239666918.68 | 100.0 | 475247148.09 | 50.4 |
High impact interventions that reduce these leading causes of DALYs in this age group are described below.
Overall, fifty-five ($$n = 55$$) studies were identified. Health promotion ($$n = 14$$); Disease Prevention ($$n = 28$$); Curative ($$n = 13$$); Rehabilitative/Palliative Care ($$n = 0$$)
## Disease burden and interventions in late childhood (5–9 years)
The age group 5–9 (taken as a proxy for the 6–11 years old cohort), contributes the least DALYs, $3.4\%$ for the African region and $2\%$ globally among the six age cohorts. The major DALY contributors in this age group in 2019 were malaria, dietary iron deficiency, diarrheal diseases, and HIV. Lower respiratory tract infections, Invasive Non-typhoidal Salmonella (iNTS) and meningitis were also leading contributors of DALYs in this age group. In this age group, $80.1\%$ of malaria, $86.7\%$ of HIV/AIDS and $83.5\%$ of Invasive Non-typhoidal Salmonella (iNTS) of the global DALYs were contributed by the African region [Table 5]. Due to the lack of data that measures effective interventions using DALYs averted as the outcome measure, invasive Non-typhoidal *Salmonella is* not addressed here.
**Table 5**
| Rank | Conditions | DALYs, African Region | % DALYs | DALYs, Global | Africa Contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | Malaria | 2088923.73 | 9.8 | 2607727.98 | 80.1 |
| 2.0 | Dietary iron deficiency | 2026596.59 | 9.5 | 5432574.74 | 37.3 |
| 3.0 | Diarrheal diseases | 1707585.99 | 8.0 | 3898530.44 | 43.8 |
| 4.0 | HIV/AIDS | 938149.16 | 4.4 | 1081973.66 | 86.7 |
| 5.0 | Lower respiratory infections | 904461.15 | 4.3 | 2297525.32 | 39.4 |
| 6.0 | Invasive Non-typhoidal Salmonella (iNTS) | 673233.82 | 3.2 | 806327.59 | 83.5 |
| 7.0 | Meningitis | 652749.2 | 3.1 | 1172487.85 | 55.7 |
| | Subtotal | 8991699.63 | 42.3 | 17297147.58 | 52.0 |
| | Total DALYs, 5–9 years | 21263276.1 | 100.0 | 62227042.6 | 34.2 |
Overall, forty-four ($$n = 44$$) studies were identified to address these leading conditions. Health promotion ($$n = 16$$); Disease Prevention ($$n = 15$$); Curative ($$n = 13$$); Rehabilitative/Palliative Care ($$n = 0$$)
## Disease burden and interventions in Adolescents (12–24 years)
In the age group 10–24 (taken as a proxy for the 12–24 years old cohort), the major DALY contributors were HIV/AIDS, malaria, diarrheal diseases and TB. Three non-communicable diseases such as migraine, road injuries and major depressive disorders appeared as leading conditions in this age group. In 2019, this age group was responsible for $9.2\%$ of the total DALYs in the African region and $7.8\%$ of the global DALYs. In this age group, HIV/AIDS and malaria in the African region contributed $83.2\%$ and $81.0\%$, respectively of the Global DALYs for these diseases [Table 6].
**Table 6**
| Rank | Conditions | DALY, African Region | % | DALYs, Global | Africa Contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | HIV/AIDS | 4901169.97 | 8.6 | 5892211.31 | 83.2 |
| 2.0 | Malaria | 3351548.39 | 5.9 | 4137889.76 | 81.0 |
| 3.0 | Diarrheal diseases | 2142777.42 | 3.8 | 5888885.3 | 36.4 |
| 4.0 | Drug-susceptible tuberculosis | 1847994.64 | 3.3 | 4441577.08 | 41.6 |
| 5.0 | Migraine | 1752739.9 | 3.1 | 10776503.35 | 16.3 |
| 6.0 | Motor vehicle road injuries | 1615664.15 | 2.8 | 6022887.08 | 26.8 |
| 7.0 | Major depressive disorder | 1511328.66 | 2.7 | 7170287.25 | 21.1 |
| | Subtotal | 17123223.12 | 30.2 | 44330241.12 | 38.6 |
| | Total DALY, 10–24 Years | 56755881.8 | 100.0 | 237718695.9 | 23.9 |
Overall, fifty-seven ($$n = 57$$) studies were identified that address the leading conditions. Health promotion ($$n = 11$$); Disease Prevention ($$n = 24$$); Curative ($$n = 22$$); Rehabilitative/Palliative Care ($$n = 0$$)
## Disease burden and interventions in adults (25–59 years)
In Africa, this age group is ravaged by diseases of both infectious and non-communicable nature. HIV/AIDS is the leading condition, with almost $17\%$ of DALYs in this cohort attributed to this single disease. This same population suffers from cardiovascular diseases, diabetes mellitus type 2 and cancers, which contribute $13.6\%$ of the total DALYs in the African region. Almost one out of five ($24\%$) of the DALYs in the African region are contributed by this age group. In this age group, two communicable diseases such as HIV/AIDS and malaria contributed $72\%$ and $87\%$ of the global DALYs, while cardiovascular diseases, diabetes mellitus and total cancers contributed only $16.6\%$ of the global DALYs [Table 7]. This indicates that even in the adult population of the African region HIV/AIDS, tuberculosis and malaria remain significant public health problems despite the increasing NCD trends.
**Table 7**
| Rank | Conditions | DALYs, African Region | % DALYs | DALYs, Global | Africa Contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | HIV/AIDS | 24801031.79 | 16.7 | 34446312.72 | 72.0 |
| 2.0 | CVD*+DM | 12068261.73 | 8.1 | 139514387.87 | 8.7 |
| 3.0 | Drug-susceptible tuberculosis | 8201866.33 | 5.5 | 23401190.59 | 35.0 |
| 4.0 | Total cancers | 8121182.55 | 5.5 | 102362972.15 | 7.9 |
| 5.0 | Malaria | 5220940.27 | 3.5 | 5997964.72 | 87.0 |
| 6.0 | Diarrheal diseases | 3942411.44 | 2.6 | 11767585.64 | 33.5 |
| 7.0 | Major depressive disorder | 3135647.97 | 2.1 | 22389053.94 | 14.0 |
| | Subtotal | 65491342.08 | 44.0 | 339879467.63 | 19.3 |
| | Total DALYS in 25–59 Years | 148786945.64 | | 1069431992.99 | |
Overall, seventy-eight ($$n = 78$$) studies were identified that address the leading conditions in this age group. Health promotion ($$n = 16$$); Disease Prevention ($$n = 44$$); Curative ($$n = 17$$); Rehabilitative/Palliative Care ($$n = 1$$).
## Rehabilitative and palliative care
Basic palliative care (palliative care volunteers training programme + home-based visits by volunteers every fortnight + pain treatment through morphine, laxatives and palliative radiotherapy for eligible patients) and extended that replaces the volunteers with nurses and strengthened pain management with antidepressants, anti-emetics and zelodronic acid averted breast cancer DALYs in Ghana [90].
Basic palliative care (palliative care volunteers training programme + home-based visits by volunteers every fortnight + pain treatment through morphine, laxatives and palliative radiotherapy for eligible patients) and extended that replaces the volunteers with nurses and strengthened pain management with antidepressants, anti-emetics and zelodronic acid averted breast cancer DALYs in Ghana [90].
Using QALY as an outcome measure for rehabilitative and palliative care interventions, we found interventions for the adult and elderly groups and the control of cardiovascular diseases and COPD. In combating cardiovascular disease, including diabetes among adults, home-based cardiac rehabilitation was an effective intervention than usual care alone in patients with reduced ejection fraction heart failure [117]. Similarly, person-centred care services delivered at the primary or secondary level, when added to usual care in chronic heart failure or COPD for individuals, can improve patients’ quality of life [118]. Among the elderly, in addition to the interventions listed above, in the palliative care setting, two additional interventions that resulted in QALY gains were person-centred home-based palliative care and home-based care for advanced heart failure by medical personnel [119, 120].
## Disease burden and interventions in Elderly (Above 60 years)
The leading causes of DALYs in this age group (60+ years) were the non-communicable diseases such as cardiovascular diseases (CVDs), cancers and type 2 diabetes mellitus. Almost $35\%$ of the DALYs for this age group in the African region were contributed by cardiovascular diseases, diabetes mellitus and cancers. Another non-communicable disease with significant contributions to the DALYs in the region was chronic obstructive pulmonary diseases. The dominance of non-communicable diseases in this cohort demonstrates well the effects of epidemiological transition that *Africa is* experiencing and the opportunities to avert future burden by applying effective interventions in the younger populations. In 2019, leading non-communicable conditions such as CVD with diabetes mellitus, total cancers and COPD in the African region contributed $6.1\%$, $4.4\%$ and $3.3\%$ of the global DALYs, respectively, while almost $90.8\%$ of the global malaria burden for this age group was contributed by the African region [Table 8].
**Table 8**
| Rank | 60+ Conditions | DALYs, African Region | % DALYs | DALYs, Global | Africa Contribution (%) |
| --- | --- | --- | --- | --- | --- |
| 1.0 | CVD+DM | 16411642.68 | 26.1 | 268943113.46 | 6.1 |
| 2.0 | Total cancers | 5864268.24 | 9.3 | 134113016.07 | 4.4 |
| 3.0 | Lower respiratory infections | 3389037.25 | 5.4 | 20745385.95 | 16.3 |
| 4.0 | Drug-susceptible tuberculosis | 2339599.19 | 3.7 | 9730629.95 | 24.0 |
| 5.0 | Diarrheal diseases | 2199841.57 | 3.5 | 13818136.55 | 15.9 |
| 6.0 | Malaria | 1906745.18 | 3.0 | 2100198.19 | 90.8 |
| 7.0 | Chronic obstructive pulmonary disease | 1904720.0 | 3.0 | 57582989.97 | 3.3 |
| | Subtotal | 34015854.11 | 54.2 | 507033470.13 | 6.7 |
| | Total DALY, 60+ years | 62772598.3 | 100.0 | 981472243.67 | 6.4 |
Overall, sixty-two ($$n = 62$$) studies were identified. Health promotion ($$n = 9$$); Disease Prevention ($$n = 33$$); Curative ($$n = 16$$); Rehabilitative/Palliative Care ($$n = 4$$).
## Discussion
This comprehensive review has documented the leading causes of DALYs and identified effective interventions for the leading conditions in the African region. Based on this analysis, a menu of effective interventions that could avert most of the DALYs for the major diseases and conditions were identified by age cohort, public health functions and levels of care.
Most of these interventions are almost compatible with those generated by UHC compendium database [121], WHO AFRO essential health interventions [122], WHO-CHOICE [2] and DCP3 [123]. This is aligned to the ongoing regional shift, placing emphasis on the continuum of care across the public health function and throughout the life course. The structural alignment also builds on the foundation that interventions are most effective when targeted to the population of interest. In this review, however, the cost component was not considered and the focus was mainly on DALYs averted by these interventions. Based on this analysis, several factors were identified that affect the performance of these interventions in averting DALYs. The following six are some of the factors that need due consideration by programmers during the implementation of prevention or control activities at national or subnational levels: i) implementation coverage of the interventions; ii) degree of combination of multiple interventions across public health functions; iii) subsidization or providing free interventions; iv) implementation in constituencies of high disease burden or prevalence v) early identification and management of risk factors and vi) applicability of some interventions to specific age cohorts only;.
The coverage and combination of interventions across public health functions demonstrated a significant impact in averting DALYs. Interventions implemented at higher coverage were more effective in averting DALYs than those with low coverage. When ITN was implemented as an intervention for malaria control at $50\%$ coverage, it averted 4,961,812 DALYs but averted 7,629,171 and 8,872,378 average yearly DALYs when coverage was increased to $80\%$ and $95\%$, respectively [48]. Implementing multiple health promotive, preventive and curative interventions in combination was found to avert more DALYs than a single intervention. The provision of only pneumococcal vaccine for the prevention of lower respiratory tract infections in children averted 3,465,493 DALYs. In contrast, a combination of breastfeeding promotion, pneumococcal vaccine, *Haemophilus influenzae* type B vaccine, use of cleaner liquid fuels, community-based case management and zinc supplementation in combination averted 23,081,510 DALYs [29]. In road safety, the combined implementation of helmet use, safety belt use, and breath tests for alcohol were effective in averting more DALYs than the implementation of a single intervention independently. Similarly, interventions were more effective in improving newborn survival and averting newborn deaths when combined than when considered separately. The combination of multiple promotive and preventive interventions is particularly relevant in repivoting health systems towards UHC as the implication is instituting deliberate efforts to shift service delivery to strengthening primary care [124, 125].
Another important observation noted was the effectiveness of similar interventions applied with the same coverage may differ depending on disease burden or the population exposed to the risk of a particular condition. Case management with artemisinin-based combination therapy with $95\%$ coverage averts 9,254,473 average yearly DALYs in West Africa, while the similar intervention with the same coverage averts only 5,886,159 average yearly DALYs in East Africa [48]. Effectiveness studies of some interventions primarily related to NCDs, have been performed in the context of high-income countries; thus, the question arises whether such studies apply to health care delivery settings in developing countries. For example, the higher the background incidence of a condition, the more effective a prevention or screening intervention is likely to be. Therefore, the relative incidence patterns for disease intervention in HICs and LMICs/LICs must be carefully considered. A good example is the relatively high incidence of cervical cancer in developing countries compared to developed countries. Interventions for cervical cancer prevention and screening are likely to be more cost-effective in developing countries than in developed countries, all else being equal. Therefore, the critical contributions varying contexts contribute to the effectiveness of different interventions, the invaluable role of determinants of health in contributing to health and well-being is underlined. In this regard, the social circumstances in a setting will influence service coverage, access and acceptability [126, 127].
The impact of price subsidies or provision of free interventions resulting in higher effectiveness was another important finding of this study. Price subsidies on fortified packaged infant cereals (F-PICs) were effective interventions for addressing dietary iron deficiency. It averted 268,301 DALYs at 50g/d when given free; averted 95,599 DALYs at $80\%$ discount and 60,840 at $50\%$ discount among individuals of poor socioeconomic status [28]. In the case of epilepsy and mental disorders, public financing for epilepsy treatment and the enhanced availability of medications demonstrated high effectiveness in averting the DALYs for these conditions [128]. In HIV prevention, sustained free HIV voluntary counselling and testing was more effective in averting DALYS than paid services. Relatedly, evidence has shown how using payment incentives supports positive patient and provider behaviour changes and how expanding insurance coverage opens up healthcare to many populations. Furthermore, social health determinants such as non-financial barriers related to indirect costs (e.g., food, fear of loss of income), also influence acceptability and access e.g., treatment with dignity and non-discrimination, and ought to be considered [129].
In preventing NCDs, combination drug treatment for individuals with a low absolute risk of a cardiovascular event was more effective in averting more DALYs than when treatment was given to individuals with high absolute risk. While the $35\%$ absolute risk averts only 264 716 DALYs per year, combination drug treatment for individuals with a $5\%$ absolute risk averts 404 684 DALYs per year, indicating the need for early risk detection and its management.
This review also identified that these interventions are not equally effective across all age cohorts as some of the interventions are limited to some population groups. In contrast, many others demonstrated effectiveness across the life course. Taking malaria-related interventions as an example, intermittent preventive treatment in infants (IPTi) and intermittent preventive treatment in pregnant women (IPTp) were some of the specific interventions that demonstrated effectiveness in infants and pregnant women, respectively. On the other hand, insecticide-treated bed nets (ITNs) and indoor residual spraying (IRS) were effective interventions for malaria control across all age cohorts. Regulations for tobacco use, road transport use and healthy diets such as salt, sugar and unhealthy fat consumption were interventions that demonstrated effectiveness across several age cohorts. Thus, the important role of multi-sectoral collaborations, both within and outside of the health sector, also influences effective coverage and needs to be closely coordinated and regulated [130]. The indirect effects of strengthened multi-sectoral collaborations have been shown to provide additional benefits of better health risk preparedness [131].
Most interventions across public health functions that averted more DALYs were easily implementable at community or lower levels of care. For example, condom promotion and Voluntary HIV counselling and Testing for HIV prevention, antiretroviral therapy for the management of HIV/AIDS were effective when implemented at the community and lower levels of care. Routine blood pressure and blood glucose screening and treatment for the prevention of cardiovascular diseases, and instituting school-based psychological interventions to prevent the onset of depression among students were effective interventions that could be easily implemented at the community and lower levels of care to avert thousands of DALYs. This means countries can improve access to effective interventions without causing unnecessary financial hardship to their population, especially those residing in remote areas. Most interventions assessed were effective; however, a few interventions showed low effectiveness. For example, Zehtab’s et al. [ 98], in analysing the cost-effectiveness of breast cancer screening using mammography in 35-69year-old women in an Iranian setting, found low effectiveness of annual breast cancer screening due to the short interval of the screening program. The focus on community-level interventions places communities at the centre, ensuring what the systems provide is aligned to population needs and that the care provided is person-centred. However, there need to be more efforts to ensure no groups are left behind, particularly the disadvantaged and that the equity lens used mirrors the social gradient and the complexity of social stratification of communities [132, 133].
Even though this study generated effective interventions for the different age cohorts and public health functions, ranking the interventions using the averted DALYs was difficult due to varied methodologies across studies. To allow comparison across the interventions and ranking of the most effective ones using DALYs, an extra step involving modelling all the identified interventions and standardizing the methodology, including assumptions, would be required. Therefore, further analysis using standardised methods would enhance the potential generalizability and applicability of the study findings to comparable settings within the African continent.
Our analysis of the burden of disease indicated that, despite the declining trends in CMNN conditions and increasing trend of non-communicable diseases in the African Region, neonatal conditions and communicable diseases remain the major contributors of DALYs. It also indicated that the burden distribution varies across the life course, with some conditions dominating specific age cohorts more than others. However, some conditions such as communicable diseases (lower respiratory tract infections, diarrheal diseases, malaria, HIV and tuberculosis) are not limited to specific age groups but remain challenges across several age cohorts. In the burden of disease study of 2019, neonatal disorders were the leading contributors ($32\%$) of global DALY in the under ten age group. Our findings also confirmed that neonatal conditions remain the leading contributors of DALYs in neonates and early childhood. This was reflected in the high neonatal mortality rate of 27 deaths for 1000 live births in sub–Saharan Africa, making it the highest in the world [134]. Though the contribution of NCD DALYs in the African region was only $10\%$, three NCD conditions (cerebrovascular diseases, cancers and diabetes mellitus) were responsible for almost $35.5\%$ of the DALYs among the cohort of elderly. Nonetheless, it should also be noted the contribution of communicable diseases to the development of NCDs and specifical cancers in the African region as up to one-fourth of cancers in developing countries are associated with chronic infections such as Hepatitis B Virus (HBV), Human Papillomavirus (HPV) and Helicobacter Pylori (H pylori) infections [123]. Hence, countries that strive to achieve universal health coverage need to focus their prioritization and resources on conditions that contribute to the largest contributors of DALYs.
The analysis of disease burden also showed that some conditions may not rank as leading causes of DALYs when considered individually but may cause significant disease burden and become important public health problems when grouped, further exposing risk for health security. Significant among these groups were maternal conditions, NTDs, vaccine-preventable and epidemic-prone diseases. According the disease burden data base, approximately $46.6\%$ of the 196,471 global maternal deaths in 2019 were in the African Region. The four leading conditions (Maternal haemorrhage, abortion and miscarriage, hypertensive disorders, sepsis and infection were responsible for almost $54\%$ of deaths that occurred in the region. With 542 deaths per 100 000 live births, sub-Saharan Africa has the highest maternal mortality ratio (MMR) in the world, but the main causes of death were not among the top contributors of DALYs in the region. Even though they didn’t present as leading causes individually, a group of 20 diseases designated by WHO as neglected tropical diseases (NTDs) [135] were responsible for an estimated 16.5 million DALYs in 2019, with over $39\%$ of the DALYs contributed by the Africa region. Likewise, vaccine-preventable diseases such as whooping cough and measles were responsible for 9,676,707 DALYs in the cohort of early childhood. The impact of these conditions on the poor communities and their amenability to effective interventions that could lead to their elimination is a valid justification for countries to include them as priority diseases when aiming to achieve UHC and Health security. Similarly, conditions of public health concern such as sickle cell disorders and epilepsy fuel the disease burden disproportionately, with the poor and vulnerable population is most affected due to unavailability and inaccessibility of services. Some conditions such as Invasive Non-typhoidal *Salmonella have* shown an increasing trend of almost $40\%$ over the years in late childhood; however, studies that measure the effectiveness of interventions using DALYs averted for this condition were not found. The inclusion of effective interventions for these conditions in a health benefits package, say instituting public financing for epilepsy treatment; mass drug administration for NTDs; vaccination of children, or scaling up mandatory newborn screening for sickle cell disorder; serves the purpose of UHC and Health security for most African nations. There are also diseases that tally low as the DALYs contributed is a small fraction of the total (Ebola, 195,394.11, $100\%$ in Africa, Yellow fever, 249,956.6, $86\%$ in Africa) that need due consideration as their impact is catastrophic when it happens. To ensure the soundness of their health security and prepare for any shock that could disrupt the health system, countries also need to consider epidemic-prone diseases such as Ebola, haemorrhagic fevers etc. in addition to the leading causes of DALYs in their prioritization.
## Limitations
The findings for this study were anchored on published articles of a high standard and with a defined methodology; therefore, there was a marked under-representation of publications for the elderly age cohort and for some public health functions like the rehabilitation and palliative services, especially in the African setting. These are important areas that future research directions and countries need to think about as we move toward re-pivoting health systems.
This review used DALYs as its measure of effectiveness, and the cost component of cost-effectiveness was not considered within the scope of this study. Cost-effectiveness evaluation of an intervention is critical as it gives a more accurate depiction of whether an intervention will be feasible within a context or not. The affordability of a selected effective intervention would require careful consideration within the African context and the availability of the recommended interventions. While some interventions may be effective, they may not be cost-effective and affordable for a recommendation. Hence a further review that examines the full economic evaluation of the evidence is highly recommended to ensure recommended interventions are both cost-effective and affordable.
Studies showing whether an intervention is effective seldom cover the entire potential beneficiary population, and service providers in the private sector are often not recorded. Most studies, especially in developing countries, were purely modelling studies. Whereas heuristic extrapolation is an important first analytical step, indicating the need for more direct and realistic studies of benefit would be derived from clinical evaluation studies of interventions, specifically tailored to developing countries’ needs and conditions, including controlled clinical trials where possible.
## Conclusion
The evidence from the data generated clearly indicates that disease burden distribution varies across the life course, with some conditions dominating specific age cohorts more than others. Future research would benefit in further deciphering this pattern across sub-regions and countries, for tailored and targeted analysis Despite an increasing trend in NCDs, infectious diseases and neonatal conditions remain the major overall DALY contributors in the African region. It also demonstrated that effective health promotion, disease prevention and curative interventions exist for the major conditions contributing to Africa’s disease burden across the life course. Whilst some interventions have already been implemented at scale, and others have great potential to achieve a significant reduction in disease burden when customised to the country’s unique setting.
Therefore, to benefit from these effective interventions and achieve the triple aims of UHC, health security and coverage of health determinants, national and regional authorities and planners need to clearly identify the major conditions across the life course and prioritize interventions according to population needs and characteristics. Furthermore, they need to mobilize resources and engage multiple health and health-related stakeholders for a combination of interventions across public health, achieve high coverages, strive to identify and manage risk factors early, prioritize high disease burden areas and high-risk populations for the greatest possible impact during the implementation phase. With this available evidence aimed at UHC, countries can design and customise effective health benefits packages and leverage each other’s resources based on their unique socio-economic, demographic, and geo-political standing to benefit their populations.
## References
1. Smith PG, Morrow RH, Ross DA. *Field Trials of Health Interventions: A Toolbox* (2015.0)
2. Bertram MY, Edejer TTT. **Introduction to the Special Issue on “The World Health Organization Choosing Interventions That Are Cost-Effective (WHO-CHOICE) Update.”**. *Int J Health Policy Manag* (2021.0) 1. PMID: 32610734
3. Jamison Dean T, Gelband Hellen, Horton, Susan Jha, Prabhat Laxminarayan, Ramanan Mock, Charles N. *Improving Health and Reducing Poverty* (2017.0) **9**
4. 4World Malaria Report 2021 | Medicines for Malaria Venture [Internet]. [cited 2022 Apr 14]. Available from: https://www.mmv.org/newsroom/publications/world-malaria-report-2021?gclid=CjwKCAjw6dmSBhBkEiwA_W-EoJx3xSnSnXQlOH-xG_LIKpO7WgDlv9BOSuFWpeVlTjlM2ELnur9olBoCdWgQAvD_BwE
5. **NCD Countdown 2030: efficient pathways and strategic investments to accelerate progress towards the Sustainable Development Goal target 3.4 in low-income and middle-income countries**. *The Lancet* (2022.0) **399** 1266-78. DOI: 10.1016/S0140-6736(21)02347-3
6. Black RE, Laxminarayan R, Temmerman M. *Reproductive, Maternal, Newborn, and Child Health: Disease Control Priorities, Third Edition (Volume 2)* (2016.0)
7. Murray CJ, Acharya AK. **Understanding DALYs (disability-adjusted life years)**. *J Health Econ* (1997.0) **16** 703-30. DOI: 10.1016/s0167-6296(97)00004-0
8. Haagsma J. A., Polinder S., Cassini A., Colzani E., Havelaar A. H.. **Review of disability weight studies: comparison of methodological choices and values**. *Population health metrics* (2014.0) **12** 1-14. PMID: 24479861
9. 9Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2019 (GBD 2019) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2020. Available from http://ghdx.healthdata.org/gbd-results-tool.. *Global Burden of Disease Study 2019 (GBD 2019) Results* (2020.0)
10. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. *BMJ* (2021.0) 29-n71
11. Group PRISMA-P, Moher D, Shamseer L, Clarke M, Ghersi D, Liberati A. **Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement**. *Syst Rev* (2015.0) **4** 1. DOI: 10.1186/2046-4053-4-1
12. Ofman JJ, Sullivan SD, Neumann PJ, Chiou CF, Henning JM, Wade SW. **Examining the value and quality of health economic analyses: implications of utilizing the QHES**. *J Manag Care Pharm JMCP* (2003.0) **9** 53-61. DOI: 10.18553/jmcp.2003.9.1.53
13. Goodman DM, Ramaswamy R, Jeuland M, Srofenyoh EK, Engmann CM, Olufolabi AJ, Abimiku AG. **The cost effectiveness of a quality improvement program to reduce maternal and fetal mortality in a regional referral hospital in Accra, Ghana**. *PLOS ONE* (2017.0) **12** e0180929. DOI: 10.1371/journal.pone.0180929
14. McLaren ZM, Sharp A, Hessburg JP, Sarvestani AS, Parker E, Akazili J. **Cost effectiveness of medical devices to diagnose pre-eclampsia in low-resource settings**. *Dev Eng* (2017.0) **2** 99-106. DOI: 10.1016/j.deveng.2017.06.002
15. Memirie ST, Tolla MT, Desalegn D, Hailemariam M, Norheim OF, Verguet S. *May 1* (2019.0) **34** 289-97
16. Downing J, El Ayadi A, Miller S, Butrick E, Mkumba G, Magwali T. **Cost-effectiveness of the non-pneumatic anti-shock garment (NASG): evidence from a cluster randomized controlled trial in Zambia and Zimbabwe**. *BMC Health Serv Res* (2015.0) **15** 37. DOI: 10.1186/s12913-015-0694-6
17. Bahamondes L, Bottura BF, Bahamondes MV, Goncalves MP, Correia VM, Espejo-Arce X. **Estimated disability-adjusted life years averted by long-term provision of long acting contraceptive methods in a Brazilian clinic**. *Hum Reprod* (2014.0) **29** 2163-70. DOI: 10.1093/humrep/deu191
18. Joshi BN, Shetty SS, Moray KV, Sachin O, Chaurasia H. **Cost-effectiveness of uterine balloon tamponade devices in managing atonic post-partum hemorrhage at public health facilities in India. Ferket B, editor**. *PLOS ONE* (2021.0) **16** e0256271. PMID: 34407132
19. Sutherland T, Downing J, Miller S, Bishai DM, Butrick E, Fathalla MMF, Young RC. **Use of the Non-Pneumatic Anti-Shock Garment (NASG) for Life-Threatening Obstetric Hemorrhage: A Cost-Effectiveness Analysis in Egypt and Nigeria**. *PLoS ONE* (2013.0) **8** e62282. DOI: 10.1371/journal.pone.0062282
20. Dalziel K, Segal L, Katz R. **Cost-effectiveness of mandatory folate fortification v. other options for the prevention of neural tube defects: results from Australia and New Zealand**. *Public Health Nutr* (2010.0) **13** 566-78. DOI: 10.1017/S1368980009991418
21. Tobe RG, Martin GR, Li F, Moriichi A, Wu B, Mori R. **Cost-effectiveness analysis of neonatal screening of critical congenital heart defects in China**. *Medicine (Baltimore)* (2017.0) **96** e8683. DOI: 10.1097/MD.0000000000008683
22. Ali A, Nudel J, Heberle CR, Santorino D, Olson KR, Hur C. **Cost effectiveness of a novel device for improving resuscitation of apneic newborns**. *BMC Pediatr* (2020.0) **20** 46. DOI: 10.1186/s12887-020-1925-5
23. Willcox M, Harrison H, Asiedu A, Nelson A, Gomez P, LeFevre A. **Incremental cost and cost-effectiveness of low-dose, high-frequency training in basic emergency obstetric and newborn care as compared to status quo: part of a cluster-randomized training intervention evaluation in Ghana**. *Glob Health* (2017.0) **13** 88
24. Ahmed N, Giorgakoudi K, Usuf E, Okomo U, Clarke E, Kampmann B. **Potential cost-effectiveness of a maternal Group B streptococcal vaccine in The Gambia**. *Vaccine* (2020.0) **38** 3096-104. DOI: 10.1016/j.vaccine.2020.02.071
25. Chow J, Darley S, Laxminarayan R. *Cost-effectiveness of Disease Interventions in India* (2007.0) **131**
26. Chola L, Fadnes LT, Engebretsen IMS, Nkonki L, Nankabirwa V, Sommerfelt H, van Wouwe J. **Cost-Effectiveness of Peer Counselling for the Promotion of Exclusive Breastfeeding in Uganda**. *PLOS ONE* (2015.0) **10** e0142718. DOI: 10.1371/journal.pone.0142718
27. Niessen L.. **Comparative impact assessment of child pneumonia interventions**. *Bull World Health Organ* (2009.0) **87** 472-80. DOI: 10.2471/blt.08.050872
28. Goodman C.. **The cost-effectiveness of improving malaria home management: shopkeeper training in rural Kenya**. *Health Policy Plan* (2006.0) **21** 275-88. DOI: 10.1093/heapol/czl011
29. Plessow R, Arora NK, Brunner B, Wieser S, Reyes ONE. **Cost-Effectiveness of Price Subsidies on Fortified Packaged Infant Cereals in Reducing Iron Deficiency Anemia in 6-23-Month-Old-Children in Urban India**. *Apr 13* (2016.0) **11** e0152800. DOI: 10.1371/journal.pone.0152800
30. Wieser S, Brunner B, Tzogiou C, Plessow R, Zimmermann MB, Farebrother J. **Reducing micronutrient deficiencies in Pakistani children: are subsidies on fortified complementary foods cost-effective?**. *Public Health Nutr* (2018.0) **21** 2893-906. DOI: 10.1017/S1368980018001660
31. Trenouth L, Colbourn T, Fenn B, Pietzsch S, Myatt M, Puett C. **The cost of preventing undernutrition: cost, cost-efficiency and cost-effectiveness of three cash-based interventions on nutrition outcomes in Dadu, Pakistan**. *Health Policy Plan* (2018.0) **33** 743-54. DOI: 10.1093/heapol/czy045
32. Pasricha SR, Gheorghe A, Sakr-Ashour F, Arcot A, Neufeld L, Murray-Kolb LE. **Net benefit and cost-effectiveness of universal iron-containing multiple micronutrient powders for young children in 78 countries: a microsimulation study**. *Lancet Glob Health* (2020.0) **8** e1071-80. DOI: 10.1016/S2214-109X(20)30240-0
33. Svefors P, Selling KE, Shaheen R, Khan AI, Persson LÅ, Lindholm L, Wieringa F. **Cost-effectiveness of prenatal food and micronutrient interventions on under-five mortality and stunting: Analysis of data from the MINIMat randomized trial, Bangladesh**. *PLOS ONE* (2018.0) **13** e0191260. DOI: 10.1371/journal.pone.0191260
34. Kasteng F, Murray J, Cousens S, Sarrassat S, Steel J, Meda N. **Cost-effectiveness and economies of scale of a mass radio campaign to promote household life-saving practices in Burkina Faso**. *BMJ Glob Health* (2018.0) **3** e000809. DOI: 10.1136/bmjgh-2018-000809
35. Sacks G, Veerman JL, Moodie M, Swinburn B. **Traffic-light’ nutrition labelling and ‘junk-food’ tax: a modelled comparison of cost-effectiveness for obesity prevention**. *Int J Obes* (2011.0) **35** 1001-9. DOI: 10.1038/ijo.2010.228
36. Sinha A, Levine O, Knoll MD, Muhib F, Lieu TA. **Cost-effectiveness of pneumococcal conjugate vaccination in the prevention of child mortality: an international economic analysis**. *The Lancet* (2007.0) **369** 389-96. DOI: 10.1016/S0140-6736(07)60195-0
37. Tasslimi A, Nakamura MM, Levine O, Knoll MD, Russell LB, Sinha A. **Cost effectiveness of child pneumococcal conjugate vaccination in GAVI-eligible countries.**. *Int Health* (2011.0) **3** 259-69. DOI: 10.1016/j.inhe.2011.08.003
38. Kim SY, Lee G, Goldie SJ. **Economic evaluation of pneumococcal conjugate vaccination in The Gambia**. *BMC Infect Dis* (2010.0) **10** 260. DOI: 10.1186/1471-2334-10-260
39. Khan AI, Levin A, Chao DL, DeRoeck D, Dimitrov DT, Khan JAM, Maheu-Giroux M. **The impact and cost-effectiveness of controlling cholera through the use of oral cholera vaccines in urban Bangladesh: A disease modeling and economic analysis**. *PLoS Negl Trop Dis* (2018.0) **12** e0006652. DOI: 10.1371/journal.pntd.0006652
40. Seo MK, Baker P, Ngo KNL. **Cost-effectiveness analysis of vaccinating children in Malawi with RTS,S vaccines in comparison with long-lasting insecticide-treated nets**. *Malar J* (2014.0) **13** 66. DOI: 10.1186/1475-2875-13-66
41. Ndeketa L, Mategula D, Terlouw DJ, Bar-Zeev N, Sauboin CJ, Biernaux S. **Cost-effectiveness and public health impact of RTS,S/AS01 E malaria vaccine in Malawi, using a Markov static model**. *Wellcome Open Res* (2020.0) **5** 260. DOI: 10.12688/wellcomeopenres.16224.2
42. Pecenka C, Usuf E, Hossain I, Sambou S, Vodicka E, Atherly D. **Pneumococcal conjugate vaccination in The Gambia: health impact, cost effectiveness and budget implications**. *BMJ Glob Health* (2021.0) **6** e007211. DOI: 10.1136/bmjgh-2021-007211
43. Sinha A, Constenla D, Valencia JE, O’Loughlin R, Gomez E, de la Hoz F. **Cost-effectiveness of pneumococcal conjugate vaccination in Latin America and the Caribbean: a regional analysis**. *Rev Panam Salud Publica Pan Am J Public Health* (2008.0) **24** 304-13. DOI: 10.1590/s1020-49892008001100002
44. Goudet S, Jayaraman A, Chanani S, Osrin D, Devleesschauwer B, Bogin B, Gopichandran V. **Cost effectiveness of a community-based prevention and treatment of acute malnutrition programme in Mumbai slums, India**. *PLOS ONE* (2018.0) **9** e0205688. DOI: 10.1371/journal.pone.0205688
45. Alonzo González M, Menéndez C, Font F, Kahigwa E, Kimario J, Mshinda H. **Cost-effectiveness of iron supplementation and malaria chemoprophylaxis in the prevention of anaemia and malaria among Tanzanian infants**. *Bull World Health Organ* (2000.0) **78** 97-107. PMID: 10686744
46. Tan-Torres Edejer T, Aikins M, Black R, Wolfson L, Hutubessy R, Evans DB. **Cost effectiveness analysis of strategies for child health in developing countries**. *BMJ* (2005.0) **331** 1177. DOI: 10.1136/bmj.38652.550278.7C
47. Hutton G.. **Cost-effectiveness of malaria internittent preventive treatment in infants (IPTi) in Mozambique and the United Republic of Tanzania**. *Bull World Health Organ* (2009.0) **87** 123-9. PMID: 19274364
48. Stuckey EM, Stevenson J, Galactionova K, Baidjoe AY, Bousema T, Odongo W, Snounou G. **Modeling the Cost Effectiveness of Malaria Control Interventions in the Highlands of Western Kenya**. *PLoS ONE* (2014.0) **9** e107700. DOI: 10.1371/journal.pone.0107700
49. Morel CM, Lauer JA, Evans DB. **Cost effectiveness analysis of strategies to combat malaria in developing countries**. *BMJ* (2005.0) **331** 1299. DOI: 10.1136/bmj.38639.702384.AE
50. Hansen KS, Ndyomugyenyi R, Magnussen P, Clarke SE. **Cost-effectiveness analysis of three health interventions to prevent malaria in pregnancy in an area of low transmission in Uganda**. *Int Health* (2012.0) **4** 38-46. DOI: 10.1016/j.inhe.2011.10.001
51. Gilmartin C, Nonvignon J, Cairns M, Milligan P, Bocoum F, Winskill P. **Seasonal malaria chemoprevention in the Sahel subregion of Africa: a cost-effectiveness and cost-savings analysis**. *Lancet Glob Health* (2021.0) **9** e199-208. DOI: 10.1016/S2214-109X(20)30475-7
52. Mueller DH, Wiseman V, Bakusa D, Morgah K, Daré A, Tchamdja P. **Cost-effectiveness analysis of insecticide-treated net distribution as part of the Togo Integrated Child Health Campaign**. *Malar J* (2008.0) **7** 73. DOI: 10.1186/1475-2875-7-73
53. Hailu A, Lindtjørn B, Deressa W, Gari T, Loha E, Robberstad B. **Cost-effectiveness of a combined intervention of long lasting insecticidal nets and indoor residual spraying compared with each intervention alone for malaria prevention in Ethiopia**. *Cost Eff Resour Alloc* (2018.0) **16** 61. DOI: 10.1186/s12962-018-0164-1
54. Maheu-Giroux M, Castro MC. **Cost-effectiveness of larviciding for urban malaria control in Tanzania**. *Malar J* (2014.0) **13** 477. DOI: 10.1186/1475-2875-13-477
55. Alonso S, Chaccour CJ, Wagman J, Candrinho B, Muthoni R, Saifodine A. **Cost and cost-effectiveness of indoor residual spraying with pirimiphos-methyl in a high malaria transmission district of Mozambique with high access to standard insecticide-treated nets**. *Malar J* (2021.0) **20** 143. DOI: 10.1186/s12936-021-03687-1
56. Pillarisetti A, Jamison DT, Smith KR, Mock CN, Nugent R, Kobusingye O, Smith KR. *Injury Prevention and Environmental Health* (2017.0)
57. Zhang S, Incardona B, Qazi SA, Stenberg K, Campbell H, Nair H. **Cost–effectiveness analysis of revised WHO guidelines for management of childhood pneumonia in 74 Countdown countries**. *J Glob Health* (2017.0) **7** 010409. DOI: 10.7189/jogh.07.010409
58. Kortz TB, Herzel B, Marseille E, Kahn JG. **Bubble continuous positive airway pressure in the treatment of severe paediatric pneumonia in Malawi: a cost-effectiveness analysis**. *BMJ Open* (2017.0) **7** e015344. DOI: 10.1136/bmjopen-2016-015344
59. Tozan Y, Klein EY, Darley S, Panicker R, Laxminarayan R, Breman JG. **Prereferral rectal artesunate for treatment of severe childhood malaria: a cost-effectiveness analysis**. *The Lancet* (2010.0) **376** 1910-5. DOI: 10.1016/S0140-6736(10)61460-2
60. Lubell Y, Riewpaiboon A, Dondorp AM, von Seidlein L, Mokuolu OA, Nansumba M. **Cost-effectiveness of parenteral artesunate for treating children with severe malaria in sub-Saharan Africa**. *Bull World Health Organ* (2011.0) **89** 504-12. DOI: 10.2471/BLT.11.085878
61. Moodie ML, Herbert JK, de Silva-Sanigorski AM, Mavoa HM, Keating CL, Carter RC. **The cost-effectiveness of a successful community-based obesity prevention program: the be active eat well program**. *Obes Silver Spring Md* (2013.0) **21** 2072-80. DOI: 10.1002/oby.20472
62. Ramachandran A, Manabe Y, Rajasingham R, Shah M. **Cost-effectiveness of CRAG-LFA screening for cryptococcal meningitis among people living with HIV in Uganda**. *BMC Infect Dis* (2017.0) **17** 225. DOI: 10.1186/s12879-017-2325-9
63. Hogan DR, Baltussen R, Hayashi C, Lauer JA, Salomon JA. **Cost effectiveness analysis of strategies to combat HIV/AIDS in developing countries**. *BMJ* (2005.0) **331** 1431-7. DOI: 10.1136/bmj.38643.368692.68
64. Aldridge RW, Iglesias D, Cáceres CF, Miranda JJ. **Determining a cost effective intervention response to HIV/AIDS in Peru**. *BMC Public Health* (2009.0) **9** 352. DOI: 10.1186/1471-2458-9-352
65. Dandona L, Kumar SP, Kumar GA, Dandona R. **Cost-effectiveness of HIV prevention interventions in Andhra Pradesh state of India**. *BMC Health Serv Res* (2010.0) **10** 117. PMID: 20459755
66. Sharma M, Farquhar C, Ying R, Krakowiak D, Kinuthia J, Osoti A. **Modeling the Cost-Effectiveness of Home-Based HIV Testing and Education (HOPE) for Pregnant Women and Their Male Partners in Nyanza Province, Kenya**. *JAIDS J Acquir Immune Defic Syndr* (2016.0) **72** S174-80. DOI: 10.1097/QAI.0000000000001057
67. Thielman NM, Chu HY, Ostermann J, Itemba DK, Mgonja A, Mtweve S. **Cost-Effectiveness of Free HIV Voluntary Counseling and Testing Through a Community-Based AIDS Service Organization in Northern Tanzania**. *Am J Public Health.* (2006.0) **96** 114-9. DOI: 10.2105/AJPH.2004.056796
68. Vassall A, Chandrashekar S, Pickles M, Beattie TS, Shetty G, Bhattacharjee P, Beck EJ. **Community Mobilisation and Empowerment Interventions as Part of HIV Prevention for Female Sex Workers in Southern India: A Cost-Effectiveness Analysis**. *PLoS ONE* (2014.0) **9** e110562. DOI: 10.1371/journal.pone.0110562
69. Holm AL, Veerman L, Cobiac L, Ekholm O, Diderichsen F, van Baal PHM. **Cost-Effectiveness of Preventive Interventions to Reduce Alcohol Consumption in Denmark**. *PLoS ONE* (2014.0) **9** e88041. DOI: 10.1371/journal.pone.0088041
70. Chisholm D, Naci H, Hyder AA, Tran NT, Peden M. **Cost effectiveness of strategies to combat road traffic injuries in sub-Saharan Africa and South East Asia: mathematical modelling study**. *BMJ* (2012.0) **344** e612-e612. DOI: 10.1136/bmj.e612
71. Ditsuwan V, Lennert Veerman J, Bertram M, Vos T. **Cost-Effectiveness of Interventions for Reducing Road Traffic Injuries Related to Driving under the Influence of Alcohol**. *Value Health* (2013.0) **16** 23-30. DOI: 10.1016/j.jval.2012.08.2209
72. Gortmaker SL, Long MW, Resch SC, Ward ZJ, Cradock AL, Barrett JL. **Cost Effectiveness of Childhood Obesity Interventions: Evidence and Methods for CHOICES**. *Am J Prev Med* (2015.0) **49** 102-11. DOI: 10.1016/j.amepre.2015.03.032
73. Galárraga O, Shah P, Wilson-Barthes M, Ayuku D, Braitstein P. **Cost and cost-effectiveness of voluntary medical male circumcision in street-connected youth: findings from an education-based pilot intervention in Eldoret, Kenya**. *AIDS Res Ther* (2018.0) **15** 24. DOI: 10.1186/s12981-018-0207-x
74. Quaife M, Terris‐Prestholt F, Eakle R, Cabrera Escobar MA, Kilbourne‐Brook M, Mvundura M. **The cost‐effectiveness of multi‐purpose**. *J Int AIDS Soc* (2018.0)
75. Vogelzang M, Terris-Prestholt F, Vickerman P, Delany-Moretlwe S, Travill D, Quaife M. **Cost-Effectiveness of HIV Pre-exposure Prophylaxis Among Heterosexual Men in South Africa: A Cost-Utility Modeling Analysis**. *Acquir Immune Defic Syndr* (2020.0) **84** 173-81. DOI: 10.1097/QAI.0000000000002327
76. Coffey P, Mvundura M, Nundy N, Kilbourne-Brook M. **Estimating the hypothetical dual health impact and cost-effectiveness of the Woman’s Condom in selected sub-Saharan African countries**. *Int J Womens Health* (2015.0) **271**. DOI: 10.2147/IJWH.S75040
77. Shin H, Jo Y, Chaisson RE, Turner K, Churchyard G, Dowdy DW. **Cost‐effectiveness of a 12 country‐intervention to scale up short course TB preventive therapy among people living with HIV**. *J Int AIDS Soc* (2020.0). PMID: 33107219
78. Bishai DM, Hyder AA. **Modeling the cost effectiveness of injury interventions in lower and middle income countries: opportunities and challenges**. *Cost Eff Resour Alloc* (2006.0) **4** 2. DOI: 10.1186/1478-7547-4-2
79. Gureje O, Chisholm D, Kola L, Lasebikan V, Saxena S. **Cost-effectiveness of an essential mental health intervention package in Nigeria**. *World Psychiatry* (2007.0) **104**. PMID: 17342226
80. Langley I, Lin HH, Egwaga S, Doulla B, Ku CC, Murray M. **Assessment of the patient, health system, and population effects of Xpert MTB/RIF and alternative diagnostics for tuberculosis in Tanzania: an integrated modelling approach**. *Lancet Glob Health* (2014.0) **2** e581-91. DOI: 10.1016/S2214-109X(14)70291-8
81. Mohan CI, Bishai D, Cavalcante S, Chaisson RE. *The cost-effectiveness of DOTS in urban Brazil.* **6**
82. Baltussen R, Floyd K, Dye C. **Cost effectiveness analysis of strategies for tuberculosis control in developing countries**. *BMJ* (2005.0) **331** 1364. DOI: 10.1136/bmj.38645.660093.68
83. Shah L, Rojas M, Mori O, Zamudio C, Kaufman JS, Otero L. **Cost-effectiveness of active case-finding of household contacts of pulmonary tuberculosis patients in a low HIV, tuberculosis-endemic urban area of Lima, Peru**. *Epidemiol Infect* (2017.0) **145** 1107-17. DOI: 10.1017/S0950268816003186
84. CHISHOLM D.. **Choosing cost-effective interventions in psychiatry: results from the CHOICE programme of the World Health Organization**. *World Psychiatry* (2005.0) **4** 37-44. PMID: 16633504
85. Chisholm D.. **Schizophrenia treatment in the developing world: an interregional and multinational cost-effectiveness analysis**. *Bull World Health Organ* (2008.0) **86** 542-51. DOI: 10.2471/blt.07.045377
86. Smit F, Lokkerbol J, Riper H, Majo MC, Boon B, Blankers M. **Modeling the Cost-Effectiveness of Health Care Systems for Alcohol Use Disorders: How Implementation of eHealth Interventions Improves Cost-Effectiveness.**. *J Med Internet Res* (2011.0) **13** e56. DOI: 10.2196/jmir.1694
87. Sampaio F, Barendregt JJ, Feldman I, Lee YY, Sawyer MG, Dadds MR. **Population cost-effectiveness of the Triple P parenting programme for the treatment of conduct disorder: an economic modelling study**. *Eur Child Adolesc Psychiatry* (2018.0) **27** 933-44. DOI: 10.1007/s00787-017-1100-1
88. Webb M, Fahimi S, Singh GM, Khatibzadeh S, Micha R, Powles J. **Cost effectiveness of a government supported policy strategy to decrease sodium intake: global analysis across 183 nations**. *BMJ* (2017.0) i6699. DOI: 10.1136/bmj.i6699
89. Hendriksen MA, Hoogenveen RT, Hoekstra J, Geleijnse JM, Boshuizen HC, van Raaij JM. **Potential effect of salt reduction in processed foods on health**. *Am J Clin Nutr* (2014.0) **99** 446-53. DOI: 10.3945/ajcn.113.062018
90. Zelle SG, Nyarko KM, Bosu WK, Aikins M, Niëns LM, Lauer JA. **Costs, effects and cost-effectiveness of breast cancer control in Ghana**. *Trop Med Int Health TM IH* (2012.0) **17** 1031-43. DOI: 10.1111/j.1365-3156.2012.03021.x
91. Suraratdecha C, Stuart RM, Manopaiboon C, Green D, Lertpiriyasuwat C, Wilson DP. **Cost and cost-effectiveness analysis of pre-exposure prophylaxis among men who have sex with men in two hospitals in Thailand**. *J Int AIDS Soc* (2018.0) **21** e25129. DOI: 10.1002/jia2.25129
92. Terris-Prestholt F.. **Is antenatal syphilis screening still cost effective in sub-Saharan Africa**. *Sex Transm Infect* (2003.0) **79** 375-81. DOI: 10.1136/sti.79.5.375
93. Azadi M, Bishai DM, Dowdy DW, Moulton LH, Cavalcante S, Saraceni V. *Cost-Effectiveness of Tuberculosis Screening and Isoniazid Treatment in the TB/HIV in Rio de Janeiro (THRio) Study* (2017.0) **11**
94. Kim JJ, Campos NG, O’Shea M, Diaz M, Mutyaba I. **Model-Based Impact and Cost-Effectiveness of Cervical Cancer Prevention in Sub-Saharan Africa**. *Vaccine* (2013.0) **31** F60-72. DOI: 10.1016/j.vaccine.2012.07.093
95. Woo PPS, Kim JJ, Leung GM. **What Is the Most Cost-Effective Population-Based Cancer Screening Program for Chinese Women?**. *J Clin Oncol* (2007.0) **25** 617-24. DOI: 10.1200/JCO.2006.06.0210
96. Portnoy A, Abbas K, Sweet S, Kim JJ, Jit M. **Projections of human papillomavirus (HPV) vaccination impact in Ethiopia, India, Nigeria and Pakistan: a comparative modelling study**. *BMJ Glob Health* (2021.0) **6** e006940. DOI: 10.1136/bmjgh-2021-006940
97. Davidović M, Zielonke N, Lansdorp-Vogelaar I, Segnan N, de Koning HJ, Heijnsdijk EAM. **Disability-Adjusted Life Years Averted Versus Quality-Adjusted Life Years Gained: A Model Analysis for Breast Cancer Screening**. *Value Health* (2021.0) **24** 353-60. DOI: 10.1016/j.jval.2020.10.018
98. Zehtab N, Jafari M, Barooni M, Nakhaee N, Goudarzi R, Zadeh MHL. **Cost-Effectiveness Analysis of Breast Cancer Screening in Rural Iran.**. *Asian Pac J Cancer Prev* (2016.0) **17** 609-14. DOI: 10.7314/apjcp.2016.17.2.609
99. Finkelstein EA, Krishnan A, Naheed A, Jehan I, de Silva HA, Gandhi M. **Budget impact and cost-effectiveness analyses of the COBRA-BPS multicomponent hypertension management programme in rural communities in Bangladesh, Pakistan, and Sri Lanka**. *Lancet Glob Health* (2021.0) **9** e660-7. DOI: 10.1016/S2214-109X(21)00033-4
100. Krishnan A, Finkelstein EA, Kallestrup P, Karki A, Olsen MH, Neupane D. **Cost-effectiveness and budget impact of the community-based management of hypertension in Nepal study (COBIN): a retrospective analysis**. *Lancet Glob Health* (2019.0) **7** e1367-74. DOI: 10.1016/S2214-109X(19)30338-9
101. Basu S, Flood D, Geldsetzer P, Theilmann M, Marcus ME, Ebert C. **Estimated effect of increased diagnosis, treatment, and control of diabetes and its associated cardiovascular risk factors among low-income and middle-income countries: a microsimulation model**. *Lancet Glob Health* (2021.0) **9** e1539-52. DOI: 10.1016/S2214-109X(21)00340-5
102. Cadilhac DA, Magnus A, Sheppard L, Cumming TB, Pearce DC, Carter R. **The societal benefits of reducing six behavioural risk factors: an economic modelling study from Australia**. *BMC Public Health* (2011.0) **11** 483. DOI: 10.1186/1471-2458-11-483
103. Angell B, Lung T, Praveen D, Maharani A, Sujarwoto S, Palagyi A. **Cost-effectiveness of a mobile technology-enabled primary care intervention for cardiovascular disease risk management in rural Indonesia**. *Health Policy Plan.* (2021.0) **36** 435-43. DOI: 10.1093/heapol/czab025
104. Kasaie P, Weir B, Schnure M, Dun C, Pennington J, Teng Y. **Integrated screening and treatment services for HIV, hypertension and diabetes in Kenya: assessing the epidemiological impact and cost‐effectiveness from a national and regional perspective**. *J Int AIDS Soc* (2020.0) **23**
105. Phillips AN, Bansi-Matharu L, Cambiano V, Ehrenkranz P, Serenata C, Venter F. **The potential role of long-acting injectable cabotegravir–rilpivirine in the treatment of HIV in sub-Saharan Africa: a modelling analysis**. *Lancet Glob Health* (2021.0) **9** e620-7. DOI: 10.1016/S2214-109X(21)00025-5
106. Sun D, Dorman S, Shah M, Manabe YC, Mischka V, Nicol MP. *Cost-Utility of Lateral-Flow Urine Lipoarabinomannan for Tuberculosis Diagnosis in HIV-infected African Adults* (2014.0) **13**
107. Shah M, Dowdy D, Joloba M, Ssengooba W, Manabe YC, Ellner J. **Cost-effectiveness of novel algorithms for rapid diagnosis of tuberculosis in HIV-infected individuals in Uganda**. *AIDS* (2013.0) **27** 2883-92. DOI: 10.1097/QAD.0000000000000008
108. Hunchangsith P, Barendregt JJ, Vos T, Bertram M. **Cost-Effectiveness of Various Tuberculosis Control Strategies in Thailand**. *Value Health* (2012.0) **15** S50-5. DOI: 10.1016/j.jval.2011.11.006
109. Lu X, Smare C, Kambili C, El Khoury AC, Wolfson LJ. **Health outcomes of bedaquiline in the treatment of multidrug-resistant tuberculosis in selected high burden countries**. *BMC Health Serv Res* (2017.0) **17** 87. DOI: 10.1186/s12913-016-1931-3
110. Ha DA, Chisholm D. **Cost-effectiveness analysis of interventions to prevent cardiovascular disease in Vietnam**. *Health Policy Plan* (2011.0) **26** 210-22. DOI: 10.1093/heapol/czq045
111. Pham QD, Wilson DP, Kerr CC, Shattock AJ, Do HM, Duong AT, Chung MH. **Estimating the Cost-Effectiveness of HIV Prevention Programmes in Vietnam, 2006–2010: A Modelling Study**. *PLOS ONE.* (2015.0) **10** e0133171. DOI: 10.1371/journal.pone.0133171
112. Jensen J, Saxe H, Denver S. **Cost-Effectiveness of a New Nordic Diet as a Strategy for Health Promotion**. *Int J Environ Res Public Health* (2015.0) **12** 7370-91. DOI: 10.3390/ijerph120707370
113. Etchie TO, Sivanesan S, Adewuyi GO, Krishnamurthi K, Rao PS, Etchie AT. **The health burden and economic costs averted by ambient PM2.5 pollution reductions in Nagpur, India**. *Environ Int* (2017.0) **102** 145-56. DOI: 10.1016/j.envint.2017.02.010
114. Tolla MT, Norheim OF, Memirie ST, Abdisa SG, Ababulgu A, Jerene D. **Prevention and treatment of cardiovascular disease in Ethiopia: a cost-effectiveness analysis.**. *Cost Eff Resour Alloc* (2016.0) **14** 10. DOI: 10.1186/s12962-016-0059-y
115. Pearson CAB, Bozzani F, Procter SR, Davies NG, Huda M, Jensen HT. **COVID-19 vaccination in Sindh Province, Pakistan: A modelling study of health impact and cost-effectiveness. Nichols BE, editor**. *PLOS Med* (2021.0) **18** e1003815. PMID: 34606520
116. Stanciole AE, Ortegon M, Chisholm D, Lauer JA. **Cost effectiveness of strategies to combat chronic obstructive pulmonary disease and asthma in sub-Saharan Africa and South East Asia: mathematical modelling study**. *BMJ* (2012.0) **344** e608-e608. DOI: 10.1136/bmj.e608
117. Taylor RS, Sadler S, Dalal HM, Warren FC, Jolly K, Davis RC. **The cost effectiveness of REACH-HF and home-based cardiac rehabilitation compared with the usual medical care for heart failure with reduced ejection fraction: A decision model-based analysis**. *Eur J Prev Cardiol* (2019.0) **26** 1252-61. DOI: 10.1177/2047487319833507
118. Pirhonen L, Gyllensten H, Olofsson EH, Fors A, Ali L, Ekman I. **The cost-effectiveness of person-centred care provided to patients with chronic heart failure and/or chronic obstructive pulmonary disease**. *Health Policy OPEN* (2020.0) **1** 100005
119. Sahlen KG, Boman K, Brännström M. **A cost-effectiveness study of person-centered integrated heart failure and palliative home care: Based on a randomized controlled trial**. *Palliat Med* (2016.0) **30** 296-302. DOI: 10.1177/0269216315618544
120. Maru S, Byrnes J, Carrington MJ, Chan YK, Thompson DR, Stewart S. **Cost-effectiveness of home versus clinic-based management of chronic heart failure: Extended follow-up of a pragmatic, multicentre randomized trial cohort—The WHICH? study (Which Heart Failure Intervention Is Most Cost-Effective & Consumer Friendly in Reducing Hospital Care)**. *Int J Cardiol* (2015.0) **201** 368-75. PMID: 26310979
121. 121WHO. Universal health coverage/UHC Compendium/. 2021.. *Universal health coverage/UHC Compendium/* (2021.0)
122. AFRO. *Digital Menu of Essential Interventions* (2021.0)
123. Jamison DT, Breman JG, Measham AR, Alleyne G, Claeson M, Evans DB. *In: Disease Control Priorities in Developing Countries (2nd Edition)* (2006.0) 569-90
124. Verrecchia R., Thompson R., Yates R.. **Universal health coverage and public health: a truly sustainable approach**. (2019.0) **4** e10-e11. DOI: 10.1016/S2468-2667(18)30264-0
125. Sanders D, Nandi S, Labonté R, Vance C, Van Damme W. **From primary health care to universal health coverage–one step forward and two steps back**. *Lancet* (2019.0) **394** 619-20. DOI: 10.1016/S0140-6736(19)31831-8
126. Vega J., Frenz P.. **Integrating social determinants of health in the universal health coverage monitoring framework**. (2013.0) **34** 468-472. PMID: 24569977
127. Valentine N.B., Bonsel G.J.. **Exploring models for the roles of health systems’ responsiveness and social determinants in explaining universal health coverage and health outcomes**. (2016.0) **9** 29329. DOI: 10.3402/gha.v9.29329
128. Megiddo I, Colson A, Chisholm D, Dua T, Nandi A, Laxminarayan R. **Health and economic benefits of public financing of epilepsy treatment in India: An agent‐based simulation model**. *Epilepsia* (2016.0) **57** 464-74. DOI: 10.1111/epi.13294
129. McCarthy D., Klein S.. *The triple aim journey: improving population health and patients’ experience of care, while reducing costs* (2010.0)
130. 130World Bank Group, 2019. Lessons Learned in Financing Rapid Response to Recent Epidemics in West and Central Africa: A Qualitative Study. World Bank.. (2019.0)
131. Alam U., Nabyonga-Orem J., Mohammed A., Malac D.R., Nkengasong J.N., Moeti M.R.. **Redesigning health systems for global heath security**. (2021.0) **9** e393-e394. DOI: 10.1016/S2214-109X(20)30545-3
132. Christine Masong M., Ozano K., Tagne M.S., Tchoffo M.N., Ngang S., Thomson R.. **Achieving equity in UHC interventions: who is left behind by neglected tropical disease programmes in Cameroon?**. (2021.0) **14** 1886457. DOI: 10.1080/16549716.2021.1886457
133. 133World Health Organization, 2017. WHO community engagement framework for quality, people-centred and resilient health services (No. WHO/HIS/SDS/2017.15). World Health Organization.. (2017.0)
134. 134WHO. Newborns: improving survival and well-being. 2020.. *Newborns: improving survival and well-being* (2020.0)
135. 135WHO. Control of Neglected Tropical Diseases. 2022.. *Control of Neglected Tropical Diseases* (2022.0)
|
---
title: Community characteristics and the risk of non-communicable diseases in Ghana
authors:
- Winfred A. Avogo
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021620
doi: 10.1371/journal.pgph.0000692
license: CC BY 4.0
---
# Community characteristics and the risk of non-communicable diseases in Ghana
## Abstract
Non-communicable Diseases (NCDs) are rising quickly in low- and middle- income countries. In Ghana, chronic diseases are major causes of morbidity and mortality, yet data and the evidence- base for awareness, detection, and management of NCDs are lacking. Using data from the 2014 Ghana Demographic and Health Survey (GDHS), the first national study with information on hypertension and other risk factors, we examine the correlates and community characteristics associated with the risk of hypertension, obesity, and anemia among women. We find that hypertension prevalence in Ghana was 16 percent and 17 percent were overweight/obese, while 41 percent had anemia of any form. On community characteristics, the level of poverty in a community was significantly associated with lower risks of all three NCDs, while the aggregate level of employment had higher risks. On individual characteristics, the wealth of a household, women’s educational level and urban residence were significant predictors of NCDs. We interpret the findings within the literature on neighborhood characteristics, the social gradient of health and in the context of speeding up the attainment of the Sustainable Development Goals (SGDS) to reduce premature deaths by one-third by 2030.
## Introduction
According to research by World Health Organization (WHO), non-communicable diseases (NCDs), which includes cardiovascular diseases, diabetes, cancer, chronic respiratory illnesses, and mental health disorders, are the leading cause of death worldwide, collectively responsible for 74 percent of global mortality [1]. An estimated 41 million people die each year from NCDs. Nearly 75 percent of all NCD deaths occur in low-and middle-income countries. Similarly, 86 percent of people who die from NCDs before reaching their seventieth birthday, die in low-and middle-income countries [1]. In sub-Saharan Africa (SSA), deaths from NCDs are rising faster than anywhere else in the world and we now know that the on-going COVID-19 pandemic has the highest morbidity and mortality risks among older adults especially those with underlying conditions such as NCDs [2].
In Ghana, NCDs account for 43 percent of all deaths, with cardiovascular diseases accounting for 19 percent of NCD deaths [3]. The few population-based surveys available on Ghana have also shown that NCDs are growing among the urban poor with a dual burden of infectious and chronic diseases. The prevalence of hypertension (raised blood pressure), for example, has been increasing over several decades and has significant impact on cardiovascular disease morbidity and mortality, especially in urban Ghana [4–8]. Similarly, there is a high and rising prevalence of obesity among Ghanaian adults–over 43 percent of the adult population are either overweight or obese [9]. Also, anemia which disproportionately affects children, women, and individuals from low-income areas of the country is associated with many chronic conditions, including Human Immunodeficiency Virus (HIV) and sickle cell disease as well as malaria and other cognitive and physical performance conditions [10]. All three chronic risk factors of NCDs (hypertension, obesity, and anemia) are asymptomatic in nature (silent killers), especially during the early stages when interventions and treatments are most effective.
The economic and social impact of NCDs is attributable to rapid demographic, epidemiological and nutritional transitions that are occurring in SSA and pose a threat to the attainment of the 2030 Agenda for Sustainable Development (SGDs), which includes a target (3.4) to reduce by one third, [relative to 2015 levels] premature mortality from NCDs and to promote mental health and well-being [11]. Currently, more than half of countries in the world are likely to miss SDG target 3.4 [12]. NCDs also threaten the attainment of the demographic dividend- the eventual aging of the population which results in fewer dependent children and the elderly and therefore, greater economic productivity through labor force participation. Yet, despite these threats, routine data collection systems and the broad evidence- base for increased awareness, detection, management, and control of NCDs are lacking in Ghana and other resource constrained settings.
In this paper, we draw on the first nationally representative sample of women aged 15–49 from the Demographic and Health Surveys (DHS) that include high-quality data on NCDs to examine the influence of neighborhood/community-level characteristics and conditions and the risks of NCDs with special focus on prevalence of hypertension, overweight/obesity, and anemia in Ghana. We interpret the results within the literature on neighborhood characteristics and health and in the context of interventions to accelerate the attainment of SDG target 3.4. to improve health, reduce death and disability due to NCDs and to improve the prospects of a demographic dividend in sub-Saharan Africa.
The epidemiological, demographic, and sociological literature has seen an explosion in research on the effect of neighborhoods and the community context on health [13–15]. This is, in part, a recognition that social influences on health operate through several structural conditions that shape individual lives and opportunities [16].
Studies on the relationship between neighborhood characteristics and NCD risk factors in higher income countries suggest a possible influence of community characteristics on health [17]. For example, living in communities with higher socio-economic disadvantage is associated with higher Body mass index (BMI), net of individual characteristics [18]. Similarly, individuals living in zip codes with the greatest percentage of happy and physically active tweets had lower obesity and diabetes prevalence net of individual characteristics [19]. Indeed, these findings are supported by the theory of the social gradient of health which posits that people of lower socio-economic status (SES) suffer a heightened health risk for nearly all diseases (including NCDs) compared to those of higher SES [20–23]. However, the social gradient of health hypothesis may be less consistent in low-to middle-income countries at the initial-middle stages of the demographic and epidemiological transitions as lifestyle and behavior changes of the emerging middle class may predispose them to diseases including NCDs. Thus, some studies in developing countries have shown a negative association (higher SES increase the risk of NCDs) [24–26] while others have demonstrated a positive association (higher SES decrease the risk of NCDs) [27,28]. Given these emerging socio-economic changes and disparities in individual risk factors of NCDs, as well as the dearth of broad evidence on the impact of neighborhoods on NCDs in low-middle-income countries, there is the need to explore the mechanisms that link neighborhood characteristics to NCD risk in these settings.
We draw on two mechanisms to conceptualize this relationship. First, diffusion theory highlights the role of social learning and social influence through individual social interaction in the community [29–33]. Social learning and social influence through education and employment levels in a community may afford women the information and awareness of NCDs, lifestyle changes and appropriate health seeking behaviors to avoid the risk of NCDs. Social influence can offer social and emotional support that buffers the risk of stress and NCDs [34]. Likewise, healthy behaviors such as access to and consumption of healthy food, health screening, not smoking, moderate alcohol consumption etc. are known to spread through social learning and social influence within social networks [35]. These spillover effects from other women’s education and awareness of NCDs and behaviors may affect the behavior of uneducated women in communities with high aggregate levels of education compared to uneducated women elsewhere [36].
On the other hand, neighborhood effects could impact NCD risks negatively through the globalization and diffusion of unhealthy lifestyles. The broader social and economic transformation which has led to a rise in average community education and employment may have sped up the ongoing nutritional and epidemiological transition thereby leading to poor dietary habits from processed Western diets and foods high in calories, sugar, salt, and total fat [37] as well as the lack of physical exercise and immoderate alcohol consumption. These habits, which are peculiar to the affluent, educated, middle class in urban centers are known to increase the prevalence of obesity, hypertension, and diabetes. Thus, women living in communities with high levels of aggregate education, wealth and employment will be impacted by the risk of NCDs independent of their individual education and employment.
Finally, Ghana’s health care system despite the decentralization of health services to the community-level through the Community-based Health Planning and Services (CHPS) is still relatively weak- basic healthcare services for the detection and screening of NCDs are inaccessible nor are those services targeted towards the poor [38]. Thus aggregate-level health coverage in a community in terms of the National Health Insurance Scheme (NHIS) and health seeking behavior in the community will impact NCD risks at the population level. Poor women who reside in communities with high aggregate-level health coverage will have low risks factor burdens of NCDs.
## Data
Data are drawn from the cross-sectional study of the Ghana Demographic and Health Survey (GDHS) conducted in 2014, the first survey in Ghana to include several anthropometric and blood pressure measurements, Body Mass Index (BMI) and anemia testing. The analysis relies on a total of 9396 women aged 15–49. Women who did not have information on their blood pressure readings, as well as pregnant women were excluded from the measurement of hypertension, obesity, and anemia. The 2014 GDHS was conducted by the Ghana Statistical Service (GSS), the Ghana Health Service (GHS), and the National Public Health Reference Laboratory (NPHRL) of the GHS in collaboration with the Measure DHS+ program of the United States Aid for International Development (USAID). The survey was designed to assist program managers, researchers and policymakers involved in planning, managing, and coordinating strategies for improving the health of Ghanaians.
The MEASURE DHS+ program also has experience with addressing ethical issues related to the protection of human subjects during biological sampling. Ethical clearance was provided by the Ghana Health Service Ethical Review Committee, Research and Development Division, Ghana Health Service; and the Institutional Review Board of ICF International. Descriptions of the protocols for biological testing can be found in a separate documentation on the DHS website (e.g., Anemia Testing Manual for Population-Based Surveys [39].
The DHS surveys used a two-stage sample design to produce estimates of key indicators at the national level. In the first stage, stratified sampling techniques are used to select clusters, delineated from 2010 Ghana Population and Housing Census (PHC), as the primary sampling unit (PSU). This resulted in a total of 427 clusters selected in the entire country (216 clusters in urban areas and 211 clusters in rural areas).
The second stage involved a systematic random sampling of households within each PSU. About 30 households were selected from each cluster and a total of 12,831 households were selected throughout the country. Since the sample is not self-weighting at the national and sub-sample levels, weighting factors (provided in the data) are used to produce results that are proportional at the national level and account for unequal probability of selection. On the response rate, 97 percent of the eligible women in each household were interviewed [40].
## Dependent outcomes
A. Hypertension. Blood pressure measurements were taken from consenting women aged 15–49 at intervals of 10 minutes during the individual interview. Blood pressure was measured using the LIFE SOURCE UA-767 Plus blood pressure monitor: a digital oscillometric blood pressure measuring device with automatic upper-arm inflation and automatic pressure release. In this paper, hypertension was defined as an average of the second and third measurements of systolic blood pressure (SBP) > = 140 mmHg and/or an average diastolic blood pressure (DBP) > = 90 mmHg according to internationally recommended categories [41]. Blood pressure measurements excluded pregnant women at the time of the survey due to the possibility of gestational hypertension.
B. Obesity. The GDHS survey also includes height (in cm) and weight (in kg) measurements for women aged 15–49. Weight and height measurements were used to estimate body mass index (BMI = weight in kg ÷ [height in meters, 2). BMI was further classified as underweight (BMI < 18.5 kgm−2), normal weight (BMI = 18.5–24.9 kgm2), overweight (BMI = 25.0–29.9 kgm−2), and obese (BMI ≥ 30.0 kgm−2) according to WHO recommendations. Pregnant women at the time of the survey were excluded since their BMI, comparable to other women in the survey, should be based on their weight before pregnancy.
C. Anaemia testing. Blood specimens for anemia testing were collected in half of the selected households from women aged 15–49 who voluntarily consented to be tested. Blood samples were drawn from a drop of blood taken from a finger prick and the haemoglobin concentration was measured using the HemoCue photometer system. Hemoglobin levels were further categorized into severe anemia if less than 7.0 g/dl and moderate between 7 and 9.9 g/dl and mild 10 and 12.9 g/dl and not anemic (13 g/dl and above), according to WHO recommendations [42]. Anemia was recoded as a dichotomous outcome (1- any form of anemia vs 0- not anemic). Pregnant women were also excluded from the measurement of anemia.
## Community variables
We use three variables to capture neighborhood/aggregate levels of education, employment, and poverty in the community. The first is the percentage of educated women in the community. This variable was constructed by aggregating the individual and household level variables at the primary sampling unit (PSU). The second aggregate level variable is the percentage of working women in the community. Finally, we include the percentage of poor women in the community. These three community variables have been demonstrated to be associated with reproductive outcomes [36,43].
## Covariates
The DHS surveys collected information on the characteristics of respondents at the time of the survey. We used individual educational attainment (1. No education, 2. Primary education, Secondary or higher education) as a covariate to examine the impact of community variables on the dependent outcomes. Individual women’s education has been shown in numerous studies to have a strong effect on health outcomes [44]. We included variables on the place of residence (urban vs. rural) to capture the effect of urbanization, the age-group of respondents is also included as the risk of NCDs varies by age [37,38]. The region of residence (Southern vs. Northern Ghana) is accounted for to capture inequalities in socio-economic development across the country, ethnicity (Akan, which is the largest ethnic group, vs other tribes) is included as a covariate to account for differences in morbidity and access to healthcare that are tied to ethnic background. The wealth index of the household, which is constructed using Principal Component Analysis, is included as a composite measure of household wealth. It combines information on household water supply and sanitation, floor type and electricity as well as other household goods into an index [45]. The index is recoded from quintiles into two categories of 1. Poorest and poorer respondents and 2. Middle, richer and richest respondents. We also include National Health Insurance coverage (no, yes) as a proxy for healthcare access and visits to health facilities in the last 6 months (yes, no).
## Analytical methods
Descriptive statistics were used to produce cross tabulations and chi-square tests of independence to describe variations in the dependent outcomes. Multilevel binary logistic regression was used to examine the impact of community variables and individual covariates on hypertension, obesity, and anemia. Due to the hierarchical nature of the data, (individual women are nested within clusters and households), failure to control for the correlation resulting from characteristics of women within the same cluster and household as well as aggregating the community variables at the cluster level will result in biased standard errors and estimates. We applied appropriate sampling weights (level weights) that correspond to each stage of sampling in DHS surveys at both the bivariate level and multivariate level. Results show that omitting level-weights may lead to underestimating the variation between level-2 units (at the cluster level and household level).
At the multivariate level, we fitted multilevel logistic regression models using XTMELOGIT command in STATA. Two-level random intercept models from the two-stage sampling design with the PSU and sampling weights were applied. We also fit Survey Logistic procedure in STATA: “svy: logit” to compare the results. Both procedures produced identical results. For more information on multilevel modelling using DHS data see [46]. A check for multicollinearity was conducted using the variance inflation factor (VIF) and all the variable used had less than 10 (VIF) and tolerance value lower than 1.0 (1/VIF) suggesting the absence of multicollinearity.
## Ethical considerations
Procedures and questionnaires for standard DHS surveys have been reviewed and approved by ICF Institutional Review Board (IRB). Additionally, country-specific DHS survey protocols are reviewed by the ICF IRB and typically by an IRB in the host country. ICF IRB ensures that the survey complies with the U.S. Department of Health and Human Services regulations for the protection of human subjects (45 CFR 46), while the host country IRB ensures that the survey complies with laws and norms of the nation. Written informed consent was obtained from the parent/guardian of each participant under 18 years of age.
## Descriptive results
Table 1 shows the percent distribution of NCDs, community variables and background characteristics of respondents. The table shows 16 percent of women were considered hypertensive (16.6 percent) and a similar percentage were considered overweight/obese (17 percent) while 41 percent reported anemia of any form.
**Table 1**
| Characteristics | Number | Percent |
| --- | --- | --- |
| Prevalence of Non-Communicable Diseases | | |
| Hypertension | 1240.0 | 16.16 |
| Obesity | 543.0 | 17.19 |
| Anemia | 1692.0 | 41.05 |
| Community-level variables | | |
| Percent of women educated | | |
| Low | 2778.0 | 24.24 |
| Medium and High | 5517.0 | 75.76 |
| Percent of poor women | | |
| Low | 2820.0 | 44.59 |
| Medium and High | 5475.0 | 55.41 |
| Percent of working women | | |
| Low | 4306.0 | 47.13 |
| Medium and High | 3744.0 | 52.27 |
| Individual-level characteristics | | |
| Respondent’s Education | | |
| | 2231.0 | 20.93 |
| Primary | 1431.0 | 16.72 |
| Secondary/higher | 4633.0 | 62.36 |
| Wealth Index | | |
| Poorest/Poorer | 3510.0 | 32.21 |
| Middle/Richer/Richest | 4785.0 | 67.79 |
| Ethnicity | | |
| Akan | 3444.0 | 50.73 |
| Other Tribes | 4758.0 | 49.27 |
| Region | | |
| Northern | 2289.0 | 14.16 |
| Southern | 5809.0 | 85.84 |
| Place of Residence | | |
| Urban | 4131.0 | 45.51 |
| Rural | 4164.0 | 54.49 |
| Age group | | |
| 18–24 | 2226.0 | 26.56 |
| 25–34 | 2907.0 | 35.49 |
| 35–49 | 3162.0 | 37.95 |
| Health Insurance Coverage | 5504.0 | 62.55 |
| Visited Health facility in last 12 months | 4543.0 | 54.08 |
On community level characteristics, 76 percent of women resided in communities with medium to high levels of education (that is with primary education and above) while 55 percent resided in communities considered poor using the wealth index of the households within that community. Similarly, 52 percent of women resided in communities with middle to high levels of women working outside the household.
Table 1 also highlights the individual level characteristics of the respondents. Sixty-two percent had a secondary or higher education, 32 percent resided in households considered poor according to the wealth index. There was an even split in ethnicity between Akan and non-Akans (51 vs 49 percent) and majority of respondents resided in the Southern part of the country (86 percent) and in rural areas (55 percent). Finally, 63 percent had access to health insurance coverage while 54 percent had visited a health facility in the last 12 months.
Table 2 presents bivariate analysis of NCDs by community level variables and individual level characteristics. The Table reports column percentages (for easy comparison across categories of NCDs) and 95 percent Confidence Intervals of the proportions. In communities with medium to high levels of poverty, 12 percent were hypertensive compared to 18 percent in communities with medium to high levels of education. Similarly, 21 percent of women in communities with medium to high levels of education were either overweight or obese compared to only 9 percent in poor communities. These relationships were statistically significant, using the Chi-square. The results of the percent of working women in the community and NCDs were not statistically significant, although in the expected direction. For example, 17 percent were hypertensive in communities with a medium to high percentage of working women (compared to 16 percent in communities with a low percentage of working women) and 22 percent of women in communities with a high percentage of working women were obese (compared to 16 percent in communities with a low percentage of working women). All the community level variables had similar risk of reporting anemia of any form ($40\%$ in communities with high levels of education (compared to 44 precent in communities with low levels of education), 42 percent in poor communities (compared to 39 percent with low levels of poverty) and 42 in communities with high percentage of working women, although as indicated, this relationship was not statistically significant). ( See Table 2).
**Table 2**
| Unnamed: 0 | Hypertension | Unnamed: 2 | Obesity | Unnamed: 4 | Anemia | Unnamed: 6 |
| --- | --- | --- | --- | --- | --- | --- |
| Community-level variables | % (95% CI) | p-value | % (95% CI) | p-value | % (95% CI) | p-value |
| Percent of educated women | | | | | | |
| Low | 10.36 (0.087, 0.122) | | 5.17 (0.036, 0.075) | | 44.12 (0.409, 0.474) | |
| Medium and High (Versus Low Levels) | 18.02 (0.166, 0.196) | <0.001 | 20.96 (0.189, 0.232) | <0.001 | 40.05 (0.38, 0.430) | 0.050 |
| Percent of poor women | | | | | | |
| Low | 20.9 (0.192, 0.228) | | 26.63 (0.236, 0.299) | | 38.71 (0.355, 0.410) | |
| Medium and High (Versus Low Levels) | 12.32 (0.110, 0.138) | <0.001 | 9.28 (0.080, 0.107) | <0.001 | 42.94 (0.406, 0.453) | 0.050 |
| Percent of working women | | | | | | |
| Low | 15.94 (0.146, 0.174) | | 16.2 (0.144, 0.181) | | 40.8 (0.387, 0.430) | |
| Medium and High (Versus Low Levels) | 17.16 (0.149, 0.196) | 0.390 | 21.83 (0.164, 0.284) | 0.064 | 42.20 (0.376, 0.469) | 0.593 |
| Individual-level characteristics | | | | | | |
| Respondent’s Education | | | | | | |
| | 14.36 (0.125, 0.16) | | 9.03 (0.069, 0.118) | | 45.29 (0.411, 0.486) | |
| Primary | 17.22 (0.148, 0.199) | | 18.15 (0.146, 0.224) | | 42.28 (0.380, 0.468) | |
| Secondary/higher | 16.49 (0.151, 0.18) | 0.145 | 19.66 (0.176, 0.219) | <0.001 | 39.25 (0.369, 0.417) | 0.016 |
| Wealth Index | | | | | | |
| Poorest/Poorer | 10.49 (0.092, 0.119) | | 4.33 (0.031, 0.059) | | 45.77 (0.431, 0.484) | |
| Middle/Richer/Richest | 18.86 (0.174, 0.204) | <0.001 | 23.23 (0.298, 0.256) | <0.001 | 38.76 (0.362, 0.414) | <0.001 |
| Ethnicity | | | | | | |
| Akan | 17.02 (0.154, 0.188) | | 18.63 (0.164, 0.211) | | 38.76 (0.360, 0.416) | |
| Other Tribes | 15.42 (0.140, 0.169) | 0.135 | 15.68 (0.010) | 0.105 | 43.23 (0.475, 0.457) | 0.018 |
| Region | | | | | | |
| Northern | 10.01 (0.080, 0.124) | | 47.9 (0.025, 0.070) | | 43.03 (0.391, 0.470) | |
| Southern | 17.45 (0.1611, 0.189) | <0.001 | 19.39 (0.175, 0.215) | <0.001 | 40.73 (0.385, 0.430) | 0.317 |
| Place of Residence | | | | | | |
| Urban | 19.81 (0.183, 0.214) | | 22.97 (0.23, 0.258) | | 40.67 (0.378, 0.436) | |
| Rural | 11.79 (0.103, 0.135) | <0.001 | 9.95 (0.084, 0.1175) | <0.001 | 41.5 (0.390, 0.441) | 0.671 |
| Age group | | | | | | |
| 18–24 | 5.03 (0.040, 0.063) | | 4.69 (0.035, 0.063) | | 44.59 (0.412, 0.480) | |
| 25–34 | 12.09 (0.103, 0.142) | | 16 (0.135, 0.189) | | 39.15 (0.360, 0.425) | |
| 35–49 | 27.77 (0.256, 0.300) | <0.001 | 26.62 (0.236, 0.299) | <0.001 | 40.33 (0.374, 0.434) | 0.058 |
| Health Insurance Coverage | 17.15 (0.158, 0.185) | <0.001 | 17.6 (0.155, 0.198) | 0.563 | 41.04 (0.387, 0.4341) | |
| Visited Health facility in last 12 months | 17.81 (0.163, 0.194) | <0.001 | 18.48 (0.1637, 0.208) | 0.061 | 39.61 (0.371, 0.421) | 0.086 |
On individual women’s characteristics, women in wealthier households reported higher levels of NCD risk factors than poorer households (e.g., 19 percent vs. 11 percent reported hypertension and 23 percent vs. 4 percent reported obesity). Residing in Southern Ghana was associated with higher risks of hypertension than Northern Ghana (18 percent vs. 10 percent). However, the risk was reversed for obesity-48 percent were obese or overweight in Northern Ghana compared to 20 percent in Southern Ghana. Urban areas are also associated with higher NCD risks than rural areas (e.g., 20 percent vs. 12 percent for hypertension and 23 percent vs. 10 percent for obesity). Older women were also at a higher risk of NCDs compared to younger women (e.g., 28 percent of women aged 35–49 were hypertensive (27 percent were obese/overweight) compared to only 5 percent of respondents aged 18–24 for both hypertension and obesity). Finally, women with access to health insurance (17 percent) and women who visited a health facility in the last 12 months (18 percent) reported hypertension. All the above-described relationships were statistically significant given the chi-square value.
## Multivariate results
In Table 3, we start with the multilevel logistic regression results for hypertension. The aggregate-level of education, although positive, was not a significant predictor of the prevalence of hypertension. However, the aggregate-levels of poverty and employment in the community were significantly associated with hypertension, holding all other covariates constant. The odds of hypertension were lower if aggregate levels of poverty are high in the community (compared to lower levels of poverty)—(Odds Ratio (OR) = 0.78, $p \leq 0.05$). Following the same pattern, a high aggregate level of women’s employment in the community was associated with higher odds of hypertension (Odds Ratio (OR) = 1.12, $p \leq 0.05$).
**Table 3**
| Unnamed: 0 | Hypertension | Hypertension.1 | Obesity | Obesity.1 | Anemia | Anemia.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Community-level variables | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value |
| Percent of women educated | | | | | | |
| Low | 1 | | 1 | | 1 | |
| Medium and High | 1.03 (0.857, 1.248) | 0.726 | 1.05 (0.815, 1.36) | 0.697 | 0.977 (0.837, 1.142) | 0.772 |
| Percent of poor women | | | | | | |
| Low | 1 | | 1 | | 1 | |
| Medium and High | 0.78 (0.659, 0.924) | 0.004 | 0.59(0.443, 0.808) | <0.001 | 1.10 (0.910, 1.319) | 0.293 |
| Percent of working women | | | | | | |
| Low | 1 | | 1 | | 1 | |
| Medium and High | 1.12 (1.010, 1.237) | 0.031 | 1.17(1.022, 1.348) | 0.024 | 1.032 (0.939, 1.136) | 0.504 |
| Individual-level characteristics | | | | | | |
| Respondent’s Education | | | | | | |
| None/primary | 1 | | 1 | | 1 | |
| Secondary/higher | 1.01 (0.897, 1.137) | 0.864 | 1.20(1.030, 1.420) | 0.021 | 0.94 (0.839, 1.046) | 0.246 |
| Wealth Index | | | | | | |
| Poorest/Poorer | 1 | | 1 | | 1 | |
| Middle/Richer/Richest | 1.30(1.038, 1.625) | 0.022 | 3.09(1.946, 4.913) | <0.001 | 0.73(0.556, 0.964) | 0.027 |
| Ethnicity | | | | | | |
| Other Tribes | 1 | | 1 | | 1 | |
| Akan | 0.91 (0.778, 1.075) | 0.280 | 0.76(0.580, 0.991) | 0.043 | 0.89 (0.741, 1.074) | 0.229 |
| Region | | | | | | |
| Northern | 1 | | 1 | | 1 | |
| Southern | 1.15+ (0.879, 1.50) | 0.310 | 1.45 (0.909, 2.314) | 0.119 | 1.36 (1.069, 1.727) | 0.012 |
| Place of Residence | | | | | | |
| Rural | 1 | | 1 | | 1 | |
| Urban | 0.85 (0.655, 1.094) | 0.202 | 0.89 (0.804, 1.591) | 0.479 | 0.78(0.630, 0.955) | 0.017 |
| Age group | | | | | | |
| 18-24/25-34 | 1 | | 1 | | 1 | |
| 35–49 | 2.83(2.499, 3.209) | <0.001 | 2.85 (2.430, 3.358) | <0.001 | 0.91 (0.820, 1.005) | 0.063 |
| Health Insurance Coverage | | | | | | |
| No | 1 | | 1 | | 1 | |
| Yes | 1.16 (0.986, 1.372) | 0.730 | 0.89 (0.681, 1.178) | 0.430 | 1.072 (0.907, 1.268) | 0.411 |
| Visited Health facility in last 12 months | | | | | | |
| No | 1 | | 1 | | 1 | |
| Yes | 1.23 (1.051, 1.437) | <0.001 | 1.16 (0.940, 1.420) | 0.168 | 0.878 (0.745, 1.04) | 0.118 |
On individual and household correlates of hypertension, individual women’s education, ethnicity, and region of residence (South vs. the North) were not statistically significant predictors of hypertension. The odds of hypertension, however, were higher with the level of wealth of a household. The odds of hypertension among rich compared to poor households were 1.3 times ($p \leq 0.05$), confirming the earlier results that poverty reduces your odds of hypertension. Older women (aged 35–49) and visiting a health facility in the last 12 months were associated with higher odds of hypertension (Odds Ratio (OR) = 2.83, $p \leq 0.001$ and 1.23, $p \leq 0.001$; respectively).
Next, we consider the multilevel results of the prevalence of obesity. Like hypertension, the aggregate level of education in the community did not significantly predict of obesity. However, the aggregate-level of poverty and women’s employment in the community as observed on the model on hypertension was significantly associated with obesity. The odds of obesity were lower if aggregate levels of poverty are high in the community (Odds Ratio (OR) = 0.59, $p \leq 0.001$) and odds of obesity were higher if the aggregate level of women employed outside the household was higher (Odds Ratio (OR) = 1.17, $p \leq 0.05$).
On covariates, unlike hypertension, an increase in individual education had significantly higher odds of obesity (Odds Ratio (OR) = 1.20, $p \leq 0.05$). The odds of obesity of women with a secondary or higher education were 1.2 times that of women who never went to school or have only a primary education. Similarly, household wealth status and women’s age, as observed in the model on hypertension, had higher odds of obesity. The odds of obesity if you reside in a richer household were 3.09 times compared to poorer households. That of older women were similar; 2.85 times. Ethnicity (Akan vs other tribes) had lower odds of obesity (Odds Ratio (OR) = 0.76, $p \leq 0.05$). Urban residence, having health insurance and visit to a health facility in the last 12 months were not significant predictors of obesity.
Finally, we consider the models for anemia. Community-level variables did not significantly predict the prevalence of anemia. On covariates of anemia, richer households had lower odds of anemia (Odds Ratio (OR) = 0.73, $p \leq 0.05$). Similarly urban residence had lower odds (Odds Ratio (OR) = 0.78, $p \leq 0.05$). Surprisingly, residence in Southern Ghana, which is socio-economically well-off than Northern Ghana increases the odds of anemia (Odds Ratio (OR) = 1.36, $p \leq 0.05$).
## Discussion
NCDs are rising rapidly in Africa and undermining social and economic development. Ghana, like many middle- to low-income countries, is dealing with the triple burden of infectious diseases, malnutrition/overweight obesity, and chronic diseases. This study probed if neighborhood effects, or community variables played a role in the risks of NCDs. In both bivariate and multivariate analysis, women living in poorer communities reported lower prevalence of NCDs (hypertension and obesity), while women residing in communities with a high level of employment reported higher risks of NCDs (hypertension and obesity). These results were reinforced with the results of the wealth index of a household and the individual educational level of women. Wealthier households had higher odds of hypertension and obesity and lower odds of anemia. Women with higher education (secondary and higher level) had higher odds of obesity.
First, the results of this study are largely consistent with the literature on neighborhood effects on one hand and the nutritional and epidemiological transition as well as the globalization of unhealthy lifestyles on the other. The results confirmed that there is considerable social inequality among communities and NCD risks tend to be bundled together at the neighborhood/community level, but this is slightly inconsistent with the social gradient of health hypotheses. In developing countries like Ghana, the current NCD disease burden is greater among the middle class compared to the lower class in higher income countries. As stated above, there is a high risks of NCDs in communities with a high levels of women employed and in wealthier households, while there are lower risks in communities with high levels of poverty. It is thus urgent to develop targeted primary intervention strategies in both communities (rich and poor). As we know from the nutritional and epidemiological transitions soon, if not already, NCDs will be concentrated among both the rich and poor as rapid urbanization brings people from rural communities to the cities, like what prevails in higher income countries [47].
Second, within the context of the epidemiological and nutritional transition–which focuses on the complex change in patterns of health and disease and on the interactions between these patterns and the demographic, economic, and sociological determinants and consequences [48,49]. It is clear that developing countries like Ghana have scarcely completed the previous stages of the transition (where infectious diseases prevail) before gradually shifting to the age of degenerative and lifestyle diseases with NCDs emerging as the major causes of death.
Globalization and diffusion of unhealthy human diet and lifestyle behaviors such as eating processed Western diets and foods high in calories, sugar, salt, and total fat [37,50,51] as well as the lack of physical exercise and immoderate alcohol consumption has occurred in parallel with the increasing prevalence of obesity and NCDs in low-middle-income countries [51,52]. There is thus a need to design and fund effective awareness and early intervention programs to mitigate the human resource and economic burden of NCDs while tackling the continuing scourge of infectious diseases and malnutrition. This will accelerate the attainment of the 2030 Agenda for Sustainable Development (SGDs) and a demographic dividend. This study by evaluating the significance of key community variables and individual determinants of NCD risks in low-middle-income countries, has underscored the importance of country-tailored assessments that are needed for awareness, detection, and management of NCDs.
Finally, our study is not without limitations. Apart from its cross-sectional design, the differential selection of individuals into communities and other indirect pathways of neighborhood/community factors or simultaneity biases prevent us from drawing causal inferences [53–55]. Similarly, our study is unable to clarify if neighborhood/community differences are due to characteristics of the areas or differences in the types of people living in different areas. Thus, there is the need for further research to distinguish between context and composition when examining neighborhood effects on health. Nonetheless, neighborhood-level mechanisms have proven effective in other outcomes and can be usefully applied to NCDs with important policy implications for health promotion and reduction of health disparities in low-middle-income countries.
## References
1. 1World Health Organization. Global Status Report on Non-Communicable Diseases (2022). Geneva: World Health Organization; 2020 [cited 2022 Dec 04]. Available from: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases.. *Global Status Report on Non-Communicable Diseases (2022)* (2020.0)
2. Katzmarzyk PT, Salbaum JM, Heymsfield SB. **Obesity, non-communicable diseases, and COVID‐19: A perfect storm**. *American Journal of Human Biology* (2020.0)
3. 3World Health Organization. Non-communicable Diseases Country Profile 2018.
Geneva: World Health Organization; 2018 [cited 2022 Dec 04] p.95. Available from: https://www.who.int/publications/i/item/9789241514620.. *Non-communicable Diseases Country Profile 2018.* (2018.0) 95
4. Sanuade OA, Boatemaa S, Kushitor MK. **Hypertension prevalence, awareness, treatment and control in Ghanaian population: Evidence from the Ghana demographic and health survey.**. *PloS one.* (2018.0) **13** e0205985. DOI: 10.1371/journal.pone.0205985
5. Aikins AD, Kushitor M, Koram K, Gyamfi S, Ogedegbe G. **Chronic non-communicable diseases, and the challenge of universal health coverage: insights from community-based cardiovascular disease research in urban poor communities in Accra, Ghana.**. *BMC public health.* (2014.0) **14** 1-9. PMID: 24383435
6. Donkor ES. **Stroke in the century: a snapshot of the burden, epidemiology, and quality of life.**. *Stroke research and treatment.* (2018.0)
7. Addo J, Agyemang C, Smeeth L, Aikins AD, Adusei AK, Ogedegbe O. **A review of population-based studies on hypertension in Ghana**. *Ghana medical journal* (2012.0) **46** 4-11. PMID: 23661811
8. Agyemang C, van den Born BJ. **Non-communicable diseases in migrants: an expert review**. *Journal of travel medicine* (2019.0) **26** 107. DOI: 10.1093/jtm/tay107
9. Ofori-Asenso R, Agyeman AA, Laar A, Boateng D. **Overweight, and obesity epidemic in Ghana—a systematic review and meta-analysis.**. *BMC public health.* (2016.0) **16** 1-8. DOI: 10.1186/s12889-016-3901-4
10. Anderson AK. **Prevalence of anemia, overweight/obesity, and undiagnosed hypertension and diabetes among residents of selected communities in Ghana.**. *International journal of chronic diseases* (2017.0). DOI: 10.1155/2017/7836019
11. Lee BX, Kjaerulf F, Turner S, Cohen L, Donnelly PD, Muggah R. **Transforming our world: implementing the 2030 agenda through sustainable development goal indicators**. *Journal of public health policy* (2016.0) **37** 13-31. DOI: 10.1057/s41271-016-0002-7
12. Countdown NC. **NCD Countdown 2030: pathways to achieving Sustainable Development Goal target 3.4**. *The Lancet* (2020.0). DOI: 10.1016/S0140-6736(20)31761-X
13. Decker MJ, Isquick S, Tilley L, Zhi Q, Gutman A, Luong W. **Neighborhoods matter. A systematic review of neighborhood characteristics and adolescent reproductive health outcomes.**. *Health & place.* (2018.0) **54** 178-90. PMID: 30312942
14. Pickett KE, Pearl M. **Multilevel analyses of neighborhood socioeconomic context and health outcomes: a critical review**. *Journal of Epidemiology & Community Health* (2001.0) **55** 111-22. PMID: 11154250
15. Mohnen SM, Schneider S, Droomers M. **Neighborhood characteristics as determinants of healthcare utilization–a theoretical model**. *Health economics review* (2019.0) **9** 1-9. PMID: 30656503
16. Diez Roux AV. **Investigating neighborhood and area effects on health**. *American journal of public health* (2001.0) **91** 1783-9. DOI: 10.2105/ajph.91.11.1783
17. Morland K, Wing S, Roux AD, Poole C. **Neighborhood characteristics associated with the location of food stores and food service places**. *American journal of preventive medicine* (2002.0) **22** 23-9. DOI: 10.1016/s0749-3797(01)00403-2
18. Robert SA, Reither EN. **A multilevel analysis of race, community disadvantage, and body mass index among adults in the US.**. *Social science & medicine* (2004.0) **59** 2421-34. DOI: 10.1016/j.socscimed.2004.03.034
19. Nguyen QC, Brunisholz KD, Yu W, McCullough M, Hanson HA, Litchman ML. **Twitter-derived neighborhood characteristics associated with obesity and diabetes**. *Scientific reports* (2017.0) **7** 1-0. PMID: 28127051
20. Kosteniuk JG, Dickinson HD. **Tracing the social gradient in the health of Canadians: primary and secondary determinants.**. *Social science & medicine* (2003.0) **57** 263-76. DOI: 10.1016/s0277-9536(02)00345-3
21. Whitehead M, Dahlgren G. **Concepts, and principles for tackling social inequities in health: Levelling up Part 1.**. *World Health Organization: Studies on social and economic determinants of population health.* (2006.0) **2** 460-74
22. Kanjilal S, Gregg EW, Cheng YJ, Zhang P, Nelson DE, Mensah G. **Socioeconomic status and trends in disparities in 4 major risk factors for cardiovascular disease among US adults, 1971–2002.**. *Archives of internal medicine* (2006.0) **166** 2348-55. DOI: 10.1001/archinte.166.21.2348
23. Marmot M, Bell R. **Social determinants, and non-communicable diseases: time for integrated action**. *Bmj* (2019.0) 364. DOI: 10.1136/bmj.l251
24. Miszkurka M, Haddad S, Langlois ÉV, Freeman EE, Kouanda S, Zunzunegui MV. **Heavy burden of non-communicable diseases at early age and gender disparities in an adult population of Burkina Faso: World Health Survey.**. *BMC public health.* (2012.0) **12** 1-10. PMID: 22214479
25. Ploubidis GB, Mathenge W, De Stavola B, Grundy E, Foster A, Kuper H. **Socioeconomic position and later life prevalence of hypertension, diabetes and visual impairment in Nakuru, Kenya.**. *International journal of public health.* (2013.0) **58** 133-41. DOI: 10.1007/s00038-012-0389-2
26. Tenkorang EY, Kuuire VZ. **Noncommunicable diseases in Ghana: Does the theory of social gradient in health hold?**. *Health Education & Behavior.* (2016.0) **43** 25S-36S. DOI: 10.1177/1090198115602675
27. Negin J, Cumming R, de Ramirez SS, Abimbola S, Sachs SE. **Risk factors for non‐communicable diseases among older adults in rural Africa**. *Tropical Medicine & International Health* (2011.0) **16** 640-6. DOI: 10.1111/j.1365-3156.2011.02739.x
28. Hosseinpoor AR, Bergen N, Kunst A, Harper S, Guthold R, Rekve D. **Socioeconomic inequalities in risk factors for non-communicable diseases in low-income and middle-income countries: results from the World Health Survey.**. *BMC public Health.* (2012.0) **12** 1-3. DOI: 10.1186/1471-2458-12-912
29. Montgomery MR, Casterline JB. **The diffusion of fertility control in Taiwan: Evidence from pooled cross-section time-series models.**. *Population studies.* (1993.0) **47** 457-79. DOI: 10.1080/0032472031000147246
30. Montgomery MR, Casterline JB. **Social learning, social influence, and new models of fertility**. *Population and development review* (1996.0) **22** 151-75
31. Bongaarts J, Watkins SC. **Social interactions, and contemporary fertility transitions**. *Population and development review* (1996.0) **1** 639-82
32. Kohler HP, Behrman JR, Watkins SC. **The density of social networks and fertility decisions: Evidence from South Nyanza District, Kenya.**. *Demography* (2001.0) **38** 43-58. DOI: 10.1353/dem.2001.0005
33. Benefo KD. **The community-level effects of women’s education on reproductive behavior in rural Ghana.**. *Demographic research.* (2006.0) **14** 485-508
34. Kravdal Ø. **Education and fertility in sub-Saharan Africa: Individual and community effects.**. *Demography* (2002.0) **39** 233-50. DOI: 10.1353/dem.2002.0017
35. DeRose LF, Kravdal Ø. **Educational reversals and first-birth timing in sub-Saharan Africa: A dynamic multilevel approach.**. *Demography* (2007.0) **44** 59-77. DOI: 10.1353/dem.2007.0001
36. Johns LE, Aiello AE, Cheng C, Galea S, Koenen KC, Uddin M. **Neighborhood social cohesion and posttraumatic stress disorder in a community-based sample: findings from the Detroit Neighborhood Health Study.**. *Social psychiatry and psychiatric epidemiology.* (2012.0) **47** 1899-906. DOI: 10.1007/s00127-012-0506-9
37. Rosenquist JN, Murabito J, Fowler JH, Christakis NA. **The spread of alcohol consumption behavior in a large social network**. *Annals of internal medicine* (2010.0) **152** 426-33. DOI: 10.7326/0003-4819-152-7-201004060-00007
38. Anderson AK. **Prevalence of anemia, overweight/obesity, and undiagnosed hypertension and diabetes among residents of selected communities in Ghana.**. *International journal of chronic diseases* (2017.0) **2017**. DOI: 10.1155/2017/7836019
39. Sharmanov A.. *Anemia testing manual for population-based surveys* (2000.0)
40. 40Ghana Statistical Service—GSS, Ghana Health Service—GHS, and ICF International.
2015. Ghana Demographic and Health Survey 2014. Rockville, Maryland, USA: GSS, GHS, and ICF International. [Cited Dec 4]. Available at http://dhsprogram.com/pubs/pdf/FR307/FR307.pdf.. *Ghana Demographic and Health Survey 2014* (2015.0)
41. Chalmers JO, MacMahon S, Mancia G, Whitworth J, Beilin L, Hansson L. **1999 World Health Organization-International Society of Hypertension Guidelines for the management of hypertension. Guidelines sub-committee of the World Health Organization**. *Clinical and experimental hypertension (New York, NY: 1993).* (1999.0) **21** 1009-60. DOI: 10.3109/10641969909061028
42. Freeman AM, Rai M, Morando DW. *Anemia Screening.* (2022.0)
43. Adedini SA, Odimegwu C, Bamiwuye O, Fadeyibi O, Wet ND. **Barriers to accessing health care in Nigeria: implications for child survival.**. *Global health action.* (2014.0) **7** 23499. DOI: 10.3402/gha.v7.23499
44. Kebede E, Striessnig E, Goujon A. **The relative importance of women’s education on fertility desires in sub-Saharan Africa: A multilevel analysis.**. *Population Studies.* (2022.0) **76** 137-56. DOI: 10.1080/00324728.2021.1892170
45. Rutstein SO, Rojas G. *Guide to DHS statistics* (2006.0) **38**
46. Elkasabi M, Ren R, Pullum TW. *Multilevel modeling using DHS Surveys: a framework to approximate level-weights*
47. Engelgau M, Rosenhouse S, El-Saharty S, Mahal A. **The economic effect of non-communicable diseases on households and nations: a review of existing evidence**. *Journal of health communication* (2011.0) **16** 75-81. PMID: 21916715
48. Sanders JW, Fuhrer GS, Johnson MD, Riddle MS. **The epidemiological transition: the current status of infectious diseases in the developed world versus the developing world**. *Science Progress* (2008.0) **91** 1-37. DOI: 10.3184/003685008X284628
49. Omran AR. **The epidemiologic transition: a theory of the epidemiology of population change.**. *The Milbank Memorial Fund Quarterly* **49** 509-538. PMID: 5155251
50. Schmidhuber J, Shetty P. **The nutrition transition to 2030. Why developing countries are likely to bear the major burden.**. *Acta agriculturae scan section c.* (2005.0) **2** 150-66
51. Popkin BM, Adair LS, Ng SW. **Global nutrition transition and the pandemic of obesity in developing countries.**. *Nutrition reviews.* (2012.0) **70** 3-21. DOI: 10.1111/j.1753-4887.2011.00456.x
52. Katzmarzyk PT, Mason C. **The physical activity transition**. *Journal of Physical activity and Health* (2009.0) **6** 269-80. DOI: 10.1123/jpah.6.3.269
53. Popkin BM, Gordon-Larsen P. **The nutrition transition: worldwide obesity dynamics and their determinants.**. *International journal of obesity* (2004.0) **28** S2-9. DOI: 10.1038/sj.ijo.0802804
54. Duncan GJ, Raudenbush SW. **Assessing the effects of context in studies of child and youth development**. *Educational psychologist* (1999.0) **34** 29-41
55. Winship C, Morgan SL. **The estimation of causal effects from observational data**. *Annual review of sociology* (1999.0) **25** 659-706
|
---
title: 'Prevention and management of type 2 diabetes mellitus in Uganda and South
Africa: Findings from the SMART2D pragmatic implementation trial'
authors:
- David Guwatudde
- Peter Delobelle
- Pilvikki Absetz
- Josefien Olmen Van
- Roy William Mayega
- Francis Xavier Kasujja
- Jeroen De Man
- Mariam Hassen
- Elizabeth Ekirapa Kiracho
- Juliet Kiguli
- Thandi Puoane
- Claes-Goran Ostenson
- Stefan Peterson
- Meena Daivadanam
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021626
doi: 10.1371/journal.pgph.0000425
license: CC BY 4.0
---
# Prevention and management of type 2 diabetes mellitus in Uganda and South Africa: Findings from the SMART2D pragmatic implementation trial
## Abstract
Health systems in many low- and middle-income countries are struggling to manage type 2 diabetes (T2D). Management of glycaemia via well-organized care can reduce T2D incidence, and associated morbidity and mortality. The primary aim of this study was to evaluate the effectiveness of facility plus community care interventions (integrated care), compared to facility only care interventions (facility care) towards improvement of T2D outcomes in Uganda and South Africa. A pragmatic cluster randomized trial design was used to compare outcomes among participants with T2D and those at high risk. The trial had two study arms; the integrated care arm, and the facility care arm; and in Uganda only, an additional usual care arm. Participants were enrolled at nine primary health facilities in Uganda, and two in South Africa. Participants were adults aged 30 to 75 years, and followed for up to 12 months. Primary outcomes were glycaemic control among participants with T2D, and reduction in HbA1c > = 3 mmol/mol among participants at high risk. Secondary outcomes were retention into care and incident T2D. Adjusted analysis revealed significantly higher retention into care comparing integrated care and facility care versus usual care in Uganda and integrated care versus facility care in South Africa. The effect was particularly high among participants at high risk in Uganda with an incident rate ratio of 2.46 [1.33–4.53] for the facility care arm and 3.52 [2.13–5.80] for the integrated care arm. No improvement in glycaemic control or reduction in HbA1c was found in either country. However, considerable and unbalanced loss to follow-up compromised assessment of the intervention effect on HbA1c. Study interventions significantly improved retention into care, especially compared to usual care in Uganda. This highlights the need for adequate primary care for T2D and suggest a role for the community in T2D prevention.
Trial registration number: ISRCTN11913581.
## Introduction
The global burden of type 2 diabetes mellitus (T2D) and associated morbidity and mortality continues to rise, especially in low- and middle-income counties (LMICs) [1]. The 2019 global burden of disease estimated that diabetes among adults would increase from an estimated prevalence of $8.8\%$ in 2015, to a prevalence of $10.4\%$ by 2040; and most of the increase would occur in LMICs. The same study further showed that in 2019, diabetes ranked 8th as the leading cause of deaths and Disability Adjusted Life Years (DALYs) lost, up from 20th in 1990 [2]. Global estimates also show that sustained glycemic control among patients with diabetes is only $50\%$ [3]. Data from STEPS surveys conducted in 40 LMICs showed that of all patients with T2D in these countries, only $38\%$ are on treatment, and $23\%$ achieve glycemic control [4]. Furthermore, data from the South African National Health and Nutrition Examination Survey (SANHANES-1 [2011–2012]) found that among individuals with T2D, $45.4\%$ were unscreened, $14.7\%$ were screened but undiagnosed, $2.3\%$ were diagnosed but untreated, and $18.1\%$ were treated but uncontrolled; suggesting that $80.5\%$ of the population with diabetes had unmet need for care [5].
To manage the disease adequately, people living with T2D do not only need timely care but also diabetes self-management education and ongoing diabetes self-management support [6]. Similarly, to prevent or delay onset of T2D, people at high risk should receive preventive care, education and support to manage risk for T2D. There is strong evidence demonstrating that in the absence of risk reduction interventions, approximately $5\%$– $10\%$ of individuals with pre-diabetes progress to diabetes annually [7]. Several randomized controlled trials conducted in high income countries (HICs), including the Diabetes Prevention Program (DPP) [8], and the Finnish Diabetes Prevention Study (DPS) [9], also demonstrate that lifestyle/ behavioral interventions are highly effective in preventing T2D.
Given the increasing burden of diabetes in LMICs, and the lack of targeted prevention strategies coupled with inadequate care for T2D, there is an urgent need to identify context relevant interventions with the potential to work across different income settings, aimed at preventing progress to diabetes among individuals at high risk, as well as improving diabetes management outcomes among patients with diabetes.
The ‘Self-Management Approach and Reciprocal learning for Type 2 Diabetes’ (SMART2D) was a collaborative project implemented between January 2015 and December 2019 in two different settings: a rural area in Uganda (a low-income country), and a semi-urban township in Cape Town, South Africa (a middle-income country). The main objective of the project was to formulate and implement a contextually appropriate self-management strategy for the prevention and control of T2D in each setting and to evaluate its outcomes [10, 11]. As part of the project, a pragmatic cluster randomized trial was designed and implemented to evaluate the effectiveness and implementation of integrated facility plus community care interventions, compared to facility care only, in improving T2D management outcomes among patients with diabetes, and individuals at high risk of T2D in Uganda and South Africa. The Self-determination theory (SDT) [12] was used as part of the theoretical framework for our study and its relevance was tested using the baseline data of the SMART2D trial. The aim of the analysis reported in this article was to evaluate the effectiveness of the interventions towards improvement of T2D prevention and management outcomes in the two countries.
## Ethics statement
In Uganda, approval for the study was obtained from the Research and Ethics Committee of the Makerere University School of Public Health (reference number 426), and from the Uganda National Council for Science and Technology (reference number HS 2118). In South Africa, approval was obtained from the Biomedical Science Research Ethics Committee of the University of the Western Cape (reference number BM/$\frac{17}{1}$/36). Written informed consent was obtained from eligible subjects before enrollment in the study. Subjects not enrolled in the study but had a fasting plasma glucose test reading of at least 6.1 mmol/L, and were not into care were advised to as soon as possible report to the nearest government owned health facility for further evaluation.
## Study design and setting
The study used a pragmatic cluster randomized trial design, to allocate clusters to study arms. Clusters were primary health care facilities and their respective catchment areas. In Uganda, a health facility was eligible for inclusion into the trial if it provided primary health care, and was at least 5 kilometers away from an already selected one. Using these two criteria, nine health primary health care facilities were selected with the aim of being able to randomize three facilities to each of the three study arms. In South Africa, facilities were assessed for eligibility based on facilities being managed by the same governance structure and existence of a community-based platform for T2D services support. Based on these criteria, two primary health care facilities were selected. The primary study units were patients with diabetes, and individuals at high risk of T2D recruited within each of the study clusters. Recruitment of potential participants was then conducted within the catchment areas, and at the health facilities as described below.
In South Africa, the trial included two study arms; a facility plus community care (integrated care) arm, and a facility only care (facility care) arm. Because in South Africa facility care provides the required standard of care for patients with diabetes, the facility care arm served as the control arm with no study interventions; whereas the integrated care arm served as the intervention arm. In Uganda where primary care processes for diabetes are not regularly fully provided at primary health care facilities, the trial had three study arms including the two study arms as in South Africa, plus a usual care arm that served as the control arm. We made no interventions in the usual care arm except provision of diabetes medications to avoid stock outs.
## Eligibility criteria
Subjects were eligible for participation in the study if they fulfilled the following criteria: aged between 30 and 75 years, had resided in their respective communities for at least 6 months, had no plans of out-migrating from the study area over the next 12 months; able to provide written informed consent; allow home visits and follow-up contacts; and a diagnosis of diabetes of no longer than 12 months, or if classified as being at high risk of T2D. A subject was classified as having diabetes if he/she had already been diagnosed with diabetes and was in care for diabetes at the study health facility, or if they had at least two fasting plasma glucose test readings of greater than 7.0 millimoles per liter (mmol/L) conducted by our study staff on two separate days within two weeks prior to enrollment into the study. In Uganda, a subject was classified as being at high risk of T2D if they had pre-diabetes, defined as having at least two fasting plasma glucose test readings of between 6.1–6.9 mmol/L conducted by our study staff on two separate days within two weeks prior to enrollment into the study. In South Africa, a subject was classified as being at high risk of T2D if they had a random plasma glucose test reading ≤ 11 mmol/L, a BMI of at least 25, and having one or more of the following: hypertension, a cardiovascular disease, physical inactivity, family history of T2D, or previous gestational diabetes. Pregnancy and serious mental disability were exclusion criteria. Once eligibility was confirmed, written informed consent was obtained and the subject enrolled into the study.
## Interventions
The facility care arm comprised two intervention strategies including: 1) organization of care at facility level, and 2) strengthening of the patient role in self-management. The integrated care arm included the two intervention strategies in the facility care arm, plus three community intervention strategies including: 1) community mobilization (including dissemination of messages on healthy lifestyles to community members and key stakeholders); 2) creating a supportive environment by establishing peer support groups and identifying a care companion for each participant; and, 3) establishing a community extension link between the facility and the community for the care of patients. Interventions in this trial are hence related to clusters. In Uganda, study health facilities were supported with medication and diabetes diagnostic equipment, reagents and strips to address critical gaps in usual care that would otherwise influence trial implementation. Details of the various elements within each intervention strategy have been described elsewhere [13]. Participants were exposed to the intervention strategies for at least 9 months. Most of the intervention strategies were developed based on principles of the Self Determination Theory (SDT) [14]. The theoretical framework and development of the care interventions have been described elsewhere [11, 15].
## Measurements
Following enrollment, trained study staff administered a standardized questionnaire to obtain baseline data that included: socio-demographic characteristics (age, sex, level of education, marital status) and behavioral characteristics (alcohol use, tobacco use, self-reported levels of physical activity, food consumption patterns, foot care, and medical and medication history). Physical measurements included weight, height, waist circumference, blood pressure measurements; and a baseline glycated haemoglobin A1c (HbA1c).
Blood pressure (BP) was measured using a digital upper arm sphygmomanometer (Omron® M3 series, Omron Healthcare Inc.), with three measurements taken at least five minutes apart. The mean of the last two BP measurements was used in the analysis. Weight was measured using a digital weighing scale (Seca 813, Hanover, USA) and height was measured using a roll-up stadiometer (Seca 213, Hanover, USA). Body mass index (BMI) was calculated by dividing the participant’s weight in kilograms by the height in meters squared (kg/m2). Waist circumference was measured using a measuring tape with the participant in light clothing. Plasma glucose was measured using a point of care glucometer (Accucheck®) by Roche, which uses 10 microliters (μL) of capillary blood derived from a finger prick, and the Cobas b101 point-of-care analyzer (in Uganda) and the Alfinion AS100 Analyser (in South Africa) were used to measure HbA1c levels in millimoles/mole (mmol/mol) using 5 μL of capillary blood to provide results that are free from hemoglobin variant interference, which is National Glycohemoglobin Standardization Program (NGSP) certified.
Each participant was followed for up to 12 months or until death, withdrawal of consent, or loss to follow-up. Follow-up of participants was completed at all country sites by the end of December 2019.
## Quality control of interventions administration
In both countries, a contextualized structured training program guided intervention administration including pre-intervention training as well as quality control during the implementation phase. In Uganda, regular support supervision visits were conducted to the health facilities to ensure that the interventions were implemented according to protocol and with sufficient dose of exposure, fidelity and reach. During the visits, study team members checked and re-enforced health worker compliance to key elements of the intervention e.g. ensuring no stock out of key medicines and testing equipment, compliance with treatment guidelines, task shifting, maintenance of the patient information system, and supporting the patient role in self-care. Similar support was given to the community level interventions. In South Africa, where the intervention focused mainly on the community interventions, quality assurance visits were implemented to supervise peer support groups to check for intervention fidelity; monthly mock peer group sessions were organized and refresher training in diabetes management. Record keeping practices were also supported to improve reporting feedback.
## Outcomes
The primary outcome among participants with diabetes was glycemic control at month 12, and among participants at high risk of T2D was reduction in HbA1c levels of at least 3 mmol/mol between baseline and month 12. Glycemic control was defined using the American Diabetes Association definition, i.e. having an HbA1c reading below 53 mmol/mol ($7.0\%$ using the Diabetes Control and Complications Trial method) [16] at month 12. Regarding reduction in HbA1c levels of at least 3 mmol/mol by month 12 among participants at high risk of T2D, we decided to use this cutoff based on findings from an intervention trial reported by Katula et al [2021] that aimed at reducing HbA1c levels among people with pre-diabetes [17]. In this study, investigators observed significant reduction in HbA1c levels, an average of -2.52 mmol/mol [$95\%$ CI = -2.89 –-2.16]. Based on this, we decided to use a cutoff reduction of 3 mmol/mol by month 12. Secondary outcomes included: i) retention into care defined as having returned for the month 12 clinic appointment and endline measurements done within 14 days of the appointment date, and, ii) incident T2D diabetes defined as a participant at high risk for T2D whose baseline HbA1c test reading was less than 48 mmol/mol ($6.5\%$), but had an HbA1c test reading of 48 mmol/mol ($6.5\%$) or greater at month 12.
## Sample size
A detailed description of sample size calculations has been reported elsewhere, showing the number of clusters with unequal sizes, the coefficient of intra-cluster correlation (ICC), and the other parameters used in the calculations [13].
## Data management and analysis
At each country site, research data was managed by trained data managers, with regular data cleaning, and quarterly upload into RedCap software onto a server via secure links. The database password was protected at all levels, with access only to authorized study staff.
Important to note is that although the units of allocation to study arms were the health facilities, the units of analysis were the individual participants. We report the percentage of participants with T2D that achieved glycemic control, the percentage of participants at high risk that achieved reduction in HbA1c of at least 3 mmol/mol between baseline and month 12, the percentage of participants at high risk that progressed to T2D by month 12 (incident T2D), and the percentage of participants with T2D and those at high risk that were retained into care. For the two primary outcomes (glycaemic control and reduction in HbA1c of at least 3 mmol/mol), participants that were lost to follow-up, their missing data were imputed using multiple imputations based on a "missing at random assumption". Imputations were created based on a fully conditional specification which imputes multivariate missing data on a variable-by-variable basis in an iterative fashion [18]. For the calculation of the imputation values, we used predictive mean matching, an implicit model which is assumed robust to misspecification. Standard Error (SE) estimates were pooled based on Rubin’s rules across 10 imputed data sets. All model variables and additional extraneous variables (i.e. marital status, employment and income) were used as predictors. The procedure was done using the “Mice” package in R [19]. A sensitivity analysis was run in which missing outcomes were set equal to "not having the outcome of interest". For the secondary outcome retention into care, analyses were conducted using the intention-to-treat analysis strategy [20]. In this strategy, all participants enrolled in the trial are included in the analysis as recommended by CONSORT guidelines.
Adjusted analyses were conducted separately for each country. By design, within each country the data had a two-level structure, i.e. individual level (level 1) and health facility level (level 2). Multilevel modified Poisson regression modeling was used to take into account this hierarchical structure, with individual participants nested within the health facility of enrollment. The health facility indicator was included in the model as a random effect, except for South Africa where only one facility per arm was included. The modified Poisson regression model provides estimated incident rate ratios (IRR) of the outcome of interest, comparing rates of the outcome of interest across the different categories of independent variables. Modified Poisson regression modeling was preferred over logistic regression modeling to avoid underestimation of the standard errors for the estimated risk ratios that is usually encountered with logistic regression modeling when the prevalence of the outcome is greater than $10\%$ [21, 22]. To adjust for imbalances of relevant population characteristics at baseline, adjusted analysis modeling was weighted using propensity scores. These scores were calculated using gradient boosted logistic regression, a supervised nonparametric machine learning technique available in R’s ’twang’ package [23], with the trial arm as the outcome and relevant baseline characteristics of participants as the covariates [24, 25].
Furthermore, we controlled for the various baseline covariates including socio-demographic characteristics (age, sex, education level, marital status), physical characteristics (BMI), behavioral characteristics (tobacco use, alcohol use, physical activity levels), and health status indicators (hypertension status and baseline HbA1c test reading). The covariates were pre-determined during the planning of the study, and selection of these was based on theory, and previous similar research. All statistical analyses were performed using STATA version 14 [26] and R version 4.1.2 [2021-11-01] [27].
## Patient and public involvement
Local community leaders and other key stakeholder were involved in the design and development of the SMART2D intervention elements as described in detail elsewhere [10, 11]. At the end of the trial, the local community leaders, selected study participant representatives and care companions participated in a workshop of dissemination of findings from the trial to policy makers.
## Enrollment and follow-up
Between 11th January 2017 and 30th November 2018 health care facilities and individuals were screened for eligibility to participate in the study. In Uganda, out of the 101 facilities screened nine were selected, whereas in South Africa, out of the 9 facilities, two were selected. In Uganda, 28,175 subjects residing in the nine cluster areas were screened, out of which 801 were enrolled (424 with T2D and 377 at high risk). In South Africa, 1,584 subjects were screened at the selected health facilities out of which 566 were enrolled (281 with T2D and 285 at high risk. Figs 1 and 2 summarize the trial schema for Uganda and South Africa, respectively; giving details of the distribution of numbers from screening through enrollment, loss to follow-up, and number of participants analyzed by study arm; and Table 1 summarizes the number of participants by cluster.
**Fig 1:** *Uganda CONSORT flow chart.* **Fig 2:** *South Africa CONSORT flow chart.* TABLE_PLACEHOLDER:Table 1
## Baseline characteristics of participants
Details of the baseline characteristics of participants, stratified by country site and by study arm are presented in Table 2. Among participants with diabetes, significant differences between the study arms in Uganda were observed only in the percentage of participants with hypertension, which was significantly lower in the integrated care arm than the other two arms. In South Africa, participants in the integrated care arm had significantly lower mean age, BMI and percentage of participants with hypertension but higher mean HbA1c than participants in the facility care arm.
**Table 2**
| Characteristic | Participants with T2D | Participants with T2D.1 | Participants with T2D.2 | Participants with T2D.3 | Participants with T2D.4 | Participants at high risk of T2D | Participants at high risk of T2D.1 | Participants at high risk of T2D.2 | Participants at high risk of T2D.3 | Participants at high risk of T2D.4 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Characteristic | Uganda | Uganda | Uganda | South Africa | South Africa | Uganda | Uganda | Uganda | South Africa | South Africa |
| Characteristic | Facility care (n = 141) | Integrated care (n = 142) | Usual care (n = 141) | Facility care (n = 141) | Integrated care (n = 140) | Facility care (n = 126) | Integrated care (n = 126) | Usual care (n = 125) | Facility care (n = 140) | Integrated care (n = 145) |
| Sex: | | | | | | | | | | |
| Females | 78 (55.3%) | 85 (60.0%) | 75 (53.2%) | 100 (70.9%) | 87 (61.1%) | 103 (81.8%) | 98 (77.8%) | 89 (71.2%) | 98 (70.0%) | 126 (86.9%) |
| | P = 0.440 § | P = 0.440 § | P = 0.513 Ω | P = 0.130 § | P = 0.130 § | P = 0.433 § | P = 0.433 § | P = 0.135 Ω | P = 0.001 § | P = 0.001 § |
| Age in years: | | | | | | | | | | |
| 30–45 | 38 (27.0%) | 31 (21.8%) | 29 (20.6%) | 32 (22.7%) | 48 (34.3%) | 38 (30.2%) | 43 (34.1%) | 34 (27.2%) | 31 (22.1%) | 51 (35.2%) |
| 46–60 | 79 (56.0%) | 84 (59.2%) | 80 (56.7%) | 85 (60.3%) | 66 (47.1%) | 62 (49.2%) | 54 (42.9%) | 61 (48.8%) | 79 (56.4%) | 66 (45.5%) |
| 60–75 | 24(17.0%) | 27 (19.1%) | 32 (2.7%) | 24 (17.0%) | 26 (18.6%) | 26 (20.6%) | 29 (23.0%) | 30 (24.0%) | 30 (21.4%) | 28 (19.3) |
| | P = 0.596 § | P = 0.596 § | P = 0.606 Ω | P = 0.059 § | P = 0.059 § | P = 0.599 § | P = 0.599 § | P = 0.735 Ω | P = 0.046 § | P = 0.046 § |
| Mean ± SD | 52.5 ± 9.9 | 52.9 ± 9.6 | 52.6 ± 9.9 | 52.7 ± 9.3 | 50.1 ± 11.0 | 51.9 ± 10.9 | 50.6 ± 11.7 | 52.5 ± 11.2 | 53.5 ± 9.9 | 49.7 ± 10.6 |
| | P = 0.754 § | P = 0.754 § | P = 0.298 Ω | P = 0.029 § | P = 0.029 § | P = 0.364 § | P = 0.364 § | P = 0.660 | P = 0.002 § | P = 0.002 § |
| Schooling attained: | | | | | | | | | | |
| | 43 (30.5%) | 34 (23.9%) | 32 (22.7%) | 6 (4.3%) | 3 (2.1%) | 44 (34.9%) | 38 (30.2%) | 38 (30.4%) | 3 (2.1%) | 1 (0.7%) |
| Grade 1–7 | 69 (48.9%) | 78 (54.9%) | 73 (51.8%) | 41 (29.1%) | 28 (20.0%) | 58 (46.0%) | 65 (51.6%) | 71 (56.8%) | 38 (27.1%) | 28 (19.3%) |
| Above grade 7 | 29 (20.6%) | 30 (21.1%) | 35 (25.0%) | 94 (66.7%) | 109 (77.9%) | 24 (19.1%) | 23 (18.3%) | 16 (12.8%) | 99 (70.7%) | 116 (80.0%) |
| | P = 0.446 § | P = 0.446 § | P = 0.542 Ω | P = 0.075 § | P = 0.075 § | P = 0.651 § | P = 0.651 § | P = 0.584 Ω | P = 0.179 § | P = 0.179 § |
| Marital status | | | | | | | | | | |
| Never married | 2 (1.4%) | 0 (0.0%) | 3 (2.1%) | 27 (19.2%) | 26 (18.6) | 1 (0.8%) | 0 (0.0%) | 1 (0.8%) | 13 (9.3%) | 35 (24.1%) |
| Married/ cohabiting | 105 (74.5%) | 103 (72.5%) | 103 (73.1%) | 82 (58.2%) | 81 (57.9%) | 85 (67.5%) | 90 (71.4%) | 82 (65.6%) | 78 (55.7%) | 69 (47.6%) |
| Separated/ Divorced | 21 (14.9%) | 19 (13.4%) | 19 (13.5%) | 15 (10.6%) | 11 (7.9%) | 18 (14.3%) | 11 (8.7%) | 20 (16.0%) | 16 (11.4%) | 15 (10.3%) |
| Widowed | 13 (9.2%) | 20 (14.1%) | 16 (11.4%) | 17 (12.1%) | 22 (15.7%) | 22 (17.5%) | 25 (19.8%) | 21 (16.8%) | 33 (23.6%) | 26 (17.9%) |
| | P = 0.358 § | P = 0.358 § | P = 0.628 Ω | P = 0.734 § | P = 0.734 § | P = 0.376 § | P = 0.376 § | P = 0.613 Ω | P = 0.009 § | P = 0.009 § |
| Employment status | | | | | | | | | | |
| Employed in public or private sector | 9 (6.4%) | 13 (9.2%) | 14 (9.9%) | 42 (29.8%) | 56 (40.0%) | 5 (4.0%) | 9 (7.1%) | 8 (6.4%) | 65 (46.4%) | 31 (21.4%) |
| Self-employed | 24 (17.0%) | 23 (16.2%) | 24 (17.0%) | 11 (7.8%) | 8 (5.7%) | 13 (10.3%) | 21 (16.7%) | 21 (16.8%) | 10 (7.1%) | 7 (4.8%) |
| Unemployed | 9 (6.4%) | 3 (2.1%) | 9 (6.4%) | 65 (46.1%) | 61 (43.6%) | 8 (6.4%) | 5 (4.0%) | 5 (4.0%) | 47 (33.6%) | 86 (59.3%) |
| Peasant farmer | 92 (65.3%) | 93 (65.5%) | 87 (61.2%) | - | - | 98 (77.8%) | 87 (69.1%) | 88 (70.4%) | - | - |
| Retired | - | - | 3 (2.1%) | 18 (12.8%) | 13 (9.3%) | - | - | - | 16 (11.4%) | 16 (11.0%) |
| Other | 7 (5.0%) | 10 (7.0%) | 7 (5.0%) | 5 (3.6%) | 2 (1.4%) | 2 (1.6%) | 4 (3.2%) | 3 (2.4%) | 2 (1.4%) | 5 (3.5%) |
| | P = 0.381 § | P = 0.381 § | P = 0.660 Ω | P = 0.329 § | P = 0.329 § | P = 0.285 § | P = 0.285 § | P = 0.638 Ω | P< 0.001 § | P< 0.001 § |
| BMIa (kg/m2) | | | | | | | | | | |
| < 25.0 | 67 (47.5%) | 57 (40.1%) | 67 (47.5%) | 12 (8.5%) | 14 (10.0%) | 62 (49.2%) | 71 (56.4%) | 66 (52.8%) | 16 (11.4%) | 16 (11.2%) |
| 25.0–29.9 | 49 (34.8%) | 53 (37.3%) | 47 (33.3%) | 22 (15.6%) | 44 (31.4%) | 38 (30.2%) | 31 (12.3%) | 38 (30.4%) | 27 (19.3%) | 36 (24.8%) |
| ≥ 30.0 | 25 (17.7%) | 32 (22.5%) | 27 (19.2%) | 107 (75.9%) | 82 (58.6%) | 26 (20.6%) | 24 (19.1%) | 21 (16.8%) | 97 (69.3%) | 93 (64.1%) |
| | P = 0.403 § | P = 0.403 § | P = 0.673 Ω | P = 0.004 § | P = 0.004 § | P = 0.497 § | P = 0.497 § | P = 0.725 Ω | P = 0.527 § | P = 0.527 § |
| Mean ± SD | 26.1 ± 7.4 | 26.6 ± 5.6 | 26.2 ± 8.1 | 36.1 ± 8.3 | 32.8 ± 7.0 | 26.8 ± 16.1 | 25.9 ± 6.6 | 25.2 ± 4.7 | 33.7 ± 7.0 | 34.1 ± 8.5 |
| | P = 0.533 § | P = 0.533 § | P = 0.343 Ω | P< 0.001 § | P< 0.001 § | P = 0.570 § | P = 0.570 § | P = 0.630 Ω | P = 0.640 § | P = 0.640 § |
| Current tobacco use n (%) | 3 (2.1%) | 5 (3.5%) | 5 (3.6%) | 21 (14.9%) | 11 (7.9%) | 1 (0.8%) | 2 (1.6%) | 3 (2.4%) | 19 (13.6%) | 26 (17.9%) |
| | P = 0.723 § | P = 0.723 § | P = 0.826 Ω | P = 0.090 § | P = 0.090 § | P = 0.992 § | P = 0.992 § | P = 0.596 Ω | P = 0.313 § | P = 0.313 § |
| Alcohol use, n (%) | | | | | | | | | | |
| 1–7 days/week | 10 (7.1%) | 13 (9.2%) | 11 (7.8%) | 9 (6.4%) | 18 (12.9%) | 11 (8.7%) | 18 (14.3%) | 11 (8.8%) | 13 (9.3%) | 37 (25.5%) |
| < 4 days/ mth | 2 (1.4%) | 5 (3.5%) | 3 (2.1%) | 18 (12.8%) | 17 (12.1%) | 1 (0.8%) | 4 (3.2%) | 1 (0.8%) | 16 (11.4%) | 21 (14.5%) |
| Never | 129 (91.5%) | 123 (86.6%) | 126 (89.4% | 114 (80.9%) | 104 (74.3%) | 114 (90.5%) | 104 (82.5%) | 113 (90.4%) | 111 (79.3%) | 87 (60.0%) |
| | P = 0.395 § | P = 0.395 § | P = 0.760 Ω | P = 0.176 § | P = 0.176 § | P = 0.142 § | P = 0.142 § | P = 0.248 Ω | P< 0.001 § | P< 0.001 § |
| Hypertensivec, n (%) | 93 (64.0%) | 71 (50.0%) | 86 (61.0%) | 62 (44.0%) | 35 (25.0%) | 77 (61.1%) | 75 (59.5%) | 53 (42.4%) | 57 (40.7%) | 47 (32.4%) |
| | P = 0.007 § | P = 0.007 § | P = 0.020 Ω | P = 0.001 § | P = 0.001 § | P = 0.797 § | P = 0.797 § | P = 0.004 Ω | P = 0.146 § | P = 0.146 § |
| Equivalent of at least 150 mins of moderate physical activity/ wk? | | | | | | | | | | |
| Yes | 111 (78.7) | 122 (85.9%) | 118 (83.7%) | 111 (78.7%) | 99 (70.7%) | 109 (86.5%) | 116 (92.1%) | 111 (88.8%) | 114 (81.4%) | 118 (81.4%) |
| | P = 0.142 § | P = 0.142 § | P = 0.323 Ω | P = 0.122 § | P = 0.122 § | P = 0.221 § | P = 0.221 § | P = 0.373 Ω | P = 0.991 § | P = 0.991 § |
| HbA1c (%) | | | | | | | | | | |
| Mean ± SD | 9.6 ± 2.4 | 9.5 ± 2.3 | 9.5 ± 2.5 | 8.3 ± 2.1 | 8.9 ± 2.5 | 5.8 ± 0.73 | 5.7 ± 0.85 | 5.9 ± 0.8 | 5.9 ± 0.79 | 5.6 ± 0.83 |
| | P = 0.755 § | P = 0.755 § | P = 0.954 Ω | P = 0.042 § | P = 0.042 § | P = 0.434 § | P = 0.434 § | P = 0.155 Ω | P = 0.008 § | P = 0.008 § |
Among participants at high risk for T2D in Uganda, significant differences were again observed between the study arms only in the percentage with hypertension, however in this group the usual care arm had the lowest percentage (Table 2). In South Africa, significant differences between the study arms were observed with more female, younger and never married, and unemployed participants with higher level of alcohol use and lower mean HbA1c in the integrated care arm.
## Effectiveness of the trial interventions on glycemic control among participants with T2D
Among participants with T2D, bi-variable analysis showed a significantly higher percentage of glycemic control in Uganda at $29.2\%$, compared to $16.4\%$ in South Africa ($p \leq 0.001$). We observed no significant differences in the crude percentages of participants with glycemic control between the study arms in either of the two countries. Adjusted analysis also revealed no significant differences in the estimated rates of glycemic control between any of the study arms in either of the two countries, except for the facility versus usual care arm in Uganda with an incidence rate ratio (IRR) of 0.71 [0.52–0.96] (Table 3).
**Table 3**
| Country | Study arm | Glycaemic control among participants with T2D | Glycaemic control among participants with T2D.1 | Glycaemic control among participants with T2D.2 | Glycaemic control among participants with T2D.3 | Reduction in HbA1c of at least 3 mmol/mol among participants at high risk of T2D | Reduction in HbA1c of at least 3 mmol/mol among participants at high risk of T2D.1 | Reduction in HbA1c of at least 3 mmol/mol among participants at high risk of T2D.2 | Reduction in HbA1c of at least 3 mmol/mol among participants at high risk of T2D.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Country | Study arm | number enrolled | Number with glycemic control (%) | Crude IRR [95% CI] | Adjusted IRR [95% CI] | number enrolled | Number with reduction in HbA1c ≥ 3 mmol/mol (%) | Crude IRR [95% CI] | Adjusted IRR [95% CI] |
| | Usual care | 141 | 46 (32.6%) | 1.0 | 1.0 | 125 | 22 (17.6%) | 1.0 | 1.0 |
| Uganda | Facility care | 141 | 42 (29.8%) | 0.67 [0.47–0.95] | 0.71 [0.52–0.96]a | 126 | 47 (37.3%) | 0.92 [0.69–1.22] | 0.92 [0.70–1.22]a |
| | Integrated care | 142 | 36 (25.4%) | 0.64 [0.44–0.94] | 0.66 [0.42–1.03]a | 126 | 69 (54.8%) | 0.95 [0.71–1.27] | 0.97 [0.74–1.26]a |
| | sub-total | 424 | 124 (29.2%) | | | 377 | 138 (36.6%) | | |
| | ICC = 0.028 | ICC = 0.028 | ICC = 0.028 | | | ICC = 0.043± | ICC = 0.043± | ICC = 0.043± | |
| South Africa | Facility care | 141 | 25 (17.7%) | 1.0 | 1.0 | 140 | 9 (6.4%) | 1.0 | 1.0 |
| | Integrated care | 140 | 21 (15.0%) | 1.01 [0.71–1.62] | 1.08 [0.71–1.62]b | 145 | 7 (4.8%) | 1.00 [0.35–2.84] | 1.08 [0.38–3.10]b |
| | sub-total | 281 | 46 (16.4%) | | | 285 | 16 (5.6%) | | |
## Effectiveness of the trial interventions on reduction in HbA1c levels among participants at high risk of T2D
Bi-variable analysis revealed a significantly higher percentage of participants at high risk for T2D achieved a reduction in HbA1c of at least 3 mmol/mol in Uganda at $36.6\%$, compared to $5.6\%$ in South Africa ($p \leq 0.001$). In South Africa, we observed no statistically significant difference in the crude percentages between the two study arms. However in Uganda, the crude percentage of participants with reduction in HbA1c of at least 3 mmol/mol was significantly higher in the integrated care arm at $54.8\%$, followed by $37.3\%$ in the facility care arm, and significantly lower in the usual care arm at $17.6\%$ ($p \leq 0.001$). The crude percentage with reduction in HbA1c of at least 3 mmol/mol of $54.8\%$ was significantly higher from the percentage with reduction in HbA1c of at least 3 mmol/mol in the facility care arm of $37.3\%$ ($$p \leq 0.005$$).
Adjusted analysis revealed no significant differences in the estimated rates of reduction in HbA1c of at least 3 mmol/mol between the study arms in either of the two countries (Table 3).
## Effectiveness of the trial interventions on retention into care among participants with T2D
Overall among participants with T2D, bi-variable analysis revealed a significantly higher percentage of retention into care in Uganda at $75.0\%$, compared to $48.0\%$ in South Africa ($p \leq 0.001$). In neither country did we observe a significant difference in the percentage of participants retained into care between the facility care arm and the integrated care arm. In Uganda, there was a significantly lower retention into care in the usual care arm at $58.9\%$, compared to the integrated care and facility care arms combined ($83.0\%$) ($p \leq 0.001$) (Table 4).
**Table 4**
| Country | Study arm | Retention into care among participants with T2D | Retention into care among participants with T2D.1 | Retention into care among participants with T2D.2 | Retention into care among participants with T2D.3 | Retention into care among participants at high risk of T2D | Retention into care among participants at high risk of T2D.1 | Retention into care among participants at high risk of T2D.2 | Retention into care among participants at high risk of T2D.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Country | Study arm | Number enrolled | Number retained into care (%) | Crude IRR [95% CI] | Adjusted IRR [95% CI] | Number enrolled | Number retained into care (%) | Crude IRR [95% CI] | Adjusted IRR [95% CI] |
| Uganda | Usual care | 141 | 83 (58.9%) | 1.0 | 1.0 | 125 | 31 (24.8%) | 1.0 | 1.0 |
| | Facility care | 141 | 116 (82.3%) | 1.40 [1.19–1.64] | 1.41 [1.08–1.84]g | 126 | 77 (61.1%) | 2.46 [1.76–3.45] | 2.46 [1.33–4.53]m |
| | Integrated care | 142 | 119 (83.8%) | 1.42 [1.22–1.66] | 1.41 [1.09–1.83]g | 126 | 109 (86.5%) | 3.49 [2.55–4.77] | 3.52 [2.13–5.80]m |
| | sub-total | 424 | 318 (75.0%) | | | 377 | 217 (57.6%) | | |
| | ICCΣ = 0.09306 (SE = 0.05142) | ICCΣ = 0.09306 (SE = 0.05142) | ICCΣ = 0.09306 (SE = 0.05142) | | | ICCΦ = 0.33952 (SE = 0.11861) | ICCΦ = 0.33952 (SE = 0.11861) | ICCΦ = 0.33952 (SE = 0.11861) | |
| South Africa | Facility care | 141 | 73 (51.8%) | 1.0 | 1.0 | 140 | 68 (48.6%) | 1.0 | 1.0 |
| | Integrated care | 140 | 62 (44.3%) | 0.86 [0.67–1.09] | 1.15 [1.10–1.19]h | 145 | 74 (51.0%) | 1.05 [0.83–1.33] | 1.02 [1.02–1.03]n |
| | sub-total | 281 | 135 (48.0%) | | | 285 | 142 (49.8%) | | |
| | ICC Ω = 0.00408 (SE = 0.01576) | ICC Ω = 0.00408 (SE = 0.01576) | ICC Ω = 0.00408 (SE = 0.01576) | | | ICCψ < 0.00001 (SE = 0.00994) | ICCψ < 0.00001 (SE = 0.00994) | ICCψ < 0.00001 (SE = 0.00994) | |
Adjusted analysis showed that in Uganda compared to the usual care arm, rates of retention into care were significantly higher in the facility care arm (IRR = 1.41 [1.08–1.84]), and the integrated care arm (IRR = 1.41 [1.09–1.83]). In South Africa, rates of retention into care were significantly higher in the integrated care arm compared to those in the facility care arm (IRR = 1.15 [1.10–1.19]) (Table 4).
## Effectiveness of the trial interventions on retention into care among participants at high risk of T2D
Among participants at high risk of T2D, a significantly higher percentage of retention into care was observed in Uganda at $57.6\%$, compared to $49.5\%$ in South Africa ($$p \leq 0.039$$). No significant differences were observed in South Africa between the two study arms. However in Uganda, a significantly higher percentage of retention into care was observed in the integrated care arm at $86.5\%$ compared to that in the facility care arm at $61.1\%$ ($p \leq 0.001$). Moreover, a significantly lower percentage of retention into care was observed in the usual care arm at $24.8\%$, compared to the other two arms combined ($p \leq 0.001$) (Table 4).
Adjusted analysis showed that in Uganda, compared to the usual care arm, rates of retention into care were significantly higher in the facility care arm with an IRR = 2.46 [1.33–4.43], and in the integrated care arm with an IRR = 3.52 [2.13–5.80]. However there was no significant difference in rates of retention into care in the integrated care arm compared to the facility care arm.
In South Africa, rates of retention into care were significantly higher in the integrated care arm compared to the facility care arm, with an IRR = 1.02 [1.02–1,03] (Table 4).
## Effectiveness of the trial interventions on Incident T2D among participants at high risk
A total of 328 participants at high risk for T2D had an HbA1c reading of less than $6.5\%$ at baseline. Of these, 15 ($4.6\%$) progressed to T2D; of which 9 out of 217 ($4.1\%$) were in Uganda, and 6 out of 142 ($4.2\%$) were in South Africa ($$p \leq 0.971$$). We found no significant difference in the percentage of participants at high risk that progressed to T2D between the study arms, neither in Uganda, nor in South Africa. Because of the small number of incident T2D cases observed in the study, adjusted analysis comparing incident rates of T2D between the study arms was not possible.
## Intra-Cluster Correlation coefficients
The intra-cluster correlation coefficients, in respect to each study outcome of interest are presented in Tables 3 and 4.
## Discussion
We found significant improvements in retention into care comparing improved facility care to current usual care, and an added value of integrated care that included peer support in both countries. Overall, the effect size estimates of integrated care on retention into care were larger in Uganda, especially among participants at high risk of T2D.
The significant effect of the interventions on retention into care implies that a considerably less proportion of participants was lost to follow-up in the intervention arms of the study, that is, in South Africa the integrated care arm; and in Uganda the facility care and integrated care arms. This effect was relatively high in some study arms compared to others, resulting in highly unbalanced data between control (i.e. usual care) and intervention arms (i.e. facility only or integrated care). Because of this considerable and unbalanced loss to follow-up, the validity of our findings with regards to the effect the study intervention on the primary outcomes of glycaemic control and reduction in HbA1c might be questionable, despite having used multiple imputations [28] to address the missing data during analysis. Sensitivity analysis comparing the results of the multiple imputation analysis with those of intention-to-treat in which all cases lost to follow-up were set ’not having the outcome of interest’ [20] (see S1 Table), reveals strong differences between both approaches which may further question the validity of our results. The unexpectedly high number of missingness also reduced the power of our study which may explain why we did not find significant differences between the study arms for the primary outcomes. We conclude that in pragmatic implementation trials such as the present study with the objective to enhance retention into care and glycemic control, may result in unbalanced missingness, hence compromising further assessment of the effectiveness.
Although there is evidence for community-based peer support interventions in high-income countries, research from LMICs have shown inconsistent associations with improvements in clinical, behavioral, and psychological outcomes [6]. In our study, all positive findings were similar for both the integrated care arm and the facility care arm in Uganda, suggesting a stronger role for facility care. Still in Uganda, whereas community-based peer support strategies added no value for participants with T2D, participants at high risk benefited from integrated care in terms of retention into care. Put together, our findings point to the need for an adequately functioning primary care system for T2D management and a role of the community for T2D prevention.
In the only previous study on peer support for diabetes self-management we were able to find in the Uganda context [29], Baumann et al [2015] conducted a small quasi-experimental study and their findings indicated improvements in HbA1c, diastolic blood pressure, and dietary patterns. In our study, the intervention did not show a significant effect on reduction in HbA1c of at least 3 mmol/nol, but as mentioned before, the unbalanced loss-to-follow-up compromised valid analysis of the data. It is notable, however, that minimal changes at the facility level such as ensuring no medication stock-outs, availability of clinical monitoring equipment, etc. in Uganda, led to significant improvements in retention into care; especially among participants at high risk of T2D. Stock-outs and poor availability of essential medication and diagnostic tests are known barriers to diabetes and non-communicable disease management and prevention in Sub-Saharan Africa [30, 31].
Further in Uganda where the standard of usual care is low for patients with T2D, and participants at high risk remain largely unidentified and untreated [32], our study was able to significantly improve retention into care in both participant groups, and in both intervention arms. However, people without a T2D diagnosis but with a high-risk diagnosis are a new patient group for health facilities and it seems community involvement was able to further improve their retention into care and HbA1c outcomes. Taken together, the findings from Uganda call for further research on potential mediating factors related to self-management behaviors as well as implementation outcomes.
The non-significant findings in South Africa in regards to reduction in HbA1c were similar to an earlier peer support intervention in the same context [33] which showed negligible effect in terms of primary and secondary outcomes. In that study, most intervention participants had not attended a single intervention session, and a qualitative process evaluation [34] revealed that intervention fidelity was also constrained. Despite continued efforts to monitor intervention fidelity in our study, it is probable that failure to achieve a significant difference in HbA1c in South Africa could be attributed to inadequate intervention implementation rather than intervention efficacy. Findings of an extensive process evaluation are currently being analyzed and already point to several issues, such as a low absorptive capacity among Community Health Workers (CHWs), insufficient supervision of CHWs and rapid staff turnover and organizational change at the level of the implementing partner; security concerns, and low engagement of study participants which proved difficult without incentivization (e.g. FPG and BP measurement).
In their systematic review on peer support interventions for diabetes self-management Werfalli et al [6] pointed at several reasons for inconsistent findings: low quality of studies, implementation issues, and lack of sufficient (contextualized) underlying theory to build the intervention. As discussed above, our study faced many challenges related to implementation especially in South Africa. Another challenge was related to selection of health care facilities as trial clusters. In Uganda, the study area was large enough to avoid contamination between trial arms and to have enough clusters for successful randomization, which resulted in minor differences only between the study arms at baseline. In South Africa, where many other studies were conducted simultaneously in the same geographically restricted area, only two facilities could be selected for feasibility reasons, with large differences in demographic catchment areas as reflected in the baseline differences between both study arms. Among participants with T2D and those at high risk, the integrated care arm had significantly younger participants with a better risk factor profile. These differences imply that participants in the integrated care arm might have been less likely to adhere to the intervention due to e.g., work and family obligations, and less likely to show benefits of the intervention due to their risk profile.
In settings like the Uganda rural study setting where the current standard of care for people at risk of or with T2D is still poor, basic improvements in facility care improves retention into care. In this context we also see added value of community-based peer support to provide integrated care. We therefore recommend implementation of such measures, as minimal resources are needed. However, in South Africa, a middle-income urban setting with higher facility standard of care and highly mobile populations, we could not demonstrate added value of integrated interventions beyond facility care. Here, more studies are called for to identify strategies that can further increase retention and glycemic control.
## Strengths and limitations
The SMART2D pragmatic cluster randomized trial evaluated the added benefit of community intervention strategies on type 2 diabetes outcomes, beyond optimized health facility strategies in Uganda and South Africa, exemplifying a low- and middle-income setting that are facing challenges in tackling the type 2 diabetes burden. The study was designed with a solid evidence-base including, for example, the Self-Determination Theory (SDT), and utilized evidence-based and contextualized strategies contextualized based upon formative research and collaboration with local and sub-national stakeholders, to ensure relevance, acceptability and feasibility for scale-up. The study was able to demonstrate that when minimum level quality care is lacking, even minimal improvements in the facility care like training nurses to follow-up patients and encourage participants to keep appointments; can provide significant improvements in retention into care and, potentially, in diabetes management outcomes. However, since the interventions were implemented as a package of intervention elements, we were unable to identify specific intervention elements that contributed towards improved T2D prevention and diabetes management outcomes.
A major strength of our study was its theoretical underpinning. Many intervention studies either lack any explicit theory of change or they utilize a theory that has not been created or tested in a LMIC context. Our formative research provided evidence for SDT with regards to physical activity in the Uganda site and healthy eating and physical activity in the South Africa site [35–37]. Furthermore, SDT formed an essential part of the SMART2D self-management framework developed for this study [15]. Other strengths include the comprehensive situational analysis in all participating country sites [10, 15, 35], participatory intervention development including research teams and key stakeholders [11], and a comprehensive process evaluation aiming for a deeper understanding of the effect of the intervention components through the analyses of secondary outcomes as potential mediators of change, as well as enablers and barriers for implementation. Together, these analyses can further explain which, why and how specific components were implemented and this may explain change in secondary and main outcomes.
Our study had some limitations that are important to point out. First, due to several ongoing studies in the South African study site, selection of clusters was limited and resulted in several baseline differences between the study arms that might have influenced the findings. Furthermore, ours was a complex health intervention and each intervention strategy had one or more intervention elements as described in the trial protocol publication [13]. Intervention elements differed slightly between both countries because of contextual adaptation. Complex health interventions are not usually generalizable except for their main ingredients, as implementation in different settings should always take context into account.
## References
1. 1WHO. Global Report on Diabetes [Internet]. Vol. 978, Isbn. 2016. Available from: https://sci-hub.si/https://apps.who.int/iris/handle/10665/204874%0Ahttps://apps.who.int/iris/bitstream/handle/10665/204874/WHO_NMH_NVI_16.3_eng.pdf?sequence=1%0Ahttp://www.who.int/about/licensing/copyright_form/index.html%0Ahttp://www.who.int/about/licens. *Global Report on Diabetes [Internet]* (2016) **978**
2. Theo Lim SS, Abbafati C, Abbas KM, Abbasi M, Abbasifard M, Abbasi-Kangevari M. **Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019**. *Lancet (London, England) [Internet]* (2020) **396** 1204-22. PMID: 33069326
3. 3ADA. Advancements For Life: 2020 Research Report [Internet]. 2020. Available from: https://www.diabetes.org/reports. *Advancements For Life: 2020 Research Report [Internet]* (2020)
4. Manne-Goehler J, Geldsetzer P, Agoudavi K, Andall-Brereton G, Aryal KK, Bicaba BW, Wareham NJ. **Health system performance for people with diabetes in 28 low- and middle-income countries: A cross-sectional study of nationally representative surveys**. *PLOS Med [Internet]* (2019) **16** e1002751. DOI: 10.1371/journal.pmed.1002751
5. Stokes A, Berry KM, Mchiza Z, Parker W-A, Labadarios D, Chola L. **Prevalence and unmet need for diabetes care across the care continuum in a national sample of South African adults: Evidence from the SANHANES-1, 2011–2012**. *PLoS One [Internet]* (2017) e0184264. PMID: 28968435
6. Werfalli M, Raubenheimer PJ, Engel M, Musekiwa A, Bobrow K, Peer N. **The effectiveness of peer and community health worker-led self-management support programs for improving diabetes health-related outcomes in adults in low- and-middle-income countries: a systematic review**. *Syst Rev [Internet]* (2020) **9** 133. PMID: 32505214
7. Tabák AG, Herder C, Rathmann W, Brunner EJ, Kivimäki M. **Prediabetes: a high-risk state for diabetes development**. *Lancet (London, England) [Internet]* (2012) **379** 2279-90. DOI: 10.1016/S0140-6736(12)60283-9
8. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *N Engl J Med [Internet]* (2002) **346** 393-403. DOI: 10.1056/NEJMoa012512
9. Lindström J, Ilanne-Parikka P, Peltonen M, Aunola S, Eriksson JG, Hemiö K. **Sustained reduction in the incidence of type 2 diabetes by lifestyle intervention: follow-up of the Finnish Diabetes Prevention Study**. *Lancet (London, England) [Internet]* (2006) **368** 1673-9. DOI: 10.1016/S0140-6736(06)69701-8
10. Van Olmen J, Delobelle P, Guwatudde D, Absetz P, Sanders D, Alvesson HM. **Using a cross-contextual reciprocal learning approach in a multisite implementation research project to improve self-management for type 2 diabetes**. *BMJ Glob Heal* (2018) **3** 1-9
11. Absetz P, Van Olmen J, Guwatudde D, Puoane T, Alvesson HM, Delobelle P. **SMART2D - Development and contextualization of community strategies to support self-management in prevention and control of type 2 diabetes in Uganda, South Africa, and Sweden**. *Transl Behav Med.* (2020) **10** 25-34. DOI: 10.1093/tbm/ibz188
12. Richard R, Edward D. **Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being**. *Am Psychol [Internet]* (2000) **55** 68-78. PMID: 11392867
13. Guwatudde D, Absetz P, Delobelle P, Östenson C-G, Olmen Van J, Alvesson HM. **Study protocol for the SMART2D adaptive implementation trial: a cluster randomised trial comparing facility-only care with integrated facility and community care to improve type 2 diabetes outcomes in Uganda, South Africa and Sweden**. *BMJ Open [Internet]* (2018) **8** e019981
14. Ryan RM, Deci Unwersity EL, Rochestu J. **Overview of Self-Determination Theory: An Organismic Dialectical Perspective**. *HANDBOK OF SELF-DETERMINATION [Internet]* (2014) 1-33
15. De Man J, Aweko J, Daivadanam M, Alvesson HM, Delobelle P, Mayega RW. **Diabetes self-management in three different income settings: Cross-learning of barriers and opportunities**. *PLoS One [Internet]* (2019) **14** e0213530. DOI: 10.1371/journal.pone.0213530
16. Davies MJM, D’Alessio DA, Fradkin J, Kernan WN, Mathieu C, Mingrone G. **Management of Hyperglycemia in Type 2 Diabetes, 2018. A Consensus Report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD)**. *Diabetes Care [Internet]* (2018) **41** 2669-701. PMID: 30291106
17. Katula JA, Dressler E V, Kittel CA, Harvin LN, Almeida FA, Wilson KE. **Effects of a Digital Diabetes Prevention Program: An RCT**. *Am J Prev Med [Internet]* (2022)
18. Van Buuren S, Brand JPL, Groothuis-Oudshoorn CGM, Rubin DB. **Fully conditional specification in multivariate imputation**. *J Stat Comput Simul [Internet]* (2006) **76** 1049-64
19. van Buuren S, Groothuis-Oudshoorn K. **mice: Multivariate Imputation by Chained Equations in R**. *J Stat Softw [Internet]* (2011) **45** 1-67
20. Montori VM, Guyatt GH. **Intention-to-treat principle**. *Cmaj* (2001) **165** 1339-41. PMID: 11760981
21. Zou G.. **A modified poisson regression approach to prospective studies with binary data**. *Am J Epidemiol* (2004) **159** 702-6. DOI: 10.1093/aje/kwh090
22. McNutt L-A, Wu C, Xue X, Hafner JP. **Estimating the relative risk in cohort studies and clinical trials of common outcomes**. *Am J Epidemiol [Internet]* (2003) **157** 940-3. DOI: 10.1093/aje/kwg074
23. Cefalu M, Ridgeway G, McCaffrey D, Morral A, Griffin B. **;, Burgette L. twang: Toolkit for Weighting and Analysis of Nonequivalent Groups.**. *R package version 2.4 [Internet]* (2021)
24. Guo S, Fraser MW. *Propensity score analysis : statistical methods and applications [Internet].* (2014) 421
25. Lalani N, Jimenez RB, Yeap B. **Understanding Propensity Score Analyses**. *Int J Radiat Oncol Biol Phys [Internet]* (2020) **107** 404-7. DOI: 10.1016/j.ijrobp.2020.02.638
26. 26Statacorp. Stata Statistical Software: Release 14 [Internet]. Texas, USA: StataCorp LP, 4905 Lakeway Drive, College Station, Texas 77845 USA; 2015. Available from: https://www.stata.com/. *Stata Statistical Software: Release 14 [Internet]* (2015)
27. 27R-Core-Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. [Internet]. 2021 [cited 2022 Feb 23]. Available from: https://www.google.com/search?client=firefox-b-d&q=R+Core+Team+%282021%29.+R%3A+A+language+and+environment+for+statistical+computing.+R+Foundation+for+Statistical+Computing%2C+++Vienna%2C+Austria. *R: A language and environment for statistical computing* (2021)
28. Jakobsen JC, Gluud C, Wetterslev J, Winkel P. **When and how should multiple imputation be used for handling missing data in randomised clinical trials–a practical guide with flowcharts**. *BMC Med Res Methodol [Internet]* (2017) **17** 162. PMID: 29207961
29. Baumann LC, Frederick N, Betty N, Jospehine E, Agatha N. **A demonstration of peer support for Ugandan adults with type 2 diabetes**. *Int J Behav Med [Internet]* (2015) **22** 374-83. DOI: 10.1007/s12529-014-9412-8
30. Kuupiel D, Tlou B, Bawontuo V, Drain PK, Mashamba-Thompson TP. **Poor supply chain management and stock-outs of point-of-care diagnostic tests in Upper East Region’s primary healthcare clinics, Ghana**. *PLoS One* (2019) **14** 1-15. DOI: 10.1371/journal.pone.0211498
31. Mukundiyukuri JP, Irakiza JJ, Nyirahabimana N, Ng’ang’a L, Park PH, Ngoga G. *Ann Glob Heal* (2020) **86** 1-15. DOI: 10.5334/aogh.2729
32. 32Uganda Ministry of Health. UGANDA CLINICAL GUIDELINES 2016—National Guidelines on Management of Common Conditions [Internet]. Uganda Ministry of Health; 2016. p. 449–455 (Section 8.1.3 Diabetes Mellitus). Available from: http://library.health.go.ug/publications/guidelines/uganda-clinical-guidelines-2016. *UGANDA CLINICAL GUIDELINES 2016—National Guidelines on Management of Common Conditions [Internet]* (2016) 449-455
33. Mash RJ, Rhode H, Zwarenstein M, Rollnick S, Lombard C, Steyn K. **Effectiveness of a group diabetes education programme in under-served communities in South Africa: a pragmatic cluster randomized controlled trial**. *Diabet Med [Internet]* (2014) **31** 987-93. DOI: 10.1111/dme.12475
34. Botes AS, Majikela-Dlangamandla B, Mash R. **The ability of health promoters to deliver group diabetes education in South African primary care**. *African J Prim Heal Care Fam Med [Internet]* (2013) **5**
35. De Man J, Wouters E, Absetz P, Daivadanam M, Naggayi G, Kasujja FX. **What Motivates People With (Pre)Diabetes to Move? Testing Self-Determination Theory in Rural Uganda**. *Front Psychol [Internet]* (2020) **11** 404. DOI: 10.3389/fpsyg.2020.00404
36. De Man J, Wouters E, Delobelle P, Puoane T, Daivadanam M, Absetz P. **Testing a Self-Determination Theory Model of Healthy Eating in a South African Township.**. *Front Psychol [Internet]* (2020) **11** 2181. PMID: 32982885
37. De Man J, Kasujja FX, Delobelle P, Annerstedt KS, Alvesson HM, Absetz P. **Motivational determinants of physical activity in disadvantaged populations with (pre)diabetes: a cross-cultural comparison**. *BMC Public Health [Internet]* (2022) **22** 164. PMID: 35073882
|
---
title: '‘The broker also told me that I will not have problems after selling because
we have two and we can survive on one kidney’: Findings from an ethnographic study
of a village with one kidney in Central Nepal'
authors:
- Bijaya Shrestha
- Bipin Adhikari
- Manash Shrestha
- Ankit Poudel
- Binita Shrestha
- Dev Ram Sunuwar
- Shiva Raj Mishra
- Luechai Sringernyuang
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021627
doi: 10.1371/journal.pgph.0000585
license: CC BY 4.0
---
# ‘The broker also told me that I will not have problems after selling because we have two and we can survive on one kidney’: Findings from an ethnographic study of a village with one kidney in Central Nepal
## Abstract
Kidney selling is a global phenomenon engraved by poverty and governance in low-income countries with the higher-income countries functioning as recipients and the lower-income countries as donors. Over the years, an increasing number of residents in a village near the capital city of Nepal have sold their kidneys. This study aims to explore the drivers of kidney selling and its consequences using ethnographic methods and multi-stakeholder consultations. An ethnographic approach was used in which the researcher lived and observed the residents’ life and carried out formal and informal interactions including in-depth interviews with key informants, community members and kidney sellers in Hokse village, Kavrepalanchok district. Participants in the village were interacted by researchers who resided in the village. In addition, remote interviews were conducted with multiple relevant stakeholders at various levels that included legal workers, government officers, non-government organization (NGO) workers, medical professionals, and policymaker. All formal interviews were audio-recorded for transcription in addition to field notes and underwent thematic analysis. The study identified processes, mechanisms, and drivers of kidney selling. Historically, diversion of a major highway from the village to another village was found to impact the livelihood, economy and access to the urban centres, ultimately increasing poverty and vulnerability for kidney selling. Existing and augmented deprivation of employment opportunities were shown to foster emigration of villagers to India, where they ultimately succumbed to brokers associated with kidney selling. Population in the village also maintained social cohesion through commune living, social conformity (that had a high impact on decision making), including behaviours that deepened their poverty. Behaviours such as alcoholism, trusting and following brokers based on the persuasion and decision of their peers, relatives, and neighbours who became the new member of the kidney brokerage also contributed to kidney selling. The other reasons that may have influenced high kidney selling were perceived to be a poor level of education, high demands of kidneys in the market and an easy source of cash through selling. In Hokse village, kidney selling stemmed from the interaction between the brokers and community members’ vulnerability (poverty and ignorance), mainly as the brokers raised false hopes of palliating the vulnerability. The decision-making of the villagers was influenced heavily by fellow kidney sellers, some of whom later joined the network of kidney brokers. Although sustained support in livelihood, development, and education are essential, an expanding network and influence of kidney brokers require urgent restrictive actions by the legal authority.
## Kidney selling and its epidemiological burden
Globally, 150,000 kidneys are transplanted annually, of which the majority arrived through the kidney selling [1]. The volume of kidney selling has increased steadily over the years—accumulating $US1.7 billion transaction annually [2, 3]. Kidney selling is associated with a growing burden of End-Stage Renal Diseases (ESRD) globally. Of those with ESRD, nearly 2.5 million people (projected to double to 5.4 million by 2030) are currently in renal replacement therapy (RRT) and another 2.3–7.1 die prematurely due to lack of RRT and transplantation [4]. The growing burden of ESRD and demand for transplantation (a feasible alternative against long term dialysis) has fueled kidney selling in low-income countries—orchestrated often through an informal network by the brokers [5].
## Global scenario of kidney selling
With an exception of Iran, kidney selling is illegal around the world [6]. The phenomenon of kidney selling was first tracked by Nancy Scheper-Hughes during her field visit to the shantytowns of Brazil, which she framed as ‘body-snatching rumours’ [7] and her continued investigations in South America, Africa, Europe, and Asia [8]. Kidney selling is prevalent in countries such as Brazil, China, Moldova, the Philippines, Romania, Turkey, Egypt, Bolivia, and Peru, while the buyers come from high-income countries like Saudi Arabia, Israel, Oman, Japan, Australia, United States, Canada, and Italy [9].
In South Asia, the Indian subcontinent (e.g., India, Bangladesh, and Pakistan) in particular is a prominent hotspot for kidney selling. In these countries, brokers often persuade people to sell their kidneys through false information about the bodily function of kidneys and by enticing them with a considerable sum of financial incentive [8–11]. Countries in the Indian subcontinent are therefore considered as the ‘kidney hub’ for the rich westerners [9, 12].
## Potential reasons for kidney selling in the Indian subcontinent
Among several reasons, poverty is a significant contributor that triggers kidney selling in the Indian subcontinent [9]. In his 2001 article, anthropologist Lawrence Cohen narrated the stories of poor *Indian slum* dwellers who sold their organs to raise money for the dowry of their daughters [13]. Echoing India, countries such as Pakistan and Bangladesh have also emerged as safe places for kidney transplantation. Poor people are drawn into kidney selling in the dreams of overcoming their poverty, debt and sometimes kidney donation (selling) is framed as an act of moral obligation and altruism referring to religious beliefs [14–18]. Furthermore, the unregulated medical industry and lackadaisical regulatory measures in these low- and middle-income countries have created a favorable environment for the kidney selling [15, 19].
## Context of kidney selling in Nepal
Nepal is a small country sandwiched between two giant neighbors, China in the North and India elsewhere. The federal democratic republic of *Nepal is* divided into seven provinces. Each province has eight to 14 districts. The districts make up the local administrative units referred to as urban and rural municipalities. There are 753 local units, including six metropolitan municipalities, 11 sub-metropolitan municipalities, 276 municipalities, and 460 rural municipalities [20]. Nepal remains one of the most volatile areas for political instability with a decade-long Maoist civil war (1996−2006) and frequent political interferences [21]. In addition, unemployment, lack of infrastructure development, caste hierarchy, and illiteracy make Nepalese people vulnerable to cheap labor and organ trafficking [22, 23]. To compound the situation, the porous border between Nepal and India have facilitated many Nepalese people in succumbing to the organ trafficking [24, 25].
Hokse, a small village located in the Kavrepalanchok district situated 50 km northeast of Kathmandu, has been established as a hub for kidney selling. Over the past decades, nearly a third of the villagers have travelled to India to sell their organs. The village is infamous for being one of the cheapest sources for kidney buyers and is being stigmatized as a ‘kidney village’ in Nepal [9]. Previous accounts of kidney sellers are filled with deception by the brokers, false expectations, medical complications, and adverse consequences. The monetary incentives of kidney selling also seem to be driven by ultimate purchase capacity to buy phones, televisions, motorbikes, houses, and land Fields [26].
This research explored the reasons behind kidney selling in Hokse village, using ethnographic methods. Although kidney selling is a global phenomenon, Hokse village bears a unique social, cultural, and demographic characteristic, making it critical to explore the complex interplay of the plethora of factors behind concentrated kidney selling in the village.
## Setting
The study was carried out in the Hokse village, ward numbers 7 and 8 of Paanchkhal Municipality, Kavrepalanchok district (Fig 1). The village is about two hours mountainous drive (50 kilometers) from the capital Kathmandu to the east. People from different ethnicities such as Bahun, Chettri, Tamang, Sarki, and Danwar live in a total of 1000 households in these wards, and their livelihoods depend primarily on subsistence farming and daily wage-earning [27].
**Fig 1:** *The study site is selected Hokse village, ward numbers 7 and 8 of Paanchkhal Municipality, Kavrepalanchok district, Nepal.The map was created using ArcGIS desktop version 10.8. The shapefile of the administrative districts and location for Nepal was obtained from the Government of Nepal, Ministry of Land Management, Survey Department website and were publicly available for unrestricted use (http://www.dos.gov.np/nepal-map).*
## Study design and participants
A detailed study protocol of this research has been published elsewhere [28]. Briefly, this study utilized an ethnographic approach to understand how local social and cultural factors have affected kidney selling in the village. This study follows a standard qualitative methodology outlined by a COREQ checklist (S1 File). In-depth interviews were conducted with the primary participants who were kidney sellers aged above 18 years (Table 1). We also carried out non-participant observations and informal interactions with community members who were permanent residents of the village. In addition, we interviewed relevant medical professionals, policymakers at the ministry of health, legal workers, and non-governmental organization (NGO) staff working against organ and human trafficking for at least one year as key informants using online meeting platforms. We also interviewed two border patrol officers who were deployed at a vital border checkpoint for at least six months.
**Table 1**
| Type of participant | Study population | Number |
| --- | --- | --- |
| Primary participant | Kidney sellers of Hokse village | 10 |
| Key informants | Family members, neighbours, relatives | 15 |
| Key informants | Health care providers and local governmental officers | 6 |
| Relevant stakeholders | Transplant Unit’s medical personnel | 2 |
| Relevant stakeholders | Policymakers, legal workers, NGO/INGO workers | 8 |
| Relevant stakeholders | Border checkpoint officers | 2 |
Altruistic kidney donors were not included in this research. Informed consent was first obtained from the participants verbally. After sharing the study information, the participants were requested to sign the consent form. As this research was conducted using an ethnographic approach, the researcher familiarized himself with the community members by living in the village, observing and also participating in the village’s culture and tradition. We obtained informed consent for all formal interviews. In the case of online interviews, we obtained verbal consent from the participants.
Primary participants and key informants were asked to choose the interview site and if they would like to be interviewed or not. Respondents were assured that they could discontinue the interview at any point in time without requiring to provide any justifications. All participants were anonymized in the collected data and all personal identifiers were avoided from the transcripts.
The participants in this study is composed of a diverse group of respondents, which allowed identifying and collecting information-rich cases for an in-depth study of our topic [29]. Furthermore, as the issues around kidney trade can be culturally sensitive, we used a snowball approach to identify additional participants till we reached the data saturation [30]. However, the sampling design also evolved during the study period based on the circumstances of data collection, opportunities to enrol relevant participants and capture emerging issues.
## Data collection
The principal investigator (BS1) and co-investigator (AP) conducted the ethnographic fieldwork. The fieldwork started in February 2021 and the data analysis was completed in September 2021. BS1 conducted initial fieldwork to build a relationship with the villagers and form an initial understanding of the social and cultural fabric of the village. BS1 and AP visited subsequently and stayed in the village for various periods of duration under the supervision of BA. BA has extensive research experience around ethnography and qualitative research and was based in Nepal during the data collection period. AP and BS1 collaborated in the previous research as well and all study documents (proposal, consent form, ethical approval forms from Institutional Review Board, and the interview guide) were shared. AP and BS1 interacted regularly to share their interpretations, similarities and differences and sought suggestions and external interpretations from BA and the rest of the authors. Abiding by the national restrictions during the COVID-19 pandemic, BS1 conducted remote key informant interviews (KII) with different stakeholders using the online Zoom meeting platform. BS2 coordinated the online meeting with the KII.
Formal interviews were carried out using a topic guide to collect data. Each interview lasted from 30 minutes to two hours. Multiple in-depth interviews (IDI) were conducted with kidney sellers to understand their story and background status. A field diary was used to record important details during observation such as non-verbal communication, including the personal presentation of the participant, body expressions, gestures, facial expressions, style, and alterations in speech (silences, choking speech, blatant speech, fading speech, cringing and tremors), laughter, and other manifestations.
BS has an educational background in health and social sciences and belongs to a Newar ethnic community in western Nepal where other ethnic groups such as Bahun, Chettri, Magar, Gurung, Darai, and Bote also reside, similar to that of Hokse village [31]. AP is a graduate of Business Administration and is familiar with the local social, political, and cultural context of Nepal. Other team members included an expert medical anthropologist (LS), a clinician and social scientist with expertise in community engagement (BA), a socio-demographer familiar with the local context (DRS), a public health expert with extensive experience in non-communicable diseases including kidney diseases (SRM), and a public health professional with experience in gender-related research (MS).
## Data analysis
The audio-recorded interviews and field notes were transcribed first in the Nepalese language and then into English retaining the nuanced meanings and interpretations during translation. The interview data were sorted and labelled, followed by reiterative reading, and understanding, initial coding and theme building. Finally, the data were analyzed and interpreted using the multilevel theoretical framework of Baer, Singer and Johnsen’s Critical Medical Anthropology [32].
Content analysis was conducted independently by two investigators (BS and BA). The initial results were shared with LS, who supervised refining the codes and themes. NVivo version 10 (QSR International, Melbourne, Vic., Australia) was used to manage and analyze the data. The analysis process and findings were shared with the research team. Any disagreements between the theme during the analysis were discussed and resolved after seeking an alternative opinion from the other team member.
## Ethical approval
The study protocol was reviewed and approved by Mahidol University Central Institutional Review Board (MU-CIRB $\frac{2020}{217.1808}$), Thailand, and the Nepal Health Research Council ($\frac{716}{2020}$ PhD), Nepal.
## Participant characteristics
A total of 43 participants were enrolled in this study, including nine kidney sellers and one kidney seller who was also a broker (Tables 1 and 2). In the village, most of the kidney selling occurred between early 1990s and 2000, the kidney sellers interviewed in this study were in their mid-20s to 30s when they sold their kidneys. Higher proportion of males were engaged in kidney selling compared to females (*Male versus* female: 7:3) and it accurately reflects the kidney sellers’ population in the village.
**Table 2**
| S. No. | Age | Gender | Caste | Employment status |
| --- | --- | --- | --- | --- |
| 1 | 58 years# | Male | Bahun | 1st kidney broker from the village |
| 2 | 60 years | Male | Newar | Business |
| 3 | Early 40s | Female | Mijars | Daily wage worker |
| 4 | 45 years | Male | Mijars | Pig rearing and agriculture |
| 5 | 53 years | Male | Tamang | Daily wage worker |
| 6 | 48 years | Female | Tamang | Daily wage worker and local alcohol pub |
| 7 | 52 years | Male | Bahun | Daily wage worker/irregular |
| 8 | 48 years | Female | Tamang | Daily wage worker |
| 9 | 50 years | Male | Bahun | Daily wage worker |
| 10 | 52 years | Male | Bahun | Daily wage worker /irregular |
One of the kidneys seller who later became broker was working as a chef in an Indian restaurant who realized an opportunity to make an extra money by exploiting the familiarity and trust he had in his own village. Secondary participants comprised of key informants such as family members, neighbors, and relatives of the kidney sellers ($$n = 15$$) and local healthcare providers and governmental officers ($$n = 6$$). Two border checkpoint officers were also included in the interview. Relevant stakeholders were enrolled for remote interviews ($$n = 10$$) as indicated in Tables 1 and 3.
**Table 3**
| Key informant | Details |
| --- | --- |
| 1. Key informants from the village | |
| a. Family members, neighbours, relatives (15) | FCHV-1 |
| a. Family members, neighbours, relatives (15) | FCHV-2 |
| a. Family members, neighbours, relatives (15) | Principal |
| a. Family members, neighbours, relatives (15) | Treasurer of Ward |
| a. Family members, neighbours, relatives (15) | Local News reporter |
| a. Family members, neighbours, relatives (15) | Sellers’s mum |
| a. Family members, neighbours, relatives (15) | School teacher |
| a. Family members, neighbours, relatives (15) | Local activist |
| a. Family members, neighbours, relatives (15) | Local priest |
| a. Family members, neighbours, relatives (15) | Local people shifted to Tamaghat |
| a. Family members, neighbours, relatives (15) | Kidney sellers’ relatives—1 |
| a. Family members, neighbours, relatives (15) | Kidney sellers’ relatives—2 |
| a. Family members, neighbours, relatives (15) | Kidney sellers’ neighbor -1 |
| a. Family members, neighbours, relatives (15) | Kidney sellers’ neighbor -2 |
| a. Family members, neighbours, relatives (15) | Local person shifted to Tamaghat |
| 2. Stakeholders | |
| a. Transplant Unit’s medical personals (2) | Transplant surgeon (10 years of working experience) |
| a. Transplant Unit’s medical personals (2) | Transplant nurse (6 years of working experience) |
| b. Policymakers, legal workers, NGO/INGO workers (8) | Policy maker (former Nepal Planning Commission) |
| b. Policymakers, legal workers, NGO/INGO workers (8) | Lawyer-legal worker |
| b. Policymakers, legal workers, NGO/INGO workers (8) | Police officer- posted in the Panchkhal Municpality and registered 1st case |
| b. Policymakers, legal workers, NGO/INGO workers (8) | Government officer- Human Organ transplant center* |
| b. Policymakers, legal workers, NGO/INGO workers (8) | NGO-1 |
| b. Policymakers, legal workers, NGO/INGO workers (8) | NGO-2 |
| b. Policymakers, legal workers, NGO/INGO workers (8) | NGO-3 |
| b. Policymakers, legal workers, NGO/INGO workers (8) | NGO-4 |
| c. Border checkpoint officers (2) | NGO worker |
| c. Border checkpoint officers (2) | Government officer |
## Drivers of kidney selling in Hokse
Kidney selling in Hokse village was found to be a product of an interaction between community members’ vulnerability and contextual factors, with the brokers playing a central role (Fig 2). Below, we present our findings on how vulnerability and contextual factors contributed to the concentrated kidney selling phenomenon in the village.
**Fig 2:** *This figure outlines our broad category of themes: Vulnerability and contextual factors for kidney selling and the critical role of brokers.*
## Vulnerability
Multitude of factors contributed to the vulnerability of the villagers for kidney selling that ranged from characteristics inherent in an individual to wider social factors.
## Poor socio-economic status
Poor socioeconomic status was a major factor for kidney selling in the village. Conversations with the community members highlighted that unemployment in the village gave little options to support their daily livelihood. As a result, people often travelled to Kathmandu or India for work. Most of the community members who travelled to India ended up taking unskilled works such as construction labor, security guard, and cleaner at a restaurant. Poverty and a lack of job opportunities meant that any unexpected sum of money for a kidney was alluring enough to overlook any health issues, as one participant pointed out: In contrast, other community members were sharply critical of the kidney sellers in the village, mostly as the kidney sellers spent their money drinking alcohol. They were also disapproving of the kidney selling as the village gained substantial attention from the national and international media for being notorious for cheap kidneys followed by discrimination and stigma to the entire village. Some respondents questioned if poverty should be considered as a sole cause of kidney selling:
## Alcohol drinking
Some community members were found drunk as early as 11:00 hours. Visiting local pubs known as “Bhatti” in Nepali, for drinking at irregular times such as early in the morning was an accepted local norm. During these situations, many kidney sellers reported meeting the brokers who bought them drinks and food. These brokers used the “Bhatti” as a venue to lure innocent community members, presenting them with possibilities of job opportunities in India, and ultimately, kidney selling.
## Ignorance/Gullibility
People from the village were easily enticed by the brokers using stories of financial compensations after kidney removal. Many brokers formed false kinships and convinced sellers to sell their kidneys. Community members were gullible and got deceived by this kinship and the amount of money they were promised by the organ receiver families. One drunk kidney seller revealed:
## Social conformity
People in the village had been victim to a unique trend of kidney selling, particularly because of a tendency to follow peers, neighbors or other community members, and this was often shown to be due to high communal reliance and trust among the community members. This was also evident in community members’ decision to work in India—often they followed to look for opportunities in India rather than cities in Nepal. They had learned that people earned more money by selling their body parts.
Moreover, kidney selling thrived once the higher caste people such as Bahun and Chettri joined in. This was like a symbolic gesture to all the people that kidney selling was normal. Eventually, many community members who belonged to lower caste and were considered naive ‘Sojo’, simply followed the actions of their higher-caste peers.
## Indirect incentives
Many community members explained that kidney selling was indirectly promoted by financial and material incentives offered by various organizations supporting ‘struggling kidney sellers.’ Unfortunately, financial help from these well-intended organizations was misinterpreted and even exploited to an extent that it promoted the selling of kidneys. Many community members reported that their neighbors sold their kidneys to gain financial incentives and receive aid such as goats, water tanks, and direct cash from the supporting organizations.
## Contextual factors
Multitude of external factors in and outside the village affected as contextual influences that provided the background and aided in promoting the villagers into kidney selling:
## Shifting of business
The development of a new motorway–Araniko Highway, through Tamaghat changed the course of once-flourishing Hokse village. This shift of the highway away from Hokse village was followed by a subsequent shift in the local business center to Tamaghat, reducing the economic activities of the village. This situation led to a movement of people to Tamaghat and unemployment in the village. Community members in Hokse village then had to resort to work as cheap wage laborers. Furthermore, after their seasonal agriculture, people started travelling to Kathmandu or India for job opportunities. This shifting of local economic activities made the villagers more vulnerable to kidney selling.
## Role of medical personnel
The role of medics in the kidney selling phenomenon was also critical. Although medical personnel understood that a lot of kidney selling was being orchestrated by brokers through forged official documentation, they had little to almost no options but to accept when they were approached for kidney transplant operation.
Most authorities knew the situation of doctors and donors, but they were obliged by the legal documents produced. There were adequate legal loopholes in which kidney selling occurred. Doctors could suspect the sellers through their appearance, dress-up, follow-up arrangement, and various other clues. Nonetheless, they were constrained by the seller’s signature in the consent form.
## Fake documentations
Easy availability of fake documents and consents are another major legal loopholes that promoted kidney selling. The transplantation act permitted altruistic organ donation and brokers were able to produce an authentic-looking document that made the kidney sellers look like altruistic donors. The documents were prepared so precisely that there were no spaces for questions both legally and for medical practice.
## Policy loopholes
Inadequate awareness on human organ transplantation and a porous border between India and Nepal provided fertile ground for kidney trade. Kidney transplantations also involve a lot of patients at the end-stage of the disease such as brain death, the policies for which are vague in Nepal. Importantly, the policy fails to address the moral dilemma patients’ parties might have when seeking such consent. So, practically, seeking written informed consent from patients in critical conditions is more difficult than outlined in the law.
There are both government and non-governmental organisations working at the border between Nepal and India. One of such organizations was the Immigration Department of Nepal, which maintains all legal documents including travel documents for foreigners travelling across the border. Border officials (immigration officials and police officers) were unaware of illegal kidney selling but stressed that they were aware of human (mostly women) trafficking.
The open Nepal-India border does not require any documents from Nepalese citizens and makes it easier to cross the border. Recognizing and stopping potential kidney sellers at the border point was not possible as the kidney selling could only be identified after mutilation of body parts.
Kidney selling was only reported to the authorities when there were problems in a financial settlement. Given the high number of people cross the border without documentation, authorities explained the difficulties of spotting the illegal human trafficking; and organ trafficking was way more covert for them to identify at a border crossing.
## Role of a broker
The role of the brokers was central in kidney selling. The kidney brokers not only worked to find the sellers and buyers but were deeply involved in various steps of the illegal kidney trade. Brokers tended to identify potential sellers based on their network and contacts in the village. As most of the brokers were from the village; they knew the financial, psychological and social context of the community members. Brokers often exploited the financial difficulties faced by community members. After coaxing potential kidney sellers from the village, the brokers prepared all necessary paperwork that included a medical appointment for a regular check-up and fake documents to cross through the Nepal-India border. Finally, the brokers transported the kidney sellers to the venues where their kidneys were excised.
The connection brokers had with people at different levels made their work easier including paperwork that required time and influence. The brokers were also good at forging official papers and negotiating with medical personnel and the buyers. In addition, the brokers were adept at convincing people, often persuading community members to sell kidneys by accentuating the financial gains in such a short period, and reassuring that selling one kidney would not be a problem for their health.
The brokers were flawless in acting as caretakers of the sellers. They took the potential sellers from the village to the venues where kidneys were excised. Based on their experience, the brokers were also cognizant of medical procedures around kidney transplantation such as blood matching. Once the blood matching was done, they would negotiate the price with potential buyers.
After selling the kidney, the brokers often influenced the sellers to identify other potential sellers in the village. This was the point where new brokers were borne and their network expanded In many cases, the brokers were relatives of the kidney sellers in the village. Therefore, the sellers hesitated to report them to the legal authority, as one police officer confided to us:
## Discussion
This research identified that kidney selling is a complex social process arising from the blend of the vulnerability of community members and contextual factors which gave rise to a perfect scenario for brokers to demonstrate false hopes of alleviating the vulnerability and needs. Decision-making of community members in Hokse was affected by fellow kidney sellers with their stories of kidney selling and financial incentives. Remarkably, some of the victims of kidney selling later become brokers themselves and liaised in the system (Fig 3). One of the prominent reasons how kidney sellers later became brokers was because of their existing relationship and trust in their community where they saw an opportunity to generate an easy money. Below we discuss the study results in more detail.
**Fig 3:** *This figure outlines how Hokse village became the kidney hub.Along with the characteristics of village and community members, it shows a process of development of kidney brokers. Poverty and unemployment are exploited by a broker at first to offer them jobs in India. Further options for making more money including persuasion to sell kidneys are then implanted to the community member. As a result, the community member sells a kidney and returns to the village while remaining guided by the broker. In the pursuit of making easy money, the new kidney seller takes up a new role as a (new) broker and persuades community members to go to India for jobs and kidney selling.*
Brokers are the central element of kidney trade in Hokse village. Previous reports highlight the role of poverty, and in particular the role of the earthquakes in pushing families into poverty which ultimately prompted them to sell their kidneys [33]. Nevertheless, attributing the kidney selling only to poverty sketches an incomplete story and requires a thorough exploration of social dynamics and mechanisms that underpin the kidney selling [18]. There are equally poor and/or poorer villages in Nepal that have not acceded to the kidney trade for livelihood. Previous research on organ and human trafficking have clearly shown that poverty is a fertile ground for increased vulnerability of the population who are the victims of human trafficking, prostitution, and organ trafficking [22, 34, 35]. In all of these activities, brokers have played a critical role and their major contribution in facilitating kidney trade should not be overlooked [36]. Similar findings are evident from the research conducted in Bangladesh, which details the role of brokers in transporting, identifying buyers, and negotiating with sellers and buyers [10].
Transplant surgeons are obliged to perform surgeries under the circumstances where the medical documentation and informed consent forms meet the standard criteria, although the informed consent forms are fake [37, 38]. Often the background related to these seemingly perfect documentations are outside the scope of investigations for clinicians when they have met all the criteria. These practice of fake documentations were also evident in the other studies conducted in Bangladesh where the sellers were either prepared for the false kinship in Bangladesh or India [10]. Making up false kinship among kidney sellers and receivers was also reported in Nepal and explains a lot about the background on kidney trafficking documents [39].
One of the prominent reasons compelling community members to sell a kidney is their vulnerability due to poverty. Nonetheless, dumping all the reasons for kidney selling as ‘poverty’ is an understatement because there are poorer villages in Nepal than Hokse village which have not been victim to kidney selling. Indeed, studies have highlighted the role of poverty in kidney selling from low and middle-income countries such as Iran, Pakistan, India, the Philippines, and Bangladesh [10, 19, 40–44]. Poverty and illiteracy are highly prevalent among people of lower castes in Nepal [45]. Historically, the caste-based social hierarchy is dominated by higher castes who had more education, wealth and authority [46]. Over the years, such pre-domination remained ingrained deep into the societal dynamics and way of living leading to internalization of perceived inferiority, subjugation and dependence among lower castes [47]. Their tendency to look up to higher caste people also gradually offered them an impression that higher caste people’s decision seems superior. The majority of lower caste people took up the decision because they saw that some of the higher caste people sold their kidneys. Such a decision-making process, and a tendency to follow higher caste people are reported from the region such as India and Bangladesh [36, 40, 48].
Drinking venues referred to as local pubs (‘bhattis’) which serve homemade rice/millet wines are a popular place to socialize. The combination of vulnerability, alcoholism, local ‘bhattis’, and importantly the brokers are the perfect storm for kidney selling. The low paying non-technical jobs offered to the villagers and building relations with the buyers-sellers were performed in the local ‘bhattis’ where the brokers introduce the topic of kidney selling to make more money. A similar pattern of false hope and deception for the poor slum dwellers were evident in the Philippines and the Indian Subcontinents [10, 37, 40, 41].
Although much has been described in the literature related to the roles and contributions of non-governmental aid organizations, their roles in Hokse village are unique particularly because of how their incentives to palliate the burden seem to foster dependence, false hopes and even promote the selling of kidneys among the community members. Euphemistic aids and their disruption of social dynamics are also highlighted by literature around temporary health camps in rural Nepal [49–51]. In the wider literature, global health and aid working organizations are criticized to perpetuate the white savior mentality, white supremacy and coloniality in Low and Middle-Income Countries in the forms of aids and humanitarian acts. Their acts have been incriminated to be negligible in terms of alleviating poverty, mitigating diseases and promoting development [52, 53]. In the case of Hokse village, temporary aids simply disrupted local livelihood. The community members understood that the kidney sellers get incentives from different organizations, therefore, they followed other neighbors to sell their kidneys.
Kidney trafficking is illegal according to the organ donation policy of Nepal. The policy highlights that “No person shall operate an activity relating to organ transplantation for the purpose of the sale and purchase of an organ or similar other acts.” [ 54]. The sellers of the organ are mostly acknowledged for the deeds but the follow-up remains complex due to differences in nationality such as Nepalese selling their kidney to Indian nationalities and it is difficult to trace the sellers. Transplantation sites (being away from a person’s own country), organ receiver’s nationality and fake documents (paper tigers) add complexity to kidney trafficking. Specifically, the jurisdiction and legal liabilities become complex. Even more importantly, kidney trafficking is often reported to authorities only after serious disagreement or conflict between the organ sellers, receivers and brokers, mostly rooted in price negotiation and changes in deals at the last minute.
Legal specifications around organ donation also needs to revisit as organ donations are a sensitive topic. For instance, discussing the consent for organ donation from individuals/patients in critical conditions (e.g., brain death) are difficult. Patients or their relatives can shun away from such discussions. The clarity in policies related to organ donation among such patients and their relatives can facilitate such discussions thereby preventing last-minute organ trafficking.
## Conclusion
Kidney selling was a result of a complex interaction of vulnerability and contextual factors at the individual and community level. Brokers exploited the vulnerabilities of the community members and were able to lure them into kidney selling. Brokers used several avenues for interaction with community members often through false hopes for better livelihood. The cycle of kidney selling is perpetuated further when victims themselves become new brokers. Although increasing support for livelihood and education is critical, regulatory measures are urgently needed to curb the kidney trade. First, regulatory authorities should mandate a thorough background check of a donor such as cross-checking with their family members, local authorities, and legal representatives. Second, facilitator can always be a broker disguising as a family member and therefore needs to be suspected and explored for his/her potential role as a motivator. Third, cross-border collaboration between Nepal and India needs to scrutinize the legitimacy of consent process, documents, and potential involvement of a broker.
## References
1. Goyal M, Mehta RL, Schneiderman LJ, Sehgal AR. **Economic and health consequences of selling a kidney in India**. *JAMA* (2002.0) **288** 1589-93. DOI: 10.1001/jama.288.13.1589
2. Scheper-Hughes N.. **Perpetual scars**. *New Internationalist* (2014.0) 12-22
3. 3Donation GOo, Transplantation. Organ Donation and Transplantation Activities. 2010.. *Organ Donation and Transplantation Activities* (2010.0)
4. Bikbov B, Purcell CA, Levey AS, Smith M, Abdoli A, Abebe M. **Global, regional, and national burden of chronic kidney disease, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *The Lancet* (2020.0) **395** 709-33
5. Hannan BE. *Managed care and the acute care institution: a discussion of the impact and future implications: New York Medical College* (1998.0)
6. Zargooshi J.. **Iranian kidney donors: motivations and relations with recipients**. *J Urol* (2001.0) **165** 386-92. DOI: 10.1097/00005392-200102000-00008
7. Scheper-Hughes N.. *heft of Life: The Globalization of Organ Stealing Rumours Anthropology Today* (1996.0) **12** 3-11
8. Scheper-Hughes N.. **The Global Traffic in Human Organs**. *Current Anthropology* (2000.0) **41** 191-224. PMID: 10702141
9. Kar D, Spanjers J. *Transnational crime and the developing world* (2017.0) **30** 2019
10. Moniruzzaman M.. **"Living cadavers" in Bangladesh: bioviolence in the human organ bazaar**. *Med Anthropol Q* (2012.0) **26** 69-91. DOI: 10.1111/j.1548-1387.2011.01197.x
11. Ivanovski N, Popov Z, Cakalaroski K, Masin J, Spasovski G, Zafirovska K. **Living-unrelated (paid) renal transplantation—ten years later**. *Transplantation proceedings* (2005.0). DOI: 10.1016/j.transproceed.2004.12.139
12. Pascalev A, De Jong J, Ambagtsheer F, Lundin S, Ivanovski N, Codreanu N. **Trafficking in human beings for the purpose of organ removal: a comprehensive literature review**. *Trafficking in Human Beings for the Purpose of Organ Removal Results and Recommendations Lengerich: Pabst* (2016.0) 15-68
13. Cohen L.. **The other kidney: Biopolitics beyond recognition**. *Body & Society* (2001.0) **7** 9-29
14. Akhtar F.. **Chronic Kidney Disease, Transplantation Practices and Transplantation Law in Pakistan: Opportunity for a Global Meditation**. *Wiley-Blackwell* (2009.0) 570-6
15. Budiani‐Saberi DA, Delmonico FL. **Organ trafficking and transplant tourism: a commentary on the global realities**. *American Journal of Transplantation* (2008.0) **8** 925-9. DOI: 10.1111/j.1600-6143.2008.02200.x
16. Moniruzzaman M.. *Review: Everyday Ethics of Organ Transplant Current Anthropology* (2013.0) **54**
17. Naqvi SAA, Ali B, Mazhar F, Zafar MN, Rizvi SAH. **A socioeconomic survey of kidney vendors in Pakistan**. *Transplant International* (2007.0) **20** 934-9. DOI: 10.1111/j.1432-2277.2007.00529.x
18. Adhikari B.. **Organ and human trafficking in Nepal**. *The Lancet* (2016.0) **387** 1907. DOI: 10.1016/S0140-6736(16)30419-6
19. Mendoza RL. **Kidney black markets and legal transplants: Are they opposite sides of the same coin?**. *Health Policy* (2010.0) **94** 255-65. DOI: 10.1016/j.healthpol.2009.10.005
20. Ghimire U, Shrestha N, Adhikari B, Mehata S, Pokharel Y, Mishra SR. **Health system’s readiness to provide cardiovascular, diabetes and chronic respiratory disease related services in Nepal: analysis using 2015 health facility survey**. *BMC Public Health* (2020.0) **20** 1-15. PMID: 31898494
21. Adhikari B, Ozaki A, Marahatta SB, Rijal KR, Mishra SR. **Earthquake rebuilding and response to COVID-19 in Nepal, a country nestled in multiple crises**. *Journal of Global Health* (2020.0) **10**. DOI: 10.7189/jogh.10.020367
22. Kern A, Müller-Böker U. **The middle space of migration: A case study on brokerage and recruitment agencies in Nepal**. *Geoforum* (2015.0) **65** 158-69
23. Ray N.. **Looking at trafficking through a new lens**. *Cardozo JL & Gender* (2005.0) **12** 909
24. 24Encyclopædia Britannica. "Nepal | Culture, History, & People" [Available from: https://www.britannica.com/place/Nepal/Federal-republic.
25. 25Constitution of Nepal 2015. 20 September 2015 [Available from: https://web.archive.org/web/20190808043521/
http://www.lawcommission.gov.np/np/wp-content/uploads/2018/09/नेपालको-संविधान-1.pdf.
26. **Kidney Trafficking In Nepal**. *Kathmandu, Nepal* (2015.0)
27. 27Paanchkhal Municipalit. Paanchkhal Municipality
2020 [Available from: https://panchkhalmun.gov.np.. *Paanchkhal Municipality* (2020.0)
28. Shrestha B, Adhikari B, Shrestha M, Sringernyuang L. **Kidney Sellers From a Village in Nepal: Protocol for an Ethnographic Study**. *JMIR Res Protoc* (2022.0) **11** e29364. DOI: 10.2196/29364
29. Patton MQ. **Qualitative interviewing**. *Qualitative research and evaluation methods* (2002.0) **3** 344-7
30. Saunders B, Sim J, Kingstone T, Baker S, Waterfield J, Bartlam B. **Saturation in qualitative research: exploring its conceptualization and operationalization**. *Quality & quantity* (2018.0) **52** 1893-907. DOI: 10.1007/s11135-017-0574-8
31. 31Vyas Municipality. Ward Profile Government of Nepal
2020 [Available from: https://vyasmun.gov.np/en.. *Ward Profile Government of Nepal* (2020.0)
32. Baer HA, Singer M, Johnsen JH. **Toward a critical medical anthropology**. *Social science & medicine* (1986.0) **23** 95-8
33. Cousins S.. *Nepal: organ trafficking after the earthquake* (2016.0)
34. Fayomi OO. **Women, poverty and trafficking: A contextual exposition of the Nigerian situation**. *Journal of’Management and Social Sciences* (2009.0) **5** 65-79
35. Lundin S.. **Organ economy: Organ trafficking in Moldova and Israel**. *Public Understanding of Science* (2012.0) **21** 226-41. DOI: 10.1177/0963662510372735
36. Moniruzzaman M.. *" Living Cadavers" in Bangladesh: Ethics of the Human Organ Bazaar* (2010.0)
37. Bienstock RE. *Tales from the Organ Trade* (2013.0)
38. Ambagtsheer F, Van Balen L. **I’m not Sherlock Holmes’: Suspicions, secrecy and silence of transplant professionals in the human organ trade**. *European Journal of Criminology* (2019.0)
39. Kaiju I.. **मिर्गौला फेर्न ‘नक्कली’ नाता**. *Annapurna Post* (2018.0)
40. Cohen L.. **Where it hurts: Indian material for an ethics of organ transplantation**. *Daedalus* (1999.0) **128** 135-65. PMID: 11645873
41. Moazam F, Zaman RM, Jafarey AM. **Conversations with kidney vendors in Pakistan: An ethnographic study**. *Hastings Center Report* (2009.0) **39** 29-44. DOI: 10.1353/hcr.0.0136
42. Zargooshi J.. **Quality of life of Iranian kidney “donors”**. *The Journal of urology* (2001.0) **166** 1790-9. PMID: 11586226
43. Turner L.. **Commercial organ transplantation in the Philippines**. *Camb Q Healthc Ethics* (2009.0) **18** 192-6. DOI: 10.1017/s0963180109090318
44. Yousaf FN, Purkayastha B. **‘I am only half alive’: Organ trafficking in Pakistan amid interlocking oppressions**. *International Sociology* (2015.0) **30** 637-53
45. Measuring Acharya S., Poverty Analyzing. **(with a particular reference to the case of Nepal)**. *The European journal of comparative economics* (2004.0) **1** 195-215
46. Skinner D, Holland D. **Schools and the cultural production of the educated person in a Nepalese hill community**. *The cultural production of the educated person: Critical ethnographies of schooling and local practice* (1996.0) 273-99
47. Stash S, Hannum E. **Who goes to school? Educational stratification by gender, caste, and ethnicity in Nepal**. *Comparative Education Review* (2001.0) **45** 354-78
48. Haagen M.. *Indian Organ Trade; From the Perspective of Weak Cultural Relativism* (2005.0)
49. Citrin D.. **The anatomy of ephemeral healthcare:“health camps” and short-term medical Voluntourism in remote Nepal**. *Studies in Nepali History and Society* (2010.0) **15** 27-72
50. Sharma S.. **Trickle to torrent to irrelevance? Six decades of foreign aid in Nepal**. *Aid, Technology and Development: Routledge* (2016.0) 72-92
51. Seabrook D.. *Do no harm: perceptions of short-term health camps in Nepal* (2012.0)
52. Ashdown BK, Dixe A, Talmage CA. **The potentially damaging effects of developmental aid and voluntourism on cultural capital and well-being**. *International Journal of Community Well-Being* (2021.0) **4** 113-31
53. Forde SD. **Fear and loathing in Lesotho: An autoethno graphic analysis of sport for development and peace**. *International review for the sociology of sport* (2015.0) **50** 958-73
54. 54The Human Body Organ Transplantation (Regulation and Prohibition) Act of Nepal. 1998.. *The Human Body Organ Transplantation (Regulation and Prohibition) Act of Nepal* (1998.0)
|
---
title: 'Community purchases of antimicrobials during the COVID-19 pandemic in Uganda:
An increased risk for antimicrobial resistance'
authors:
- Agnes N. Kiragga
- Leticia Najjemba
- Ronald Galiwango
- Grace Banturaki
- Grace Munyiwra
- Idd Iwumbwe
- James Atwine
- Cedric Ssendiwala
- Anthony Natif
- Damalie Nakanjako
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021632
doi: 10.1371/journal.pgph.0001579
license: CC BY 4.0
---
# Community purchases of antimicrobials during the COVID-19 pandemic in Uganda: An increased risk for antimicrobial resistance
## Abstract
Self-Medication (SM) involves the utilization of medicines to treat self-recognized symptoms or diseases without consultation and the irrational use of over-the-counter drugs. During the COVID-19 pandemic, the lack of definitive treatment led to increased SM. We aimed to estimate the extent of SM for drugs used to treat COVID-19 symptoms by collecting data from pharmacy sale records. The study was conducted in Kampala, Uganda, where we extracted data from community pharmacies with functional Electronic Health Records between January 2018 and October 2021 to enable a comparison of pre-and post-COVID-19. The data included the number of clients purchasing the following drugs used to treat COVID-19 and its symptoms: Antibiotics included Azithromycin, Erythromycin, and Ciprofloxacin; Supplements included Zinc and vitamin C, while Corticosteroids included dexamethasone. A negative binomial model was used to estimate the incident rate ratios for each drug to compare the effect of COVID-19 on SM. In the pre- COVID-19 period (1st January 2018 to 11th March 2020), 19,285 customers purchased antibiotics which included; Azithromycin ($$n = 6077$$), Ciprofloxacin ($$n = 6066$$) and Erythromycin ($$n = 997$$); health supplements including Vitamin C [430] and Zinc ($$n = 138$$); and Corticosteroid including Dexamethasone ($$n = 5577$$). During the COVID-19 pandemic (from 15th March 2020 to the data extraction date in October 2021), we observed a $99\%$ increase in clients purchasing the same drugs. The number of clients purchasing Azithromycin increased by $19.7\%$ to 279, Ciprofloxacin reduced by $58.8\%$ to 96 clients, and those buying Erythromycin similarly reduced by $35.8\%$ to 492 clients. In comparison, there were increases of $170\%$, $181\%$, and $377\%$ for Vitamin C, Zinc, and Dexamethasone, respectively. The COVID-19 pandemic underscored the extent of SM in Uganda. We recommend future studies with a representation of data from pharmacies located in rural and urban areas to further study pandemics’ effect on antimicrobials prescriptions, including obtaining pharmacists’ perspectives using mixed methods approaches.
## Introduction
The increasing trend of self-medication and inappropriate drug purchases are becoming major public health concerns [1], driving antimicrobial resistance (AMR). According to the World Health Organization’s (WHO) definition, self-medication (SM) is the selection and use of medicines by individuals to treat self-recognized illnesses or symptoms without the consultation of a physician [2]. The most commonly self-prescribed medications are analgesics, antipyretics, antitussives, antidiarrheals, calcium and vitamin supplements, anabolic steroids, sedatives, certain antibiotics, and many herbal and homeopathic remedies [3].
Self-medication is a reasonably widespread practice globally, particularly in deprived communities in low and middle-income countries, and studies have estimated SM in Africa ranging between 30–$45\%$ [4–6]. If correctly practiced and under scenarios of proper governing laws on drug prescription, self-medication can self-empower patients to manage their health and improve their health care. In economically deprived communities with overstretched health systems, prescribed and non-prescribed drug purchases may save time for healthcare professionals and resources spent on managing minor and chronic ailments [7]. In addition, drug purchases among community dwellers have been shown to increase survival and reflect changes in healthcare practices [8, 9]. On the contrary, unlimited community drug purchase causes misuse of antimicrobials for viral infections such as colds and influenza [10], which increases the risk of antimicrobial resistance in the respective communities.
In Uganda, the second wave of the SARS-CoV2 Delta Variant led to a spike in COVID-19 cases. Approximately 162,000 cases and 3,557 deaths were reported by April 25, 2022, [11]. During the pandemic’s peak, the increasing case fatality rate and the global reports about the disease caused fears that could have led to excessive and rushed purchases of various drugs, including antibiotics, to treat symptoms of COVID-19 disease. Similarly, the desire to receive updated information on the treatment of COVID-19 led to increases in trends in google searches on self-medication for COVID-19 [12] and or self-medication with drugs such as hydroxychloroquine and dexamethasone which may have paused several outcomes, particularly among person with diabetes [13–15]. However, the magnitude to which antimicrobials were used or misused in the community has not been quantified. There remains a dearth of empirical evidence of increased drug purchases and self-medication. A recent study demonstrated an exponential increase in Azithromycin supply during the COVID-19 pandemic but did not provide proof of consumption at the community level [16]. In this study, our objective was to determine the effect of COVID-19 on customer drug purchases in urban Ugandan communities. We mined data from unconventional data sources, such as pharmacy records for purchases of unprescribed drugs as a proxy for community self-medication and drug use during the COVID-19 pandemic. The results of the trends of purchased self-medication before, during, and after the peak COVID-19 pandemic, as presented in this paper, present a potential risk of increased anti-microbial resistance in the respective communities.
## Study design and setting
This was a cross-sectional study conducted in Kampala, Uganda, in collaboration with proprietors of private chain pharmacies, including the Ecopharm Pharmacy Uganda Limited and Vine Pharmaceuticals, one of the largest pharmacy chains in Uganda [17, 18]. Pharmacies provide unprescribed drugs and over-the-counter medication to patients with fever, cough, flu, and other common symptoms.
## Data sources
Electronic drug sale data was obtained from 13 branches of Ecopharm and 35 pharmacies from Vine pharmaceuticals before and after the start of the COVID-19 pandemic in Uganda on 15th March 2020).
## Data variables
We extracted drug sale data between 1st January 2018 to 31st October 2021. The data extracted included monthly customer sales, quantity, and dates of purchases for three main drug classes: antibiotics, supplements, and corticosteroids. These drug classes were purposively selected as they comprised the most common drugs purchased in community pharmacies during the pandemic. Antibiotics included Azithromycin, Erythromycin, and Ciprofloxacin; Supplements included Zinc and vitamin C, while Corticosteroids included dexamethasone. In addition, to relate the drug purchases with the COVID-19 epidemic in Uganda, we extracted daily cases of COVID-19 from the Ugandan Ministry of Health’s COVID-19 dashboard [19].
## Data analysis
We determined the proportion of purchasers (clients in the community), the types of drugs (prescribed and non-prescribed) purchases, and the trends in drug purchases before and during the COVID-19 pandemic and during the different waves accompanying SARS-Cov-2 strains. A single group interrupted time-series analysis using segmented regression was used to estimate the effect of the COVID-19 pandemic on drug purchases [18]. The number of clients purchasing selected drugs over time (the outcome) was modeled using a negative binomial distribution for accounting for overdispersion (variance exceeds the mean). The observation time was divided into five segments that matched the period before and after COVID-19, particularly the ascents and descents of Uganda’s first and second waves S1 Fig. In the model, we estimated five slopes described as slope 1(β1) (change in drug purchases during the pre-COVID pandemic); slope 2 (β3) represents the change in drug purchases during the ascent of the 1st COVID-19 wave; slope 3 (β5) represents the change in drug purchases during the descent of the 1st COVID-19 wave; slope 4 (β5) represents the change in drug purchases during the ascent of the 2nd COVID-19 wave; and slope 5 (β7) represents the change in drug purchases during the descent of the 2nd COVID-19 wave.
## The model equation
The standard Interrupted time series analysis negative binomial regression model assumed the following form: Yit=β0+β1T+β2Z1+β3X1+β4Z2+β5X2+β6Z3+β7X3+β8Z4+β9X4.
Yit is the aggregated outcome variable measured at each equally spaced time point t, and T is a continuous variable indicating the number of days since the start of follow–up. This variable enables us to understand the temporal change in the antibiotic sales pattern and to control for secular trends. Zi’s (Z1, Z2, Z3, Z4) are dummy (indicator) variables representing the first, second, third, and fourth ascent and descent periods (preintervention periods 0, otherwise 1) that happened in March 2020, December 2020, March 2021 and June 2021, respectively. Xi’s (X1, X2, X3, X4) are continuous variables counting the number of days since each ascent or descent period in our study (equates to 0 for the pre–intervention (COVID-19) period).
In the case of a single-group method, β0 represents the intercept or starting number of clients purchasing the drugs the variable. β1 is the slope or trajectory or change in the number of clients purchasing the medications during the pre-COVID-19 era until the start of the epidemic in Uganda. β2 represents the change in the number of drug purchases during the ascent of the first COVID-19 wave in the period immediately following the introduction of the ascent of the first COVID-19 wave. β3 represents the difference in the trend after the increasing peak of the ascent of the first COVID-19 wave. β4 represents the change in the level of the outcome immediately after the descent of the first COVID-19 wave. β5 represents the difference in the trend of the ascent of the second COVID-19 wave. β6 represents the change in the outcome level immediately after the ascent of the second COVID-19 wave. β7 represents the difference in the trend after the rise (ascent) of the second COVID-19 wave. Β8 represents the change in the level of the outcome immediately after the descent of the second COVID-19 wave. β9 represents the difference in the trend of the decline of the second COVID-19 wave. These terms are displayed in S2 Fig, and all data used in the manuscript has been shared as S1 Data.
The coefficients (β) (incidence rates) from the negative binomial regression models were transformed into exponents to obtain the incident rate ratios (IRR) presented in the tables. The rates allowed us to compare changes between groups during the various periods. The final model adjusted for seasonality by including a term for the two major rainfall seasons: March to May and September to December. This was done to account for the effect of rainfall seasons on the occurrence of upper respiratory infections in Uganda. All the analysis and data visualization were carried out using the R statistical software version 4.1.1 using the Dplyr package for data management and the Mass package for Negative binomial regression.
## Ethical considerations
The study received local approval from the Joint Clinical Research Center Institutional Review and Ethics Committee and the Uganda National Council of Science and Technology (SIR61ES).
## Results
During the pre-*Covid era* (1st January 2018 and 11th March 2020), a total of 19,285 customers purchased the selected drugs, which included: the antibiotics Azithromycin ($$n = 6077$$), Ciprofloxacin ($$n = 6066$$) and Erythromycin ($$n = 997$$); health supplements including Vitamin C [430] and Zinc ($$n = 138$$); and Corticosteroid including Dexamethasone ($$n = 5577$$) S1 Fig. After the onset of the COVID-19 era between 15th March 2020 and the data extraction date in October 2021, we observed a $99\%$ increase in the number of clients during 20 months. The number of clients purchasing Azithromycin increased by $19.7\%$ to 279, Ciprofloxacin reduced by $58.8\%$ to 96 clients, and those buying Erythromycin similarly reduced by $35.8\%$ to 492 clients. In comparison, there were increases of $170\%$, $181\%$, and $377\%$ for Vitamin C, Zinc, and Dexamethasone, respectively.
## Effect of COVID-19 on drug purchases before and during the COVID-19 era
We estimated the impact of the COVID-19 pandemic on purchases of the selected drugs by calculating the change in the number of customers each month Fig 1. The changes were assessed before COVID-19 and along the slopes (ascent and descent) of Uganda’s two first epidemic waves (S2 Fig).
**Fig 1:** *Shows the trends in purchasing individual drugs pre- and during the COVID-19 pandemic in Uganda (black line).The red line shows the national monthly number of reported COVID-19 cases (see secondary vertical axis). The time points correspond to the months and years before and during the COVID-19 epidemic in Uganda. For example, March 2020 corresponds to the month in which the first COVID-19 case was diagnosed in Uganda.*
## Antibiotics
During the pre-pandemic era, an average of 218.98 customers purchased Azithromycin medications. During the COVID-19 period, however, at the ascent and decline of cases in the first wave, the number of customers purchasing Azithromycin reduced by $79\%$ with a risk ratio (RR) ($95\%$ CI) of β3 = 0.21 ($95\%$ CI: 0.05–0.92) and β5 = 0.35 ($95\%$ CI: 0.16–0.77). However, towards the peak of the second wave, we observed a 2-fold increase in the number of clients with RR ($95\%$ CI) of β7 = 2.33 (1.00–5.41), $$P \leq 0.015$$, and as the second wave subsided, the purchase of Azithromycin dropped significantly by $85\%$ with RR ($95\%$ CI) β9 = 0.15 (0.07–0.30), $P \leq 0.001.$ Before the onset of COVID-19, an average of 206.29 customers purchased Erythromycin medication which reduced gradually by $17\%$ over time, RR ($95\%$ CI), β1 = 0.83 (0.81–0.84). During the first wave of the COVID-19 era, there were no significant changes in the number of customers buying Erythromycin, as the number of cases increased and decreased. However, during the second wave, as the number of cases peaked, we observed a six-fold increase with RR of β7 = 6.88 ($95\%$ CI: 3.66–13.43) and later a decline that corresponded with a reducing number of reported COVID-19 cases, RR ($95\%$) β9 = 0.04 (0.03–0.07), all $P \leq 0.001.$ A similar trend was observed for Cipfloxacin with no significant changes before COVID-19 and during the first wave, but with an increase as the second wave peaking with RR ($95\%$ CI) of β7 = 2.85 (1.29–6.32) and the number of purchases reduced by $64\%$ with RR β9 = 0.36 ($95\%$ CI: 0.20–0.67), as the second wave subsided. The use of Augmentin similarly increased by $43\%$ during the start of the pandemic RR (1.43 ($95\%$: 0.54–3.81), and this later reduced as the second wave declined by $56\%$ RR $95\%$ CI: 0.44 (0.21–0.93) “Table 1.”
**Table 1**
| Drug | β0 (Constant) | p | Β1 (Slope 1) | p.1 | Β3 (Slope 2) | p.2 | Β5 (Slope 3) | p.3 | Β7 (Slope 4) | p.4 | Β9 (Slope 5) | p.5 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Drug | | p | Pre-COVID period | p | Ascent of wave 1 | p | Descent of wave 1 | p | Ascent of wave 2 | p | Descent of wave 2 | p |
| Drug | (95% CI) | p | RR (95% CI) | p | RR (95% CI) | p | RR (95% CI) | p | RR (95% CI) | p | RR (95% CI) | p |
| Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics | Antibiotics |
| Azithromycin | 218.98(156.59–312.55) | <0.001 | 1.00(0.98–1.03) | 0.686 | 0.21(0.05–0.92) | 0.563 | 0.35(0.16–0.77) | 0.108 | 2.33(1.00–5.41) | 0.015 | 0.15(0.07–0.30) | <0.001 |
| Erythromycin | 206.29(187.31–226.87) | <0.001 | 0.83(0.80–0.84) | <0.001 | 0.99(0.62–1.46) | 0.961 | 1.56(0.87–2.93) | 0.149 | 6.88(3.66–13.43) | <0.001 | 0.04(0.03–0.07) | <0.001 |
| Ciprofloxacin | 267.56(192.80–378.54) | <0.001 | 0.99(0.97–1.01) | 0.338 | 0.98(0.89–1.08) | 0.718 | 0.69(0.39–1.23) | 0.223 | 2.85(1.29–6.32) | 0.013 | 0.36(0.20–0.67) | 0.001 |
| Augmentin | 133.04(88.85–204.90) | <0.001 | 1.02(1.00–1.05) | 0.068 | 0.95(0.84–1.07) | 0.454 | 0.87(0.42–1.77) | 0.688 | 1.43(0.54–3.81) | 0.462 | 0.44(0.21–0.93) | 0.025 |
| Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements | Supplements |
| Zinc | 4.74(2.65–8.62) | <0.001 | 1.01(0.97–1.05) | 0.673 | 1.08(0.91–1.28) | 0.383 | 4.84(1.35–24.67) | 0.012 | 0.22(0.04–0.99) | 0.048 | 0.26(0.09–0.71) | 0.011 |
| Vitamin C | 9.06(4.26–22.07) | <0.001 | 1.04(0.99–1.09) | 0.192 | 0.95(0.63–1.37) | 0.731 | 0.58(0.06–4.98) | 0.513 | 35.15(1.65–187.21) | 0.008 | 0.01(0.00–0.07) | <0.001 |
| Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids | Corticosteroids |
| Dexamethasone | 413.31(219.09–840.79) | <0.001 | 0.95(0.91–0.99) | 0.014 | 1.12(0.91–1.38) | 0.302 | 0.47(0.14–1.56) | 0.215 | 3.07(0.61–15.52) | 0.185 | 0.17(0.05–0.60) | 0.005 |
## Supplements
The purchase of supplements, including Zinc and Vitamin C, was affected by the COVID-19 epidemic. Before the onset of COVID-19, there was an average of β0 = 4.74 persons purchasing Zinc supplements, which remained relatively stable in the months leading up to the start of the pandemic. However, as the number of cases of COVID-19 decreased during the first wave, the number of clients increased 4-fold with RR of β5 = 4.84 ($95\%$ CI: 1.35–24.67), $$P \leq 0.012.$$ The decreases in the number of clients purchasing Zinc were sustained with a significant reduction at the rise and fall of the second wave of the pandemic with RR ($95\%$ CI) of β7 = 0.22 (0.04–0.99), $$P \leq 0.048$$ and β9 = 0.26 (0.09–0.71), $$P \leq 0.011$$, respectively. We observed remarkable increases in the number of persons purchasing Vitamin C during the second wave of the pandemic, with a 35-fold increase from the pre-pandemic era, RR ($95\%$ CI) of β7 = 35.15 (1.65–187.2), $$P \leq 0.008$$, and the purchases reduced with reducing COVID-19 cases, RR ($95\%$ CI) β9 = 0.01 (0.00–0.07), $P \leq 0.001$ “Table 1”.
## Corticosteroids
We observed many people purchasing Dexamethasone before the pandemic, with an average of β0 = 413.31 monthly customers, which reduced slightly over time. However, during the second wave of COVID-19, there was a non-statistically significant three-fold increase in the number of customers, RR ($95\%$ CI) β7 = 3.07 (0.61–15.52), and after that, a substantial decrease in purchases at the end of the second wave, β9 = 0.17 (0.05–0.60), $$P \leq 0.005$$ “Table 1”.
## Discussion
We observed an increase in drug purchases and an indicative increase in self-medication during the COVID-19 epidemic in Uganda. There were significant increases in the purchase of antibiotics, including Azithromycin and Erythromycin. Moreover, the increases in the number of drug purchases correlated with the different waves in the country, especially during the second wave of the pandemic, which was driven by the (SARS-CoV-2) B.1.617.2 (delta) variant [20].
In Uganda, like many other parts of the world, the alarming infodemic, rising self-medication, and stockpiling of drugs, including antibiotics, led to elevated purchases of critical medications believed to treat COVID-19 [21, 22]. In addition, the exponential increase in uptake of antibiotics such as Azithromycin for the misinformed treatment of viral infections, including COVID-19, likely led to heightened antimicrobial resistance [16, 23]. Similarly, we demonstrated significant increases in people purchasing health supplements, including vitamin C.
This could have been influenced by beliefs spread through social media or rumors about the efficacy of particular drugs, particularly during the lockdown, when access to health facilities was reduced. In addition, the increase in Vitamin C was driven by recommendations from scientific evidence, health workers, and social media about its role in increasing immunity and improving the survival of COVID-19 patients [24]. Finally, the self-medication of the selected drugs could have been stimulated by media and influencers who encouraged people in the community to stockpile medicines and supplements, as depicted in our results.
Unfortunately, individuals who self-medicate might misinterpret the recommended dosage and frequency, thus leading to an increased likelihood of severe side effects, drug interactions, and limited efficacy. The unregulated use of corticosteroids such as dexamethasone, which is only recommended among persons with severe conditions related to cytokines, poses a serious risk of severe adverse outcomes among people with underlying comorbidities, including diabetes [25].
Our results demonstrated a significant increase in the use of Azithromycin, particularly during the second wave. This steep increase in the use of antibiotics was similar to what was reported in India [16, 26], a country that continues to lead in drug manufacturing. In India, the increase was noted as early as the first wave, while in Uganda, the growth was higher in the second wave, with the most significant consumption among presumptive or confirmed COVID-19 cases. These findings seriously affect antimicrobial resistance in low- and middle-income countries. During the pandemic, the overstretched health systems, fear of contracting the disease, and non-pharmaceutical prevention measures, such as the suspension of public transport, made it impossible to access treatment and foreclosed the option of self-medication at a nearby community pharmacy [27]. Like many other resource-limited settings, Uganda’s access to proper health care was limited during the pandemic. Community pharmacies presented a potential opportunity to access treatment for COVID-19-related symptoms. Community pharmacies offered clients convenience (e.g., short waiting times), privacy, and efficiency, characteristics that made them desirable destinations during the peak of the second wave in Uganda, driven by the Delta variant [28, 29]. In June 2021, during the second wave of the pandemic in Uganda, a widely distributed flyer recommended using a COVID-19 treatment package that included Azithromycin, Augmentin, Dexamethasone, Vitamin C, and Zinc, leading to massive increases in drug purchases.
Proper communication about the harmful effects of irrational drug use is essential during pandemics. More importantly, the community should be sensitized about seeking advice from health experts before purchasing and eventually using any medication. In Uganda, social media was critical in promoting self-medication for COVID-19. The drug stockpiling and use of unprescribed medication revealed the public’s vulnerability during the global pandemic threat [30]. The fear of contracting and the experience of being diagnosed with COVID-19 caused members of communities to buy drugs to feel secure and treat COVID-19 symptoms. However, the consequential effect on antimicrobial resistance in the community is yet to be evaluated. Therefore, pharmacists and drug shops had a vital role in the community response to the COVID-19 pandemic. Regular critical analysis of pharmacy purchase data may be relevant to inform appropriate guidance on self-medication in the community [31].
## Strengths and limitations
The major strength of our study is its originality and innovativeness in using customer purchase data. First, due to the need for studies using similar data sources in Africa, this research could be used as a benchmark for future studies that utilize unconventional data sources such as community pharmacy records. Such unexplored data sources can address essential health outcomes and activities in challenging times, such as pandemics. Secondly, unlike previous studies that compared drug use pre- and post-COVID-19, our unique approach of splitting the post-pandemic era into different waves to correct special drug purchases with the increasing and reducing reported cases of COVID-19 provides additional information on the effect of each of the waves on community self-medication based on the severity of the variants.
Our study was limited to pharmacies with electronic receipting services from which the data could be extracted. Therefore, our data is likely an underestimation of the increase in self-medication during the pandemic because most pharmacies and drug shops in the community still need electronic receipting systems. In Cameroon, mathematical modeling approaches have been used to estimate the effect of self-medication on the COVID-response [32]. Among participants in a Nigerian study, nearly $80\%$ reported purchasing drugs from community pharmacies [6], thus emphasizing the need for innovative ways to estimate self-medication following data mining pharmacy drug records. We postulate that customer purchase data and drug stockpiling during the epidemic are proxy indicators of increased antimicrobial consumption in the absence of measures of direct consumption. We, therefore, recommend interventions to measure antimicrobial consumption and resistance patterns in the community. Our data lacked demographic characteristics of the individuals who purchased the medication and is therefore limited in multivariable modeling of the drivers of drug purchases. The absence of unique identifiers in the pharmacy database made it impossible to identify repeated purchases from a single individual. Therefore any correlation among clients was not accounted for in the modes. With increased advocacy for data sharing and the use of data science methods to mine pharmacy records, the value of such unconventional data sources will go a long way in informing public health responses in our settings of interest. Finally, we are aware that there could be structural differences in clients who purchase drugs from private and public pharmacies that might be driven by their spending ability and differences in drug pricing in rural and urban areas in Africa [33]. Due to the high prices of drugs, the increase in purchases might have been more in urban Uganda than in rural areas [34]. Therefore, we recommend that future studies be undertaken with a larger sample size from pharmacies located in rural and urban areas. Studies should consider pharmacists’ perspectives using a mixed methods approach.
## Conclusion
Analysis of drug purchase records pharmacies revealed significant increases in antibiotics and supplements purchases correlated with the trends in COVID-19 cases during the pandemic. The unregulated or unrestricted availability of over-the-counter drugs and their use for the treatment of non-bacterial infections such as SARS-CoV-2 remains of great concern for consequential antimicrobial resistance patterns. More data is needed to understand the impact of COVID-19 self-prescription practices on antimicrobial resistance patterns in the community.
## References
1. Quincho-Lopez A, Benites-Ibarra CA, Hilario-Gomez MM, Quijano-Escate R, Taype-Rondan A. **Self-medication practices to prevent or manage COVID-19: A systematic review.**. *PLoS One* (2021.0) **16** e0259317. DOI: 10.1371/journal.pone.0259317
2. 2World Health Organization. WHO Collaborating Centre for Drug Statistics Methodology. Guidelines for ATC classification and DDD assignment 2013. 2012.. *WHO Collaborating Centre for Drug Statistics Methodology. Guidelines for ATC classification and DDD assignment 2013* (2012.0)
3. **Guidelines for the regulatory assessment of medicinal products for use in self-medication.**. (2000.0)
4. Achienge OL. *The Pandemic Within the Pandemic* (2021.0)
5. Sadio AJ, Gbeasor-Komlanvi FA, Konu RY, Bakoubayi AW, Tchankoni MK, Bitty-Anderson AM. **Assessment of self-medication practices in the context of the COVID-19 outbreak in Togo.**. *BMC Public Health* (2021.0) **21** 58. DOI: 10.1186/s12889-020-10145-1
6. Wegbom AI, Edet CK, Raimi O, Fagbamigbe AF, Kiri VA. **Self-Medication Practices and Associated Factors in the Prevention and/or Treatment of COVID-19 Virus: A Population-Based Survey in Nigeria.**. *Front Public Health.* (2021.0) **9** 606801. DOI: 10.3389/fpubh.2021.606801
7. Eticha T, Mesfin K. **Self-medication practices in Mekelle, Ethiopia.**. *PLoS One* (2014.0) **9** e97464. DOI: 10.1371/journal.pone.0097464
8. Curtin D, O’Mahony D, Gallagher P. **Drug consumption and futile medication prescribing in the last year of life: an observational study.**. *Age Ageing.* (2018.0) **47** 749-53. PMID: 29688246
9. Pulkki J, Aaltonen M, Raitanen J, Rissanen P, Jylha M, Forma L. **Purchases of medicines among community-dwelling older people: comparing people in the last 2 years of life and those who lived at least 2 years longer.**. *Eur J Ageing* (2020.0) **17** 361-9. PMID: 32904873
10. 10World Health Organization. https://www.euro.who.int/en/health-topics/disease-prevention/antimicrobial-resistance/about-amr. 2022. [Accessed 5th January, 2023]
11. 11COVID-19 Data Explorer. https://ourworldindata.org/explorers/coronavirus-data-explorer. 2022. [Accessed 5th January, 2023]. *COVID-19 Data Explorer* (2022.0)
12. Onchonga D.. **A Google Trends study on the interest in self-medication during the 2019 novel coronavirus (COVID-19) disease pandemic**. *Saudi Pharm J.* (2020.0) **28** 903-4. DOI: 10.1016/j.jsps.2020.06.007
13. Gautret P, Lagier JC, Parola P, Hoang VT, Meddeb L, Mailhe M. **Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an open-label non-randomized clinical trial.**. *Int J Antimicrob Agents* (2020.0) **56** 105949. DOI: 10.1016/j.ijantimicag.2020.105949
14. Rayman G, Lumb AN, Kennon B, Cottrell C, Nagi D, Page E. **Dexamethasone therapy in COVID-19 patients: implications and guidance for the management of blood glucose in people with and without diabetes**. *Diabet Med* (2021.0) **38** e14378. DOI: 10.1111/dme.14378
15. Noreen S, Maqbool I, Madni A. **Dexamethasone: Therapeutic potential, risks, and future projection during COVID-19 pandemic**. *Eur J Pharmacol* (2021.0) **894** 173854. DOI: 10.1016/j.ejphar.2021.173854
16. Bagcchi S.. *COVID 19 boosted antibioticsales in India* (2022.0)
17. 17Ecopharm Uganda Limited. https://ecopharmug.com/ 2022 [Accessed 5th January, 2023]
18. 18Vine Pharmaceuticals Uganda Limited. https://www.yellow.ug/company/8622/vine-pharmaceuticals-ltd 2022 [Accessed 5th January, 2023]
19. Linden A.. **Challenges to validity in single-group interrupted time series analysis**. *J Eval Clin Pract.* (2017.0) **23** 413-8. DOI: 10.1111/jep.12638
20. Lopez Bernal J, Andrews N, Gower C, Gallagher E, Simmons R, Thelwall S. **Effectiveness of Covid-19 Vaccines against the B.1.617.2 (Delta) Variant.**. *N Engl J Med* (2021.0) **385** 585-94. DOI: 10.1056/NEJMoa2108891
21. Ahmad S, Babar MS, Essar MY, Sinha M, Nadkar A. **Infodemic, self-medication and stockpiling: a worrying combination.**. *East Mediterr Health J* (2021.0) **27** 438-40. DOI: 10.26719/emhj.21.010
22. Al Zoubi S, Gharaibeh L, Jaber HM, Al-Zoubi Z. **Household Drug Stockpiling and Panic Buying of Drugs During the COVID-19 Pandemic: A Study From Jordan.**. *Front Pharmacol.* (2021.0) **12** 813405. DOI: 10.3389/fphar.2021.813405
23. Sulis G, Batomen B, Kotwani A, Pai M, Gandra S. **Sales of antibiotics and hydroxychloroquine in India during the COVID-19 epidemic: An interrupted time series analysis.**. *PLoS Med.* (2021.0) **18** e1003682. DOI: 10.1371/journal.pmed.1003682
24. Bae M, Kim H. **Mini-Review on the Roles of Vitamin C, Vitamin D, and Selenium in the Immune System against COVID-19.**. *Molecules* (2020.0) **25**. DOI: 10.3390/molecules25225346
25. Soy M, Keser G, Atagunduz P, Tabak F, Atagunduz I, Kayhan S. **Cytokine storm in COVID-19: pathogenesis and overview of anti-inflammatory agents used in treatment.**. *Clin Rheumatol.* (2020.0) **39** 2085-94. DOI: 10.1007/s10067-020-05190-5
26. Afridi MI, Rasool G, Tabassum R, Shaheen M, Siddiqullah M.. **Prevalence and pattern of self-medication in Karachi: A community survey.**. *Pak J Med Sci.* (2015.0) **31** 1241-5. PMID: 26649022
27. Malik M, Tahir MJ, Jabbar R, Ahmed A, Hussain R. **Self-medication during Covid-19 pandemic: challenges and opportunities.**. *Drugs Ther Perspect* (2020.0) **36** 565-7. DOI: 10.1007/s40267-020-00785-z
28. Okereke M.. **Spread of the delta coronavirus variant: Africa must be on watch**. *Public Health Pract (Oxf)* (2021.0) **2** 100209. DOI: 10.1016/j.puhip.2021.100209
29. Nations United. *Delta variant drives Africa COVID threat to “whole new level”: WHO warns; ‘dominant’ in Europe by August* (2021.0)
30. Kiwanuka MAN. **Institutional vulnerabilities, COVID-19, resilience mechanisms and societal relationships in developing countries**. *International Journal of Discrimination and the Law* (2021.0) **21**
31. 31National Drug Authority. Good Distribution Practice for Pharmaceutical Products. https://wwwndaorug. 2018
[Accessed 5th January, 2023]. *Good Distribution Practice for Pharmaceutical Products* (2018.0)
32. Kong JD, Tchuendom RF, Adeleye SA, David JF, Admasu FS, Bakare EA. **SARS-CoV-2 and self-medication in Cameroon: a mathematical model.**. *J Biol Dyn.* (2021.0) **15** 137-50. DOI: 10.1080/17513758.2021.1883130
33. Russo G, McPake B. **Medicine prices in urban Mozambique: a public health and economic study of pharmaceutical markets and price determinants in low-income settings.**. *Health Policy Plan* (2010.0) **25** 70-84. DOI: 10.1093/heapol/czp042
34. Suh GH. **High medicine prices and poor affordability.**. *Curr Opin Psychiatry* (2011.0) **24** 341-5. DOI: 10.1097/YCO.0b013e3283477b68
|
---
title: Effect of fixed 7.5 minutes’ moderate intensity exercise bouts on body composition
and blood pressure among sedentary adults with prehypertension in Western-Kenya
authors:
- Karani Magutah
- Grace Mbuthia
- James Amisi Akiruga
- Diresibachew Haile
- Kihumbu Thairu
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021634
doi: 10.1371/journal.pgph.0000806
license: CC BY 4.0
---
# Effect of fixed 7.5 minutes’ moderate intensity exercise bouts on body composition and blood pressure among sedentary adults with prehypertension in Western-Kenya
## Abstract
Prehypertension is a modifiable risk factor for cardiovascular disease observed to affect an estimated 25–$59\%$ of global population and closely associated with body composition. Without appropriate interventions, one-third of individuals with prehypertension would develop full-blown hypertension within 4 years. The existing exercise recommendations need substitutes that appeal more yet accord similar or better outcomes in desire to halt this progression. This study evaluated the effect of Fixed 7.5-minute Moderate Intensity Exercise (F-7.5m-MIE) bouts on Body Composition and Blood Pressure (BP) among sedentary adults with prehypertension in Western-Kenya in a Randomized Control Trial (RCT) performed throughout the day compared to the single-continuous 30-60-minute bouts performed 3 to 5 times weekly. This RCT, with three arms of Experimental Group1 (EG1) performing the F-7.5m-MIE bouts, Experimental Group 2 (EG2) performing current World Health Organization (WHO) recommendation of ≥30-min bouts, and, control group (CG), was conducted among 665 consenting pre-hypertensive sedentary adults enrolled from western Kenya. EG1 and EG2 performed similar weekly cumulative minutes of moderate intensity exercises. Adherence was determined using activity monitors and exercise logs. Data regarding demographic characteristics, heart rate, BP, and anthropometric measures were collected at baseline and 12th week follow-up. Data regarding univariate, bivariate and multivariate (repeated measurements between and within groups) analysis were conducted using STATA version 13 at $5\%$ level of significance. The study revealed that males ($92.1\%$ in EG1, $92\%$ in EG2 and $96.3\%$ in CG) and females ($94.6\%$ in EG1, $89.3\%$ in EG2 and $95\%$ in CG) in the three arms completed the exercise at follow-up respectively. At 12th week follow-up from all exercise groups, males’ and females’ measurements for waist-hip-ratio, waist-height-ratio, systolic BP (SBP), heart rate and pulse pressure showed significant drops from baseline, while diastolic BP (DBP) and body mass index (BMI) reported mixed results for males and females from the various treatments. Both treatments demonstrated favourable outcomes. However, differences in the change between baseline and endpoint yielded mixed outcomes (SBP; $p \leq 0.05$ for both males and females, DBP; $p \leq 0.05$ for males and females, waist-height-ratio; $$p \leq 0.01$$ and <0.05 for males and females respectively, waist-hip-ratio; $$P \leq 0.01$$ and >0.05 for males and females respectively, BMI; $p \leq 0.05$ for both males and females, heart rate; $p \leq 0.05$ for males and females and pulse pressure; $$p \leq 0.01$$ and >0.05 for males and females respectively). The study design however could not test for superiority. The study demonstrated that the F-7.5m- MIE treatment programme and the WHO recommended 3–5 times weekly bouts of 30–60 minutes regime produced comparably similar favourable outcomes in adherence and BP reductions with improved body composition.
Trial registration: Trial registered with Pan African Clinical Trial Registry (www.pactr.org): no. PACTR202107584701552. ( S3 Text)
## Introduction
Sedentary lifestyles has contributed to the recent increase in non-communicable diseases in Kenya, causing a heavy health and economic burden [1]. Hypertension, the commonest cardiovascular disease (CVD) affects 1 in 7 people globally, contributing the most mortality [2–4]. Traditionally defined as blood pressure (BP) ≥$\frac{140}{90}$, revised definition has included lower values ≥$\frac{130}{80}$ [5]. A third ($33\%$) of individuals with higher-normal BP develop hypertension with age [6, 7]. Identifying at-higher-risk individuals early and intervening before full-blown disease is critical [6–8]. Prehypertension, the pre-disease state, is defined as systolic BP (SBP) ≥120–139 and/or diastolic BP (DBP) ≥80–89 in individuals aged 18 years and above on ≥2 consecutive measurements. Although recently the term “elevated” BP (SBP 120–129 and DBP <80 mmHg) was proposed, these proposals are not yet adopted in Kenya [5–7, 9, 10] and literature based on this new definition lacks. Prehypertension affects 25–$59\%$ of the global population [6, 9]. Independently, it is a modifiable risk factor for CVD.
At $26\%$ prevalence in sub-Saharan Africa (SSA), hypertension burden and relationship to CVD is a growing concern [11]. In 2018, $25\%$ of Kenyans had hypertension, highlighting extent of the burden locally [4, 12]. The prevalence of prehypertension in SSA ranges from 21–$33\%$ [3, 11] and was $47\%$ ($51\%$ for males; $46\%$ for females) in Kenya in 2018 [4], higher than neighbouring Uganda ($33\%$; $42\%$ for males, $29\%$ for females) [13]. Prevalence peaks at 69 years, dropping thereafter coinciding with increasing hypertension prevalence as individuals transition to full-blown hypertension [4, 7]. Without interventions, 20–$33\%$ of individuals with prehypertension develop hypertension within 4 years especially if they have higher DBP before age 50, and the risk doubles at BP ranges 130-$\frac{139}{85}$-89 mmHg as opposed to 120-$\frac{129}{80}$-84 mmHg [6, 14, 15].
Pre-hypertension also poses direct risk to CVD by associating with chronic cardiac and vascular changes like arterial stiffness and decreasing intima-media thickness, left ventricular hypertrophy, coronary heart disease, chronic kidney disease and end-stage-renal-disease [7, 16, 17]. Modifiable factors such as smoking, being sedentary, obesity, dyslipidaemia, and dietary issues are associated with development of prehypertension, although it is unclear how these factors contribute in progression from prehypertension to hypertension [6, 13, 18]. Waist-Hip ratio (WHR), body mass index (BMI) and weight-height ratio (WHtR) have all been linked with development of prehypertension and full-blown hypertensive disease for all ages [19–22].
The cost of screening and treating a hypertensive individual in *Kenya is* USD 178 monthly [23]. This is worrying in an economy where while $25\%$ are hypertensive and therefore likely to spend heavily on treatment, $36\%$ of them live on <1 USD a day [23, 24]. Currently, first line management of prehypertension is lifestyle change targeting modifiable factors as opposed to pharmacologic interventions, unless there is concurrent diabetes, kidney, or cardiac disease [4, 9]. Evidence supporting pharmacological intervention is inconclusive [25]. Exercise independently lowers BP in both hypertensive and non-hypertensive individuals [10, 26, 27]. Aerobic exercise in younger individuals or those with prehypertension not only lowers BP but also mitigates occurrence of full blown hypertensive disease even independently of dietary adjustment and weight loss interventions. Further, it has better results in all forms of hypertension where drug-therapy has failed [7, 9, 28, 29].
The search for a feasible way to prevent transition of prehypertension to hypertension therefore is necessary. We recently found that moderate-intensity exercise regimes involving bouts of <10 minutes but whose cumulative weekly time equals current World Health Organization (WHO) recommendations of 150 minutes has higher appeal and, yet, confer appreciable health benefits on sedentary normotensive individuals aged ≥50 years [30–32]. Existing guidelines of moderate intensity exercise for adults have traditionally been achieved by performing 30–60 minute bouts of exercise for 3–5 days weekly, and, for hypertension, there is advocacy to do this daily [27, 32, 33]. Despite our recent findings on beneficial health outcomes of shorter bouts, [30, 31], it is unclear if these benefits would translate similarly for sedentary individuals with prehypertension. Longer regimes of ≥30 minutes in 3–5 days weekly as currently recommended are beneficial but lack appeal [27, 34]. Studies on optional regimes are scanty, with pockets of emerging literature showing accumulating short exercise bouts may impact BP. Emerging knowledge points that accumulating running time of 30 minutes daily, in frequent short bouts of ≥10 minutes lowers SBP in non-hypertensive individuals in 24 hours to a few days [35]. This, however, not only remains inconclusive, but evidence/literature on longer term effect of short-bouts on BP is lacking. Further, evidence on effect of short-intermittent exercise on DBP is minimal, but in a study using 10-minutes-walking exercise tests reduced SBP but not DBP [36]. We are not aware of any randomized-controlled trial that has examined the effect of sub-10 minutes’ moderate intensity exercise on sedentary individuals with prehypertension, a neglected subpopulation.
The current work tested for equally or more appealing exercise regimes amongst individuals with prehypertension as a way to control their BP. We evaluated adherence to and BP benefits of cumulative fixed bouts of 7.5-minutes’ moderate intensity exercises (F-7.5m-MIE) performed throughout the day compared to the single-continuous 30–60 minute bouts among sedentary adults with prehypertension.
## Ethics statement
This work was approved by Moi Teaching and Referral Hospital (MTRH) / Moi University Institutional Research Ethics Committee on 17th March 2020 (approval no. 0003551) (S1 Text), received national licensure on 15th May 2020 (NACOSTI/P/$\frac{20}{4938}$) (S2 Text), and availed a physician throughout implementation phase to handle any adverse events. Following explanation of the objectives and procedures of the study, subjects were medically screened and gave written consent to participate in the study.
## Design
This was a randomized controlled field trial amongst residents from western region of Kenya where >$80\%$ are sedentary [37]. This followed an improved adherence and appreciable body composition and cardio-metabolic finding in an older cohort [38].
## Trial registration
The authors confirm that all ongoing and related trials for this intervention are registered. This trial was retrospectively registered with Pan African Clinical Trial Registry (www.pactr.org) (no. PACTR202107584701552) (S3 Text). The investigators earlier felt that being a behavioral as opposed to substance/drug trial, registration would not have been necessary. Upon learning that it met the WHO definition of a clinical trial being a prospective study that assigned participants to behavioural interventions and measuring associated effects, and with the trial already done, retrospective registration became necessary and letter S4 Text was obtained.
## Study population and sampling
We studied 665 sedentary adults who had prehypertension (≥18 (range 18–79) years; weekly metabolic equivalent minutes (MET-minutes) <600; SBP ≥120–139 mmHg and/or DBP ≥80–89 mmHg) following a local print advertisement. For motivation, screened volunteers received a full physical examination. The study had 3 arms each comprising males and females. ANOVA sample computation using expected DBP means of 72.9±1.4 mmHg for participants performing our trial regime (described below) and 72.2±1.8 mmHg for those on the traditional regime as found in our previous data [39], and, 82 mmHg for the non-interventional group expected to maintain baseline values at end point, and, further, a considered drop out estimated at $25\%$ yielded 750 participants. For this computation and using our quoted previous data, DBP gave larger sample than using SBP. We did individual-level randomization for each sex into trial (F-7.5m-MIE) arm EG1, current standard WHO recommendation (30–60 min bouts) arm EG2, and the non-intervention group CG (no guidelines exist for prehypertension care). After signing an informed consent, participants picked sealed envelopes they personally shuffled, randomly grouping themselves. Thereafter, each was explained to what their regime entailed. The associated progress from recruitment of participants and through all phases for the follow-up period is shown in the flow chart in Fig 1.
**Fig 1:** *Participants recruitment and follow up.*
## Protocol description
EG1 participants performed 3 bouts of F-7.5m-MIE each daily, their weekly cumulative exercise time reaching 150 minutes. Correspondingly, EG2 participants performed current recommendation of 30–60 minutes’ sessions for 3–5 days weekly, similarly yielding 150 minutes. The prescribed home-based moderate intensity exercises (jogging and related activities as per WHO guidelines) were such that they raised heart rate (HR) to 50–$70\%$ of participant’s maximal expected HR (220 –(minus) age in years, or where one could talk while performing but not sing) [33]. This was performed for 12 weeks. Participants wore activity monitors (Vivoactive 4 Garmin smart watches, Garmin International Inc 1200 East 51st Street, Olathe, Kansas, 66062 USA) on select days but additionally kept exercise logs analysed for adherence and quality control weekly, and, additionally, for data safety and monitoring board reporting and assessments, and for discontinuation and/or medical referral if indicated. Phone calls and text reminders for exercise performance follow-up were done weekly to ensure observance of prescriptions thus reducing attrition and cross-over effects of adopting non-prescribed exercise regimes. Participants meeting 150 minutes of exercise weekly and whose BP dropped or, at worst, remained in prehypertension ranges were retained in the follow-up. Those whose BP rose to hypertensive levels were reviewed by study physician and referred for pharmacotherapy. The CG continued normal lifestyles but were similarly followed up for BP measurements alongside intervention arms after 12 weeks.
## Data collection
The first participant in the current trial was recruited on 31st August 2020 and the follow-up for the last participant ended on 21st January 2021. Baseline data collected included bio-demographic characteristics, exercise patterns and HR and BP recordings (described below) using Omron M2 Basic (HEM-7120-E) automatic BP monitor (Omron Healthcare Co. Ltd, Kyoto, Japan). We also measured weight using a mechanical scale (CAMRY, BR9012, Shanghai, China), height, and hip and waist circumferences using a tape measure (also described below). The HR, BP, weight, hip and weight circumferences were repeated after 12 weeks.
## Measures
The Omrom M2 BP monitor described was used for HR and BP measurements. For both, measurements were taken as the average of two recordings done 2 minutes apart with participants awake and in a sitting position, and at rest. The cuff was held at the mid-upper arm and at level with the heart, and a normalization measurement that was discarded initially done to test the cuff and allay anxiety for the participant. The next two recorded measurements were adopted as per guidelines provided by American Heart Association [40]. From their average, we got the HR and BP measurements used for this study. Body composition measurements were taken in centimetres with participants upright standing, with feet positioned close together and arms at the side as provided in the criterion defined by WHO [41]. Specifically, height was taken with participants without shoes and facing straight ahead. This was done from the highest point of the head to the plantar of the foot. Where they had to have shoes, the shoe-sole width was subtracted from final height. For waist circumferences, the tape was run in direct contact to skin at the umbilicus level or point that yielded the least circumference. The widest portion of the buttocks was used for hip circumference. Weight was measured in kilograms with participants on light clothing and without shoes. The WHtR and WHR were computed through dividing the waist circumference by the height and by the hip circumferences respectively. The BMI was computed by dividing the weight in kilograms with the height in metres squared. The pulse pressure (PP) was gotten as the difference between SBP and DBP.
## Analysis
Data were analysed using STATA v.13 at univariate (means and standard deviations) for baseline and week 12, and bivariate level (t-tests; ANOVA) comparing data between groups. Multivariate analysis (mixed type MANOVA; RM ANOVA) comparing data between and within groups (repeated measurements) was also performed. For effect sizes, Cohen’s “d” was computed between groups. A P value of ≤0.05 signified a difference in BP between and within groups.
## Results
Male ($$n = 294$$) and female ($$n = 371$$) participants had mean age of 35.3±12.2 and 34.4±11.8 years respectively. For the males, $39\%$ and $50\%$ had highest level of education as secondary and tertiary respectively. For the females, this was $42\%$ and $35\%$ respectively. For the males and females respectively, $38.7\%$ and $20.6\%$ were in formal employment, $35.3\%$ and $49.6\%$ were in business or farming, and $25.7\%$ and $21.4\%$ were still in college while $8.5\%$ of the females were housewives.
For the males and females respectively, $97.3\%$ ($$n = 294$$) and $94.9\%$ ($$n = 371$$) had mean SBP ≥120 mmHg (129.9±5.4 mmHg and 128.7±4.9 mmHg) at baseline, and $83.7\%$ of the males ($$n = 294$$) and $83\%$ of the females ($$n = 371$$) had DBP ≥80 mmHg (84.0±3.4 mmHg and 84.1±3.3 mmHg respectively). Analysis of variance showed SBP at baseline was not different among males and females allocated into the different exercise regimes ($$p \leq 0.48$$ and $$p \leq 0.40$$ respectively). Similarly, DBP was not different for the males and females in the different regimes at baseline ($$p \leq 0.90$$ and $$p \leq 0.88$$ respectively). Among the males, BMI, WHR, WHtR, PP and HR means were 24.2±3.8, 0.92±0.08, 0.49±0.07, 98.3±4.0 and 74.1±9.6 beats per minute (bpm) respectively. For the females, the same, respectively, were 25.6±4.7, 0.91±0.08, 0.52±0.08, 97.8±3.89 and 77.3±10.7. Baseline demographic characteristics based on randomised groups are presented in Table 1.
**Table 1**
| Unnamed: 0 | M-EG1 (n = 101) | M-EG2 (n = 112) | M-CG (n = 81) | F-EG1 (n = 130) | F-EG2 (n = 140) | F-CG (n = 101) |
| --- | --- | --- | --- | --- | --- | --- |
| Age (years) | 35.8±12.8 | 34.5±11.9 | 35.7±12.2 | 35.0±12.8 | 34.1±11.3 | 34.2±11.1 |
| WHtR | 0.49±0.06 | 0.49±0.06 | 0.51±0.07 | 0.53±0.08 | 0.53±0.08 | 0.52±0.07 |
| WHR | 0.92±0.08 | 0.93±0.09 | 0.92±0.06 | 0.90±0.08 | 0.90±0.09 | 0.92±0.07 |
| BMI (Kg/M2) | 23.8±3.4 | 24.2±3.8 | 24.8±4.22 | 26.2±4.9 | 25.9±4.8 | 24.5±4.3 |
| SBP (mmHg) | 129.5±6.1 | 130.0±5.7 | 129.0±5.4 | 128.6±5.8 | 127.8±5.7 | 127.6±5.6 |
| DBP (mmHg) | 82.5±5.1 | 82.7±4.7 | 82.7±3.8 | 82.8±5.0 | 82.5±4.9 | 82.5±4.3 |
| Pulse Pressure | 98.1±4.4 | 98.5±4.2 | 98.1±3.2 | 98.0±4.2 | 97.6±3.9 | 97.5±3.4 |
| Pulse rate (b/m) | 75.2±9.2 | 73.1±10.0 | 74.2±9.5 | 77.5±11.4 | 77.0±9.2 | 77.5±11.9 |
For the follow-up period, activity data were similar for the two interventional groups with cumulative exercise minutes 155.6+2.9 versus 154.5+3.2 minutes (M-EG1 and M-EG2 respectively) for the males ($$p \leq 0.1$$). Similarly, among the females, it was 154.6+3.1 versus 154.7+3.1 minutes for F-EG1 and F-EG2 respectively ($$p \leq 0.70$$). For the overall adherence, $92.9\%$ of the participants completed the 12 weeks follow-up with $92.1\%$ in M-EG1, $92\%$ in M-EG2 and $96.3\%$ in M-CG for the male groups and $94.6\%$ in F-EG1, $89.3\%$ in F-EG2 and $95\%$ in F-CG for the female groups.
After 12 weeks adherence to prescribed exercises for all groups (M-EG1 $$n = 93$$, M-EG2 $$n = 103$$, M-CG $$n = 78$$, F-EG1 $$n = 123$$, F-EG2 $$n = 125$$, F-CG $$n = 96$$), there were varied effects on the cardiovascular and body composition measurements as shown in Table 2.
**Table 2**
| Variable | Group | Week 0 | Week 12 | Mean Δ (wk12-wk0) | P value (Δ from wk 0) |
| --- | --- | --- | --- | --- | --- |
| Male | | | | | |
| WHtR | M_EG1 | 0.49±0.06 | 0.48±0.06 | -0.01±0.02 | 0.0 |
| | M_EG2 | 0.48±0.06 | 0.48±0.06 | -0.01±0.02 | 0.0 |
| | M_CG | 0.51±0.07 | 0.51±0.07 | 0.00±0.02 | 0.82 |
| WHR | M_EG1 | 0.92±0.08 | 0.90±0.08 | -0.01±0.03 | 0.0 |
| | M_EG2 | 0.93±0.09 | 0.91±0.12 | -0.02±0.1 | 0.0 |
| | M_CG | 0.92±0.06 | 0.93±0.06 | 0.01±0.05 | 0.06 |
| BMI (Kg/M2) | M_EG1 | 24.1±3.4 | 23.7±3.4 | -0.4±0.7 | 0.41 |
| | M_EG2 | 24.0±3.8 | 23.5±3.6 | -0.5±0.7 | 0.99 |
| | M_CG | 24.6±3.7 | 24.7±3.7 | 0.1±0.5 | 0.0 |
| SBP (mmHg) | M_EG1 | 129.7±6.1 | 123.1±9.2 | -6.6±8.0 | 0.0 |
| | M_EG2 | 129.9±5.6 | 121.9±9.5 | -8.0±8.3 | 0.0 |
| | M_CG | 129.3±5.1 | 129.9±5.7 | 0.6±6.1 | 0.08 |
| DBP (mmHg) | M_EG1 | 82.3±5.1 | 77.0±5.5 | -5.3±6.8 | 0.04 |
| | M_EG2 | 82.6±4.5 | 77.5±5.9 | -5.1±6.5 | 0.01 |
| | M_CG | 82.6±3.8 | 81.7±5.4 | -0.9±5.7 | 0.31 |
| Pulse | M_EG1 | 98.1±4.4 | 92.4±6.1 | -5.7±6.7 | 0.7 |
| Pressure | M_EG2 | 98.4±4.1 | 92.3±6.4 | -6.1±6.1 | 0.79 |
| | M_CG | 98.2±3.2 | 97.8±4.5 | -0.4±4.8 | 0.97 |
| Pulse (b/m) | M_EG1 | 75.1±9.1 | 72.2±7.3 | -2.9±6.4 | 0.0 |
| | M_EG2 | 72.7±9.9 | 70.1±7.9 | -2.6±6.1 | 0.0 |
| | M_CG | 73.8±9.4 | 75.1±16.6 | 1.2±12.6 | 0.0 |
| Female | | | | | |
| WHtR | F_EG1 | 0.53±0.08 | 0.52±0.07 | -0.01±0.02 | 0.0 |
| | F_EG2 | 0.52±0.08 | 0.52±0.07 | -0.01±0.02 | 0.0 |
| | F_CG | 0.51±0.07 | 0.51±0.07 | 0.00±0.03 | 0.0 |
| WHR | F_EG1 | 0.91±0.08 | 0.9±0.08 | -0.02±0.04 | 0.0 |
| | F_EG2 | 0.9±0.09 | 0.88±0.08 | -0.02±0.03 | 0.0 |
| | F_CG | 0.92±0.07 | 0.91±0.11 | -0.01±0.1 | 0.04 |
| BMI (Kg/M2) | F_EG1 | 26.3±4.9 | 25.7±4.6 | -0.6±0.9 | 0.01 |
| | F_EG2 | 25.5±4.5 | 25.1±4.0 | -0.5±1.2 | 0.06 |
| | F_CG | 25.5±4.1 | 24.8±4.1 | 0.3±0.7 | 0.92 |
| SBP (mmHg) | F_EG1 | 128.6±5.7 | 121.3±8.4 | -7.3±6.5 | 0.0 |
| | F_EG2 | 127.7±5.7 | 120.5±7.5 | -7.2±7.5 | 0.01 |
| | F_CG | 127.7±5.4 | 127.0±7.7 | -0.7±6.9 | 0.0 |
| DBP (mmHg) | F_EG1 | 82.9±4.8 | 77.4±5.5 | -5.5±5.9 | 0.12 |
| | F_EG2 | 82.5±4.8 | 77.4±6.0 | -5.1±6.1 | 0.01 |
| | F_CG | 82.4±4.3 | 82.1±5.9 | -0.3±6.5 | 0.26 |
| Pulse | F_EG1 | 98.1±4.2 | 92.0±5.6 | -6.1±5.3 | 0.16 |
| Pressure | F_EG2 | 97.6±3.9 | 91.8±5.6 | -5.8±5.6 | 0.48 |
| | F_CG | 97.5±3.5 | 97.1±5.7 | -0.4±5.9 | 0.35 |
| Pulse (b/m) | F_EG1 | 77.4±11.0 | 73.3±8.5 | -4.0±6.4 | 0.0 |
| | F_EG2 | 77.1±9.4 | 74.1±7.7 | -3.0±5.2 | 0.0 |
| | F_CG | 77.0±11.0 | 75.8±9.3 | -1.2±7.3 | 0.0 |
Regression models for differences in the change between baseline and endpoint for the various groups all for males and females respectively showed that SBP ($p \leq 0.001$; $F = 31.39$ and $p \leq 0.001$; $F = 41.90$), DBP ($p \leq 0.001$; $F = 18.44$ and $p \leq 0.001$; $F = 34.45$), WHtR ($$p \leq 0.18$$; $F = 1.78$ and $$p \leq 0.69$$; $F = 0.16$), WHR ($$p \leq 0.01$$; $F = 6.83$ and $$p \leq 0.44$$; $F = 0.60$), BMI ($$p \leq 0.28$$; $F = 1.16$ and $$p \leq 0.45$$; $F = 0.58$), HR ($p \leq 0.001$; $F = 9.42$ and $p \leq 0.001$; $F = 10.72$) and PP ($$p \leq 0.01$$; $F = 6.47$ and $$p \leq 0.11$$; $F = 2.56$) all yielded mixed outcomes.
At baseline, more participants had higher values for various cardiovascular and body composition than observed at endpoint. Fig 2 shows the percentage dropping to respective variables’ cut-offs after the 12 weeks follow-up. It is noteworthy that the mean values changed as shown in Table 2 even where percentage drop of participants to below cut-offs for the various variables was marginal. The change difference for the two interventional groups were also minimal, with Cohen’s d for effect sizes for the various variables between baseline and week 12 for males and females respectively being: WHtR (0.02 and 0.2), WHR (0.1 and 0.3), BMI (0.04 and 0.2), SBP (0.2 and 0.01), DBP (0.03 and 0.1), HR (0.1 and 0.2) and pulse (0.05 and 0.1).
**Fig 2:** *Percentage attaining recommended cut-offs after 12 weeks.*
## Discussion
All the 665 participants of the current study had prehypertension at baseline, and were sedentary as per the WHO global physical activity questionnaire. The fact that majority had at least secondary level education made it easier to communicate what having prehypertension meant, the importance of exercise interventions, the prescribed exercise instructions, and how to monitor exercise activity. Similarly, the participants were predominantly young adults which allowed ease in exercise participation. However, almost all participants were in formal employment, in business or still in college and therefore had to purposely make time to follow their respective exercise prescriptions.
After the 12 weeks follow-up, those retained in the intervention groups had similar cumulative exercise time when groups per sex were compared. This allowed comparisons of the various variables yielded. Among the males from the two interventional groups, adherence rates were similar while females on shorter bouts of exercise had a 5-percentage-points higher adherence compared to those in the longer bouts’ group. Previous work on similar exercise regimes from the same setting have equally shown higher adherence in both males and females performing shorter bouts in an older population, suggesting that such regimes may be more appealing [30]. The current study equally suggests that across all adult ages, adoption of shorter exercise bouts may improve exercise adherence since individuals maintaining their exercise prescriptions matched those performing the WHO standard exercise regimes. Where individuals have prehypertension like in the current study, this may yield more benefits in control of their BP.
At the end of 12 weeks for body composition measurements, there was a significant reduction in both WHtR and WHR in the two experimental groups for both sexes compared to the baseline. The manner of this reduction was similar for the two regimes when compared between themselves and with the control group. To provide for body fuel during the prescribed exercises, the body mainly breaks down fatty stores. When these fats are in the abdominal region, and coupled with fact that height minimally changes over such a short follow-up period, and, further, that the hip region is mainly muscular as opposed to fatty tissue, the breakdown of the waist-region fats from the two interventions lowered these ratios. For BMI among males, recorded reduction was insignificant for both interventional groups but noteworthy was that the control group had a significant increase. Females in the short-bouts arm had a reduction in their BMI but the changes in the longer bouts and the control groups were insignificant. It is likely that the 12 weeks follow-up period could not allow clear and appreciable changes as weight reduction which affects BMI has been shown to be slower than other anthropometric measurements such as waist circumference change that affects WHR and WHtR, explaining the differences observed in these measures [42]. We previously observed positive changes in body composition from the same setting on healthy-sedentary individuals by use of comparable exercise regimes [31]. Elsewhere, similar follow-up periods for accumulated shorter bouts of exercise have yielded favourable body compositions [43, 44], although they were not specific for individuals with prehypertension.
There were mixed outcomes regarding cardiovascular effects of the 12 weeks’ exercise interventions. For both males and females, there was significant reduction in SBP for both experimental groups but not for the control. The manner of this reduction was similar for both regimes. The DBP also dropped in both regimes for the males and in the longer bouts for the female, and there was no change in the control group. There was a reduction of the resting HR by 3 to 4 BPM for the females and 2.5 to 3 BPM for males in all regimes, with no difference inter-regimes. For males and females, the 12 weeks follow-up reduced PP by 6 mmHg for both exercise regimes, although the difference was not significant. While these show that the prescribed exercises improved these cardiovascular measures, probably a longer follow-up is necessary to reduce this further. Still, the absolute PP reduction alludes to the narrowing of the pressure difference between SBP and DBP, which is an independent risk factor for cardiovascular disease related to stiffening of blood vessels [45]. In this study, short bouts of aerobic exercises reduced this stiffening in a similar manner to the currently prescribed bouts. A recent meta-analysis showed that there was no difference between accumulated shorter bouts and the traditionally longer-continuous bouts of exercise in BP modulation in the general population [46]. The present work shows similar benefits among a sub-population with prehypertension.
When the mean change in value of the various variables between baseline and endpoint was modelled for difference between regimes, there were significant differences observed between the two interventional groups on the one hand and the control group on the other. For SBP, DBP and HR, all had differences in mean change between the two interventional groups for both sexes. For WHR and PP, there a difference between the short and the long bouts among the males only. In both males and females, there was no observable difference between the two regimes for BMI and WHtR mean change. When we considered the effect sizes in the differences between the two regimes, computed Cohen’s “d” seemed to suggest that variances were negligible since the effects sizes were mostly below 0.2. Considering the outcomes, the two interventional regimes were largely similar showing that the trial regime was comparable to the existing WHO standard for moderate intensity exercises as currently recommended. The mechanisms in body composition and cardiovascular measurements change following an F-7.5m-MIE is somewhat same as that in the longer regime. Previous studies from our set up and elsewhere have shown that shorter prescribed bouts could actually yield better outcomes on body composition and cardiovascular measures [30, 31, 39, 47], which differences, although apparent but minimal in the current work, could not be pursued further for direction because our design could not support testing for superiority.
The percentage drop for WHR means ≥0.9 and ≥0.85 for males and females respectively was higher for short and long bouts for males and females respectively, but these differences were marginal. A similar mixed picture was observed for WHtR >0.5 and also for BMI but here, the longer regime appeared superior in the percentage drop in both categories for both males and females. These observations are replicated from an older population in the same set up [31], but one who although sedentary were not prehypertensive. We were unable to find any studies that have looked at effect of such short exercise regimes on similar cardiovascular and body composition variables as ours, and the current attempt is the first we are aware of. Still, probably a longer follow-up would illuminate these differences better, and, additionally, reduce those remaining with prehypertensive BPs or with body composition measurements above their respective cut-offs by an even larger proportion for both sexes and regimes. Of importance is that even with these mixed results, all variables had mean values drop between baseline and endpoint, suggesting similar value in the two experimental regimes. Evidence from previous studies on the effect of short and long bouts of exercises on body composition is inconclusive. While Alizadeh et al. [ 48] showed significant reduction in BMI and weight among females on shorter bouts of exercises compared to the long bouts of exercises, other studies [49–51] found both intermittent short bouts and continuous exercise programs to be effective in weight loss and improving body composition with no significant difference between the programs. On the contrary, a study involving middle aged obese women [52] showed that long bouts exercise are superior in the reduction of BMI, weight and fat mass. Further, Chung et al. [ 52] concluded that multiple short bout exercises are better than prolonged exercise when the goal is to reduce waist circumference. However, a meta-analysis on the effects of continuous compared to accumulated exercise on health showed no statistical differences between short and long exercise for any anthropometric or body composition outcome except body weight [46]. Further studies on the long term effect of short bouts exercises on body composition are recommended.
In the current study, both the short and the long exercise regimes had at least an 8 percentage drop in males who, separately, had SBP ≥120 and DBP ≥80 mmHg at baseline. Among the females, the percentage drop was higher among the long bouts’ group than in the shorter for SBP, but similar at $8.5\%$ for DBP. This underscores the similarity in effect for the two exercise regimes among individuals needing to regulate their BPs. While the two regimes had similar percentage drops for SBP and DBP in males and also DBP in females, it was unclear from the current study why this differed in SBP for the females. While we are not aware of any study that has looked at effect of aerobic exercises among individuals with prehypertension that lasted a similar period, one study has found regular moderate-intensity exercises lasting 10 minutes per session have yielded BP reductions of up to 5 mmHg [35]. High intensity exercises performed for a period slightly longer than in the current study have shown SBP and DBP decreases by about 8.7 mmHg and 5.4 mmHg respectively, similar to values in our current moderate-intensity study [53]. This is further supported by a systematic review and meta-analysis of randomized trials that showed comparable BP changes between high and moderate intensity individuals with pre- to established hypertension [54]. Given that in BP control adherence to prescriptions of moderate intensity exercises is higher than that for high intensity exercises [55], and, further, now that shorter regimes of moderate intensity exercises are comparable to the longer regimes of similar intensity in adherence, then prescriptions of shorter-bouts moderate intensity exercises here demonstrated as similarly beneficial in BP modulation for individuals with prehypertension become an important intervention for this sub-population.
Given that the drop-outs in each arm were less than $30\%$ as the set criterion for follow-ups exceeding 4 weeks [56], and, further, the drop in SBP exceeding 7 mmHg in 12 weeks for the two interventional groups as has been shown elsewhere using higher intensities and for longer periods [53], and this a significant drop when compared with the control group, we consider the current study a success. It demonstrates that F-7.5m-MIE performed thrice daily and whose cumulative exercise time reaches the 150 minutes mark as currently advocated for by WHO to be achieved in bouts of 30–60 minutes is equally beneficial among individuals with prehypertension. Exercise prescriptions involving shorter regimes which could be more appealing for some could play a crucial role in prevention of full-blown hypertension for individuals who have prehypertension. The shorter bouts in the current study had an insignificantly higher adherence rate but probably in a longer follow-up as demonstrated in our previous studies using normotensive individuals [31] or in a cross-over design, this would differ more significantly and/or shed more light. With the shorter bouts easier to implement, adhere to and yiedling similar cardiovascular disease protection as the longer regimes in sedentery individuals who have prehypertension, they should be considered in quest to reduce progression of prehypertension to hypertension, and, thus, providing more feasibe options that positively impact outcomes associated with exercise prescriptions practice. The current study therefore offers an additional yet equally appealing exercise regime that individuals with prehypertension could chose from and adopt in mitigating progression to full-blown hypetensive states.
## Limitations
Blinding of participants was impossible in the study and as such, interventional group participants may have had peer interactions that affected adherence, and likely also influenced the control group. Dietary records, smoking and use of non-medical drugs/stimulants, known confounders of BP and body compositions were not controlled for. Additionally, activity monitors were limited for the 665 participants and therefore not available for everyone throughout the follow-up period, with the control group completely missing out. These limitations may have affected the quality of our data and this may affect the generalisability of the current results.
## Conclusion
In sedentary individuals with prehypertension, F-7.5m-MIE performed over 12 weeks reduce BP and improve body composition to a similar magnitude as do the traditional longer bouts that last 30–60 minutes as currently recommended by WHO. The high adherence to this shorter bouts’ regime offer additional prescriptive options that could be used for individuals who are sedentary and have prehypertension.
## References
1. Ssewanyana D, Abubakar A, van Baar A, Mwangala PN, Newton CR. **Perspectives on Underlying Factors for Unhealthy Diet and Sedentary Lifestyle of Adolescents at a Kenyan Coastal Setting**. *Front Public Health* (2018.0) **6** 11. DOI: 10.3389/fpubh.2018.00011
2. Poulter NR, Prabhakaran D, Caulfield M. **Hypertension.**. *Lancet* (2015.0) **386** 801-812. DOI: 10.1016/S0140-6736(14)61468-9
3. Ataklte F, Erqou S, Kaptoge S, Taye B, Echouffo-Tcheugui JB, Kengne AP. **Burden of undiagnosed hypertension in sub-saharan Africa: a systematic review and meta-analysis**. *Hypertension* (2015.0) **65** 291-298. DOI: 10.1161/HYPERTENSIONAHA.114.04394
4. Ogola E, Yonga G, Njau K. **A14369 High Burden of Prehypertension in Kenya: Results from the Healthy Heart Africa (HHA) program**. *Journal of Hypertension* (2018.0) **36** e330
5. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: Executive Summary.**. *A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines* (2017.0)
6. Egan BM, Stevens-Fabry S. **Prehypertension—prevalence, health risks, and management strategies**. *Nat Rev Cardiol* (2015.0) **12** 289-300. DOI: 10.1038/nrcardio.2015.17
7. Redwine KM, Falkner B. **Progression of prehypertension to hypertension in adolescents**. *Curr Hypertens Rep* (2012.0) **14** 619-625. DOI: 10.1007/s11906-012-0299-y
8. Niiranen TJ, Larson MG, McCabe EL, Xanthakis V, Vasan RS, Cheng S. **Prognosis of Prehypertension Without Progression to Hypertension**. *Circulation* (2017.0) **136** 1262-1264. DOI: 10.1161/CIRCULATIONAHA.117.029317
9. Kanegae H, Oikawa T, Kario K. **Should Pre-hypertension Be Treated**. *Curr Hypertens Rep* (2017.0) **19** 91. DOI: 10.1007/s11906-017-0789-z
10. Whelton PK, Carey RM, Aronow WS, Casey DE, Collins KJ, Dennison Himmelfarb C. **2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults.**. *A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines* (2018.0) **71** e127-e248
11. Guwatudde D, Nankya-Mutyoba J, Kalyesubula R, Laurence C, Adebamowo C, Ajayi I. **The burden of hypertension in sub-Saharan Africa: a four-country cross sectional study**. *BMC Public Health* (2015.0) **15** 1211. DOI: 10.1186/s12889-015-2546-z
12. Mohamed SF, Mutua MK, Wamai R, Wekesah F, Haregu T, Juma P. **Prevalence, awareness, treatment and control of hypertension and their determinants: results from a national survey in Kenya.**. *BMC Public Health* (2018.0) **18** 1219. DOI: 10.1186/s12889-018-6052-y
13. Nuwaha F, Musinguzi G. **Pre-hypertension in Uganda: a cross-sectional study**. *BMC Cardiovascular Disorders* (2013.0) **13** 101. DOI: 10.1186/1471-2261-13-101
14. Vasan RS, Beiser A, Seshadri S, Larson MG, Kannel WB, D’Agostino RB. **Residual lifetime risk for developing hypertension in middle-aged women and men: The Framingham Heart Study**. *JAMA* (2002.0) **287** 1003-1010. DOI: 10.1001/jama.287.8.1003
15. Kanegae H, Oikawa T, Okawara Y, Hoshide S. **Which blood pressure measurement, systolic or diastolic, better predicts future hypertension in normotensive young adults?**. *J Clin Hypertens (Greenwich)* (2017.0) **19** 603-10. PMID: 28444926
16. Huang Y, Cai X, Liu C, Zhu D, Hua J, Hu Y. **Prehypertension and the risk of coronary heart disease in Asian and Western populations: a meta-analysis**. *J Am Heart Assoc* (2015.0) **4**
17. Li Y, Xia P, Xu L, Wang Y, Chen L. **A Meta-Analysis on Prehypertension and Chronic Kidney Disease**. *PLoS One* (2016.0) **11** e0156575. DOI: 10.1371/journal.pone.0156575
18. Song J, Chen X, Zhao Y, Mi J, Wu X, Gao H. **Risk factors for prehypertension and their interactive effect: a cross- sectional survey in China**. *BMC Cardiovascular Disorders* (2018.0) **18** 182. DOI: 10.1186/s12872-018-0917-y
19. Huxley R, Mendis S, Zheleznyakov E, Reddy S, Chan J. **Body mass index, waist circumference and waist:hip ratio as predictors of cardiovascular risk—a review of the literature**. *Eur J Clin Nutr* (2010.0) **64** 16-22. DOI: 10.1038/ejcn.2009.68
20. Chaudhary S, Alam M, Singh S, Deuja S, Karmacharya P, Mondal M. **Correlation of Blood Pressure with Body Mass Index, Waist Circumference and Waist by Hip Ratio**. *J Nepal Health Res Counc* (2019.0) **16** 410-413. PMID: 30739931
21. Xiao YQ, Liu Y, Zheng SL, Yang Y, Fan S, Yang C. **Relationship between hypertension and body mass index, waist circumference and waist-hip ratio in middle-aged and elderly residents**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2016.0) **37** 1223-1227. DOI: 10.3760/cma.j.issn.0254-6450.2016.09.008
22. Wieniawski P, Werner B. **Prediction of the hypertension risk in teenagers**. *Cardiol J* (2020.0). DOI: 10.5603/CJ.a2020.0079
23. Oti SO, van de Vijver S, Gomez GB, Agyemang C, Egondi T, Kyobutungi C. **Outcomes and costs of implementing a community-based intervention for hypertension in an urban slum in Kenya**. *Bull World Health Organ* (2016.0) **94** 501-509. DOI: 10.2471/BLT.15.156513
24. 24report Wb (https://data.worldbank.org/country/kenya; accessed on 2/12/2020).
25. Marta M, Zanchetti A, Wong ND, Malyszko J, Rysz J, Banach M. **Patients with prehypertension—do we have enough evidence to treat them**. *Curr Vasc Pharmacol* (2014.0) **12** 586-597. DOI: 10.2174/15701611113119990126
26. Borjesson M, Onerup A, Lundqvist S, Dahlof B. **Physical activity and exercise lower blood pressure in individuals with hypertension: narrative review of 27 RCTs**. *Br J Sports Med* (2016.0) **50** 356-361. DOI: 10.1136/bjsports-2015-095786
27. Pescatello LS, MacDonald HV, Lamberti L, Johnson BT. **Exercise for Hypertension: A Prescription Update Integrating Existing Recommendations with Emerging Research**. *Curr Hypertens Rep* (2015.0) **17** 87. DOI: 10.1007/s11906-015-0600-y
28. Chan J L, Ji Y K, Eugene S, Sung H H, MiKyung L, Justin Y J. **The Effects of Diet Alone or in Combination with Exercise in Patients with Prehypertension and Hypertension: a Randomized Controlled Trial**. *Korean Circ J* (2018.0) **48** 637-651. DOI: 10.4070/kcj.2017.0349
29. Lopes S, Mesquita-Bastos J, Alves AJ, Ribeiro F. **Exercise as a tool for hypertension and resistant hypertension management: current insights**. *Integr Blood Press Control* (2018.0) **11** 65-71. DOI: 10.2147/IBPC.S136028
30. Magutah K, Meiring R, Patel NB, Thairu K. **Effect of short and long moderate-intensity exercises in modifying cardiometabolic markers in sedentary Kenyans aged 50 years and above**. *BMJ Open Sport Exerc Med* (2018.0) **4** e000316. DOI: 10.1136/bmjsem-2017-000316
31. Magutah K, Patel NB, Thairu K. **Effect of moderate-intensity exercise bouts lasting <10 minutes on body composition in sedentary Kenyan adults aged >/ = 50 years**. *BMJ Open Sport Exerc Med* (2018.0) **4** e000403. DOI: 10.1136/bmjsem-2018-000403
32. 32World Health Organization (2010) Global Recommendations on Physical Activity for Health.. *Global Recommendations on Physical Activity for Health* (2010.0)
33. 33CDC (2015) https://www.cdc.gov/physicalactivity/basics/measuring/heartrate.htm accessed on 14/9/2018.
34. Linke SE, Gallo LC, Norman GJ. **Attrition and adherence rates of sustained vs. intermittent exercise interventions**. *Ann Behav Med* (2011.0) **42** 197-209. DOI: 10.1007/s12160-011-9279-8
35. Alpsoy Ş. **Exercise and Hypertension**. *Adv Exp Med Biol* (2020.0) **1228** 153-167. DOI: 10.1007/978-981-15-1792-1_10
36. Park S, Rink L, Wallace J. **Accumulation of physical activity: blood pressure reduction between 10-min walking sessions**. *J Hum Hypertens* (2008.0) **22** 475-482. DOI: 10.1038/jhh.2008.29
37. Nambakai JE, Kamau J, Amusa LO, Goon DT, Andanje M. **Factors influencing participation in physical exercise by the elderly in Eldoret West District, Kenya.**. *African Journalfor Physical, Health Education, Recreation and Dance (AJPHERD)* (2011.0) **17** 462-472
38. Magutah K, Patel NB, Thairu K. **Majority of Elderly Sedentary Kenyans Show Unfavorable Body Composition and Cardio-Metabolic Fitness**. *J Aging Sci* (2016.0) **4** 160
39. Magutah K, Thairu K, Patel N. **Effect of short moderate intensity exercise bouts on cardiovascular function and maximal oxygen consumption in sedentary older adults**. *BMJ Open Sport Exerc Med* (2020.0) **6** e000672. DOI: 10.1136/bmjsem-2019-000672
40. Muntner P, Shimbo D, Carey RM, Charleston JB, Gaillard T, Misra S. **Measurement of Blood Pressure in Humans: A Scientific Statement From the American Heart Association**. *Hypertension* (2019.0) **73** e35-e66. DOI: 10.1161/HYP.0000000000000087
41. 41WHO (2011) Waist circumference and waist–hip ratio: report of a WHO expert consultation, Geneva, 8–11 December 2008.
42. O’Donoghue G BC, Cunningham C, Lennon O, Perrotta C. **What exercise prescription is optimal to improve body composition and cardiorespiratory fitness in adults living with obesity? A network meta-analysis**. *Obes Rev.* (2021.0) **22** e13137. DOI: 10.1111/obr.13137
43. Kim H, Reece J, Kang M. **Effects of Accumulated Short Bouts of Exercise on Weight and Obesity Indices in Adults: A Meta-Analysis.**. *Am J Health Promot* (2020.0) **34** 96-104. DOI: 10.1177/0890117119872863
44. Lee HS, Lee J. **Effects of Exercise Interventions on Weight, Body Mass Index, Lean Body Mass and Accumulated Visceral Fat in Overweight and Obese Individuals: A Systematic Review and Meta-Analysis of Randomized Controlled Trials.**. *Int J Environ Res Public Health* (2021.0) **18**. DOI: 10.3390/ijerph18052635
45. White WB. **Systolic versus diastolic blood pressure versus pulse pressure**. *Curr Cardiol Rep* (2002.0) **4** 463-467. DOI: 10.1007/s11886-002-0107-4
46. Murphy MH, Lahart I, Carlin A, Murtagh E. **The Effects of Continuous Compared to Accumulated Exercise on Health: A Meta-Analytic Review.**. *Sports Med (Auckland, NZ)* (2019.0) **49** 1585-1607. DOI: 10.1007/s40279-019-01145-2
47. Madjd A TM, Delavari A, Malekzadeh R, Macdonald IA, Farshchi HR. **Effect of a Long Bout Versus Short Bouts of Walking on Weight Loss During a Weight-Loss Diet: A Randomized Trial**. *Obesity (Silver Spring).* (2019.0) **27** 551-558. DOI: 10.1002/oby.22416
48. Alizadeh Z, Kordi R, Rostami M, Mansournia MA, Hosseinzadeh-Attar SMJ, Fallah J. **Comparison between the effects of continuous and intermittent aerobic exercise on weight loss and body fat percentage in overweight and obese women: a randomized controlled trial**. *International journal of preventive medicine* (2013.0) **4** 881-888. PMID: 24049613
49. Schmidt WD, Biwer CJ, Kalscheuer LK. **Effects of long versus short bout exercise on fitness and weight loss in overweight females**. *Journal of the American College of Nutrition* (2001.0) **20** 494-501. DOI: 10.1080/07315724.2001.10719058
50. Jakicic JM, Winters C, Lang W, Wing RR. **Effects of intermittent exercise and use of home exercise equipment on adherence, weight loss, and fitness in overweight women: a randomized trial**. *Jama* (1999.0) **282** 1554-1560. DOI: 10.1001/jama.282.16.1554
51. Sykes K, Choo LL, Cotterrell M. **Accumulating aerobic exercise for effective weight control**. *The journal of the Royal Society for the Promotion of Health* (2004.0) **124** 24-28. DOI: 10.1177/146642400312400109
52. Chung J, Kim K, Hong J, Kong H-J. **Effects of prolonged exercise versus multiple short exercise sessions on risk for metabolic syndrome and the atherogenic index in middle-aged obese women: a randomised controlled trial**. *BMC women’s health* (2017.0) **17** 1-9. PMID: 28049464
53. Delgado-Floody P, Izquierdo M, Ramírez-Vélez R, Caamaño-Navarrete F, Moris R, Jerez-Mayorga D. **Effect of High-Intensity Interval Training on Body Composition, Cardiorespiratory Fitness, Blood Pressure, and Substrate Utilization During Exercise Among Prehypertensive and Hypertensive Patients With Excessive Adiposity**. *Front Physiol* (2020.0) **11** 558910. DOI: 10.3389/fphys.2020.558910
54. Costa EC, Hay JL, Kehler DS, Boreskie KF, Arora RC, Umpierre D. **Effects of High-Intensity Interval Training Versus Moderate-Intensity Continuous Training On Blood Pressure in Adults with Pre- to Established Hypertension: A Systematic Review and Meta-Analysis of Randomized Trials**. *Sports Med* (2018.0) **48** 2127-2142. DOI: 10.1007/s40279-018-0944-y
55. Ávila-Gandía V, Sánchez-Macarro M, Luque-Rubia A, García-Sánchez E, Cánovas F, López-Santiago A. **High versus Low-Moderate Intensity Exercise Training Program as an Adjunct to Antihypertensive Medication: A Pilot Clinical Study**. *J Pers Med* (2021.0) **11**. DOI: 10.3390/jpm11040291
56. Furlan AD, Malmivaara A, Chou R, Maher CG, Deyo RA, Schoene M. **2015 Updated Method Guideline for Systematic Reviews in the Cochrane Back and Neck Group**. *Spine (Phila Pa 1976)* (2015.0) **40** 1660-1673. DOI: 10.1097/BRS.0000000000001061
|
---
title: 'Understanding networks in rural Cambodian farming communities and how they
influence antibiotic use: A mixed methods study'
authors:
- Jane Mingjie Lim
- Sokchea Huy
- Ty Chhay
- Borin Khieu
- Li Yang Hsu
- Clarence C. Tam
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021636
doi: 10.1371/journal.pgph.0001569
license: CC BY 4.0
---
# Understanding networks in rural Cambodian farming communities and how they influence antibiotic use: A mixed methods study
## Abstract
Biosecurity and preventive animal health services in Cambodian smallholder backyard farming systems are often limited, leading to an over-reliance on antibiotics. However, data on factors influencing antibiotic use in these settings are lacking. We conducted a study in two rural Cambodian farming communities to investigate how social and contextual influences affect both human and animal antibiotic use behaviours. Data were collected in three phases: a baseline household census to enumerate village residents, a social network survey to understand village-level social ties, and in-depth interviews to elicit information about the influence of social ties on their decision-making processes. Primary outcome measures included knowledge, attitudes and practices surrounding antibiotic use, and awareness of issues relating to antibiotic resistance. Participants commonly accessed antibiotics or learned animal antibiotic use practices through village-level informal sources such as pharmacies or animal health workers. While most participants reported not using antibiotics for animal growth promotion or illness prevention, misconceptions surrounding both antibiotic effectiveness and resistance were common. Social networks capturing informal, work-related and health-related social ties showed that familial connections and geographic proximity were of primary importance for information sharing. Using exponential random graph models, we demonstrated that familial ties, and closer geographic and geodesic distance, were associated with similarity in overall antibiotic knowledge and attitudes. The informal private sector plays a major role in provision of antibiotics and antibiotic-related information in backyard farming communities, but such information is maintained within close social groups. This demonstrates the importance of engaging village-level informal sources in the provision of antibiotic-related information for both human and animal health, as well as in potential interventions to encourage appropriate antibiotic behaviours in lower-resourced settings.
## Background
Smallholder backyard farming systems comprise a significant portion of Cambodia’s gross domestic product and serve a substantial economic role in many rural Cambodian households–more than half of all households keep some poultry [1]. Biosecurity and preventive animal health services in these settings are limited, while antibiotics to treat both human and animal infections are widely available without prescription through formal and informal sellers. To ensure quality of feed and veterinary drugs, the Ministry of Agriculture, Fisheries, and Forestry and the General Directorate of Animal Health and Production jointly established regulation in 2018 requiring the registration of imported animal feed and veterinary drugs. However, there are limited data on antibiotic use in the agricultural sector and the pathways through which agricultural communities access antibiotics [2].
While village and animal health agents advise on livestock health for most backyard farming communities in Cambodia, unrestricted access to antibiotics for self-medication is common via pharmacies and other drug outlets [3,4] such as untrained pharmacists, animal feed stores and grocery stores. Further, the interconnectedness of humans and animals in communities with high concentrations of backyard farming can facilitate inappropriate antibiotic use and sharing of antibiotics between humans and animals. Understanding the scale and impact of these practices is essential, to design effective interventions addressing setting-specific inappropriate use of antibiotics [5].
To understand the flow of antibiotics and health-related information in smallholder farming operations, and how this influences human and animal antibiotic use behaviours, we conducted a mixed methods, social network study within two Cambodian farming communities.
## Methods
We recruited participants from two farming villages in Takeo, a primarily rural province in southern Cambodia bordering Vietnam (Fig 1). Each village comprised approximately 100 households. Data were collected in three phases: a baseline household census to enumerate village residents, a social network survey, and in-depth interviews with key informants in the network. Participants were reimbursed with a token of USD3 and USD8 for the household census and in-depth interviews respectively.
**Fig 1:** *Villages 1 and 2 in Takeo, Cambodia.This map was created in R using the package rnaturalearth [6] using maps from Natural Earth (http://www.naturalearthdata.com/) where all maps are public domain.*
## Patient and public involvement
No members of the public were involved in the study. To inform the content of our study tools, we pilot tested the study materials with members of the public external to our research. Key findings will be disseminated to study respondents through the administration team at CelAgrid.
## Household census
We first conducted a photographic census (S1 Appendix) of adults residing in both villages to facilitate nomination of social ties in the subsequent network survey. A community leader accompanied the research team to enumerate occupied households. Households were eligible for inclusion if 1) they owned farm animals, and 2) farming was a main source of household income. In each eligible household, we invited all adults (aged ≥ 21 years) to participate. Individuals who provided verbal consent had their photograph taken using a smartphone. Photographs were immediately printed using portable wireless printers, labelled with a unique participant identifier embedded in a QR code, and placed in a photo album on a page labelled with a unique QR code for the household (Fig 2). Participating household coordinates were also recorded using global positioning system (GPS) receivers on electronic tablets.
**Fig 2:** *Placement of participants’ household and individual unique IDs on a photo album page.Author Clarence Tam is the photographer for this figure.*
## Social network survey
Following the photographic census, we returned to participating households to conduct a social network survey (S2 Appendix). In each household, we interviewed one adult using a standardised questionnaire. We collected information about participants’ demographic and socioeconomic details, animal ownership, knowledge, attitudes and practices surrounding antibiotic use, and awareness of issues relating to antibiotic resistance. We also asked participants if they had leftover antibiotics at home at the time of interview. For those who did, we requested permission to take photographs of the antibiotics for verification.
Participants were then shown the photo albums compiled during the census and asked to nominate up to three key social ties across different domains of their life, including who they spent the most time talking to day-to-day (talking network), who they discussed work-related issues with (work-related network), who they obtained health advice from (health network), and who they consulted for animal health matters.
Participants’ social tie nominations were recorded by scanning the relevant QR codes in the photo album identifying the household and individual. If the participant nominated a village resident not captured in our census, we added a unique QR code to the affiliated household and used that as a proxy for that nomination. The network survey was conducted by trained field staff in Khmer and took approximately 20–30 minutes per household. Data were collected using ODK Collect (v.1.21.0) software [7] on electronic tablets running the Android operating system.
## In-depth interviews
We then selected 28 key informants from participants’ talking networks. These individuals were invited to take part in an in-depth interview to elicit information about the influence of social ties on their decision-making processes related to personal and animal health, and antibiotic use (S3 Appendix). Key informants represented nodes in the network that were well-connected (high in-degree, or receiving nominations from many different households), that served as bridges between different parts of the network (high betweenness) and isolated nodes in the network (low in-degree and low betweenness) (Fig 3). We also selected for in-depth interviews six individuals who were identified as sources of antibiotics in the two villages, including a health centre manager, village health animal worker, pharmacy owner, an animal feed store owner and two grocery store owners.
**Fig 3:** *Selection of key informant sources for in-depth interviews.Nodes represent individual households. Key informants selected for in-depth interviews are shown in red.*
Interviews were conducted in Khmer by trained data collectors (CT, SH) using a semi-structured interview guide designed to elicit information about participants’ antibiotic decision-making in infection prevention or treatment for themselves and their animals. To facilitate discussion about the importance of social relationships, we used a tiered semi-circular mat with coloured sectors of varying radius to represent increasing social distance from the interviewee (Fig 4). Interviewees were then asked to place photographs of individuals they had nominated in the network survey on the mat at different distances relative to themselves, representing how important nominees were as sources of information for their own health and for animal health matters. Interviews lasted 45 minutes on average and were audio-recorded with permission from the interviewee. Audio recordings were transcribed and translated into English verbatim for analysis.
**Fig 4:** *Using an interactive activity, key informants place photographs of their nominated contacts on a mat at varying distances from themselves representing the strength of social connection across talking, work-related and health-related networks.Author Clarence Tam is the photographer for this figure.*
## Data analysis
We first generated directed social networks to visualise the types of connections formed within each village. Next, we derived a score for each participant from ten variables that captured information about specific antibiotic use practices (S1 Table); higher scores indicated more favourable behaviours. Additionally, we assigned each participant a binary data string based on their responses to these ten questions. We then assessed the similarity between individuals’ antibiotic practices by computing the pairwise matrix of Hamming distances for all dyads in the social network. The Hamming distance measured the number of positions at which these binary strings varied between two individuals.
We conducted univariable and multivariable linear (Gaussian) regression analyses to investigate the relationship between participants’ antibiotic scores and their sociodemographic characteristics. We informed these findings with thematic analysis of qualitative data from key informant in-depth interviews, to provide contextual information about individuals’ decision-making processes in relation to animal health and antibiotic use, and the influence of social ties on health behaviours.
We used exponential random graph modelling (ERGM) to evaluate the extent to which network parameters and covariates adequately represented the talking network structures observed in the study. This also allowed us to explore if ties were more likely to exist between certain nodes that shared similar attributes, such as antibiotic scores or low Hamming distances. Univariable analysis of each variable was conducted to select variables for the final ERGMs. Variables were selected based on a p-value cut-off point of 0.25.
Social network data were analysed using the igraph [8] (https://igraph.org) and statnet [9] (http://www.statnetproject.org) packages in R version 3.6.1 [10]. In-depth interview transcripts were imported into Dedoose Version 8.0.35 [11] to facilitate data coding, retrieval and analysis.
## Ethics statement
This study was approved by the institutional review board of the National University of Singapore (reference number: S-18-161) and the National Ethics Committee for Health Research, Cambodia (reference number: 203NECHR). Participants gave verbal consent to participate in each phase of the study, as well as permission for in-depth interviews to be audio-recorded. No participants’ personal identifiers were audio recorded. Participant quotations are depicted by a study identifier to maintain anonymity. We returned the photographs to the individual participants once all data collection was completed.
## Participant and network characteristics
In villages 1 and 2, 143 and 105 participants from 97 and 67 households were included in the photographic census, while 89 and 56 participants completed the network survey respectively. Most participants were female, married or living with a partner, and most commonly owned buffaloes and chickens (Table 1).
**Table 1**
| Participant characteristicsn (%) | Village 1n = 143 | Village 2n = 105 |
| --- | --- | --- |
| Sex | Sex | Sex |
| Male | 48 (33.6) | 31 (29.5) |
| Female | 95 (66.4) | 74 (70.5) |
| Age | Age | Age |
| 21–29 | 18 (12.6) | 10 (9.5) |
| 30–39 | 36 (25.2) | 13 (12.4) |
| 40–49 | 13 (9.1) | 15 (14.3) |
| 50–59 | 37 (25.9) | 18 (17.2) |
| 60–69 | 22 (15.4) | 27 (25.7) |
| 70–79 | 12 (8.4) | 16 (15.2) |
| 80–89 | 5 (3.4) | 6 (5.7) |
| Marital status | Marital status | Marital status |
| Married or living with partner | 76 (53.1) | 78 (74.3) |
| Separated or divorced | 51 (35.7) | 21 (20.0) |
| Never married | 9 (6.3) | 6 (5.7) |
| Widowed | 7 (4.9) | 0 (0) |
| Education | Education | Education |
| No formal education | 31 (21.7) | 31 (29.5) |
| Up to primary education | 39 (27.3) | 35 (33.3) |
| Up to secondary education | 47 (32.8) | 18 (17.2) |
| More than secondary education | 26 (18.2) | 21 (20.0) |
| Animal ownership * | Animal ownership * | Animal ownership * |
| Cow | 80 (82.5) | 43 (64.2) |
| Chicken | 80 (82.5) | 43 (64.2) |
| Dog/Cat | 73 (75.3) | 39 (58.2) |
| Duck | 23 (23.7) | 20 (29.9) |
| Pig | 8 (8.2) | 2 (2.9) |
| Buffalo | 1 (1.0) | 0 (0) |
## Network information
In terms of who they spent the most time talking to on a typical day, participants from the network survey generally nominated family members and contacts from other households who were geographically close to them (Fig 5), most commonly citing convenience. Although individuals generally felt comfortable making casual conversation with others, they would not necessarily trust them with more sensitive information, especially if they were a non-familial tie:
**Fig 5:** *Graphic representations of directed social networks observed in both villages.*
Participants with identical antibiotic response patterns (Hamming distance = 0) had shorter geodesic distances in their talking networks ($r = 0.66$, $p \leq 0.001$) and lived closer to each other ($r = 0.68$, $$p \leq 0.02$$). Participants who lived closer to each other were also more likely to have shorter geodesic distances ($r = 0.74$, $$p \leq 0.02$$), indicating that social connections and antibiotic perceptions and practices in these communities correlate strongly with shorter geographic distance (Fig 6).
**Fig 6:** *Box plots of Hamming, geodesic and geographic distances.*
Work-related matters were also primarily discussed with familial connections or neighbouring households. In choosing whom to discuss work-related issues with, participants valued trustworthiness, availability, experience and convenience. Work-related issues discussed with others typically related to animal farming or agriculture.
In contrast, general health advice was typically sought from personnel at (government-run) health centres and (private sector) pharmacies. Those who turned to familial connections or other villagers for health-related advice tended to do so if their family members had some healthcare training, had experienced a similar illness previously, or had good recommendations for a healthcare provider. Most participants were also willing to share this information with others:
## Antibiotic use behaviours
Key informant interviews identified pharmacies and health centres in nearby villages as the most common sources of antibiotics for human and animal use, as well as grocery stores in the village. Other sources of veterinary antibiotics identified in the survey included the local animal health worker ($50\%$) and the animal feed store ($27.4\%$).
In terms of human health, antibiotics were most commonly used to treat wounds ($54.8\%$ of survey respondents) and fever ($53.4\%$), and less commonly for respiratory symptoms such as coughs ($33.6\%$), sore throats ($38.4\%$) and runny nose ($24.0\%$). This was corroborated by key informant interviews: When deciding whether to use antibiotics for their animals, participants often relied on their own knowledge ($80.8\%$) or information from television, radio and the internet ($41.1\%$). A quarter ($25.9\%$) reported not using antibiotics to treat any animal diseases. Instead, they used alternative methods such as paracetamol or increased biosecurity practices such as improved hygiene.
When purchasing antibiotics, participants typically described their symptoms to the seller, asked for antibiotics to manage symptoms from a self-diagnosis, or asked for specific antibiotics by name. Participants also reporting feigning symptoms in order to obtain antibiotics from human health sources for use in animals: Almost half of the participants ($43.4\%$) had shared antibiotics with family and friends experiencing similar illness, but sharing antibiotics for unrelated symptoms was uncommon ($4.1\%$). About a quarter ($28.3\%$) said that they had shared their own antibiotics with their sick animals but were less likely to use human antibiotics for growth promotion ($2.8\%$) or illness prevention ($11.0\%$). A fifth ($22.1\%$) and a third ($33.1\%$) of participants usually got a prescription before buying antibiotics for themselves and their animals, respectively.
## Attitudes towards antibiotics
About a quarter ($28.3\%$) of the participants in both villages said that they kept a supply of antibiotics at home whether they were sick or not; $18.6\%$ of households surveyed were verified to have antibiotics based on photographs. Common types of leftover antibiotics found in participants’ homes included amoxicillin, ampicillin, cephalosporin, tetracycline and lincomycin.
## Antibiotic scores and regression analyses
Participants’ median antibiotic scores were 7.01 (range: 0–10). In multivariable regression analyses (Table 2) exploring the association between participants’ sociodemographic characteristics and their antibiotic scores, we found that participants in village 2 had slightly lower antibiotic scores (β = -0.77; $95\%$ CI: -1.31 –-0.23). When compared to participants with no formal education, participants with secondary (β = -1.29; $95\%$ CI: -2.24 –-0.33) and more than secondary education (β = -1.48; $95\%$ CI: -2.53 –-0.42) tended to have lower antibiotic scores. Additionally, compared to participants who make their own decisions about animal care, participants whose spouses (β = -0.77; $95\%$ CI: -1.40 –-0.13) and children (β = -1.29; $95\%$ CI: -2.40 –-0.18) made decisions about their household’s animal care had lower antibiotic scores.
**Table 2**
| Unnamed: 0 | Coefficient | 95% CI |
| --- | --- | --- |
| Estimate | 8.47*** | 7.11–10.35 |
| Village | Village | Village |
| Village 1# | | |
| Village 2 | -0.77** | -1.31–-0.23 |
| Sex | Sex | Sex |
| Male# | | |
| Female | 0.45 | -0.17–1.06 |
| Age (years) | -0.01 | -0.03–0.02 |
| Education | Education | Education |
| No education# | | |
| Primary | -0.52 | -1.49–0.45 |
| Secondary | -1.29** | -2.24–-0.33 |
| More than secondary | -1.48** | -2.53–-0.42 |
| Marital status | Marital status | Marital status |
| Married# | | |
| Never married | -0.25 | -1.22–0.71 |
| Separated/Divorced | -0.31 | -1.40–0.77 |
| Widowed | -0.46 | -1.11–0.20 |
| Makes decisions about animals | Makes decisions about animals | Makes decisions about animals |
| Myself# | | |
| Spouse | -0.77* | -1.40–-0.13 |
| Children | -1.29* | -2.40–-0.18 |
| Parent | -0.49 | -1.63–0.66 |
| Other | -0.73 | -2.04–0.57 |
| Get health advice | Get health advice | Get health advice |
| Not health professional# | | |
| Health professional | 0.14 | -0.38–0.67 |
ERGM analysis of the talking network highlighted the importance of reciprocity and closed triads (Table 3), indicating that nominations in the network were likely to be reciprocated, and that participants who shared a social time in common were also more likely to have nominated each other (β: 1.04, $95\%$ CI: 0.51–1.57; β = 0.99, $95\%$ CI:0.64–1.34). In village 1, age homophily was a strong determinant of network structure–the log odds of a tie existing between two nodes were higher if they were closer in age (β: 0.60; $95\%$ CI: 0.15–1.06). In village 2, network structure was strongly dependent on age and sex homophily. In addition, social ties were more likely to exist between individuals who had the same antibiotic scores (β: 0.45; $95\%$ CI:0.02–0.88), or shared similar practices in relation to obtaining antibiotic prescriptions for their animals (β: 0.84; $95\%$ CI:0.29–1.39).
**Table 3**
| Unnamed: 0 | Talking network | Talking network.1 | Talking network.2 | Talking network.3 | Work-related network | Work-related network.1 | Work-related network.2 | Work-related network.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Village 1 | Village 1 | Village 2 | Village 2 | Village 1 | Village 1 | Village 2 | Village 2 |
| | Estimate | 95% CI | Estimate | 95% CI | Estimate | 95% CI | Estimate | 95% CI |
| Network structural measures | Network structural measures | Network structural measures | Network structural measures | Network structural measures | | | | |
| Intercept (Edges) | -5.34*** | -5.89 –-4.79 | -6.00*** | -6.78 –-5.22 | -6.01*** | -6.85 –-5.17 | -4.89*** | -5.50 –-4.28 |
| Reciprocity | 2.00** | 0.80–3.20 | 1.84*** | 1.04–2.64 | 1.87*** | 0.77–2.97 | - | - |
| Triad closure | 1.04*** | 0.51–1.57 | 0.99*** | 0.64–1.34 | 0.86** | 0.33–1.39 | 1.49** | 0.51–2.47 |
| Homophily | Homophily | Homophily | Homophily | Homophily | Homophily | Homophily | Homophily | Homophily |
| Age | 0.60** | 0.15–1.06 | 0.45* | 0.02–0.88 | 0.50* | 0.01–0.99 | 0.29 | -0.33–0.92 |
| Antibiotic scores | 0.26 | -0.21–0.73 | 0.45* | 0.02–0.88 | 0.46 | -0.01–0.93 | - | - |
| Sex | - | - | 0.47* | 0.08–0.86 | 0.26 | -0.23–0.75 | - | - |
| Family | - | - | 0.28 | -0.15–0.71 | 0.47 | -0.02–0.96 | - | - |
| “I have shared antibiotics with my animals” | 0.38 | -0.11–0.87 | 0.47 | -0.06–0.99 | 0.49 | -0.10–1.08 | 0.46 | -0.17–1.09 |
| “I usually get a prescription before obtaining antibiotics for my animals” | 0.43 | -0.08–0.98 | 0.84** | 0.29–1.39 | 0.85** | 0.26–1.44 | 0.30 | -0.25–0.85 |
Results from both villages are presented in coefficients on the log scale and their corresponding $95\%$ confidence intervals. The intercept reflects the log odds of a tie existing between two participants if none of the other variables were included in the model. A negative effect indicates that talking ties were unlikely to be formed outside of the processes included in the model.
In the work-related networks, closed triads were important determinants of network structure in both villages (β: 0.86, $95\%$ CI: 0.33–1.39; β = 1.49, $95\%$ CI:0.51–2.47 respectively), while reciprocity was only observed in village 1 (β: 1.87, $95\%$ CI: 0.77–2.97). Additionally, ties in village 1 were more likely to exist if participants were closer in age (β: 0.50, $95\%$ CI: 0.01–0.99) or if they shared similar practices for obtaining animal antibiotic prescriptions (β: 0.85, $95\%$ CI: 0.26–1.44).
## Discussion
In this mixed methods study, we investigated how relationships and environment can influence antibiotic behaviours in Cambodian rural farming communities. Our findings indicated that participants most commonly accessed antibiotics or learned animal antibiotic practices through village-level informal sources. While most participants reported not using antibiotics for animal growth promotion or illness prevention, common misconceptions surrounding both antibiotic effectiveness and resistance existed. Additionally, participants who lived closer or had shorter geodesic distances in the network were more likely to have similar overall antibiotic knowledge and attitudes, suggesting the importance of familial ties in the community as information tends to cluster within families.
In terms of overall antibiotic behaviours, we found that most of the participants used and accessed antibiotics similarly for themselves and their animals, concomitant with general misconceptions they had about antibiotic effectiveness. For instance, antibiotics were inappropriately used in both humans and animals to treat inflammatory conditions such as wounds, pain and fever, while about a fifth of the participants had leftover antibiotics at home. Approximately a third of the participants also reported having shared antibiotics with their animals, while almost half of the participants said that they shared antibiotics with others with the same illness. The lack of clear guidelines in backyard farming drug use is not novel [12,13], and it continues to add to the selective pressure on bacterial populations in both clinical and agricultural sectors. This also has implications for food safety as well as the emergence and spread of new resistant pathogens from animals [14].
Factors that influenced participants’ decisions to obtain antibiotics for themselves and their animals included perceived quality of medicines, pricing, convenience, recommendations from others as well as trust in the personnel selling antibiotics. Antibiotics were also often acquired without a prescription, commonly from sources with no formal healthcare training, such as untrained pharmacy attendants, village animal health workers, grocery and animal feed store owners. The substantial role of informal sources of healthcare has been mirrored in studies conducted in similar rural settings [4], emphasising the importance of engaging and training this sector to increase appropriate antibiotic practice.
While some participants knew about adverse effects of taking antibiotics inappropriately, a majority were unaware about the mechanisms and consequences of antibiotic resistance. Corresponding with other findings in similar settings [15–17], participants thought that resistance occurs when the body becomes resistant to antibiotics and had generally low levels of awareness about the clinical consequences of resistance. As alternatives to antibiotics for their animals, participants practiced good hygiene and improved farm cleanliness, both of which have potential to reduce inappropriate antibiotic use on farms [16,18]. Vaccines administered by the village animal health worker were also used for specific conditions in chickens and cattle.
Family members or neighbouring households were commonly nominated as important contacts across talking, work-related and health-related networks. We also found that trust in healthcare sources was especially important in who participants nominated for health-related information. Additionally, we found that participants who lived closer or had shorter geodesic distances in the network tended to work more closely together and had more similarity in their patterns of responses to antibiotic knowledge and attitude questions. The link between social connectedness and similarity in behaviours is well-studied in other contexts [19,20] and offers some insight into how antibiotic behaviours can spread through social contacts and influence [21,22]. Our findings indicate that information in these communities tends to travel short distances. This is supported by ERGM analysis, which showed that similarity in antibiotic perceptions and practices (as measured by the Hamming distance), geodesic and geographic distances did not have significant effects on the formation of ties after controlling for general network structures such as tie reciprocity and triad closure. This suggests close-knit communities in which the formation of social ties and dissemination of health information and behaviours is largely shaped by familial relationships and geographic proximity.
Despite the significant economic role that backyard farming has in many developing countries, research on animal antibiotic use has focused on commercial farms. The use of mixed methods in this research enabled greater understanding of antibiotic use in smallholder rural farming communities from varying perspectives and provided insights on shared antibiotic interactions in both humans and animals. A number of limitations should be borne in mind. The communities included in this study were chosen based on their previous involvement with CelAgrid’s agricultural programmes. While this enabled us to establish rapport and trust with participating households, it may limit the generalisability of our results. Additionally, the language barrier between the research team and participants presented some methodological challenges in cross-language qualitative interviews. To mitigate these challenges in ensuring trustworthiness and conceptual equivalence, extensive training was conducted with interpreters and the research team was present for all in-depth interviews. Further, information on antibiotic practices was self-reporting and we were unable to validate this through other means.
The findings from our study have implications for future public health interventions and regulations to address inappropriate antibiotic use. First, recognising that patterns of antibiotic use for human and animal health are similar, the development of multi-dimensional interventions is especially pertinent in communities with high concentrations of backyard farming. Potential points of intervention include the benefits of improved sanitation and waste management, reframing antibiotic effectiveness more specifically for conditions with bacterial causes, as well as the mechanisms and clinical consequences of antibiotic resistance. Additionally, with greater understanding of the substantial roles that informal health sources and health centres take on in rural communities not only for human health but also for animal health, future interventions should engage informal providers in healthcare training and delivery, providing the appropriate resources for adequate antibiotic provision. These interventions should be complemented by information disseminated via media sources, as well as influential nodes in the community that can help to encourage adoption of appropriate antibiotic behaviours through pathways of social influence. Lastly, future interventions in these communities should explore the potential roles of weak ties as well as brokers in health-related information.
## Conclusion
Data from varying perspectives are crucial in designing more effective interventions to reduce inappropriate antibiotic use. This study utilises mixed methods conducted in different stages to explore how relationships and environment can influence antibiotic behaviours in Cambodian rural farming communities. Across talking, working and health networks in both villages, we found that participants tended to nominate familial connections or neighbouring households who were geographically close to them. This also translated to close geodesic and geographical networks having similar overall antibiotic knowledge and attitudes. Results from this study also demonstrate the importance of engaging village-level informal sources in the provision of antibiotic-related information for both human and animal health, as well as in potential interventions to encourage appropriate antibiotic behaviours in lower-resourced settings.
## References
1. 1Statistics NI of. Cambodia Demographic and Health Survey 2014. 2015.. *Cambodia Demographic and Health Survey 2014* (2015.0)
2. **Cambodia is first to organize a national high level tripartite meeting on Multi-Sectoral action plan to combat antimicrobial resistance (AMR).**. (2017.0)
3. Om C, Daily F, Vlieghe E, McLaughlin JC, McLaws ML. **Pervasive antibiotic misuse in the Cambodian community: antibiotic-seeking behaviour with unrestricted access.**. *Antimicrobial Resistance & Infection Control* (2017.0) **6** 1-8. DOI: 10.1186/s13756-017-0187-y
4. Suy S, Rego S, Bory S, Chhorn S, Phou S, Prien C. **Invisible medicine sellers and their use of antibiotics: a qualitative study in Cambodia**. *BMJ global health* (2019.0) **4** e001787. DOI: 10.1136/bmjgh-2019-001787
5. **Ministry of Health, Cambodia;**. (2014.0)
6. 6South A. rnaturalearth: world map data from natural earth. R package version 0.1. 0. The R Foundation https://CRAN/R-project/org/package=rnaturalearth. 2017.
7. Hartung C, Lerer A, Anokwa Y, Tseng C, Brunette W, Borriello G. **Open data kit: tools to build information services for developing regions.**. *In: Proceedings of the 4th ACM/IEEE international conference on information and communication technologies and development* (2010.0) 1-12
8. Csardi MG. **Package ‘igraph.’ Last accessed**. (2013.0) **3** 2013
9. Handcock MS, Hunter DR, Butts CT, Goodreau SM, Morris M. **statnet: Software tools for the representation, visualization, analysis and simulation of network data.**. *Journal of statistical software.* (2008.0) **24** 1548. DOI: 10.18637/jss.v024.i01
10. 10RStudio Team. RStudio: Integrated Development for R [Internet]. Boston, MA: RStudio, Inc.;
2015. Available from: http://www.rstudio.com/.. *RStudio: Integrated Development for R* (2015.0)
11. Version D.. *8.0. 35. Web application for managing, analyzing, and presenting qualitative and mixed method research data Los Angeles, CA: SocioCultural Research Consultants, LLC 2018* (2018.0)
12. Dyar OJ, Yin J, Ding L, Wikander K, Zhang T, Sun C. **Antibiotic use in people and pigs: a One Health survey of rural residents’ knowledge, attitudes and practices in Shandong province, China**. *Journal of Antimicrobial Chemotherapy* (2018.0) **73** 2893-9. DOI: 10.1093/jac/dky240
13. Osbjer K, Boqvist S, Sokerya S, Kannarath C, San S, Davun H. **Household practices related to disease transmission between animals and humans in rural Cambodia.**. *BMC Public Health* (2015.0) **15** 476. DOI: 10.1186/s12889-015-1811-5
14. Chang Q, Wang W, Regev‐Yochay G, Lipsitch M, Hanage WP. **Antibiotics in agriculture and the risk to human health: how worried should we be?**. *Evolutionary applications.* (2015.0) **8** 240-7. DOI: 10.1111/eva.12185
15. Resistance WA. *Multi-country public awareness survey* (2015.0) 59
16. Xu J, Sangthong R, McNeil E, Tang R, Chongsuvivatwong V. **Antibiotic use in chicken farms in northwestern China.**. *Antimicrobial Resistance & Infection Control.* (2020.0) **9** 1-9. DOI: 10.1186/s13756-019-0672-6
17. Waseem H, Ali J, Sarwar F, Khan A, Rehman HSU, Choudri M. **Assessment of knowledge and attitude trends towards antimicrobial resistance (AMR) among the community members, pharmacists/pharmacy owners and physicians in district Sialkot, Pakistan.**. *Antimicrobial Resistance & Infection Control.* (2019.0) **8** 1-7. DOI: 10.1186/s13756-019-0517-3
18. Eltayb A, Barakat S, Marrone G, Shaddad S, aalsby Lundborg C. **Antibiotic use and resistance in animal farming: a quantitative and qualitative study on knowledge and practices among farmers in Khartoum, Sudan.**. *Zoonoses and public health.* (2012.0) **59** 330-8. DOI: 10.1111/j.1863-2378.2012.01458.x
19. Lee DH, Brusilovsky P. **Social networks and interest similarity: the case of CiteULike.**. *In: Proceedings of the 21st ACM conference on Hypertext and hypermedia.* (2010.0) 151-6
20. Fan C, Liu Y, Huang J, Rong Z, Zhou T. **Correlation between social proximity and mobility similarity.**. *Scientific reports.* (2017.0) **7** 1-8. PMID: 28127051
21. Christakis NA, Fowler JH. **The spread of obesity in a large social network over 32 years**. *New England journal of medicine* (2007.0) **357** 370-9. DOI: 10.1056/NEJMsa066082
22. Poirier J, Cobb NK. **Social influence as a driver of engagement in a web-based health intervention**. *Journal of medical Internet research* (2012.0) **14** e36. DOI: 10.2196/jmir.1957
|
---
title: Associations of adverse maternal experiences and diabetes on postnatal maternal
depression and child social-emotional outcomes in a South African community cohort
authors:
- Yael K. Rayport
- Ayesha Sania
- Maristella Lucchini
- Carlie Du Plessis
- Mandy Potter
- Priscilla E. Springer
- Lissete A. Gimenez
- Hein J. Odendaal
- William P. Fifer
- Lauren C. Shuffrey
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021654
doi: 10.1371/journal.pgph.0001124
license: CC BY 4.0
---
# Associations of adverse maternal experiences and diabetes on postnatal maternal depression and child social-emotional outcomes in a South African community cohort
## Abstract
Previous literature has identified associations between diabetes during pregnancy and postnatal maternal depression. Both maternal conditions are associated with adverse consequences on childhood development. Despite an especially high prevalence of diabetes during pregnancy and maternal postnatal depression in low- and middle-income countries, related research predominates in high-income countries. In a South African cohort with or without diabetes, we investigated associations between adverse maternal experiences with postnatal maternal depression and child social-emotional outcomes. South African mother-child dyads were recruited from the Bishop Lavis community in Cape Town. Participants consisted of 82 mother-child dyads (53 women had GDM or type 2 diabetes). At 14–20 months postpartum, maternal self-report questionnaires were administered to assess household socioeconomic status, food insecurity, maternal depressive symptoms (Edinburgh Postnatal Depression Scale (EPDS)), maternal trauma (Life Events Checklist), and child social-emotional development (Brief Infant Toddler Social Emotional Assessment, Ages and Stages Questionnaires: Social-Emotional, Second Edition). Lower educational attainment, lower household income, food insecurity, living without a partner, and having experienced physical assault were each associated with postnatal maternal depressive symptoms and clinical maternal depression (EPDS ≥ 13). Maternal postnatal depression, lower maternal educational attainment, lower household income, household food insecurity, and living in a single-parent household were each associated with child social-emotional problems. Stratified analyses revealed maternal experiences (education, income, food insecurity, trauma) were associated with postnatal maternal depressive symptoms and child social-emotional problems only among dyads with in utero exposure to diabetes. Women with pre-existing diabetes or gestational diabetes in LMIC settings should be screened for health related social needs to reduce the prevalence of depression and to promote child social-emotional development.
## 1. Introduction
Metabolic disorders, such as gestational diabetes mellitus (GDM), have been linked to postnatal maternal depression in low- and middle-income countries (LMICs) [1, 2]. Research on postnatal maternal depression predominates in high-income-countries (HICs) despite depression having an especially high prevalence in LMICs. Rates of maternal depression in South Africa are as high as $35\%$ [3]. Prior research has demonstrated that in utero exposure to diabetes has long-term consequences on child emotional, behavioral, cognitive, and metabolic outcomes [4–6]. Children born to mothers with diabetes are at increased risk for specific neurodevelopmental sequelae including impaired memory and attention [7], as well as risk for autism spectrum disorder [8]. Postnatal maternal depression is independently associated with adverse childhood social-emotional developmental outcomes [9, 10], suggesting that comorbid GDM and postnatal depression could have a compounding influence on child development.
Structural social determinants of health including socioeconomic inequities and trauma have also been implicated in prenatal maternal depression. A study examining South African pregnant women from a lower socioeconomic status (SES) community found that they were more likely to be depressed if they were food insecure [11], suggesting an additive effect of multiple structural risk factors on mental health. This study also reported that women were more likely to be food insecure if they experienced suicidal ideation, had more than three children, less education, lower income, history of mental illness, experienced a threatening life event, or had been exposed to intimate partner violence [11], suggesting an interplay between household food insecurity and other maternal experiences. In a rural population from Southwestern Ethiopia, pregnant women who are food insecure were four times more likely to experience psychological distress, including depression, anxiety, or suicidality [12]. In developing countries, structural and socioeconomic inequities have linked to the risk of developing psychiatric disorders [13, 14]. Traumatic life events, such as intimate partner violence, are also associated with maternal depression. For example, in Khayelitsha, a township in the Western Cape Province of South Africa, the prevalence of depression ranged from 32–$47\%$ antenatally and 16–$32\%$ postnatally with intimate partner violence being a significant predictor of depression [15].
Structural social determinants of health including socioeconomic inequities and trauma have also been implicated in adverse childhood outcomes. Preschool aged children in food insecure households have lower literacy, numeracy, short-term memory, and self-regulation [16]. Previous research has also established associations between maternal trauma and childhood social-emotional development. For example, in a South African birth cohort study, maternal posttraumatic stress disorder was associated with poorer infant fine motor and adaptive motor development at six months of age [17].
To our knowledge, no studies have examined the relationships among structural and socioeconomic determinants, diabetes during pregnancy, postnatal maternal depression, and child social-emotional development in a low resource setting [18]. In a high-risk cohort with or without diabetes, we hypothesized that specific structural and socioeconomic determinants experienced by mothers—food insecurity and trauma—would increase a mother’s risk of developing postnatal depression and in turn, these variables would be associated with delays in social-emotional development in their children. We hypothesized maternal diabetes would moderate these associations. Specifically, we hypothesized a stronger association between trauma and food insecurity with postnatal depression in mothers who have diabetes compared to women who do not.
## 2.1. Ethics statement
All procedures complied with the ethical standards of the Institutional Review Boards and ethics review committees at Stellenbosch University (N16-08-101 and N06-10-210) and the New York State Psychiatric Institute [5338]. All participants provided informed written consent at both time points (prenatal and postnatal) before inclusion in the study. The cohort was intended to continue recruiting, however recruitment stopped due to the coronavirus pandemic.
## 2.2. Population
Between April 2018 and March of 2020, participants were recruited from the Bishop Lavis community at their Health Center or the Diabetes Clinic in Tygerberg Academic Hospital in the Western Cape Province, South Africa. This study was built on existing collaborations with the university through the Safe Passage Study [19]. Results from the Safe Passage study revealed high rates of depression and anxiety within the population: $55\%$ of mothers had a prenatal Edinburgh postnatal depression score (EPDS) ≥ 13 and $19.17\%$ of mothers had a State-Trait Anxiety Inventory (STAI) ≥ 40 [19].
In the present study, women were recruited at their first antenatal care visit. Inclusion criteria for mothers included: able to communicate fluently in Afrikaans or English, able to provide informed consent, planned to deliver at Tygerberg Hospital, pregnant with one fetus, maternal age was ≥ 18 years, gestational age at study entry 6–24 weeks, and no major congenital abnormalities in the fetus on the second trimester ultrasound scan. Participants were excluded for the following: planned abortion, planned relocation from the area prior to delivery, advice against participation from a health care provider, or skin lesion on the anterior abdominal wall that could hamper with electrode placement (for a separate study procedure).
A total of 108 pregnant women were consented to participate in the study. Of those, there was 1 intrauterine death case, 4 infant demises, 7 study withdrawals, and 7 participants were lost of follow up. Of the remaining 89 participants, 7 of the women had type I diabetes and were excluded from the present analysis. The final sample consisted of 82 infant-mother dyads: $$n = 53$$ women with GDM or type 2 diabetes and $$n = 29$$ women without diabetes. Diabetic status was abstracted from maternal medical charts. Data from this study is available in the Dryad Digital Repository: https://doi.org/10.5061/dryad.kkwh70s73 [20].
## 2.3. Structural and socioeconomic risk factors
Maternal sociodemographic and socioeconomic information was acquired through study-specific maternal self-report surveys when the children were 14–21 months of age (16 ± 1.7 months). Sociodemographic factors included education (completion of primary school or less, some high school, to high school and beyond), and partner status (married or cohabitating with a partner versus not). Socioeconomic variables included household income reported in South African Rand (R) (0–900 R, 901–5000 R, 5001–10,000 R, to more than 10,000 R per month) and food insecurity (in the last 4 months, participants have 1) enough of the kind of food they want, or 2) enough, but not always the kind of food they want, or 3) sometimes or often not enough to eat). Information on maternal history of trauma was acquired through the Life Events Checklist (LEC), a 15-item self-report measure which gathers information on whether a person has experienced potentially traumatic events [21]. The LEC has previously been used to assess PTSD in a perinatal population in South Africa [17]. Variables of trauma analyzed included whether the participant experienced physical assault (i.e., being attacked, hit, slapped, kicked, or beaten up) by selecting “happened to me” on the checklist.
## 2.4. Self-reported maternal depressive symptoms
Mental Health Questionnaires were administered along with the maternal self-report surveys. Depressive symptoms were measured with the Edinburgh Postnatal Depression Scale (EPDS), a 10-item screening tool assessing depressive symptoms in perinatal women with a higher score indicating more depressive symptoms. The EPDS has been validated as a screening instrument in South Africa in English and Afrikaans [22]. Previous reports have used a cut off score of ≥ 13 to indicate probable depression in perinatal South African women [23].
## 2.5. Toddler social-emotional assessments
Developmental assessments were conducted from 14–21 months of age (16 ± 1.7 months). Adjusted follow-up age ranged from 14–20 months of age (16 ± 1.5 months). The Brief Infant Toddler Social Emotional Assessment (BITSEA), a 42-item parental report, was used to screen for child social-emotional behavioral problems and delays in social-emotional competence [24]. The BITSEA has been validated as a screener in socioeconomically and ethnically diverse cohorts, as well as in children with autism spectrum disorder [25]. The Ages & Stages Questionnaires: Social-Emotional, Second Edition (ASQ:SE:2), a parent-completed assessment, aimed to assess child social-emotional competence across a variety of situations [26]. The ASQ is widely used in LMICs and has been validated as a screening instrument in South Africa in English and Afrikaans [27].
## 2.6. Statistical analysis
All statistical analyses were performed using R version 4.0. Primary analyses consisted of a series of minimally and fully adjusted linear regression models to examine the main effects of maternal diabetic status and structural and socioeconomic risk factors on postnatal maternal depressive symptoms. Minimally adjusted models consisted of several univariate regression analyses to examine the association of maternal diabetic status during pregnancy, educational attainment, income, partner status, food insecurity, and physical assault with postnatal maternal depressive symptoms while controlling for maternal age. Fully adjusted models consisted of multivariate linear regression models that examined the association of maternal diabetic status during pregnancy, educational attainment, income, partner status, food insecurity, and physical assault with postnatal maternal depressive symptoms while controlling for maternal age in a single model. However, due to the association between household income and physical assault, these variables were estimated in separate models. For all models, we report standardized regression coefficients (β) and the $95\%$ confidence intervals (CI) for each main effect. Secondary analyses consisted of logistic regression analyses to estimate the main effects of each risk factor with clinically relevant maternal depression, defined as having an EPDS score ≥13. We report unadjusted and adjusted odds ratios (ORs) and CI of the ORs. Linear regressions were used to estimate β and CI for correlates of the BITSEA Problem domain score, BITSEA Competence domain score, and ASQ:SE:2 score. Minimally and fully adjusted models were run as stated above but adjusted for child sex and adjusted age at assessment instead of maternal age. Tertiary analysis involved running all the previously stated regression models stratified by diabetic status.
## 3.1. Cohort characteristics
The sample included 82 mother-infant dyads. The average age of the women was 32 ± 5.1 years. Based on an EPDS cutoff of ≥ 13, $$n = 30$$ women met criteria for clinical levels of depression ($36.58\%$). $$n = 53$$ women had GDM or type 2 Diabetes ($65\%$). Most of the women had some high school education ($54\%$). The greatest percentage of women had a monthly household income of 901–5000 R ($45\%$). The predominant household food insecurity status was having enough, but not always the kind of food they wanted to eat ($40\%$). Most of the women had a spouse or partner ($63\%$). Many of the women had experienced physical assault ($45\%$). Just over half of the children were female ($57\%$). The average gestational age at birth was 38 ± 2.0 weeks. The average child age at the follow-up assessment was 16 ± 1.7 months and the average adjusted age was 16 ± 1.5 months (Table 1). Table 2 provides an overview of the demographic characteristics by clinical depression diagnosis (Table 2).
## 3.2. Association of structural, socioeconomic, and maternal health risk factors with postnatal maternal depressive symptoms
In the minimally adjusted model, an educational attainment of primary school or less (β = 5.275, $95\%$ CI = 1.010 to 9.540), lower levels of income (0–900 R: β = 9.051, $95\%$ CI = 4.497 to 13.605; 901–5000 R: β = 4.713, $95\%$ CI = 1.180 to 8.246), living without a spouse or partner (β = 4.385, $95\%$ CI = 1.674 to 7.096), higher levels of household food insecurity (enough, but not always the kind of food they want to eat: β = 3.133, $95\%$ CI = 0.529 to 5.736); sometimes or often not enough to eat: β = 6.123, $95\%$ CI = 3.296 to 8.950), and having experienced physical assault (β = 3.777, $95\%$ CI = 1.253 to 6.300) were each significantly associated with increased maternal depressive symptoms on the EPDS. In the minimally adjusted model, maternal diabetes during pregnancy was not significantly associated with increased maternal depressive symptoms (p-values > 0.05) (Fig 1).
**Fig 1:** *Betas and 95% CI for correlates of depressive symptoms (left) and odds ratios and 95% CI for correlates of clinical depression (EPDS > = 13) (right). Minimally adjusted model was adjusted for maternal age. Fully adjusted model was adjusted for maternal age as well as the other maternal experiences. Maternal physical assault and income were analyzed in separate models.*
In the fully adjusted model, an educational attainment of primary school or less (β = 4.015, $95\%$ CI = 0.048 to 7.981), living without a spouse or partner (β = 3.377, $95\%$ CI = 0.670 to 6.084), higher levels of household food insecurity (enough, but not always the kind of food they want to eat: β = 3.017, $95\%$ CI = 0.544 to 5.489); sometimes or often not enough to eat: β = 4.376, $95\%$ CI = 1.464 to 7.288), and having experienced physical assault (β = 2.549, $95\%$ CI = 0.306 to 4.791) were all significantly associated with increased maternal depressive symptoms on the EPDS. In the fully adjusted model, maternal diabetes during pregnancy and income were not significantly associated with increased maternal depressive symptoms (p-values > 0.05) (Fig 1).
## 3.3. Association of structural, socioeconomic, and maternal health risk factors with postnatal maternal depressive symptoms with clinical maternal depression
In the minimally adjusted model, an educational attainment of primary school or less (OR = 12.773, $95\%$ CI = 2.225 to 108.055), an income level of 0–900 R (OR = 13.949, $95\%$ CI = 1.833 to 171.907), living without a spouse or partner (OR = 3.408, $95\%$ CI = 1.174 to 10.427), the greatest degree of household food insecurity (sometimes or often not enough to eat: OR = 4.358, $95\%$ CI = 1.320 to 15.869), and having experienced physical assault (OR = 3.684, $95\%$ CI = 1.317 to 11.143) were each significantly associated with clinical levels of maternal depression. In the minimally adjusted model, maternal diabetes during pregnancy was not significantly associated with clinical levels of maternal depression (p-values > 0.05) (Fig 1).
In the fully adjusted model, lower levels of education (primary school or less: OR = 22.844, $95\%$ CI = 2.128 to 387.973; some high school: OR = 6.340, $95\%$ CI = 1.205 to 45.259), living without a spouse or partner (OR = 12.399, $95\%$ CI = 2.218 to 103.842), and the greatest degree of household food insecurity (sometimes or often not enough to eat: OR = 6.936, $95\%$ CI = 1.107 to 63.898) were all significantly associated with clinical levels of maternal depression. In the fully adjusted model, maternal diabetes during pregnancy, income, and experiencing physical assault were not significantly associated with clinical levels of maternal depression (p-values > 0.05) (Fig 1).
## 3.4. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional problems on the BITSEA
In the minimally adjusted model, clinical maternal depression (β = 3.043, $95\%$ CI = 0.279 to 5.808) and a maternal educational attainment of primary school or less (β = 5.497, $95\%$ CI = 0.737 to 10.256) were each associated with a higher BITSEA Problem Scores. Maternal diabetes during pregnancy, living without a spouse or partner, household food insecurity, and maternal physical assault were not associated with higher BITSEA Problem Scores in the minimally adjusted model (p-values > 0.05). In the fully adjusted model, household food insecurity (having enough, but not always the kind of food they want to eat) was also associated with increased BITSEA Problem Scores (β = 6.162, $95\%$ CI = 0.849 to 11.475). Clinical depression, maternal diabetes during pregnancy, educational attainment, income, living without a spouse or partner, and maternal physical assault were not associated with higher BITSEA Problem Scores in the fully adjusted model (p-values > 0.05) (Fig 2).
**Fig 2:** *Betas and 95% CI for correlates of BITSEA Problem Score (left), BITSEA Competence Score (middle) and ASQ:SE:2 Scored (right). Minimally adjusted model was adjusted for maternal age. Fully adjusted model was adjusted for maternal age as well as the other maternal experiences. Maternal physical assault and income were analyzed in separate models.*
## 3.5. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional competence on the BITSEA
In the minimally adjusted model, an education level of primary school or less (β = -2.208, $95\%$ CI = -4.353 to -0.064) and lower levels of income (901–5000 R) (β = –2.312, $95\%$ CI = -4.418 to -0.205) were each associated with decreased BITSEA Competence Scores. In the minimally adjusted model, maternal clinical depression, diabetes during pregnancy, household food insecurity, and maternal physical assault were not associated with decreased BITSEA Competence Scores (p-values > 0.05) (Fig 2).
In the fully adjusted model, income levels of 0–900 R (β = -3.163, $95\%$ CI = -6.207 to -0.119) and 901–5000 R per month (β = -2.953, $95\%$ CI = -5.317 to -0.590) were associated with were associated with decreased BITSEA Competence Scores. Maternal clinical depression, diabetes during pregnancy, living without a spouse or partner, household food insecurity, and maternal physical assault were not associated with decreased BITSEA Competence Scores ($p \leq 0.05$). In the fully adjusted model, maternal clinical depression, diabetes during pregnancy, educational attainment, household food insecurity, and maternal physical assault were not associated with decreased BITSEA Competence Scores ($p \leq 0.05$) (Fig 2).
## 3.6. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional development on the ASQ:SE:2
In the minimally adjusted model, lower levels of income (0–900 R) (β = 27.169, $95\%$ CI = 4.839 to 49.499) and 901–5000 R (β = 25.500, $95\%$ CI = 7.338 to 43.663) were associated with a higher ASQ:SE:2 scores, which are indicative of poorer social-emotional development. In the minimally adjusted model, maternal clinical depression, diabetes during pregnancy, educational attainment, living without a spouse or partner, household food insecurity, and maternal physical assault were not associated with higher ASQ:SE:2 scores ($p \leq 0.05$). In the fully adjusted model, living without a spouse or partner (β = 27.816, $95\%$ CI = 0.898 to 54.734) and household food insecurity were associated with higher ASQ:SE:2 scores (enough, but not always the kind of food we want to eat: β = 25.638, $95\%$ CI = 4.742 to 46.535; sometimes or often not enough to eat: β = 21.999, $95\%$ CI = 1.594 to 42.403). In the fully adjusted model, maternal clinical depression, diabetes during pregnancy, educational attainment, income, and maternal physical assault were not associated with higher ASQ:SE:2 scores ($p \leq 0.05$) (Fig 2).
## 3.7. Association of structural, socioeconomic, and maternal health risk factors with postnatal maternal depressive symptoms stratified by maternal diabetes status
In the minimally adjusted model, among women with diabetes during pregnancy, an educational attainment of primary school or less (β = 5.916, $95\%$ CI = 1.488 to 10.343), lower levels of income (0–900 R: (β = 12.777, $95\%$ CI = 7.142 to 18.413; and 901–5000 R β = 5.503, $95\%$ CI = 1.910 to 9.097), household food insecurity (sometimes or often not having enough food to eat: β = 6.482, $95\%$ CI = 3.161 to 9.804), and having experienced physical assault (β = 4.294, $95\%$ CI = 1.103 to 7.485) were each associated with increased postnatal maternal depressive symptoms on the EPDS. Living without a spouse or partner was not associated with increased maternal depressive symptoms among women with diabetes in the minimally adjusted model ($p \leq 0.05$) (Fig 3).
**Fig 3:** *Stratified by diabetes betas and 95% CI for correlates of depression.Minimally adjusted model was adjusted for maternal age. Fully adjusted model was adjusted for maternal age as well as the other maternal experiences. Maternal physical assault and income were analyzed in separate models.*
In the fully adjusted model, among women with diabetes during pregnancy, an educational attainment of primary school or less (β = 4.599, $95\%$ CI = 0.420 to 8.778), the lowest level of income (0–900 R: β = 6.827, $95\%$ CI = 0.366 to 13.287), household food insecurity (Having enough food but not always the kind of food they want to eat: β = 3.355, $95\%$ CI = 0.297 to 6.413; and Sometimes or often not having enough to eat: β = 5.060, $95\%$ CI = 1.691 to 8.430), and having experienced physical assault (β = 3.712, $95\%$ CI = 1.082 to 6.342) were associated with increased postnatal maternal depressive symptoms on the EPDS. In the fully adjusted model among diabetic women, living without a spouse or partner was not significantly associated with increased maternal depressive symptoms ($p \leq 0.05$) (Fig 3).
In the minimally adjusted model, among women without diabetes during pregnancy, living without a spouse or partner (β = 6.281, $95\%$ CI = 2.440 to 10.121) was associated with increased postnatal maternal depressive symptoms. Among, non-diabetic women, maternal educational attainment, income, household food insecurity, and experiencing physical assault were not significantly associated with increased postnatal maternal depressive symptoms in the minimally adjusted model. In the fully adjusted model, among women without diabetes during pregnancy, none of the maternal experiences were significantly associated with postnatal maternal depressive symptoms ($p \leq 0.05$) (Fig 3).
## 3.8. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional problems on the BITSEA stratified by maternal diabetic status
In the minimally adjusted model, among children with in utero exposure to diabetes, maternal educational attainment of primary school or less (β = 5.525, $95\%$ CI = 0.035 to 11.016) and lower levels of household income (901–5000 R: β = 7.361, $95\%$ CI = 1.933 to 12.788) were each associated with significantly higher BITSEA Problem Scores. Maternal clinical depression, living in a single-parent household, household food insecurity, and maternal trauma were not associated with significantly higher BITSEA Problem Scores (p-values > 0.05). In the fully adjusted model, household food insecurity (having enough but not always the kind of food they want to eat: β = 9.961 $95\%$ CI = 2.816 to 17.106; and sometimes or often not having enough to eat: β = 6.498, $95\%$ CI = 0.071 to 12.925) was associated with higher BITSEA Problem Scores. Maternal clinical depression, maternal education, household income, living in a single-parent household, and maternal trauma were not associated with significantly higher BITSEA Problem Scores (p-values > 0.05) (Fig 4).
**Fig 4:** *Stratified by diabetes betas and 95% CI for correlates of BITSEA Problem (left) and Competence Score (right). Minimally adjusted model was adjusted for maternal age. Fully adjusted model was adjusted for maternal age as well as the other maternal experiences. Maternal physical assault and income were analyzed in separate models.*
In the minimally adjusted model, among children without in utero exposure to diabetes, there were no significant associations between maternal experiences and BITSEA problem scores. However, in the fully adjusted model, among children without in utero exposure to diabetes, clinical depression was a significant predictor of higher BITSEA Problem Scores (β = 7.274, $95\%$ CI = 0.218 to 14.330). Maternal education, household income, living in a single-parent household, household food insecurity, and maternal trauma were not associated with higher BITSEA Problem Scores (p-values > 0.05) (Fig 4).
## 3.9. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional competence on the BITSEA stratified by maternal diabetic status
In the minimally and fully adjusted model, among children with in utero exposure to diabetes, maternal educational attainment of primary school or less was associated with lower BITSEA Competence Scores (minimally adjusted: β = -2.502, $95\%$ CI = -4.808 to -0.197; fully adjusted: β = -2.604, $95\%$ CI = -5.194 to -0.013). In the minimally adjusted model only, a household income of 901–5000 R was associated with lower BITSEA Competence Scores (β = -2.658, $95\%$ CI = -4.923 to -0.394). In the fully adjusted model only, household food insecurity (having enough but not always the kind of food they want to eat: β = -3.231, $95\%$ CI = -6.199 to -0.263) was associated with lower BITSEA Competence Scores. In the minimally and fully adjusted models, maternal clinical depression, a single-parent household, and maternal trauma were not significantly associated with BITSEA Competence Scores (p-values > 0.05). Household food insecurity in the minimally adjusted model and household income in the fully adjusted model were not significantly associated with BITSEA Competence Scores ($p \leq 0.05$) (Fig 4).
Among children without in utero exposure to diabetes, none of the maternal experiences were associated with lower BITSEA Competence Scores in either the minimally or fully adjusted models (p-values > 0.05) (Fig 4).
## 3.10. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional development on the ASQ:SE:2 stratified by diabetes
In the minimally adjusted model, among children with in utero exposure to diabetes, lower household income (901–5000 R) was associated with significantly higher ASQ:SE:2 scores (β = 27.974, $95\%$ CI = 7.218 to 48.730). In the minimally adjusted model, maternal clinical depression, maternal educational attainment, living in a single-parent household, household food insecurity, and maternal trauma were not associated with ASQ:SE:2 scores (p-values > 0.05). In the fully adjusted model, household food insecurity (having enough, but not always the kind of food they want to eat) was associated with higher ASQ:SE:2 scores (β = 35.677, $95\%$ CI = 7.789 to 63.564). Maternal clinical depression, maternal educational attainment, household income, living in a single-parent household, and maternal trauma were not associated with ASQ:SE:2 scores ($p \leq 0.05$) (Fig 5).
**Fig 5:** *Stratified by diabetes betas and 95% CI for ASQ:SE:2 score.Minimally adjusted model was adjusted for maternal age. Fully adjusted model was adjusted for maternal age as well as the other maternal experiences. Maternal physical assault and income were analyzed in separate models.*
Among children without in utero exposure to diabetes, none of the maternal experiences were associated with ASQ:SE:2 scores in either the minimally or fully adjusted models ($p \leq 0.05$) (Fig 5).
## 4.1. Summary
This is the first study to evaluate specific structural and socioeconomic determinants experienced by mothers–educational attainment, household income, partner status, household food insecurity, and history of trauma–in association with postnatal maternal depression in a LMIC cohort of women with or without diabetes during pregnancy. It is also the first to our knowledge to study social-emotional developmental outcomes of children born to these women. This cohort had a high prevalence of clinical depression at $36\%$ (EPDS score ≥ 13). Many of the adverse maternal experiences were associated with increased postnatal maternal depressive symptoms and clinical levels of maternal depression based on dichotomized EPDS scores. Stratified analyses revealed maternal diabetic status significantly moderated these associations. Significant associations between specific structural and socioeconomic determinants experienced by mothers and postnatal maternal depression were only observed in women with diabetes during pregnancy. When examining associations among specific structural and socioeconomic determinants experienced by mothers and maternal diabetic status with childhood social-emotional outcomes, we demonstrated that many of the structural and socioeconomic determinants were significantly associated with worse child social-emotional outcomes among children with in utero diabetes exposure, and less so among those without in utero diabetes exposure.
## 4.2. Association of structural, socioeconomic, and maternal health risk factors with postnatal maternal depression
Predictors of maternal depressive symptoms and clinical depression included lower levels of education, lower levels of income, living without a partner or spouse, higher levels of household food insecurity, and having experienced physical assault. Prior studies exploring perinatal depression in South African women have also found that specific structural and socioeconomic determinants including education, income, marital status, food insecurity, stressful life events in the last 12 months, and intimate partner violence are predictive of perinatal depression [28–30]. Other studies have shown the interconnectedness of structural and socioeconomic determinants experienced by mothers. For example, correlates of intimate partner violence experienced by mothers with depressive symptoms were associated with emotional distress and food insecurity [31]. Thus, structural and socioeconomic experienced by mothers may have an interactive effect on adverse maternal mental health outcomes.
In this study, we did not observe an association between diabetes during pregnancy and postnatal maternal depression. However, prior studies have demonstrated diabetes during pregnancy is a risk factor for postpartum depression [1, 2]. Possible explanations for our lack of replication include the small sample size of the cohort or heterogeneity of diabetes with both GDM and type 2 diabetes in the diabetes group. A systematic review examining diabetes during pregnancy and maternal depression pre-pregnancy, during pregnancy, or postpartum found different rates of depression depending on the diabetes type. The prevalence of depression among women with GDM was reported to range from $4.1\%$ to $80\%$ with a median of $14.7\%$ whereas the prevalence of depression among women with pre-existing diabetes type 1 and type 2 diabetes) ranged from $0\%$ to $60\%$ with a median of $8.3\%$ [32]. Future studies should examine the effects of diabetes on depression in a larger cohort with representation of each diabetes type in a LMIC setting.
## 4.3. Association of structural, socioeconomic, and maternal health risk factors with postnatal maternal depression stratified by maternal diabetes status
In this study, among women with diabetes, lower levels of education and income, household food insecurity, and experiencing physical assault were associated with increased postnatal maternal depressive symptoms and clinical depression. For non-diabetic women this relationship was not observed, as only living without a spouse or partner was associated with maternal postnatal depression. Given that most of the maternal experiences were associated with depression in the diabetic group only, diabetes could serve as a potential moderator between maternal experiences and depression. This finding suggests diabetes has a compounding effect on depression when comorbid with other adverse maternal experiences. In a larger cohort, we would conduct an interaction analysis to further support diabetes’ role as a moderator.
## 4.4. Association of structural, socioeconomic, and maternal physical and mental health risk factors with child social-emotional development
This study also found that predictors of increased child social-emotional problems included clinical maternal depression, lower levels of maternal education, lower household income, and household food insecurity. Correlates of decreased behavioral competencies included lower levels of maternal education and lower household income. Correlates of worse social-emotional development included lower household income, household food insecurity, and living in a single-parent household.
Other studies have also demonstrated structural and socioeconomic factors as strong predictors of child outcomes. SES has been found to be most strongly associated with child cognitive development at 5 years of age [33], supporting this study’s findings that education, income, and household food insecurity play the most influential role in child outcomes. However, we only observed an association between postnatal maternal depression and child social-emotional problems but not with social-emotional competencies. Similarly, the effects of depression on child development are mixed: one study has found that children of postnatally depressed mothers tend to be more aggressive, are hospitalized more, but have similar cognitive and social development to children born to non-depressed mothers [34]. A study examining the development of children at 2 years of age found that infant outcomes were similar despite postnatal maternal depression [28].
Among children with in utero exposure to diabetes, lower levels of maternal education, lower household income, and household food insecurity were associated with higher social-emotional problems and lower social-emotional competencies; lower household income and household food insecurity were associated with worse social-emotional development. Among children without in utero exposure to diabetes, maternal depression was correlated with higher social-emotional problems.
## 4.5. Strengths and limitations
Strengths of this study include the range of maternal structural and socioeconomic determinants analyzed, the heterogeneity in maternal postnatal depression status, and the two assessments of child social-emotional development (the BITSEA and ASQ:SE:2). The primary limitations of this study include a lack of information regarding prenatal maternal mood disorders during pregnancy and the small sample size. The small sample size limited our ability to run a more sophisticated moderation analysis for diabetic status. Additionally, we were unable to group by diabetes type (GDM vs type 2) restricting our ability to differentiate the effects of each type of diabetes on postnatal depression. In the stratified analysis, the binary variable of clinical depression was not used because of the limited power in the small sample. Other limitations include the use of maternal report measures to examine child social-emotional development. Another limitation is that participants were recruited only from two clinics in the urban South African community. Therefore, the results may not be as generalizable to the population beyond pregnant women living in urban South Africa. Nevertheless, these results are an important contribution to the literature since few studies have investigated these associations in LMIC settings.
## 4.6. Conclusion and implications
Overall, the results of this study support our hypothesis that structural and socioeconomic determinants experienced by mothers, such as lower household income, household food insecurity, and experiencing trauma increase a woman’s risk of developing postnatal depression and this relationship is influenced by maternal diabetes during pregnancy. SES factors, such as income, education, and food insecurity, as well as maternal depression also influence child emotional and behavioral development. This study suggests that women with pre-existing diabetes or gestational diabetes in LMIC settings should be screened for health related social needs to reduce the prevalence of depression and to promote child social-emotional development. Future studies should consider health related social needs as a target for intervention among pre-conception and perinatal women in LMIC settings.
## References
1. Azami M, Badfar G, Soleymani A, Rahmati S. **The association between gestational diabetes and postpartum depression: A systematic review and meta-analysis**. *Diabetes Res Clin Pract* (2019) **149** 147-155. DOI: 10.1016/j.diabres.2019.01.034
2. Muhwava LS, Murphy K, Zarowsky C, Levitt N. **Perspectives on the psychological and emotional burden of having gestational diabetes amongst low-income women in Cape Town, South Africa**. *BMC Womens Health* (2020) **20** 231. DOI: 10.1186/s12905-020-01093-4
3. Herba CM, Glover V, Ramchandani PG, Rondon MB. **Maternal depression and mental health in early childhood: an examination of underlying mechanisms in low-income and middle-income countries**. *Lancet Psychiatry* (2016) **3** 983-992. DOI: 10.1016/S2215-0366(16)30148-1
4. Ali NS, Mahmud S, Khan A, Ali BS. **Impact of postpartum anxiety and depression on child’s mental development from two peri-urban communities of Karachi, Pakistan: a quasi-experimental study**. *BMC Psychiatry* (2013) **13** 274. DOI: 10.1186/1471-244X-13-274
5. Burlina S, Dalfrà MG, Lapolla A. **Short- and long-term consequences for offspring exposed to maternal diabetes: a review**. *J Matern Fetal Neonatal Med* (2019) **32** 687-694. DOI: 10.1080/14767058.2017.1387893
6. Growth Ornoy A.. **neurodevelopmental outcome of children born to mothers with pregestational and gestational diabetes**. *Pediatr Endocrinol Rev PER* (2005) **3** 104-113. PMID: 16361984
7. Shuffrey LC, Fifer WP. (2020). DOI: 10.1016/B978-0-12-809324-5.23054-X
8. Xu G, Jing J, Bowers K, Liu B, Bao W. **Maternal diabetes and the risk of autism spectrum disorders in the offspring: a systematic review and meta-analysis**. *J Autism Dev Disord* (2014) **44** 766-775. DOI: 10.1007/s10803-013-1928-2
9. Prenoveau JM, Craske MG, West V, Giannakakis A, Zioga M, Lehtonen A. **Maternal Postnatal Depression and Anxiety and Their Association With Child Emotional Negativity and Behavior Problems at Two Years.**. *Dev Psychol.* (2017) **53** 50-62. DOI: 10.1037/dev0000221
10. Slomian J, Honvo G, Emonts P, Reginster J-Y, Bruyère O. **Consequences of maternal postpartum depression: A systematic review of maternal and infant outcomes**. *Womens Health* (2019) **15** 1745506519844044. DOI: 10.1177/1745506519844044
11. Abrahams Z, Lund C, Field S, Honikman S. **Factors associated with household food insecurity and depression in pregnant South African women from a low socio-economic setting: a cross-sectional study.**. *Soc Psychiatry Psychiatr Epidemiol* (2018) **53** 363-372. DOI: 10.1007/s00127-018-1497-y
12. Jebena MG, Taha M, Nakajima M, Lemieux A, Lemessa F, Hoffman R. **Household food insecurity and mental distress among pregnant women in Southwestern Ethiopia: a cross sectional study design**. *BMC Pregnancy Childbirth* (2015) **15** 250. DOI: 10.1186/s12884-015-0699-5
13. Patel V, Kleinman A. **Poverty and common mental disorders in developing countries**. *Bull World Health Organ* (2003) **81** 609-615. PMID: 14576893
14. Das J, Do Q-T, Friedman J, McKenzie D, Scott K. **Mental health and poverty in developing countries: Revisiting the relationship**. *Soc Sci Med* (2007) **65** 467-480. DOI: 10.1016/j.socscimed.2007.02.037
15. Jacob N.. *Mental Illness in the Western Cape* (2015) 1-35
16. Aurino E, Wolf S, Tsinigo E. **Household food insecurity and early childhood development: Longitudinal evidence from Ghana**. *PLOS ONE* (2020) **15** e0230965. DOI: 10.1371/journal.pone.0230965
17. Koen N, Brittain K, Donald KA, Barnett W, Koopowitz S, Maré K. **Maternal Posttraumatic Stress Disorder and Infant Developmental Outcomes in a South African Birth Cohort Study**. *Psychol Trauma Theory Res Pract Policy.* (2017) **9** 292-300. DOI: 10.1037/tra0000234
18. Natamba BK, Namara AA, Nyirenda MJ. **Burden, risk factors and maternal and offspring outcomes of gestational diabetes mellitus (GDM) in sub-Saharan Africa (SSA): a systematic review and meta-analysis.**. *BMC Pregnancy Childbirth* (2019) **19** 450. DOI: 10.1186/s12884-019-2593-z
19. Shuffrey LC, Sania A, Brito NH, Potter M, Springer P, Lucchini M. **Association of maternal depression and anxiety with toddler social-emotional and cognitive development in South Africa: a prospective cohort study**. *BMJ Open* (2022) **12** e058135. DOI: 10.1136/bmjopen-2021-058135
20. Rayport YK, Sania A, Lucchini M, Du Plessis C, Potter M, Springer PE. **Maternal experiences, diabetes, postnatal maternal depression, and child social-emotional outcomes in a South African community cohort, Dryad, Dataset**. (2022). DOI: 10.5061/dryad.kkwh70s73
21. Gray MJ, Litz BT, Hsu JL, Lombardo TW. **Psychometric Properties of the Life Events Checklist**. *Assessment* (2004) **11** 330-341. DOI: 10.1177/1073191104269954
22. Lawrie TA, Hofmeyr GJ, de Jager M, Berk M. **Validation of the Edinburgh Postnatal Depression Scale on a cohort of South African women.**. *South Afr Med J Suid-Afr Tydskr Vir Geneeskd.* (1998) **88** 1340-1344. PMID: 9807193
23. Redinger S, Pearson RM, Houle B, Norris SA, Rochat TJ. **Antenatal depression and anxiety across pregnancy in urban South Africa**. *J Affect Disord* (2020) **277** 296-305. DOI: 10.1016/j.jad.2020.08.010
24. Gowan B, Carter M, Carter A. **ITSEA/BITSEA: Infant-Toddler and Brief Infant-Toddler Social and Emotional Assessment.**. *The Psychological Corporation* (2006)
25. Kruizinga I, Visser JC, van Batenburg-Eddes T, Carter AS, Jansen W, Raat H. **Screening for autism spectrum disorders with the brief infant-toddler social and emotional assessment**. *PloS One* (2014) **9** e97630. DOI: 10.1371/journal.pone.0097630
26. Squires J, Bricker D, Twombly E, Murphy K, Hoselton R. **ASQ:SE--2 Technical Report.**. *Tech Rep* (2015) **28**
27. Small JW, Hix‐Small H, Vargas‐Baron E, Marks KP. **Comparative use of the Ages and Stages Questionnaires in low‐ and middle‐income countries**. *Dev Med Child Neurol.* (2019) **61** 431-443. DOI: 10.1111/dmcn.13938
28. Christodoulou J, Le Roux K, Tomlinson M, Le Roux IM, Katzen LS, Rotheram-Borus MJ. **Perinatal maternal depression in rural South Africa: Child outcomes over the first two years**. *J Affect Disord* (2019) **247** 168-174. DOI: 10.1016/j.jad.2019.01.019
29. Dewing S, Tomlinson M, le Roux IM, Chopra M, Tsai AC. **Food insecurity and its association with co-occurring postnatal depression, hazardous drinking, and suicidality among women in peri-urban South Africa**. *J Affect Disord* (2013) **150** 460-465. DOI: 10.1016/j.jad.2013.04.040
30. Pellowski JA, Bengtson AM, Barnett W, DiClemente K, Koen N, Zar HJ. **Perinatal depression among mothers in a South African birth cohort study: Trajectories from pregnancy to 18 months postpartum**. *J Affect Disord* (2019) **259** 279-287. DOI: 10.1016/j.jad.2019.08.052
31. Schneider M, Baron E, Davies T, Munodawafa M, Lund C. **Patterns of intimate partner violence among perinatal women with depression symptoms in Khayelitsha, South Africa: a longitudinal analysis**. *Glob Ment Health Camb Engl* (2018) **5** e13. DOI: 10.1017/gmh.2018.1
32. Ross GP, Falhammar H, Chen R, Barraclough H, Kleivenes O, Gallen I. **Relationship between depression and diabetes in pregnancy: A systematic review**. *World J Diabetes* (2016) **7** 554-571. DOI: 10.4239/wjd.v7.i19.554
33. Drago F, Scharf RJ, Maphula A, Nyathi E, Mahopo TC, Svensen E. **Psychosocial and environmental determinants of child cognitive development in rural south africa and tanzania: findings from the mal-ed cohort**. *BMC Public Health* (2020) **20** 505. DOI: 10.1186/s12889-020-08598-5
34. Gordon S, Rotheram-Fuller E, Rezvan P, Stewart J, Christodoulou J, Tomlinson M. **Maternal depressed mood and child development over the first five years of life in South Africa**. *J Affect Disord* (2021) **294** 346-356. DOI: 10.1016/j.jad.2021.07.027
|
---
title: 'Perceptions of diabetes patients and their caregivers regarding access to
medicine in a severely constrained health system: A qualitative study in Harare,
Zimbabwe'
authors:
- Dudzai Mureyi
- Nyaradzai Arster Katena
- Tsitsi Monera-Penduka
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021663
doi: 10.1371/journal.pgph.0000255
license: CC BY 4.0
---
# Perceptions of diabetes patients and their caregivers regarding access to medicine in a severely constrained health system: A qualitative study in Harare, Zimbabwe
## Abstract
Nearly half of all sub-Saharan African countries lack operational Diabetes Mellitus policies. This represents an opportunity to build reliable evidence to underpin such policies when they are eventually developed. Representing the interests of those with the experience of living with the condition in national diabetes policies is important, particularly the interests regarding medicine access, a key pillar in diabetes management. One way to achieve this representation is to publish patient perceptions. Patient perspectives are especially valuable in the context of diabetes in Sub-Saharan Africa, where much of the empirical work has focused on clinical and epidemiological questions. We therefore captured the challenges and suggestions around medicine access articulated by a population of diabetes patients and their caregivers. This was a qualitative interpretivist study based on data from focus group discussions with adult diabetes patients and their caregivers. Eight FGDs of 4–13 participants each whose duration averaged 13.35 minutes were conducted. Participants were recruited from diabetes outpatient clinics at two health facilities in Harare. One site was Parirenyatwa Hospital, the largest public referral and teaching hospital in Zimbabwe. The other was a private for-profit facility. Ethics approval was granted by the Joint Research Ethics Committee for University of Zimbabwe College of Health Sciences and the Parirenyatwa Group of Hospitals (Ref: JREC $\frac{295}{18}$). Diabetes patients and their caregivers are interested in affordable access to medicines of acceptable form and quality with minimum effort. Yet, they often find themselves privileging one dimension of access over another e.g. prioritising affordability over acceptability. Based on participants’ articulations, a sound diabetes policy should: 1. provide for financial and consumer protections, 2. regulate healthcare business practices and medicine prices, 3. provide for a responsive health workforce attentive to patient problems, 4. accord the same importance to diabetes that is accorded to communicable diseases, 5. decentralize diabetes management to lower levels of care, 6. limit wastage, corruption, bad macro-financial governance and a lack of transparency about how funding for health is used, and 7. provide support to strengthen patients’ and caregivers’ psychosocial networks. A diabetes policy acceptable to patients is one infused with principles of good governance, fairness, inclusiveness and humanity; characterised by: financial protection and price regulation, consumer protection, equity in the attention accorded to different diseases, decentralized service delivery, inclusion of patient voice in political decision-making, a responsive compassionate health workforce, psychosocial support for patients and their caregivers and allocative efficiency and transparency in public expenditure.
## Diabetes in Sub-Saharan Africa and the importance of patient perceptions
At least 340 million people are estimated to be living with Diabetes Mellitus (diabetes) globally and the cost of managing it exceeded half a trillion dollars as far back as 2015 [1]. In Sub-Saharan Africa (SSA) diabetes-related deaths occur mostly in people under the age of sixty [1], i.e. those in the economically active age group. In this region of the world, (where strained health systems are still grappling with high burdens of infectious disease), barriers to accessing diabetes diagnostic services, diabetes medicines and self-monitoring tools exist. Data and data collection systems on diabetes prevalence, morbidity and mortality are often unreliable, thus preventing proper estimates and hindering the crafting of appropriate responses and policies [2–4].
A national operational policy, strategy or action plan for diabetes is critical because it forms the basis for resource allocation and systematic intervention implementation in addressing this important global public health problem. Yet, by 2019, nearly half of all sub-Saharan African countries lacked operational diabetes policies or plans [5]. While this is a dire observation, it means there’s significant scope for SSA countries to improve. An opportunity exists to build a reliable evidence base that can underpin diabetes policies. These diabetes policies must be attentive to factors affecting access to medicines for diabetes patients because the key to effective diabetes management includes adherence to insulin and oral medicines [6,7]. Among the several frameworks conceptualising the access to medicines and health technologies [6–15], two of the most recently published [6,7] relate specifically to medicines for cardiovascular diseases and diabetes, suggesting the recognition of the unique salience of diabetes and cardiovascular diseases. ‘ Access to medicine’ is defined here as the multi-dimensional construct that characterises the degree to which quality and safe medicines are available, affordable, accessible and acceptable in a timely manner to those needing them [6–15]. Beran et al., [ 6], described a framework for promoting diabetic patients’ access to insulin. These authors called for the synergistic efforts of governments, the pharmaceutical private sector and a civil society that includes a robust patient voice, as the necessary ingredients for ensuring access to insulin. Yet they noted that diabetes patients are rarely represented at top levels of diabetes policymaking. Amplifying patient voice bodes well for service delivery, patient experience and health outcomes [16] and one way to achieve this amplification is to conduct research, explore and publish patient perceptions. Patient voice though, is not always exercisable [17] and in Zimbabwe (one of the countries without a substantive diabetes plan [5]), citizens’ voice was found lacking at decision-making level [18].
The objective of this study was therefore to explore the lived experiences of diabetes patients regarding access to insulin and oral diabetes medicines in Zimbabwe. These perspectives, gathered from diabetes patients and their caregivers in Zimbabwe, (a low-resource setting in SSA), provide insights into what a national diabetes policy or plan in Zimbabwe might look like. Granted, insights from Zimbabwe cannot be assumed to be applicable to the rest of SSA. However, patient perspectives such as those presented in this paper, are valuable in the broader context of diabetes in SSA. They are valuable because they point to the idea that diabetes patients and their caregivers have articulable experiences that are relevant when crafting people-centred national diabetes policies in SSA, where much of the empirical work has focused on clinical and epidemiological questions [2] perhaps, at the expense of patient perspectives. While it is urgent to improve prevention, diagnosis, and treatment of diabetes in Zimbabwe and indeed SSA, is it equally important to seek patient perspectives. This is because patient perspectives provide clues to principles that patients feel should underpin an acceptable national diabetes plan.
## Diabetes and Zimbabwe’s health system
In Zimbabwe, a country with a population of 14.86 million [19], both the public and private sectors play roles in health financing and health service delivery [20]. Following independence from colonial rule in 1980, the Zimbabwean government focused on equitable access to health and made significant investments in public health infrastructure and interventions [21]. However, Zimbabwe’s health system started to deteriorate from the mid-1990s due to economic underperformance [22]. This impacted Zimbabwe’s progress towards Universal Health Coverage, a goal Zimbabwe has been aiming for since 2009 [23]. Coupled with the economic underperformance was a monetary policy that was associated with a weak local currency. This motivated the public and businesses to transact in United States (US) Dollars instead [24]. Pricing of medicines in US Dollars occurred. By 2015, household out-of-pocket (OOP) payments ($24\%$) and external developmental assistance ($25\%$) were the biggest health financing sources, whereas government spending comprised $21\%$ of all health spending [25]. This inadequate government funding plus the mass emigration of skilled healthcare workers, have been cited as some of the causes of Zimbabwe’s critical health worker shortage. The density of physicians for instance, was recently reported to be 0.067 per 1000 population [22]. In terms of organization, the public health system comprises four levels of care. Each level of care is differentiated by the size of the facilities, the qualifications of the staff that works there, the scope of services that can be offered there and the types of medicines that can be stored or dispensed from there. The primary level consists of over 1 400 urban and rural clinics manned by nurses. The secondary and tertiary levels of care (that are designed to be accessed through a referral system that begins at the primary level), comprise forty-four district and eight provincial hospitals respectively. The fourth (quaternary) level of care comprises six referral hospitals [20]. Zimbabwe’s resource allocation system is skewed towards hospitals, with the higher-level provincial and central-level hospitals receiving more funds/subsidies than the lower-level clinics [26]. Zimbabwe has a diabetes prevalence rate estimated to be the highest among the most populous African countries. Projections indicate that Zimbabwe could have over 1.2 million diabetes patients by the year 2035 [27]. In addition, Zimbabwe exhibits the attributes of a fragile state [28–39] as described in fragility literature [40–42]. Fragility compounds health challenges because it undermines health service delivery and good health in general [40,42]. Research with diabetes patients in a context as constrained as Zimbabwe therefore satisfies the ‘the intensity criterion’, in qualitative research [43] i.e. using a case likely to produce a rich picture by virtue of the intensity of the problem or diversity of observations.
The immediate aftermath of the removal of Zimbabwe’s long-time head of government in 2017, was defined by optimism locally and internationally [44,45]. Certain corners of the global health community suggested that with the political change, a new national public health agenda was afoot [46] or ought to be [47,48], in order to resuscitate Zimbabwe’s health system marred by broader economic and health sector challenges that included disrupted medication supply [48–50]. The president of the Zimbabwe Diabetes Association was appointed Zimbabwe’s deputy health minister [51]. Afterwards, plans to implement the eponymous Novartis Access in Zimbabwe were announced [52]. Novartis Access provides selected low and middle-income countries with a basket of 15 medicines (including some diabetes ones) at the subsidised price of US$ 1 per treatment per month. An evaluation of Novartis Access after its first implementation year in Kenya showed no effect on medicine availability in households or on their price at health facilities [53]. This observation could have been partly due to a disconnect between the initiative’s implementation modalities and patient-related preferences [53]–preferences which ought to have been investigated and considered prior to implementation. It’s hoped that the articulations of diabetes patients presented in this paper will inform the Novartis Access implementation in Zimbabwe or other interventions intended to improve access to diabetes and noncommunicable disease therapies in similar contexts.
## Methods
This section describes our methods according to the requirements of the Consolidated Criteria for Reporting Qualitative Research (COREQ) [54] and the Standards for Reporting Qualitative Research (SRQR) [55].
## Ethics statement
Ethics approval was granted by the Joint Research Ethics Committee for University of Zimbabwe College of Health Sciences and the Parirenyatwa Group of Hospitals (Ref: JREC $\frac{295}{18}$). Formal written Informed consent was sought and obtained from participants before data collection commenced.
## Research philosophy
This study was grounded in the critical realism philosophy. Critical realism is the philosophy of science that conceives reality as stratified into three realms: the ‘empirical’, the ‘actual’ and the ‘deep’ [56] (ontological position). The ‘empirical’ contains observable phenomena in the positivist sense e.g. experiences and actions (e.g. diabetes patients in an outpatient clinic waiting area exhibiting signs of diabetes complications). The ‘actual’ consists of events that happen or exist but are not readily empirically accessible, (e.g. poor access to diabetes medicines). Researchers might however observe these events through focused data collection and analysis (e.g. conducting FGDs with patients) [57]. The ‘deep’ comprises things or causal pathways/mechanisms that produce phenomena in the empirical and actual realms. These things in the ‘deep’ cause observable reality either by being present or even by being absent (e.g. the absence of a sound national diabetes policy or the presence of difficult macroeconomic conditions inhabit the deep realm and cause observable poor access to diabetes medicines). Different causal pathways may be activated or inactivated by contextual factors [58–60]. Context therefore matters in critical realist research, because it determines which causal pathways are activated and which are not. Case study research, an intensive research design suited to uncovering causal explanations in contexts, [61] is therefore particularly compatible with critical realist research [62]. This was this paper’s study design.
Critical realism embraces interpretivist paradigms because it holds that the knowledge and the interpretation of reality is always influenced by the subjectivity of people (both researchers and participants) [63]. Because knowledge construction has these subjective aspects, and the activation of causal powers that produce effects is dependent on context, critical realists accept that knowledge of reality is essentially provisional and fallible. It’s subject to incremental modification as more discoveries occur [58] (epistemological position). Therefore it is accepted that the findings documented in this paper may not be applicable to other SSA health systems where different researchers and diabetes patients occupy and where different causal pathways are activated. These findings may not even be applicable to Zimbabwe in future, during a different time period when different contextual factors are activated.
Finally, human emancipation and the improvement of society are guiding values for critical realists [59,64] (axiological position). Critical realism’s focus on uncovering mechanisms of how things do happen, combined with its (poststructuralist) stance that views humans as capable of transforming society [56,65] justifies making policy recommendations for how things ought to happen in order to improve society (as this paper does by recommending crafting patient-sensitive diabetes policies) [58].
## Study design
This study was an interpretivist qualitative case study. A case is a phenomenon occurring in a spatially, temporally or demographically bounded context and is a unit of analysis in research [43]. A case study is therefore described here as the study of a phenomenon in a single (Zimbabwean) context as it unfolds/unfolded naturally (as opposed to study by experimental design). It employed the analysis of text derived from focus group discussions (FGDs) with diabetes patients and their caregivers (family members). The interpretivist paradigm, as well as qualitative research methods, embody unique qualities that made them suitable to researching the complex and varied subjective perceptions of diabetes patients in the study context. With interpretivist qualitative research, people’s lived experiences and the subjective interpretations they attach to those realities, are valid contributions to research and knowledge. No one individual’s reality is privileged over another’s. With the qualitative interpretivist approach, the goal is to provide rich (often diverse) contextualised insights as opposed to generalised laws that are upheld in every context. Interpretivism therefore utilises, as we did, in-depth formal group discussions and/or interviews to elicit rich and thick descriptions of reality from the perspectives of participants. Furthermore, analysis of findings in interpretivist research acknowledges that researchers’ subjectivity also influence their interpretations of the data. This allows researchers to be reflexive, to consciously interrogate their own biases and to seek ways of validating their conclusions [66]. One of the ways this paper’s authors validated their interpretations was to ensure that all three researchers independently analysed and coded all the transcripts, before meeting to discuss their coding and interpretations. The conclusions are based on the interpretations that all the three authors considered amply supported by the data.
## Research team and reflexivity
At the time of the study, the research team comprised three female researchers trained in qualitative research at postgraduate level. DM and TMP are registered pharmacists and NAK is a public health practitioner. TMP holds a Ph.D. in Clinical Pharmacology and at the time of data collection and analysis, DM and NAK were undertaking doctoral research looking at pharmaceutical systems and diabetes care. DM received training in Research Data Management and Data Protection. The participants had no prior acquaintance with any of the researchers and none of the researchers were involved in the care of patients at the sites where data was collected. Apart from academic interest, the researchers had no other interest in the study. There was no further contact with participants after the conclusion of each FGD. Before conducting the FGDs, based on the researchers’ familiarity with the Zimbabwean context, the researchers anticipated that participants would articulate issues to do with the poor availability, affordability and accessibility of oral medicines and insulin. This expectation was met. However, what wasn’t expected was how impassioned and vivid these articulations would be. This observation made the researchers more sensitive to themes such as empathy, justice and fairness, during data collection and analysis. It also hadn’t been expected that paradoxical testimonies about access challenges would be uncovered. For instance, as described in the results and discussion sections of this paper, authors were surprised to discover that patients with health insurance were actually more financially burdened than uninsured ones. Prior to data collection, the researcher (DM), a pharmacist by profession, had considerable preconceived ideas and prior knowledge about the research subject and what caused diabetes patients’ constrained access to medicines. In order to limit the undue influence of this prior professional knowledge on the data collection, DM sent weekly group emails to her two doctoral supervisors who were neither Zimbabwean residents nor health professionals. They were experienced qualitative researchers but were more distanced from the research phenomenon. In these long reflexive emails, DM documented key questions and observations in the FGDs. In response, the supervisors provided feedback that facilitated the identification of blind spots and biases (and good practices). These academics also reviewed the FGD guide before use and identified leading questions that suggested DM’s bias. As Mauthner and Doucet pointed out, such regular interactions with one’s research group/team significantly enhance a researcher’s ability to be reflexive [67].
## Sampling location and criteria
This study was conducted in Harare, Zimbabwe’s capital, for reasons related to logistical convenience. In addition, compared to other locations in Zimbabwe, *Harare is* the location of institutions/actors involved in Zimbabwe’s pharmaceutical system (pharmaceutical wholesalers and manufacturers, the only two universities that train pharmacists, high-level Ministry of Health officers, office bearers of pharmaceutical professional associations etc.). Therefore, Harare represented the most complete microcosm of Zimbabwe’s pharmaceutical system. Participants were recruited from diabetes outpatient clinics at two health facilities in Harare. These were: These study locations were selected for their accessibility as well as the ease of recruiting participants there. Having one study site that is a public hospital and another that is a private for-profit facility increased the possibility that a heterogenous sample (in terms of socioeconomic status), would be recruited. This was considered advantageous because diverse testimonies were being sought. These sites held diabetes outpatient clinics where diabetes patients and/or their caregivers presented, congregated and participated in support group discussions. Based on residential address data, patients who visit Parirenyatwa Referral Hospital reside in diverse geographical regions of Zimbabwe [68]. The same is true for the private for-profit facility. It was noted that several peer-reviewed studies focusing on diabetic populations in Zimbabwe recruited solely from Parirenyatwa Hospital [68–72]. Participants were eligible for inclusion in the study if they were adults on diabetes medication or caregivers/guardians of such.
## Data collection and management
Data was collected by one researcher (DM) through audio-recorded (FGDs) with patients recruited at the outpatient clinics’ waiting area. Each FGD was conducted at a secluded location away from the waiting area to protect participant privacy, protect the quality of the audio recording and pose minimum disruption to the clinic workflow. FGDs helped to triangulate and validate the testimony of participants in real-time as participants compared experiences. Sampling was purposive, based on the inclusion criteria already stated and willingness to participate. Potential participants were publicly addressed in the outpatients’ waiting areas and willing FGD participants interviewed separately. This recruitment procedure made it impossible to determine the number of eligible participants who declined to participate. After consenting, no participant withdrew from the FGD although it was made clear during the consent-seeking process that they could withdraw if they so wished. Because FGDs were conducted with patients who happened to be in an outpatient clinic on the days of data collection, who resided in different parts of the country, member-checking was impractical. However, during the FGDs, the researcher occasionally asked clarifying questions when ambiguous or surprising testimony was given. The guides used by the researcher to moderate the FGDs (attached as S1 and S2 Files) were tested in the first FGD group and no changes were subsequently made. Eight FGDs of 4–13 participants each whose duration averaged 13.35 minutes (excluding the time the moderator spent on opening and closing remarks), were conducted between 27 February 2019 and 20 March 2019. No follow-up FGDs with the same participants were conducted.
FGDs were conducted until what is commonly understood as ‘saturation’ was achieved. Saturation is a problematic concept as multiple researchers have varying ways of defining it, measuring it or determining when it’s been achieved [73]. Furthermore, it is impossible to fully rule out the discovery of new insights with continued data collection [74]. This dilemma was resolved by ceasing data collection when it was considered that further data collection would offer increasingly diminishing returns given the consistency of testimony from the FGDs already conducted (ibid). Practical and ethical considerations also contributed to the cessing of data collection when the testimony became consistent with each subsequent FGD. Such practical and ethical considerations included: the finite resources available for data collection and analysis, and the reluctance to keep disrupting the weekly outpatient clinic activities simply to recruit participants. All transcripts of audio-recordings were checked for completeness and accuracy by all the researchers. Some of the text in the transcripts was in the vernacular language of ChiShona and was analysed in this form by all the researchers, who have expert level proficiency in both English and ChiShona. Only ChiShona excerpts included in this paper were translated into English. Translations were discussed by researchers to determine, by consensus, that the original essence of each translated excerpt had been retained. No personal identifying information was contained in the audio recordings or transcripts. All project files were stored on the researchers encrypted and password-protected laptops. Transcripts of the FGDs are attached to this paper as S1 Data.
## Ethical considerations
Before being invited to sign the informed consent forms, participants had the study explained to them in ChiShona, within a group context. Then, individual participant information sheets were distributed to each participant. Those who had questions asked the researcher (DM) individually. Separate consent both for participation in the research and for audio recording of testimony was obtained from each participant. Participants were offered $5.00, which was sufficient for a modest lunch meal (according to the Zimbabwe National Statistics Office’s economic data at the time). This was offered as a token to appreciate time lost while participating in the research–time they could have been engaging in income-generating activities. While the ethical implication (coercion), and the methodological implication (selection bias) of incentivising participants are acknowledged, this token is generally a requirement in the research context. Local ethics approval would not have been granted in the absence of this provision in the study protocol. The token was kept small to minimize its persuasive power. Participants were given an option to decline this token and had to indicate on the consent form whether they had accepted or declined it. None declined.
## Analysis
FGD Transcripts were managed using NVivo 11 and MS Word computer programs. Analysis was deductive, using a pre-existing analytical framework described in the paragraph immediately following this one [9]. Each author listened to all recordings and deductively coded all transcripts solo, using the coding sheet derived from the analytical framework (S1 Table). DM then compared all three sets of coded transcripts side by side, noting differences in coding, which were subsequently discussed and resolved by consensus. Text that was unable to be categorised in any of the deductive codes was coded as ‘other’ and discussed to determine by consensus if new inductive codes were warranted. No new codes were warranted thus themes were identified deductively. How these themes resonated with extant literature on diabetes patient perspectives, is discussed in the discussion section of this paper.
## Analytical framework
Bigdeli et al. [ 9], advanced a nuanced way of conceiving ‘Access to medicine’ from a health system perspective. They considered multiple levels of organization from the household level to the international context (Fig 1). Among a set of extant access to medicines frameworks [6–15] this framework (Bigdeli et al. ’s framework [9]) was found to be the most comprehensive conception yet of what access to medicine entails. Therefore, this framework’s domains formed the basis of our coding scheme outlined in S1 Data. The four domains in that framework are as follows: 1. Individual Household and Community factors; 2. the Dimensions of Access (i.e. availability, affordability, accessibility, acceptability and quality); 3. Health System Building Blocks [75] (i.e. Financing, human resources, health information, service delivery and supply chain), and 4. Cross-cutting factors (i.e. Equity, Governance, Transparency, Innovation, Donors’ Agenda and Market forces).
**Fig 1:** *Access to medicines from a health system perspective [9].This image was made available under the Creative Commons CC-BY-NC license and permitted non-commercial use, distribution and reproduction in any medium, provided the original work is properly cited.*
## Results
All FGD participants were literate male and female adult patients or adult caregivers/guardians of diabetes patients. Focus group discussions were conducted at both a public referral hospital and at a private diabetes clinic.
Public hospital Respondents Profile: *During data* collection, with the exception of outpatients under the age of 5 or over the age of 65 who were treated for free, outpatients at Parirenyatwa hospital were charged US$15.00 in user fees [76]. This was significantly cheaper than the private sector and as a result, low-income households tended to seek care there [77]. Based on FGD testimony, some of which is contained as excerpts in the results section of this paper, participants recruited at Parirenyatwa hospital came from low-income households, were generally uninsured, used public transportation to travel to the hospital and walked on foot in search of medicines and bargain prices at various pharmacies.
Private clinic Respondents Profile: By comparison, participants recruited at the Centre for Diabetes Management were more economically well-off. The user fees there were considerably higher, (US$120 charged to first-time patients and US$70 charged for each subsequent visit). Based on FGD testimony, participants recruited at this site were generally insured. They owned and drove (or were driven in) personal vehicles when searching for medicines at pharmacies. They also understood the value of being represented at diabetes policymaking platforms.
The views expressed by the participants, such as the one illustrated below, indicate that individuals suffering from diabetes have considerable appreciation for the seriousness of diabetes as a life-threatening condition: The role of adherence to correct regimens of medicines from trusted sources, blood glucose monitoring and dietary modifications, in effectively managing diabetes was understood. See one exemplary quotes below and the rest [IHC1-IHC6] in S2 Table.
The expectation of representation at policy-making platforms (since the deputy health minister is the former president of the Zimbabwe Diabetes Association) was also well articulated by diabetes patients recruited from the private sector facility.
We suggest that taken together, these quotes indicated that diabetes patients in the study context were generally well-informed about the outpatient management of their condition. They also saw a relationship between having their interests represented at ministerial (policy) level and having their grievances addressed by government.
This sections below contain the results of our deductive analysis. Exemplary supporting quotes from the data are contained in the narrative. Results are arranged thematically in four major sections associated with the four domains in the framework depicted in Fig 1 [9]. First, individual household and community factors are described, followed by the dimensions of medicine access (Availability, Affordability, Accessibility, Acceptability, and Quality), then Health system building blocks (Financing, Human Resources, Services, Information, Supply Chain) and finally the cross-cutting factors (Donors Agenda, innovation, transparency, market forces, Governance and equity).
## Individual and household factors
Discussions with participants who take care of diabetes patients revealed that the main individual and household factor affecting access to diabetes medicines and other treatment modalities is income (or lack thereof). Furthermore, caring for people on diabetes treatment often strains household income and stretches psychosocial support networks. This can then strain relations and mental health, as the quotes below illustrate.
Another participant in the same FGD session then agreed with this sentiment: These quotes emphasise how diabetes (and possibly many other illnesses) is not merely a clinical condition that affects a patient, but can be a social issue that undermines household harmony and caregivers’ mental health.
## Availability
Regarding the availability of medicines, all discussants recruited from the public referral hospital described the erratic availability of diabetes medicines there, particularly the expensive ones.
The problems associated with the unavailability of medicines at the public hospital were compounded by the unavailability of particular oral medicines at private pharmacies. This unavailability of certain medicines in the private sector (particularly the fixed dose combination medicines), was also reported by participants recruited from the private diabetes clinic:
## Affordability
When medicines and blood glucose testing strips were available in the private sector, their unaffordability was almost universally acknowledged by participants. Some even asked the FGD moderator if she knew of places they could access cheaper medicines: What made the medicines unaffordable for many was the fact that some medicines and insulin formulations were being sold by pharmacies in foreign currency (United States Dollars), which participants could not easily acquire: It was not just the medicines themselves that made diabetes treatments unaffordable. Several of the participants resided outside Harare, the capital city, yet they had to collect their medicines at the public referral hospital in Harare (where this study was conducted). The transaction costs associated with traveling to the referral hospital to refill prescriptions made the whole endeavour unaffordable for them.
This unaffordability of medicine access fuelled the tendency for patients privilege affordability over the acceptability of the service provider:
## Accessibility
In order to find pharmacies where medicines were both available and affordable, most of the participants reported that they expended effort searching for medicines at multiple pharmacies and comparing prices. This process, in our view, compromises the accessibility of pharmacy services.
## Acceptability
Regarding the acceptability of diabetes medicines available in their context, there were isolated responses that indicated some participants’ preference for oral medicines over injectable insulin. Likewise, the relative unacceptability of insulin administered by syringe and needle compared to pen-sets was articulated.
## Quality
Concerning the quality of medicines, one participant indicated that he cared less about the quality of the medicines he procured as long as they accessed the medicines. His experience described in the quote below for instance, was recounted when the FGD moderator asked the participants about their confidence in the quality of the medicines they accessed: The majority of the participants however, expressed considerable trust in the quality of medicines dispensed at licensed pharmacies: This trust was expressed despite significant dissatisfaction with pharmacists’ business practices or in isolated cases, their clinical judgement (see results section on market forces further down in this paper’s results section). On one hand this is a positive observation; patients ought to have full confidence in the quality of medicines dispensed from health facilities. On the hand however, this unquestioning trust could be harmful if falsified, counterfeit medicines and substandard or expired medicines are
## Financing
Based on the testimony of the majority of the FGD participants, exemplified by quotes below, diabetes patients recognise the need for sustainable financing of medicine access through health insurance or government subsidies: We discovered that government spending on diabetes medicines is not evident in the wake of stock-outs at public institutions where user fees were charged. We also discovered that health insurance occasionally fails to deliver on the promise of financial protection for a variety of reasons. Some insurance plans are rejected outright by service providers. Some plans require co-payment in cash from the patient at the point of care. Some plans require that patients pay for medicines out of pocket, in forex, then claim for reimbursement in local currency, (calculated at exchange rates that are well below market ones). Furthermore, some service providers appear to fraudulently submit insurance claims to the insurers for the full cost of medicines after having collected the full amount in cash as co-payment from patients. The three quotes below, from the submissions of insured participants, illustrate all this.
## Human resources
On the human resources front, the desirability of ethical, compassionate, patient and helpful conduct by pharmaceutical service providers was frequently expressed in FGDs:
## Service delivery
Participants showed they already appreciate the value of modern medicine in the management of diabetes but are dissatisfied with the inconvenience of having to travel to the central hospital for service delivery. Also dissatisfying about service delivery was the rationing of supplies that occurs at public hospitals, which prevented patients from purchasing more than one month’s supply of medication at a time: Asked why travelling to the referral hospital was necessary, one participant revealed that it was because local health facilities nearest to patients’ homes were more expensive than the tertiary hospital:
## Health information
Access to information is an essential aspect of access to medicine. Information to do with where particular medicines can be sourced, therapeutic information and information about home management of diabetes appeared to be appreciated by nearly all participants: Surprisingly, however, information about possible side effects related to medicine regimens was unwelcome because it was viewed by many respondents as a potential deterrent to adherence: We noted that State regulatory institutions, (in particular the consumer protection and national medicines regulatory agencies), as well as diabetes patient associations can do more to provide information regarding their mandates to the lay public. We arrived at this finding because in the FGDs, participants said things that suggested that they were not aware of the existence of these agencies, what they do and how they can be accessed.
## Supply chain
Although supply chain issues were not touched on as frequently as others, a desire for price controls upstream the pharmaceutical value chain in order to mitigate the likelihood of unaffordable prices at the point of care, was expressed by one participant:
## Market forces
Several business practices of private-sector pharmacists, (motivated by profit and/or market forces), were criticised by nearly all FGD participants. An example of such practices was the pharmacist-induced demand for medicines and testing services: Objectionable business practices by pharmacists also included charging a fee for conducting monitoring tests even in emergencies: Other pharmacies rejected some health insurance schemes as valid payment and this too was considered unacceptable business practice, The pricing of pharmaceutical services and commodities in foreign currency was also reported by participants as unacceptable
Lastly, the placing of high profit mark-ups on diabetes medicines by pharmacies was also mentioned as an undesirable business practice.
## Equity
Frequently expressed was a serious dissatisfaction with the perceived inequity at policy level. This perceived inequity manifests as government’s apparent privileging of communicable diseases over diabetes, as exemplified by these quotes: Pleas were made for the government to consider that many diabetes patients may be vulnerable people who have limited means e.g. pensioners. This was followed by the suggestion that access to medicines for these vulnerable persons should not be associated with high prices: Noncommunicable diseases (NCDs) are catching up to infectious diseases as the leading cause of mortality in Africa [13]. This calls for NCD strategies that are as deliberate and effective as those designed for communicable diseases. In Zimbabwe, the privileging of communicable diseases (HIV/AIDS in particular) over NCDs is evidenced by several observations such as: the collection of an earmarked AIDS levy/tax, the existence of a dedicated statutory body dealing with HIV/AIDS and the availability of free medicines for HIV and Tuberculosis at public institutions. Furthermore, public institutions stock even the recently-developed medicines for third-line HIV treatment regimens. Whereas for diabetes, the public sector stocks and dispenses hypoglycemics developed as early as the beginning of the 20th century. Frequently, interviewed participants referenced the prioritization of communicable disease management over NCDs justifying why diabetes deserves the same level of attention. A diabetes policy must be seen to be equitable and just.
## Governance
Participants had several ideas about the governance strategies that could be included in a national diabetes policy. These included: medicine price controls, regulation of the quality of both allopathic and complementary medicines accessible to patients, the judicious governance of national resources that contribute to government revenue (e.g. diamond reserves), a return to the policy of providing senior citizens access to healthcare free of charge, and the formulation of a sound monetary policy:
## Donors
The perceived unaffordability of medicines and monitoring tests then led to the suggestions that a philanthropic entity be called upon to provide diabetes medicines for free: This apparent desire for donors to also prioritize diabetes medicines, may have been motivated by the observation many health products in Zimbabwe were ‘largely procured through donor funding’ [20].
## Innovation
Regarding innovation, it was implicitly suggested that development of reusable glucose testing strips by be looked into:
## Discussion
As expected, based on the extant literature, diabetes patients and their caregivers are interested in affordable medicines of good quality, accessed with minimum effort. Owing to the nature of the recruitment method (recruiting participants with a positive diabetes diagnosis, from major urban health institutions), it is not surprising that the views expressed by the discussants generally depict a patient population that is sensitive to the seriousness of diabetes, can articulate lived experiences with clarity and can link medicine access, a modified diet, monitoring tools and voice to positive outcomes. This is the kind of population able to provide clues about the contents of a developing country’s patient-centred-national diabetes policy. Most tenets uncovered from participant testimony relate to a recognition that diabetes is not simply an issue of pharmacotherapy but also of good governance, fairness, inclusiveness and humanity.
In this section, based on the results presented, we discuss our inductive interpretation of the ten pillars that the diabetes patients desire in a national diabetes policy. These pillars (discussed in turn below) are: financial protection, equity, a responsive workforce, consumer protection, allocative efficiency, price regulation, voice and representation, research and innovation, decentralised service delivery, and provisions for psychosocial support.
## Financial protection
Diabetes management, often being lifelong and multi-faceted, burdens households financially. Private health insurance in developing countries can cover some gaps in public spending [78]. However, Zimbabwean patients with health insurance are doubly disadvantaged when their health insurance plan is declined at the point of care. They pay out-of-pocket for medicines with no guarantee of a full reimbursement ex-post. They spend money on health insurance subscriptions and spend money again on out-of-pocket payments for medicines. Based on the aggrieved participants’ testimony it can be inferred that a sound diabetes policy ought to provide for the regulation of health insurance business practices. Predatory practices by health insurance providers and service providers doubly tax patients. Predatory practices by health insurance providers also discourage patients from purchasing health insurance.
## Responsive workforce
It has since been recommended that the number of health workers managing diabetes, as well as the level of diabetes knowledge they possess, should be increased [3]. Beyond this, the patients in our FGDs are additionally concerned with the health workers’ communication skills, business ethics and compassion. In a perfect market, patients are able to vote with their feet and seek medicines from pharmacies that meet desired standards of conduct. However, as the data suggests, patients seek medicines at outlets where prices are affordable or where their health insurance package is accepted, regardless of the quality of customer care there. A sound diabetes policy should have in-built governance safeguards that facilitate greater patient choice and enable patients to truly vote with their feet without being forced to give primacy to affordability over other dimensions of access.
## Consumer protection
In addition to financial protection, the protection of consumers of medicines is paramount. Patients lack the capacity to measure the quality of medicines at the point of care and may automatically trust pharmacies as sources of quality-verified medicines. In addition, consumers silently endure what they consider unethical business practices without recourse because of limited access to consumer protection agencies. A diabetes policy responsive to patient needs must provide for the empowerment of patients so that they may seek recourse when they feel they’ve been unfairly treated by the health system.
## Allocative efficiency
Patients who were deeply dissatisfied with the unavailability of diabetes medicines at the public hospital had ideas about the sustainable financing of medicines through earmarked taxation or the judicious use of revenue from Zimbabwe’s exportable natural resources. A strong diabetes strategy cannot escape the need for financial resources yet sub-Saharan health systems are generally inadequately funded to tackle diabetes [3]. From a patient’s perspective, corruption, bad macro-financial governance and lack of transparency about how funding for health is used, are certainly not the ingredients of a good diabetes strategy. Patients prefer that funding for health be visibly spent on items that they associate most with allocative efficiency, like medicines and rather than on items they associate with wastage, like vehicles.
## Price regulation
Although several policy options are available for low and middle-income countries who desire to control the price of pharmaceuticals [79], in Zimbabwe, the prices of pharmaceuticals are not regulated by the state through any codified framework [80]. This is thought to contribute to the variable and/or astronomical medicine prices. A diabetes policy that is acceptable to patients ought to pay attention to the prices of diabetes treatment consumables regardless of who pays for them.
## Voice and representation
The value of patient voice in matters concerning their treatment and on political forums has already been touched on in the introduction and was echoed by respondents. An advantage of including patient representatives at high levels of health system decision-making is that, patients gain a more informed understanding about how funding for health is gathered and used. A lack of transparency in this matter heightens suspicions of corruption and mismanagement.
## Research and innovation
Reusable glucose test strips were not in existence at the time of the study. However, the observation that some FGD participants suggested that reusable glucose test strips should be invented, taken together with the frequently-expressed theme of expensive glucose monitoring, shows a desire for an innovation-oriented approach to diabetes policy formulation.
## Decentralised service delivery
A fair diabetes policy ought to be cognisant of the inconveniences associated with requiring diabetes patients to regularly travel from different regions of the country to attend routine outpatient clinics at the largest referral hospital in the country. This tended to occur for several reasons. First, the user fees at lower levels of care were higher. Second, the medicines patients took were not classified as those that can be stored or dispensed at lower levels of care. Third, the referral hospital in the capital city was where diabetologists could be accessed. Decentralization of diabetes management to lower levels of care has been done with some success elsewhere in Africa [81].
## Psychosocial support
A reliance on psychosocial networks by persons living with or affected by diabetes, was evident [see IHC Quotes]. Support groups for diabetic patients have been shown to improve clinical outcomes [82]. Diabetes patients and their caregivers already congregate regularly at outpatient clinics for monitoring and prescription renewal. Building on this to form structured local support groups and social ties may improve patients’ collective voice, chances for collaboration on income-generating projects [83] and mental health [84]. Formal inclusion of community support groups within a diabetes policy framework would increase their legitimacy and possibly widen their access to funding as a consequence.
## Future research direction for a comprehensive diabetes policy
Our work in this paper documents the qualitative research conducted with people living with diabetes. While their lived experiences constitute invaluable input into any national diabetes policy, it is worth highlighting that further research with other stakeholders, utilising quantitative and mixed methods approaches is also necessary. These types of research could include: cost-effectiveness studies and other economic analyses to support the selection of one policy option over others; epidemiological studies focusing on patterns of morbidity and mortality; pharmacological studies and clinical trials to test and validate on new treatment modalities for diabetes patients.
## Conclusion
As SSA countries prepare to devise and operationalise national diabetes policies, most of which are rightly concerned with the prevention, diagnosis and treatment of diabetes and its complications, it is important to attend to the interests of those with lived experiences of diabetes. We set out to discover some of these interests by qualitatively capturing the challenges and suggestions by a population of diabetes patients in Harare, Zimbabwe. We concluded that in addition to the already documented provisions for access to affordable and universal prevention, diagnosis and treatment modalities, a diabetes policy acceptable to patients is one infused with principles of good governance, fairness, inclusiveness and humanity. Such principles manifest as provisions for: financial protection and price regulation, consumer protection, equity in the attention accorded to different diseases, decentralized service delivery, inclusion of patient voice in political decision-making for a responsive compassionate health workforce, psychosocial support for patients and their care givers and allocative efficiency and transparency in public expenditure.
## References
1. Ogurtsova K, da Rocha Fernandes J, Huang Y, Linnenkamp U, Guariguata L, Cho N. **IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040**. *Diabetes Res Clin Pract* (2017.0) **128** 40-50. DOI: 10.1016/j.diabres.2017.03.024
2. Misra A, Gopalan H, Jayawardena R, Hills A, Soares M, Reza‐Albarrán A. **Diabetes in developing countries**. *J Diabetes* (2019.0) **11** 522-39. DOI: 10.1111/1753-0407.12913
3. Atun R, Davies J, Gale E, Bärnighausen T, Beran D, Kengne A. **Diabetes in sub-Saharan Africa: from clinical care to health policy**. *Lancet Diabetes Endocrinol* (2017.0) **5** 622-67. DOI: 10.1016/S2213-8587(17)30181-X
4. van Crevel R, van de Vijver S, Moore D. **The global diabetes epidemic: what does it mean for infectious diseases in tropical countries?**. *Lancet Diabetes Endocrinol* (2017.0) **5** 457-68. DOI: 10.1016/S2213-8587(16)30081-X
5. 5World Health Organization. Global health Observatory: Policies, strategies and action plans. Available from] https://apps.who.int/gho/data/view.main.2473 Accessed on 11 May 2020.
6. Beran D, Hirsch I, Yudkin J. **Why are we failing to address the issue of access to insulin? A national and global perspective**. *Diabetes Care* (2018.0) **41** 1125-31. DOI: 10.2337/dc17-2123
7. Tran D, Njuguna B, Mercer T, Manji I, Fischer L, Lieberman M. **Ensuring patient-centered access to cardiovascular disease medicines in low-income and middle-income countries through health-system strengthening**. *Cardiol Clin* (2017.0) **35** 125-34. DOI: 10.1016/j.ccl.2016.08.008
8. Afzali M, Khorasani E, Alvandi M, Sabbagh-Bani-Azad M, Sharif Z, Saiyarsarai P. **Providing a framework for assessment of the access to medicine**. *Daru* (2019.0) **27** 243-54. DOI: 10.1007/s40199-019-00268-1
9. Bigdeli M, Jacobs B, Tomson G, Laing R, Ghaffar A, Dujardin B. **Access to medicines from a health system perspective**. *Health Policy Plan* (2013.0) **28** 692-704. DOI: 10.1093/heapol/czs108
10. 10Management Sciences for Health. MDS-3: Managing Access to Medicines and Health Technologies. Arlington, VA: Management Sciences for Health; 2012.. *MDS-3: Managing Access to Medicines and Health Technologies* (2012.0)
11. Obrist B, Iteba N, Lengeler C, Makemba A, Mshana C, Nathan R. **Access to health care in contexts of livelihood insecurity: a framework for analysis and action**. *PLoS Med* (2007.0) **4**. DOI: 10.1371/journal.pmed.0040308
12. Frost L, Reich M. *Access: how do good health technologies get to poor people in poor countries?* (2008.0)
13. 13Center for Pharmaceutical Management.
Defining and Measuring Access to Essential Drugs, Vaccines, and Health Commodities: Report of the WHO-MSH Consultative Meeting, Ferney-Voltaire, France, December 11–13, 2000.
Prepared for the Strategies for Enhancing Access to Medicines Program. Arlington, VA: Management Sciences for Health; 2003. *Prepared for the Strategies for Enhancing Access to Medicines Program* (2003.0)
14. 14World Health Organization. WHO medicines strategy: framework for action in essential drugs and medicines policy 2000–2003 (No. WHO/EDM/2000.1).
Geneva: World Health Organization; 2000.. (2000.0)
15. 15World Health Organization. Equitable access to essential medicines: a framework for collective action (No. WHO/EDM/2004.4).
Geneva: World Health Organization; 2004.. (2004.0)
16. Weiss C., Connell JP, Kubisch AC, Schorr LB, Weiss CH. *New approaches to evaluating community initiatives: Concepts, methods, and contexts* (1995.0)
17. Brinkerhoff D, Bossert T. **Health governance: principal–agent linkages and health system strengthening**. *Health Policy Plan* (2014.0) **29** 685-93. DOI: 10.1093/heapol/czs132
18. Loewenson R.. **Public participation in health systems in Zimbabwe**. *IDS Bull.* (2000.0) **31** 14-20
19. 19World Bank. (2020) Zimbabwe. Available from https://databank.worldbank.org/views/reports/reportwidget.aspx?Report_Name=CountryProfile&Id=b450fd57&tbar=y&dd=y&inf=n&zm=n&country=ZWE Accessed on 12 January 2022.
20. 20Ministry of Health and Child Care. National Health Strategy for Zimbabwe 2016–2020. Equity in Health: Leaving no one behind. Harare: Ministry of Health and Child Care; 2016.. *National Health Strategy for Zimbabwe 2016–2020* (2016.0)
21. Nyazema N.. **The Zimbabwe Crisis and the provision of social services: health and education**. *J Dev Soc* (2010.0) **26** 233-61
22. Haley C, Vermund S, Moyo P, Kipp A, Madzima B, Kanyowa T. **Impact of a critical health workforce shortage on child health in Zimbabwe: a country case study on progress in child survival, 2000–2013**. *Health Policy Plan* (2017.0) **32** 613-24. DOI: 10.1093/heapol/czw162
23. 23Ministry of Health and Child Care. The national health strategy for Zimbabwe 2009–2013. Equity and quality in health: A people’s right.
Harare: Ministry of Health and Child Care; 2009.. *Equity and quality in health: A people’s right.* (2009.0)
24. Southall R.. **Bond notes, borrowing, and heading for bust: Zimbabwe’s persistent crisis**. *Can J Afr Stud* (2017.0) **51** 389-405
25. Zeng W, Lannes L, Mutasa R. **Utilization of health care and burden of out-of-pocket health expenditure in Zimbabwe: results from a National Household Survey**. *Health Syst Reform* (2018.0) **4** 300-12. DOI: 10.1080/23288604.2018.1513264
26. Shamu S, January J, Rusakaniko S. **Who benefits from public health financing in Zimbabwe? Towards universal health coverage**. *Glob Public Health* (2017.0) **12** 1169-82. DOI: 10.1080/17441692.2015.1121283
27. Guariguata L, Whiting D, Hambleton I, Beagley J, Linnenkamp U, Shaw J. **Global estimates of diabetes prevalence for 2013 and projections for 2035**. *Diabetes Res Clin Pract* (2014.0) **103** 137-49. DOI: 10.1016/j.diabres.2013.11.002
28. 28BBC. Zimbabwe Profile -Timeline. 2019. Available from https://www.bbc.co.uk/news/world-africa-14113618 Accessed on 19 January 2019
29. 29Fund For Peace. Fragile States Index. 2018. Available from http://fundforpeace.org/fsi/wp-content/uploads/2018/04/951181805-Fragile-States-Index-Annual-Report-2018.pdf Accessed on 19 January 2019.
30. 30World Bank. CPIA Africa. 2017. Available from http://documents.worldbank.org/curated/en/850151531856335222/Assessing-Africas-policies-and-institutions-2017-CPIA-results-for-Africa Accessed on 19 January 2019.
31. Chimusoro A, Maphosa S, Manangazira P, Phiri I, Nhende T, Danda S, Claborn D. *Current Issues in Global Health* (2018.0)
32. Muchadenyika D.. **Civil society, social accountability and service delivery in Zimbabwe**. *Dev Policy Rev* (2017.0) **35** O178-95
33. Davis W, Chonzi P, Masunda K, Shields L, Mukeredzi I, Manangazira P. **Notes from the field: typhoid fever outbreak—Harare, Zimbabwe, October 2016–March 2017**. *MMWR Morb Mortal Wkly Rep* (2018.0) **67** 342. DOI: 10.15585/mmwr.mm6711a7
34. Maringira G. **Politics, privileges, and loyalty in the Zimbabwe national army**. *Afr Stud Rev* (2017.0) **60** 93-113
35. Guzura T, Ndimande J. **Electoral authoritarianism and one-party dominance in Southern Africa: The Zimbabwean case**. *Politics Gov* (2016.0) **5** 5-18
36. Muchadenyika D.. **Multi‐donor Trust Funds and Fragile States: Assessing the Aid Effectiveness of the Zimbabwe Multi‐donor Trust Fund.**. *J Int Dev* (2016.0) **28** 1337-57
37. Rogerson C.. **Responding to informality in urban Africa: Street trading in Harare, Zimbabwe**. *Urban Forum* (2016.0) **27** 229-251
38. Chirisa I, Nyamadzawo L, Bandauko E, Mutsindikwa N. **The 2008/2009 cholera outbreak in Harare, Zimbabwe: case of failure in urban environmental health and planning**. *Rev Environ Health.* (2015.0) **30** 117-24. DOI: 10.1515/reveh-2014-0075
39. Mason P.. **Zimbabwe experiences the worst epidemic of cholera in Africa**. *J Infect Dev Ctries* (2009.0) **3** 148-51. DOI: 10.3855/jidc.62
40. Baird M.. *World Development Report: Background Paper* (2011.0)
41. Cammack D, McLeod D, Menocal A, Christiansen K. *Donors and the ‘Fragile States’ Agenda: A Survey of Current Thinking and Practice* (2006.0)
42. 42DFID. Why we need to work more effectively in fragile states. London. DFID; 2005. Available from https://reliefweb.int/report/zimbabwe/why-we-need-work-more-effectively-fragile-states Accessed on 12 November 2021.. *Why we need to work more effectively in fragile states* (2005.0)
43. Miles M, Huberman A. *Qualitative Data Analysis: An Expanded Sourcebook* (1994.0)
44. Asuelime L.. **Mnangagwa’s foreign policy direction: old wine in new skin?**. *J. Afr. Foreign Aff* (2018.0) **5** 9-21
45. Rutherford B.. **Mugabe’s shadow: limning the penumbrae of post-coup Zimbabwe**. *Can J Afr Stud* (2018.0) **52** 53-68
46. Madzimbamuto F.. *An update from Zimbabwe* (2018.0)
47. Kidia K.. **The future of health in Zimbabwe**. *Glob Health Action* (2018.0) **11** 1496888. DOI: 10.1080/16549716.2018.1496888
48. Green A.. **Zimbabwe post-Mugabe era: reconstructing a health system**. *Lancet* (2018.0) **391** 17. DOI: 10.1016/S0140-6736(18)30007-2
49. Witter S, Wurie H, Chandiwana P, Namakula J, So S, Alonso-Garbayo A. **How do health workers experience and cope with shocks? Learning from four fragile and conflict-affected health systems in Uganda, Sierra Leone, Zimbabwe and Cambodia**. *Health Policy Plan.* (2017.0) **32** iii3-13. DOI: 10.1093/heapol/czx112
50. Meldrum A.. **Zimbabwe’s health-care system struggles on**. *Lancet* (2008.0) **371** 1059-60. DOI: 10.1016/s0140-6736(08)60468-7
51. 51Pindula. ‘Mnangagwa appoints new deputy minister of Health’. 2018. Available from https://news.pindula.co.zw/2018/09/10/mnangagwa-appoints-new-deputy-minister-of-health/ Accessed on 9 May 2020.
52. 52Ministry of Health Zimbabwe’s Twitter Page. Available from: https://twitter.com/MoHCCZim/status/1134061220506550273. Accessed on June 13, 2019.
53. Rockers P, Laing R, Ashigbie P, Onyango M, Mukiira C, Wirtz V. **Effect of Novartis Access on availability and price of non-communicable disease medicines in Kenya: a cluster-randomised controlled trial**. *Lancet Glob Health* (2019.0) **7** e492-502. DOI: 10.1016/S2214-109X(18)30563-1
54. Tong A, Sainsbury P, Craig J. **Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups.**. *Int J Qual Health Care* (2007.0) **19** 349-57. DOI: 10.1093/intqhc/mzm042
55. O’Brien B, Harris I, Beckman T, Reed D, Cook D. **Standards for reporting qualitative research: a synthesis of recommendations**. *Acad Med* (2014.0) **89** 1245-51. DOI: 10.1097/ACM.0000000000000388
56. Bhaskar R, Hawke G. *The order of natural necessity: a kind of introduction to critical realism* (2017.0)
57. Leca B, Naccache P. **A critical realist approach to institutional entrepreneurship**. *Organization* (2006.0) **13** 627-51
58. Haigh F, Kemp L, Bazeley P, Haigh N. **Developing a critical realist informed framework to explain how the human rights and social determinants of health relationship works**. *BMC Public Health* (2019.0) **19** 1-2. DOI: 10.1186/s12889-018-6343-3
59. Sayer A. *Realism and Social Science* (2000.0)
60. Tsoukas H.. **What is management? An outline of a metatheory**. *Br. J. Manag.* (1994.0) **5** 289-301
61. Sayer A.. *Method in Social Science: A realist approach* (1992.0)
62. Wynn D, Williams C. **Principles for conducting critical realist case study research in information systems**. *MIS Q* (2012.0) **1** 787-810
63. Maxwell J.. *A Realist Approach for Qualitative Research* (2012.0)
64. Danermark B, Ekström M, Karlsson J. (2019.0)
65. Ekström M.. **Causal explanation of social action: the contribution of Max Weber and of critical realism to a generative view of causal explanation in social science**. *Acta Sociol* (1992.0) **35** 107-22
66. Alharahsheh H, Pius A. **A review of key paradigms: Positivism VS interpretivism**. *Glob. acad. j. humanit. soc. sci* (2020.0) **2** 39-43
67. Mauthner N, Doucet A. **Reflexive accounts and accounts of reflexivity in qualitative data analysis**. *Sociology* (2003.0) **37** 413-31
68. Machingura P, Chikwasha V, Okwanga P, Gomo E. **Prevalence of and factors associated with nephropathy in diabetic patients attending an outpatient clinic in Harare, Zimbabwe**. *Am J Trop Med Hyg* (2017.0) **96** 477-82. DOI: 10.4269/ajtmh.15-0827
69. Machingura P, Macheka B, Mukona M, Mateveke K, Okwanga P, Gomo E. **Prevalence and risk factors associated with retinopathy in diabetic patients at Parirenyatwa Hospital outpatients’ clinic in Harare, Zimbabwe**. *Archives of Medical and Biomedical Research* (2017.0) **3** 104-11
70. Chako K, Phillipo H, Mafuratidze E, Zhou D. **Significant differences in the prevalence of elevated HbA1c levels for type I and type II diabetics attending the Parirenyatwa Diabetic Clinic in Harare, Zimbabwe.**. *Chin J Biol.* (2014.0) 2014
71. Mafuratidze E, Chako K, Phillipo H, Zhou D. **Over 27% of Type II Diabetic Patients Studied at Parirenyatwa Diabetic Clinic in Zimbabwe Have Evidence of Impaired Renal Function**. *Int J Sci Technol Res (New Delhi)* (2014.0) **3** 14-8
72. Mafundikwa A, Ndhlovu C, Gomo Z. **The prevalence of diabetic nephropathy in adult patients with insulin dependent diabetes mellitus attending Parirenyatwa Diabetic Clinic, Harare**. *Cent Afr J Med* (2007.0) **53** 1-6. DOI: 10.4314/cajm.v53i1-4.62599
73. Saunders B, Sim J, Kingstone T, Baker S, Waterfield J, Bartlam B. **Saturation in qualitative research: exploring its conceptualization and operationalization.**. *Qual Quant* (2018.0) **52** 1893-907. DOI: 10.1007/s11135-017-0574-8
74. Strauss A, Corbin J. *Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory* (1998.0)
75. 75World Health Organization. Everybody’s business—strengthening health systems to improve health outcomes: WHO’s framework for action. Geneva: World Health Organization; 2007.. *Everybody’s business—strengthening health systems to improve health outcomes: WHO’s framework for action* (2007.0)
76. 76Chipunza P. Public Hospitals adjust charges. https://www.herald.co.zw/public-hospitals-adjust-charges/ Accessed 12 January 2022.
77. 77Mbanje P. Desperate patients overwhelm public hospitals. Available from https://www.newsday.co.zw/2019/03/desperate-patients-overwhelm-public-hospitals/ Accessed 12 January 2022.
78. Pauly M, Zweifel P, Scheffler R, Preker A, Bassett M. **Private health insurance in developing countries**. *Health Aff* (2006.0) **25** 369-79. DOI: 10.1377/hlthaff.25.2.369
79. Nguyen T, Knight R, Roughead E, Brooks G, Mant A. **Policy options for pharmaceutical pricing and purchasing: issues for low-and middle-income countries.**. *Health Policy Plan* (2015.0) **30** 267-80. DOI: 10.1093/heapol/czt105
80. 80Ministry of Health & Child Welfare and World Health Organization.
Zimbabwe Pharmaceutical Country Profile. Harare: Ministry of Health and Child Welfare; 2011.. *Zimbabwe Pharmaceutical Country Profile* (2011.0)
81. Kengne A, Fezeu L, Sobngwi E, Awah P, Aspray T, Unwin N. **Type 2 diabetes management in nurse-led primary healthcare settings in urban and rural Cameroon**. *Prim Care Diabetes* (2009.0) **3** 181-8. DOI: 10.1016/j.pcd.2009.08.005
82. Izzah Z, Suprapti B, Aryani T, Budiatin A, Rahmadi M, Hapsari P. **Diabetes Support Groups Improve Patient’s Compliance and Control Blood Glucose Levels**. *Indones. J. Pharm* (2013.0) **2** 94-101
83. Westaway M, Seager J, Rheeder P, Van Zyl D. **The effects of social support on health, well-being and management of diabetes mellitus: a black South African perspective**. *Ethn Health* (2005.0) **10** 73-89. DOI: 10.1080/1355785052000323047
84. Pastakia S, Manyara S, Vedanthan R, Kamano J, Menya D, Andama B. **Impact of bridging income generation with group integrated care (BIGPIC) on hypertension and diabetes in rural western Kenya**. *J Gen Intern Med* (2017.0) **32** 540-8. DOI: 10.1007/s11606-016-3918-5
|
---
title: Self-care knowledge, attitude and associated factors among outpatients with
diabetes mellitus in Arsi Zone, Southeast Ethiopia
authors:
- Rahel Nega Kassa
- Hana Abera Hailemariam
- Mekdes Hailegebreal Habte
- Altayework Mekonnen Gebresillassie
journal: PLOS Global Public Health
year: 2021
pmcid: PMC10021671
doi: 10.1371/journal.pgph.0000097
license: CC BY 4.0
---
# Self-care knowledge, attitude and associated factors among outpatients with diabetes mellitus in Arsi Zone, Southeast Ethiopia
## Abstract
### Introduction
Diabetes mellitus is a chronic illness that requires continuing medical care and ongoing patient self-management, education and support to prevent acute complications and to reduce the risk of long-term complications. Therefore, this study aims to assess the self-care knowledge, attitude and associated factors among outpatients with diabetes mellitus in Arsi Zone, Southeast Ethiopia.
### Materials and methods
A cross sectional study was employed in Arsi Zone, Southeast Ethiopia from April to June 2017 among 290 patients with diabetes mellitus. Structured questionnaire was employed through face to face interview. Bivariate and multivariate regression was done and a statistical significance was declared at p value < 0.05.
### Results
Among 290 respondents, $64.8\%$ and $27.6\%$ of them had good knowledge and good attitude towards self-care in this study respectively. Being married (AOR: 3.41, $95\%$ CI: 1.480–8.095), being employed in occupation (AOR: 5.8, $95\%$ CI: 2.26–14.67) and living in higher socioeconomic status (AOR: 2.0, $95\%$ CI: 1.096–3.322) are the independent factors associated to good knowledge of respondents towards self-care whereas living in lower socioeconomic status(AOR: 0.478, $95\%$ CI: 0.262–0.874), having informal education (AOR: 4.002, $95\%$ CI: 1.941–8.254), not having family history of diabetes mellitus (AOR: 0.422, $95\%$ CI: 0.222–0.803) and having short duration of diagnosis (AOR: 3.209, $95\%$ CI: 1.537–6.779) were significantly associated factors to have poor attitude towards self-care.
### Conclusion
Majority of the study participants had good knowledge towards diabetes self-care practice while a significant number of the participants had unfavorable attitude towards diabetes self-care. Being married, being employed and living in higher socioeconomic status were the determinant factors of knowledge towards the diabetes self-care practice while having informal education and having short duration of diagnosis were the significant factors associated to the unfavorable attitude towards diabetes self-care practice. Therefore, efforts should be made on enhancing patients’ socioeconomic status and equipping with diabetic self-care centered health information particularly for those patients with short duration of diagnosis.
## Introduction
Chronic health conditions are responsible for $60\%$ of the global disease burden. Globally, approximately one in three of all adults suffer from multiple chronic conditions (MCCs) [1]. In this group of health conditions, diabetes mellitus stands out because of high morbidity and mortality rates, as well as increasing prevalence levels [2]. About 422 million people worldwide have diabetes, the majority living in low-and middle-income countries, and 1.6 million deaths are directly attributed to diabetes each year [3]. In developing countries, treatment adherence reaches a mere $20\%$, generating negative health statistics and entailing high costs for families, society and governments [4] Thus, integral care for diabetes patients should cover psychosocial and cultural aspects. Therapeutic education is fundamental to inform, motivate and strengthen patients and families to live with the chronic condition [5].
In Ethiopia, one of the top five African countries with high prevalence of diabetes in the age range of 18–99 years, there were 2,652,129 cases of diabetes in 2017 [6]. A cross sectional study conducted in Ethiopia showed that almost half of the participants ($$n = 320$$) ($49\%$) had long term diabetic complications confirmed medically [7]. Studies revealed that the diabetic care in general is below the acceptable standard [8]. Though national data on prevalence and incidence of diabetes are lacking in Ethiopia, patient attendance rates and medical admissions in major hospitals are rising. Inadequate diabetic self-management remains a significant problem facing health care providers and populations in all settings. In contrast, patients who have adequate self-management have better outcomes, live longer, enjoy a higher quality of life, and suffer fewer symptoms and minimal complications [9]. So that, Diabetes Self-*Management is* the cornerstone of care for all individuals with diabetes who want to achieve successful health-related outcomes [10]. There is an increasing amount of evidence that individuals who are educated and diligent in their diabetes self-care achieve better and durable diabetic control [11,12]. Past studies on knowledge, attitude, and practice towards the prevention of diabetes complications consistently revealed the requirements of better awareness on prevention, diagnosis, and risk factor control of diabetes [13]. A good attitude towards DM complications helps patients to change any harmful dietary and lifestyle habits [14].
*Though* general health education regarding diabetes mellitus self-care activities and its complication is being provided by health care providers working in different health settings, well organized diabetic self care education program is not yet established in Ethiopian health institutions. As to the investigators knowledge, there is no study conducted on this topic in the present study area. So, this study intended to assess self-care knowledge, attitude and associated factors among diabetic outpatients in Arsi Zone, Southeast Ethiopia.
## Materials and methods
A total of 290 adult study participants were included in this study. The study was conducted from April to June 2017. Data were collected by using pretested and structured questionnaire through face to face interview. The questionnaire was developed after reviewing relevant literature related with the problem under the study [15,16].
## Study design
A cross-sectional study design was employed.
## Study area
Facility based study which was performed among patients with diabetes who had follow up at four hospitals (AsellaReferral Hospital, Bekoji Hospital, Abomssa Hospital and Arsi Robe Hospital) found in Arsi Zone, Ethiopia.
Study period: April to June, 2017.
Sample size: 290 patients were recruited.
Sample size calculation: The study used the single population proportion sample size determination formula. The prevalence rate of good attitude towards diabetic self-care practice was taken as $78\%$ from a study conducted at Dilla UniversityReferral Hospital ([9]), with $95\%$ CI, and $5\%$ marginal error (where n is desired sample size, Z is value of standard normal variable at $95\%$ confidence interval and, p is maximum expected proportion which is $78\%$ and d is marginal error which is $5\%$).
n = Z 2 α/2 P (1-P)/d2 = (1.96)2* 0.78* 0.22 / (0.05)2 = 264 Adding $10\%$ contingency for non-response, the final total sample size was 290.
## Sampling technique
The monthly flow of the patients was determined from the Health Management and Information System (HMIS) of each hospital found in Arsi Zone. The total monthly number of the patients obtained from those four hospitals found in Arsi Zone was 507, 371, 322, and 302 respectively. The number of study participants was proportionally allocated for each hospital to maintain the representativeness of the data (99, 71, 62, and 58 respectively).Every fifth patient was selected using the systematic sampling technique for interview. Simple random sampling or lottery method was used to select the first patient to start the interview.
## Inclusion criteria
All patients with DM aged 18 years and older who were on follow up at least for the past six months prior to the study at Hospitals found in Arsi Zone.
## Exclusion criteria
Diabetes patients who were, severely ill, mentally and physically incapable for the interview.
## Operational definitions
Knowledge: study participants who score greater than $50\%$ were considered to have “good knowledge” and those who score $50\%$ or less were considered to have “poor knowledge”.
Attitude: study participants who score $50\%$ and above were considered as having a “good attitude” and respondents who score less than $50\%$ considered as having “poor attitude” for self-care.
## Data collection tool and technique
Data was collected by using pretested and structured questionnaire through face to face interview. The tool was adapted from SKILLDS (Spoken Knowledge In Low Literacy Diabetes Knowledge Assessment scale) [15] and DAS3 (Diabetes Attitude Survey) [16]. It contained socio-demographic characteristics of the respondents; clinical characteristics of the participants like type of diabetes mellitus, duration of diagnosis, presence of complications and co-morbidities were developed from different literatures. The data related to these complications and comorbidities were collected from each patient’s medical chart. Afaan Oromo language version of questionnaire was used for data collection purpose. Besides, for Amharic language only speaker study subject, in order to avoid translation bias at the spot, each data collector had one Amharic language questionnaire that was used as a dictionary.
Twelve health professionals who were fluent in speaking Amharic and Afaan Oromo were involved in the data collection process. Eight Masters of Public Health (MPH) and Masters of Science in Nursing (MSc) holder health professionals were recruited as supervisors.
## Data quality assurance
The questionnaire was translated from English language to Amharic and Afaan Oromo by different translator and back to English by second other translator who was a health professional and fluent on the respective languages to compare its consistency.
The questionnaire was pretested to determine its validity. It was pretested on $5\%$ of the total sample size in patients on follow up at Adama Hospital and necessary adjustments was made on the questionnaire before it was used for actual data collection. Data collectors and supervisors were trained for three days on the study instrument and data collection procedure. The principal investigator and the supervisors checked the collected data for completeness and corrective measures were taken accordingly.
## Statistical analysis
The data was checked for completeness and consistencies, then, it was cleaned, coded and entered in to computer using SPSS windows version 21.
Descriptive statistics was computed to describe the socio-demographic characteristics of the study participants and determine the diabetic self-care practice knowledge and attitude of the study participants. Additionally, binary and multiple logistic regression analyses were constructed to examine the existence of relationship between the diabetic self-care practice knowledge and attitude and selected variables. Statistical significance was declared at $P \leq 0.05.$ Finally, the result was presented in the form of text and using tables.
## Ethics approval and consent to participate
The approval letter was taken from the Ethical Review of College of Health Sciences of Arsi University (Ref. no: CHS/R/$\frac{0019}{2016}$/17). The final permission letter was written to the respective hospitals by the ethical review committee of Arsi Zone Health Bureau. The will of the participants to participate in the study was assured by informed written consent, which was obtained from each study participant before data collection begins. The consent was taken both for the interview and to take some clinical information from their medical records. The right and the confidentiality of study participants were assured.
## Socio-demographic characteristics of respondents
All the sampled study participants were involved in this study giving a response rate of 100 percent. The mean age of respondents was 39 ± 15.8. Most of the respondents 63 ($21.1\%$) were within 18–24 age group and majority of the respondents 142 ($47.5\%$) were Muslim in religion. Of these 290 respondents, 79 ($36.6\%$) had attained elementary school and above. Majority of the study participants, 202($67.6\%$) were married, and 177($60.1\%$) of them are poor in their socioeconomic status. Large number 158 ($53\%$) of the study participants were farmers followed by private business holders 54($19.2\%$). In addition, majority 166($55.5\%$) of the respondents were from the rural area (Table 1).
**Table 1**
| Variables | Category | Frequency | Percent |
| --- | --- | --- | --- |
| Sex | MaleFemale | 176114 | 60.739.3 |
| Age | 18–24 | 63 | 21.1 |
| Age | 25–34 | 70 | 23.4 |
| Age | 35–44 | 56 | 18.7 |
| Age | 45–54 | 48 | 16.1 |
| Age | 55–64 | 37 | 12.4 |
| Age | >65 | 25 | 8.4 |
| Religion | Muslim | 138 | 47.6 |
| Religion | Orthodox Christian | 128 | 44.1 |
| Religion | Protestant | 21 | 7.2 |
| Religion | Catholic | 3 | 1 |
| Educational status | Cannot read and write | 65 | 21.7 |
| Educational status | Can read and write | 36 | 12.4 |
| Educational status | Elementary | 40 | 13.8 |
| Educational status | High school | 19 | 6.6 |
| Educational status | College and above | 130 | 44.8 |
| Marital status | Single | 85 | 29.3 |
| Marital status | Married | 194 | 66.9 |
| Marital status | Divorced | 7 | 2.4 |
| Marital status | Widowed | 4 | 1.4 |
| Socioeconomic status | Poor | 154 | 53.1 |
| Socioeconomic status | Medium | 41 | 14.1 |
| Socioeconomic status | Rich | 95 | 32.8 |
| Occupation | Farmer | 156 | 53.8 |
| Occupation | Private business | 54 | 18.6 |
| Occupation | Student | 45 | 15.5 |
| Occupation | Employed | 35 | 12.1 |
| Place of residence | Urban | 130 | 44.8 |
| Place of residence | Rural | 160 | 55.2 |
## Clinical characteristics of the respondents
Majority 198 ($66.2\%$) of the respondents had normal body weight (18.5–24.9kg/m2). More than half, ($58.3\%$), of the respondents had type 1 diabetes. Most of the respondents, 190 ($65.5\%$) were using insulin whereas only $4.8\%$of them were taking both insulin and oral hypoglycemic agents. Most ($72\%$) of the respondents had no diabetic complications and $72.8\%$ of the respondents had no co morbidities (Table 2).
**Table 2**
| Variables | Category | Frequency | Percent |
| --- | --- | --- | --- |
| Type DM | Type 1 | 169 | 58.3 |
| Type DM | Type 2 | 121 | 41.7 |
| Family history | Yes | 61 | 21.0 |
| Family history | No | 229 | 79.0 |
| Type of treatment | Insulin Injection | 190 | 65.5 |
| Type of treatment | Oral anti hyperglycemic | 86 | 29.7 |
| Type of treatment | Both | 14 | 4.8 |
| BMI(kg/m2) | Underweight (<18.5) | 49 | 16.9 |
| BMI(kg/m2) | Normal weight1 (18.5–24.9) | 190 | 65.5 |
| BMI(kg/m2) | Overweight (25–29.9) | 47 | 16.2 |
| BMI(kg/m2) | Obese (30 and above) | 4 | 1.4 |
| Having glucometer | Yes | 40 | 13.8 |
| Having glucometer | No | 250 | 86.2 |
| Diabetic complications | Yes | 72 | 24.8 |
| Diabetic complications | No | 218 | 75.2 |
| Presence of comorbidities | Yes | 79 | 27.2 |
| Presence of comorbidities | No | 211 | 72.8 |
| Duration of diagnosis(in years) | <5 | 97 | 33.4 |
| Duration of diagnosis(in years) | 5–10 | 93 | 32.1 |
| Duration of diagnosis(in years) | >10 | 100 | 34.5 |
## Factors associated to knowledge of self-care practice among patients with diabetes mellitus
Above half ($51.2\%$) of the study participants had good knowledge in this study. Marital status, occupation, andsocio economic status were found to be a significant associated factors affecting diabetes self-care knowledge of the study participants. Those respondents who had married were 3.41 times more likely to had good diabetes self-care knowledge than the single ones (AOR: 3.41, $95\%$ CI: 1.480–8.095). On the other hand, study participants who had employed were 5.8 times more likely to have good diabetes self-care knowledge than farmers (AOR: 5.8, $95\%$ CI: 2.26–14.67). Those respondents who were living in higher socioeconomic status were 2 times more likely to had good diabetes self-care knowledge than those who were living in lower socioeconomic status (AOR: 2.0, $95\%$ CI: 1.096–3.322) (Table 3).
**Table 3**
| Variables | Self-care knowledge status | Self-care knowledge status.1 | COR (CI 95%) | AOR (CI 95%) | P value |
| --- | --- | --- | --- | --- | --- |
| | Good N% | Poor N% | COR (CI 95%) | AOR (CI 95%) | P value |
| Marital status | | | | | |
| Single | 47(55.3) | 38(44.7) | 1 | 1 | |
| Married | 135(69.6) | 59(30.4) | 2.039(1.217–3.416) | 3.461(1.480–8.095) | 0.004** |
| Divorced | 5(71.4) | 2(28.6) | 4.015(0.736–21.901) | 5.964(0.917–38.777) | 0.062 |
| Widowed | 1(25) | 3(75) | 1.606(0.216–11.957) | 3.961(0.408–38.408) | 0.235 |
| Occupation | | | | | |
| Farmer | 95(60.9) | 61(39.1) | 1 | 1 | |
| Private business | 33(61.1) | 21(38.9) | 2.3(1.223–4.324) | 3.322(1.609–6.858) | 0.001** |
| Student | 28(62.2) | 17(37.8) | 1.011(0.522–1.957) | 2.880(1.121–7.403) | 0.028** |
| Employed | 32(91.6) | 3(8.6) | 4.6(1.976–10.709) | 5.766(2.266–14.670) | 0.000** |
| Monthly income | | | | | |
| Poor | 97(63) | 57(37) | 1 | 1 | |
| Medium | 23(56.1) | 18(43.9) | 0.686(0.433–1.736) | 0.912(0.419–1.985) | 0.817 |
| Rich | 68(71.6) | 27(28.4) | 0.475(0.273–0.826) | 2(1.09–3.322) | 0.022** |
## Factors associated to attitude towards self-care among patients with diabetes mellitus
Among the respondents, about 219($73.2\%$) of them had poor attitude towards diabetes self-care activities. Monthly income, educational status family history and duration of diagnosis were found to be substantial factors associated with diabetes self-care attitude of the respondents. Accordingly, those respondents who were living in lower socioeconomic status were $47.8\%$ more likely to had poor attitude towards diabetes self-care than those who are living in higher socioeconomic status (AOR: 0.478, $95\%$ CI: 0.262–0.874). In other way, study participants who had informal education were 4.002 times more likely to have poor diabetes self-care attitude than study participants who had college and above education (AOR: 4.002, $95\%$ CI: 1.941–8.254). The other finding was that study participants who had not family history of diabetes mellitus had $42.2\%$ times more likely to have poor attitude towards diabetes self-care compared to those respondents who had family history of diabetes mellitus (AOR: 0.422, $95\%$ CI: 0.222–0.803). On the other hand, study participants who had short duration of diagnosis were 3.209 times more likely to have poor attitude than who had long duration of diagnosis (AOR: 3.209, $95\%$ CI: 1.537–6.779) (Table 4).
**Table 4**
| Variable | Self-care attitude status | Self-care attitude status.1 | COR (CI 95%) | AOR (CI 95%) | P value |
| --- | --- | --- | --- | --- | --- |
| Variable | Good N% | Poor N% | COR (CI 95%) | AOR (CI 95%) | P value |
| Monthly Income | | | | | |
| Poor | 32(20.8) | 122(79.2) | 0.486(0.275–0.857) | 0.478(0.262–0.874) | 0.017** |
| Medium | 27(65.85) | 14(34.15) | 0.956(0.445–2.052) | 0.844(0.372–1.915) | 0.684 |
| Rich | 61(64.21) | 34(35.79) | 1 | 1 | |
| Educational status | | | | | |
| Cannot read and write | 25(69.4) | 11(30.6) | 1.966(0.855–4.521) | 2.750(0.734–4.172) | 0.207 |
| Can read and write | 39(60) | 26(40) | 3.217(1.648–6.283) | 4.002(1.941–8.254) | 0.000** |
| Elementary | 25(62.5) | 15(37.5) | 2.784(1.279–6.062) | 2.866(1.255–6.546) | 0.012** |
| High school | 14(73.7) | 5(26.3) | 1.508(0.502–4.531) | 1.520(0.492–4.693) | 0.466 |
| College and above | 107(82.3) | 23(17.7) | 1 | 1 | |
| Family history | | | | | |
| Yes | 36(59) | 25(41) | 1 | 1 | |
| No | 174(76) | 55(24) | 0.433(0.239–0.783) | 0.422(0.222–0.803) | 0.017** |
| Duration of diagnosis(in Years | | | | | |
| <5 | 67(69.1) | 30(30.9) | 1.774(0.921–3.416) | 2.347(1.152–4.779) | 0.019** |
| 5–10 | 62(66.7) | 31(33.3) | 2.116(1.096–4.084) | 3.209(1.537–6.700) | 0.002** |
| >10 | 81(81) | 19(19) | 1 | 1 | |
## Discussion
This study tried to assess self-care knowledge and attitude among patients with diabetes in Arsi Zone and $64.9\%$ [$95\%$ CI (59.0–$70.2\%$)], $26.8\%$ [$95\%$ CI (22.8–33.7)] of the respondents had good knowledge and good attitude towards self-care respectively.
Regarding knowledge, the finding of this study is comparative with the study conducted at North Shewa Zone Oromia, Ethiopia ($67.8\%$) and Egypt ($52.3\%$) [17,18].
However, it is lower than the study done at Ayder Hospital Tigray and in Nigeria [19,20]. The variation from the study conducted at Ayder Hospital might be that almost $79\%$ of the study participants at Ayder Hospital were from urban area that possibly gives them the chance of getting more information than the respondents from the present study area where $55.2\%$ of the respondents were from rural area.
Besides, almost fifty percent ($49.5\%$) of the study participants included under the study conducted at Nigeria had taken tertiary level education whereas only $17.7\%$ of the study participants in the present study were attended the tertiary level or college and above. This might contribute to difference in gaining self-care related knowledge from variety of sources.
Being married, being employed and living higher socioeconomic status were significantly associated factors to had good self-care knowledge. This can be explained as married ones might have chance of detail discussion with their couple and information exchange about the disease. In addition to this support to access the awareness from different Medias might be available. Similarly, those respondents who were employed have possibility of getting awareness from different directions since they can have communication and information exchange from variety of individuals or professionals. In addition, study participants who are living in higher socioeconomic status might have probability of accessing medias as well as health professionals than those who are living in the lower level of economic status.
On the other hand, $26.8\%$ of the respondents in this study had good attitude towards diabetes mellitus self-care which is far less from the study conducted in Debre Tabor Town, Northwest Ethiopia ($39.5\%$) [21], Kenya ($49\%$) [22], Adama, Ethiopia ($81.9\%$) [23] where respondents had positive attitude towards diabetes self-care management. This variation might be due to the difference in study population, where majority of participants from Debre Tabor were civil servants ($62.1\%$), exposed to health information ($50.4\%$) unlike the participants in this study where majority of them were from rural area ($55.2\%$), under poor socio economic status ($53.1\%$), having no family history of DM ($79\%$), were farmers ($53.8\%$) and also similar in Kenya. This reveals that living in lower socioeconomic status, having informal education, not having family history of diabetes mellitus were significantly associated with having poor attitude towards self-care on diabetes mellitus.
The finding that is revealed in Bale Zone, Ethiopia [24] also confirmed that respondents in higher economic status had good attitude than in the lower economic status.
In this study, participants who had short duration of diagnosis were 3.209 times more likely to have poor attitude than who had long duration of diagnosis. This finding coincides with the study done in Brazil [25] which shows that as *Diabetes is* a chronic disease, the shorter the timing of the diagnosis, the more conflict feelings that needs to overcome in order to reach the stage of positive coping of the disease that in turn affects the attitude.
The present study revealed that most of the respondents had unfavorable attitude though most of them had good knowledge. This could be due to the fact that good knowledge does not necessarily guarantee that the patient will have good attitude. As different scientific researches [26] show people who have better knowledge are more resistance and questionable rather than implementing what they know rather than those who know less. Knowledgeable members of the public are in general, less supportive of morally contentious areas than those who are less knowledgeable.
## Conclusion
Majority of the study participants had good knowledge towards self-care in this study. Therefore, it has to be strengthening giving continual diabetic self-care centered health education for patients who have follow up on those respective hospitals found in Arsi Zone. On the contrary, many of the respondents had poor attitude towards self-care so that efforts should be made from different stakeholders like hospitals, health professionals who are following cases these patients. This can be in terms of delivering clear and brief awareness which can in turn result in change in behavior towards self-care, supporting those who are living in lower socio economic status.
Therefore, efforts should be made on equipping patients with adequate and specific diabetic self-care centered health information particularly for those patients with short duration of diagnosis. In other way enhancing the socioeconomic status of the patients is highly recommended.
## References
1. Marengoni A., Angleman S., Melis R.. **Aging with multimorbidity: a systematic review of the literature.**. *Ageing Res. Rev* (2011.0) **10** 430-439. DOI: 10.1016/j.arr.2011.03.003
2. Skolnik R.. *No communicable diseases country profiles 2018* (2018.0)
3. 3Leading causes of death and disability worldwide: 2000–2019 World Health Organization: 2020.
4. 4Ana E., Kattia O.,et al, Knowledge on diabetes mellitus in the self-care process, vol.14 no.5 RibeirãoPreto Sept./Oct. 200610.1590/S0104-11692006000500014.. DOI: 10.1590/S0104-11692006000500014
5. Mustafa A, Iqbal M, Parvez MA. **Assessment of knowledge, attitude and practices of diabetics regarding their foot care.**. *Apmc* (2017.0) **11** 43-7
6. 6International Diabetes Federation. Eighth edition 2017. IDF Diabetes Atlas, 8th edition. 2017. 1–150 p.
7. 7Kalayou K. et al Diabetes Self Care Practices and Associated Factors Among Type 2 Diabetic Patients in TikurAnbessa Specialized Hospital, Addis Ababa, Ethiopia- A Cross Sectional Study: international journal of pharmaceutical sciences and research 10.13040/IJPSR.0975-8232.3(11).4219-29.. DOI: 10.13040/IJPSR.0975-8232.3(11).4219-29
8. Yeweyenhareg. *An assessment of the health care system for diabetes in Addis Ababa, Ethiopia Ethiopian Journal of Health Development* (2005.0) **19** 203-210. DOI: 10.4314/ejhd.v19i3.9999
9. Addisu. **Assessment of Diabetic Patient Perception on Diabetic Disease and Self-Care Practice in Dilla University Referral Hospital, South Ethiopia.**. *Journal of Metabolic Syndrome* (2014.0) **03**
10. Clark M.. **Diabetes self-management education: a review of published studies.**. *Prim Care Diabetes* (2008.0) **2** 113-20. DOI: 10.1016/j.pcd.2008.04.004
11. Powers MA, Bardsley J, Cypress M, Duker P, Funnell MM, Fischl AH. **Diabetes Self-Management Education and Support in Type 2 Diabetes.**. *Diabetes Educ.* (2017.0) **43** 40. DOI: 10.1177/0145721716689694
12. Rani PK, Raman R, Subramani S, Perumal G, Kumaramanickavel G, Sharma T. **Knowledge of diabetes and diabetic retinopathy among rural populations in India, and the influence of knowledge of diabetic retinopathy on attitude and practice.**. *Rural Remote Health.* (2008.0) **8** 838. PMID: 18656993
13. Islam FM, Chakrabarti R, Dirani M, Islam MT, Ormsby G, Wahab M. **Knowledge, attitudes and practice of diabetes in rural Bangladesh: the Bangladesh Population based Diabetes and Eye Study (BPDES).**. *PLoS One.* (2014.0). DOI: 10.1371/journal.pone.0110368
14. Yitayeh B., Yonas A., Yaregal A.. **Attitude, practice and its associated factors towards Diabetes complications among type 2 diabetic patients at Addis Zemen District hospital, Northwest Ethiopia.**. *BMC Public Health* (2020.0)
15. Rothman RL, Malone R, Bryant B, Wolfe C, Padgett P, DeWalt DA. **The Spoken Knowledge in Low Literacy in Diabetes scale: a diabetes knowledge scale for vulnerable patients.**. *Diabetes Educ* (2005.0) **31** 215-24. DOI: 10.1177/0145721705275002
16. Anderson RM, Fitzgerald JT, Funnell MM, Gruppen LD. **The third version of the Diabetes Attitude Scale**. *Diabetes Care* (1998.0) **21** 1403-7. DOI: 10.2337/diacare.21.9.1403
17. Niguse H, Belay G, Fisseha G, Desale T, Gebremedhn G. **Self-care related knowledge, attitude, practice and associated factors among patients with diabetes in Ayder Comprehensive Specialized Hospital, North Ethiopia.**. *BMC Res Notes.* (2019.0) **12** 34. DOI: 10.1186/s13104-019-4072-z
18. Maged M, Alghazaly G, Alshora A. **Knowledge, practice and barriers of foot self-care among diabetic patients at tanta university hospitals, Egypt.**. *Egypt J Med* (2018.0) **36**
19. ZerihunSahile L, BenayewShifraew M, ZerihunSahile M. **Diabetic Self-Care Knowledge and Associated Factors Among Adult Diabetes Mellitus Patients on Follow-Up Care at North Shewa Zone Government Hospitals, Oromia Region, Ethiopia, 2020.**. *Diabetes MetabSyndrObes.* (2021.0) **14** 2111-2119. DOI: 10.2147/DMSO.S298336
20. Jackson, Idongesit L. **Knowledge of self-care among type 2 diabetes patients in two states of Nigeria**. **12** 404. DOI: 10.4321/s1886-36552014000300001
21. Asmamaw A, Asres G, Negese D, Fekadu A. **Knowledge and attitude about diabetes mellitus and its associated factors among people in Debre Tabor Town, Northwest Ethiopia: cross sectional study.**. *Sci J Public Health.* (2015.0) **3** 199-209. DOI: 10.11648/j.sjph.20150302.17
22. Maina K, Ndegwa Z, Njenga E, Muchemi E. **Knowledge, attitude and practices related to diabetes among community members in four provinces in Kenya: a cross-sectional study.**. *Pan Afr Med J* (2010.0) **7**
23. Adem AM, Gebremariam ET, Gelaw BK, Ahmed M, Fromsaseifu M, Thirumurugan DG. **Assessment of knowledge, attitude and practices regarding life style modification among type 2 diabetic mellitus patients attending Adama Hospital Medical College, Oromia Region, Ethiopia.**. *Glob J Med Res* (2014.0) **14** 37-48
24. Chanyalew W, Alemayehu G. **Knowledge, attitude, practices and their associated factors towards diabetes mellitus among non-diabetes community members of Bale Zone administrative towns, South East Ethiopia. A cross-sectional study.**. *PLoS One.* (2016.0) **12** e0170040. DOI: 10.1371/journal.pone.0170040
25. Karla Anna. **Knowledge and attitude about diabetes self-care of older adults in primary health care Recife, Northeastern Brazil.**. *Ciênc. saúdecolet.* (2019.0) **24**. DOI: 10.1590/1413-81232018241.35052016
26. Evans Geoffrey. **The relationship between knowledge and attitudes in the public understanding of science in Britain**. *Public Understanding of Science* (1995.0) **4**. DOI: 10.1088/0963-6625/4/2/004
|
---
title: Parental-perceived home and neighborhood environmental correlates of accelerometer-measured
physical activity among school-going children in Uganda
authors:
- Bernadette Nakabazzi
- Lucy-Joy M. Wachira
- Adewale L. Oyeyemi
- Ronald Ssenyonga
- Vincent O. Onywera
journal: PLOS Global Public Health
year: 2021
pmcid: PMC10021676
doi: 10.1371/journal.pgph.0000089
license: CC BY 4.0
---
# Parental-perceived home and neighborhood environmental correlates of accelerometer-measured physical activity among school-going children in Uganda
## Abstract
The benefits of physical activity (PA) on children’s health and well-being are well established. However, many children do not meet the PA recommendations, increasing their risk of being overweight, obese, and non-communicable diseases. Environmental characteristics of homes and neighborhoods may constrain a child’s ability to engage in PA, but evidence is needed to inform country-specific interventions in understudied low-income countries. This study assessed the associations between parental-perceived home and neighbourhood, built environment characteristics, and moderate-to-vigorous physical activity (MVPA) among children in Kampala city, Uganda. In this cross-sectional study, data were obtained from 256 children ($55.5\%$ girls) aged between 10 and 12 years and their parents. Children’s MVPA was measured using waist-worn ActiGraph accelerometers. The environments were assessed using a valid self-reported parent survey. Linear regression models with standard errors (clusters) were used to analyze the relationship between environmental variables and children’s MVPA. Sex-specific relationships were assessed using sex-stratified models. Play equipment at home (β = -2.37, $p \leq 0.001$; unexpected direction), residential density (β = 2.70, $p \leq 0.05$), and crime safety (β = -5.29, $p \leq 0.05$; unexpected direction) were associated with children’s MVPA. The sex-specific analyses revealed more inconsistent patterns of results with a higher perception of land use mix associated with less MVPA in girls (irrespective of school type attended), and higher perceptions of sidewalk infrastructure (β = -12.01, $p \leq 0.05$) and walking and cycling infrastructure (β = -14.72, $p \leq 0.05$) associated with less MVPA in girls attending public schools only. A better perception of crime safety was associated with less MVPA among boys and girls attending private schools (β = -3.80, $p \leq 0.05$). Few environmental characteristics were related to children’s MVPA in Uganda, and findings were largely inconsistent, especially among girls. Future studies are needed to understand the ecological determinants of health-related PA behaviors among children in Uganda.
## Introduction
Childhood physical activity (PA) is a lifestyle behavior that protects against various non-communicable diseases (NCDs) and tracks through adolescence and into adulthood [1, 2]. However, recent global estimates reveal that more than $80\%$ of school-going children do not engage in the recommended daily 60 minutes or more of moderate-to-vigorous physical activity (MVPA), putting them at risk of NCDs and premature mortality later in life [3]. Worldwide estimates showed that physical inactivity (PI) exceeded 53 billion dollars in direct health care costs in 2013 and was responsible for 3.2 million deaths [4]. Therefore, reducing PI is important to public health, as specified in the global efforts to reduce PI by $15\%$ by 2030 [5]. Therefore, it is necessary to identify correlates of children’s PA to guide informed intervention strategies and policies aimed at reducing PI among children [4]. PA is influenced by various factors at multiple levels, including individual, interpersonal, environmental, and policy [5]. Although there is evidence on the importance of individual characteristics (e.g., age, sex, weight status) on children’s PA [3, 4, 6–11], such socio-demographic factors cannot fully explain the insufficient PA exhibited by the majority of children. Interventions based on such evidence may only benefit a few individuals who deliberately want to be active and have generated limited effects [12].
Socio-ecological models recognize that PA behavior is influenced by the environment in which it occurs [7]. Home and neighborhood environments are two components of the PA environment relevant to children’s active lifestyle. The home environment plays a major role in shaping and promoting children’s PA through social support, enforcement of rules for PA, and creating a healthy home environment that supports PA through the provision of play equipment and limiting access to media equipment [13–15]. Parents and guardians play a vital role in influencing children’s PA because they arrange home space and determine the equipment that is available for children’s use [15]. Parents can support their children’s PA by encouraging, facilitating, modeling, spectating/supervision, and co-participation [14, 16, 17]. Rules for PA are also a means through which parents control the home environment; parental constraints in the form of rules may lower children’s MVPA [14]. In addition, playing and electronic and media equipment at home have been considered as important correlates of children’s PA [14, 15, 18–21].
The neighborhood environment includes all places built or designed by humans, such as buildings, roads, recreation facilities, walking/cycling infrastructure, and urban design, as well as fewer tangible factors, such as safety [22]. PA-friendly neighborhoods make it easy, accessible, safe, and comfortable for children to be active [23]. Review studies have concluded that the presence of recreational facilities, such as parks, playgrounds [24–28], and neighborhood walking/cycling infrastructure [27], are positively correlated with children’s PA. In a systematic review and meta-analysis on street connectivity and PA, Jia et al. found that higher street connectivity predicted higher levels of MVPA in children [29]. Moreover, low street connectivity (cul-de-sacs) and the esthetic quality of neighborhood surroundings have been associated with children’s PA [30]. A higher perception of community safety is also correlated with higher PA [31]. Nonetheless, parental concern about neighborhood safety (crime, traffic, personal, stranger danger) may lead to parents restricting their children’s use of active travel modes to neighborhood destinations, including school, outdoor play, and independent mobility, thus reducing their PA [14, 23].
However, this evidence has been largely informed by studies from high-income countries (HICs); translating findings from these studies to Sub-Saharan Africa (SSA) where the built environment and culture are different may be a challenge [6, 27, 28, 32]. Very few studies among school-going children have been conducted in SSA, and the results of these studies are inconsistent. For example, among Kenyan children aged 9 to 11 years, parental perceptions of positive neighborhood social interactions, safety, and connectivity were associated with MVPA compliance [33]. While in South Africa, among 9-to 10-year-old children, parental perceptions of the neighborhood environment were not related to children’s MVPA [34]. Kampala is transitioning to a vibrant, attractive, and sustainable city through retrofitting and expansion of the city to accommodate the rapidly growing urban population [35]. Exploring the associations between children’s PA and parental perceptions of the home and neighborhood environments in Kampala, *Uganda is* a timely issue because the results of the current study may inform the city’s on-going re-design.
Exploring the associations between children’s accelerometer-measured PA and parental perceptions of the home and neighborhood environmental attributes in Uganda may address the problem of quality PA data noted in many low-income communities (LICs) and provide country-specific evidence that can inform effective interventions and policies. Therefore, the aim of this study was to explore the relationships between the parental-perceived home and neighborhood environment with accelerometer-measured MVPA among school-going children in Uganda.
## Design and participants’ recruitment
This cross-sectional study used a multistage sampling method to recruit a gender-balanced sample of 600 children aged between 10 and 12 years from mixed-day primary schools across Uganda’s capital city, Kampala. Kampala comprises five administrative divisions (Central, Nakawa, Rubaga, Kawempe, and Makindye) covering an area of 189 km2 and a population of 1.5 million inhabitants [36]. The Uganda National Council of Science and Technology (SS 4340) and Kenyatta University Ethical Review Board (PKU/$\frac{619}{703}$) granted ethical approval for this study. Divisions were the primary sampling unit; two divisions, namely central and Nakawa, were randomly selected. The second sampling unit was schools. Because of the variability in socioeconomic status (SES) between schools in Kampala (public schools represent the lower socioeconomic strata, and private schools represent the high socioeconomic strata based on the fee structure), four private and three public schools were selected for public and private school attendance. The third sampling unit was classes in the selected schools that best corresponded to 10- to 12-year-old children. At least 70 to 100 children in grades 5–7 were selected from each school. A total of 600 child-parent pairs received survey packages that contained parent and child information sheets, consent and assent forms, and a parental questionnaire. Written informed parental consent and child assent were obtained from all participants. To be eligible, children had to be 10 to 12 years old, residents of Kampala, and without physical disability. Children attending boarding schools were not included. Only 256 child-parent pairs provided complete consent and the data required for the study (response rate, $42.6\%$). We measured the children’s MVPA using an accelerometer. Parents completed a survey that assessed home and neighborhood environment attributes and sociodemographics. The survey was adapted from the Neighborhood Impact on Kids study [18] and the Neighborhood Environment Walkability Scale for Africa (NEWS-Africa) [37] (see S1 File).
## Children’s physical activity
Children wore the ActiGraph GT3X+ accelerometer (ActiGraph LLC, Pensacola, Florida, USA) around their waist for seven consecutive days, including two weekend days for 24 hours. Data were collected at a sample rate of 80 Hz with an epoch of 1 s and then aggregated into 15s epochs. Accelerometer data were processed using ActiLife software (Version 6.13.3). The Sadeh algorithm was used to identify sleep time and determine the wake-wear time [38]. Non-wear time within a day was set at 20 consecutive minutes of 0 counts. The analysis was restricted to data from children who had at least four days, including one weekend day, with ten or more waking wear hours each day. Evenson’s age-specific cut points (≥574 counts/15 s) were used to generate the average daily minutes of MVPA [39].
## Home environment
The home environment was assessed using survey items completed by the parent on the presence of play equipment, child’s bedroom and personal media equipment, parental rules for PA, and parental social support [18]. The items demonstrated good-to-excellent test-retest reliability (ICC = 0.51–0.96) [40]. Parents reported support for their child’s PA from a series of questions, including how often they encouraged children to be active, provided transport to places to do PA or sport, watched children engage in PA, and participated in PA with their child. The questions were rated on a 5-point Likert scale ranging from “none” to “daily.” The four items were averaged to create the social support scale used for the analysis. The parental rules for PA score summed “yes” responses on 15 rules that elicited a Yes/No response from parents. For example, “stay close/within sight of the house/parent” to “Respect others (particularly adults).” Parents reported whether their child had the following play equipment at home: bike, basketball hoop, jump rope, active video game (e.g., Wii) sports equipment (e.g., balls, racquets, bats, and sticks), skateboard/roller skaters/scooter, fixed play equipment (e.g., swings), home aerobic equipment (e.g., treadmill, cycle, cross-trainer), weightlifting equipment (e.g., free weights, exercise balls), yoga/exercise mat, recreational room, trampoline, and stairs. A play equipment score was generated using the sum of all the individual play equipment (range, 0 to 13). Parents also reported whether their child’s bedroom contained the following media and electronic items: TV, computer, VCR or DVD player, and video game system (e.g., x-box, play station). A score was generated using the sum of electronic and media items in the child’s bedroom (range 0 to 4). Parents/guardians further reported whether the child had personal media and electronic items, such as a cell phone/2-way radio, handheld video game players (e.g., Sony psp), and/or a music player (e.g., MP3, I Pod). The sum of the child’s personal media and electronic equipment was generated (range, 0 to 3).
## Neighborhood environment
The NEWS-Africa [36] was used to assess parents’ perception of their neighborhood environment attributes that may support children’s PA. NEWS-Africa was adapted to the African context to help build on neighborhood-built environment research, and the 76-item instrument demonstrated excellent (ICCs >$.75\%$) or good (ICCs = 0.60 to 0.74) test-retest reliability [36]. The NEWS-*Africa is* organized into 14 subscales representing residential density (1 item), land use mix-diversity {destinations} (21 items), land use mix-diversity {recreation} (4 items), land use mix access (7 items), street connectivity (5 items), side walk infrastructure (5 items), path infrastructure (2 items), crossing infrastructure (4 items), overall walking/cycling infrastructure (12 items), esthetics (8 items), traffic safety (6 items), crime safety (4 items), personal safety (3 items), and stranger danger (3 items). All the NEWS-Africa scales, except residential density and land use mix-diversity (destination and recreation), were rated on a 4-point Likert scale ranging from “strongly agree” to “strongly disagree.” The residential density scale was rated on a single unweighted scale. Land use mix diversity (destination and recreation) was rated on a 5-point scale (i.e., 1–5 min, 6–10 min, 11–20 min, 21–30 min, and 31+ min). Items were scored as recommended by the NEWS-Africa developers and reverse-coded where necessary [36]. For analysis, all the neighborhood environment attribute scores were averaged within each scale, and higher scores were expected to be associated with more MVPA.
## Covariates
Parents reported their child’s date of birth, sex, and time spent at the current residence.
## Data analysis
Data analysis was restricted to children who had complete parent surveys and accelerometer data ($$n = 256$$). We computed descriptive statistics, including percentages, means, and standard deviations, for children’s characteristics and PA. Independent t-tests were used to compare parental perceptions of home and neighborhood environment variables by children’s sex, school type, and compliance with PA guidelines. Logistic regression analyses with robust standard errors (clusters) were used to examine the associations between parent-perceived home and neighborhood environment variables and children’s MVPA. Overall and sex-stratified logistic regression analyses were conducted. STATA (v.14.2, StataCorp, Texas, USA) was used for all analyses, and significance was set at p ≤ 0.05.
## Participant characteristics
The study sample included 256 children. The majority of the children were between 10 and 11 years old ($71.5\%$), $58.6\%$ attended private school, and $55.9\%$ were girls. The children had lived in their current address for an average of 6.1 ± 3.3 years and spent an average of 56 ± 25.7 minutes/day in MVPA. Among the study participants, the boys (60.1 ± 28.2) spent significantly more time in MVPA than the girls did (58.2 ± 23.0). Public school (low SES schools) children engaged in 26 minutes more MVPA than private school (high SES schools) children. No significant differences in children’s characteristics were observed between those included and excluded from the analysis. The results are shown in Table 1.
**Table 1**
| Variable | n | % |
| --- | --- | --- |
| School type | School type | School type |
| Private (high SES) | 150 | 58.5 |
| Public (low SES) | 106 | 41.4 |
| Sex | Sex | Sex |
| Male | 113 | 44.1 |
| Female | 143 | 55.9 |
| Age (years) | Age (years) | Age (years) |
| 10 | 88 | 34.8 |
| 11 | 94 | 36.7 |
| 12 | 74 | 28.5 |
| Time spent at the current address (Mean ±SD) | 6.1 ± 3.3 | |
| Average MVPA (minutes/day) | Average MVPA (minutes/day) | Average MVPA (minutes/day) |
| | Mean ± SD | p-Value |
| Overall | 56 ± 25.7 | |
| Sex | Sex | Sex |
| Female | 52.8 ± 23.0 | 0.023* |
| Male | 60.1 ± 28.2 | |
| School Type | School Type | School Type |
| Private (HSES) | 45.4 ± 17.8 | <0.001** |
| Public (LSES) | 71.2 ± 27.5 | |
| Age | Age | Age |
| 10 | 52.7 ± 20.9 | 0.268 |
| 11 | 56.5 ± 26.5 | |
| 12 | 59.6 ± 29.4 | |
## Home and neighborhood environment attributes and children’s characteristics (MVPA compliance, sex, and school type)
Table 2 shows the mean differences in perceived parental home and neighborhood environment attributes according to children’s MVPA categories. Parents of children who did not meet the MVPA recommendations reported more play equipment at home ($p \leq 0.001$) and higher perception of neighborhood crime safety ($$p \leq 0.022$$) but lower perception of residential density ($$p \leq 0.012$$) than their counterparts whose children met the MVPA recommendations. Differences in perceived parental home and neighborhood environment attributes by school type and sex are also provided in Table 2. Parents whose children attended private schools (high SES) reported significantly more rules for PA ($$p \leq 0.033$$), children’s personal media equipment ($$p \leq 0.002$$) and play equipment at home ($p \leq 0.001$) than parents whose children attended public schools (low SES). Attributes of the neighborhood environment also varied across school types and children’s sex. Parents of children in private (high SES) schools perceived higher crime safety relative to parents of children in public (low SES) schools ($$p \leq 0.012$$). Children in public (low SES) schools had parents who perceived a higher residential density ($p \leq 0.001$) and street connectivity ($$p \leq 0.024$$) compared to parents of children in private (high SES) schools. Parents of girls perceived a significantly higher residential density compared to parents of boys ($$p \leq 0.024$$).
**Table 2**
| Variable | School type Mean (SD) | School type Mean (SD).1 | P-value | Sex Mean (SD) | Sex Mean (SD).1 | P-value.1 | MVPA Mean (SD) | MVPA Mean (SD).1 | P- value |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Variable | Private (HSES) | Public (LSES) | | Male | Female | | Sufficient PA | Insufficient PA | |
| Home Level | Home Level | Home Level | Home Level | Home Level | Home Level | Home Level | Home Level | Home Level | Home Level |
| Parental support for PA | 2.7 (0.9) | 2.6 (1.1) | 0.411 | 2.6 (0.9) | 2.6 (1.0) | 0.725 | 2.5(1.0) | 2.7(1.0) | 0.121 |
| Parental rules for PA | 13.2(1.9) | 12.7(1.9) | 0.033* | 13.1 (1.7) | 12.9 (2.1) | 0.235 | 12.9(1.7) | 13.0(2.0) | 0.479 |
| Media equipment in child’s bedroom | 0.5 (1.0) | 0.5 (0.9) | 0.756 | 0.5 (1.0) | 0.5 (0.9) | 0.575 | 0.5(1.0) | 0.5(0.9) | 0.995 |
| Child’s personal media equipment | 0.7 (0.9) | 0.4 (0.7) | 0.002* | 0.6 (0.9) | 0.5 (0.8) | 0.203 | 0.4(0.8) | 0.6(0.9) | 0.055 |
| Play equipment at home | 4.8 (2.5) | 2.7 (2.1) | <0.001** | 4.2(2.5) | 3.6 (2.6) | 0.062 | 3.1(2.3) | 4.4(2.6) | <0.001** |
| Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level | Neighborhood Level |
| Residential density | 2.4 (1.2) | 3.8 (2.0) | <0.001** | 2.7 (1.6) | 3.2 (1.8) | 0.024* | 3.4(1.9) | 2.8(1.6) | 0.012* |
| Land use mix-diversity (destinations) | 2.7 (0.7) | 2.6 (0.8) | 0.176 | 2.6 (0.7) | 2.7 (0.7) | 0.413 | 2.7(0.8) | 2.7(0.7) | 0.618 |
| Land use mix-diversity (recreation) | 2.0 (1.0) | 2.0 (1.1) | 0.695 | 2.1 (1.0) | 2.0 (1.0) | 0.395 | 2.0(1.0) | 2.0(1.0) | 0.720 |
| Land use mix-access | 2.9 (0.6) | 2.9 (0.6) | 0.698 | 2.9 (0.6) | 2.9 (0.6) | 0.710 | 2.8(0.6) | 2.9(0.6) | 0.440 |
| Street connectivity | 2.8 (0.5) | 3.0 (0.6) | 0.024* | 2.9 (0.5) | 2.9 (0.6) | 0.834 | 2.9(0.6) | 2.8(0.6) | 0.297 |
| Sidewalks infrastructure | 2.3 (0.7) | 2.2 (0.7) | 0.077 | 2.3 (0.7) | 2.2 (0.7) | 0.069 | 2.2 (0.7) | 2.3 (0.7) | 0.236 |
| Crossing infrastructure | 2.0 (0.7) | 2.2 (0.8) | 0.086 | 2.0 (0.7) | 2.1 (0.7) | 0.697 | 2.0 (0.7) | 2.1 (0.7) | 0.458 |
| Paths infrastructure | 2.5 (0.8) | 2.6 (0.9) | 0.379 | 2.5 (0.9) | 2.6 (0.8) | 0.379 | 2.5 (0.8) | 2.5 (0.9) | 0.894 |
| Walking and cycling infrastructure | 2.2 (0.5) | 2.2 (0.6) | 0.276 | 2.2 (0.6) | 2.2 (0.6) | 0.708 | 2.1(0.5) | 2.2(0.6) | 0.357 |
| Aesthetics | 2.7 (0.6) | 2.6 (0.7) | 0.099 | 2.7 (0.6) | 2.6 (0.6) | 0.237 | 2.6(0.7) | 2.7(0.6) | 0.417 |
| Crime safety | 2.9 (0.8) | 2.6 (0.8) | 0.012* | 2.7 (0.8) | 2.8 (0.8) | 0.722 | 2.6(0.8) | 2.8(0.8) | 0.022* |
| Traffic safety | 2.5 (0.7) | 2.6 (0.7) | 0.431 | 2.6 (0.7) | 2.5 (0.7) | 0.912 | 2.6(0.7) | 2.5(0.7) | 0.931 |
| Personal safety | 2.7 (0.6) | 2.8 (0.6) | 0.491 | 2.7 (0.6) | 2.8 (0.6) | 0.523 | 2.7(0.6) | 2.8(0.6) | 0.472 |
| Stranger danger | 2.2 (0.9) | 2.0 (0.9) | 0.067 | 2.5 (0.6) | 2.5 (0.5) | 0.386 | 2.1 (0.1) | 1.9 (0.1) | 0.194 |
## Parental-perceived home and neighborhood-built environment correlates of children’s MVPA
In the overall model of the entire sample, at the home level, results showed that children spent less time in MVPA if their parents reported more play equipment at home (β = -2.37, $p \leq 0.001$). At the neighborhood level, significant positive associations were found between children’s MVPA and parental perceptions of high residential density (β = 2.70, $p \leq 0.05$). Conversely, negative associations were found between children’s MVPA and parental perceptions of high crime safety (β = -5.29, $p \leq 0.05$).
School-stratified models were created to identify correlates that may be unique to private (high SES) and public (low SES) school children (Table 3). The results of this study indicated that none of the home environment attributes were correlated with children’s MVPA in both private and public schools. At the neighborhood environment level, a higher perception of crime safety was associated with less MVPA among children in private (high SES) school (β = -3.80, $p \leq 0.05$). Table 3 also presents the sex-stratified models. Parental perception of higher land use mix accessibility was associated with less MVPA among girls, regardless of school type (school SES). Our results further revealed that for girls who attended public schools (low SES), higher parental perceptions of sidewalk infrastructure (β = -12.01, $p \leq 0.05$) and walking/cycling infrastructure (β = -14.72, $p \leq 0.05$) were associated with less MVPA. While none of the home environment variables were related to MVPA among girls, none of the home and neighborhood environment attributes were significantly related to MVPA among boys.
**Table 3**
| Overall | Overall.1 | School type (school SES) | School type (school SES).1 | Sex | Sex.1 | Sex.2 | Sex.3 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Overall | Overall | Private (HSES) | Public (LSES) | Boys | Boys | Girls | Girls |
| Overall | Overall | Private (HSES) | Public (LSES) | Private (HSES) | Public (LSES) | Private (HSES) | Public (LSES) |
| Home Environment | Home Environment | Home Environment | Home Environment | Home Environment | Home Environment | Home Environment | Home Environment |
| Parental support for PA | 5.30 (0.240) | 3.50 (0.404) | 9.88 (0.176) | 9.78 (0.140) | -3.58 (0.807) | -2.55 (0.622) | 10.48 (0.174) |
| Parental rules for PA | -1.48 (0.076) | -0.79 (0.249) | -0.34 (0.814) | -0.69 (0.585) | 0.60 (0.855) | -0.972 (0.270) | -1.00 (0.493) |
| Media equipment in child’s bedroom | -0.14 (0.934) | 0.46 (0.752) | -0.50 (0.870) | 0.77 (0.717) | 4.58 (0.353) | -0.35 (0.858) | -2.97 (0.398) |
| Child’s personal media equipment | -3.37 (0.081) | 0.46 (0.782) | -2.05 (0.577) | 1.42 (0.570) | 1.42 (0.823) | -1.11 (0.617) | -4.39 (0.283) |
| Play equipment at home | -2.37 (< 0.001) ** | 0.23 (0.698) | -1.68 (0.187) | 0.74 (0.433) | -0.27 (0.917) | -0.44 (0.545) | -2.07 (0.119) |
| Neighborhood Environment | Neighborhood Environment | Neighborhood Environment | Neighborhood Environment | Neighborhood Environment | Neighborhood Environment | Neighborhood Environment | Neighborhood Environment |
| Residential density | 2.70 (0.004) * | 1.13 (0.345) | -0.80 (0.550) | 2.04 (0.302) | -2.95 (0.191) | 0.82 (0.570) | 1.04 (0.494) |
| Land use mix-diversity (destinations) | -2.37 (0.279) | -1.24 (0.562) | -0.55 (0.870) | 1.72 (0.579) | 1.11 (0.865) | -5.45 (0.056) | 2.33 (0.530) |
| Land use mix-diversity (recreation) | -1.11 (0.487) | -0.93 (0.536) | -0.65 (0.801) | 0.07 (0.974) | 0.55 (0.901) | -2.08 (0.278) | -2.83 (0.326) |
| Land use mix-access | -4.24 (0.11) | -3.27 (0.176) | -6.86 (0.112) | 0.39 (0.916) | -1.94 (0.794) | -6.54 (0.031) * | -11.12 (0.026) * |
| Street connectivity | 2.87 (0.298) | -2.41 (0.372) | 2.11 (0.624) | -1.04 (0.826) | 13.63 (0.071) | -3.64 (0.234) | -3.50 (0.457) |
| Sidewalks infrastructure | -3.21 (0.168) | -0.48 (0.814) | -6.49 (0.108) | -0.44 (0.897) | -5.58 (0.359) | -1.06 (0.671) | -12.01 (0.015) * |
| Crossing infrastructure | -0.16 (0.942) | 0.49 (0.826) | -4.79 (0.161) | 6.37 (0.095) | -3.50 (0.510) | -3.20 (0.206) | -6.33 (0.120) |
| Paths infrastructure | -0.86 (0.652) | 0.52 (0.768) | -4.50 (0.144) | 1.92 (0.515) | -2.37 (0.603) | -0.35 (0.868) | -5.11 (0.181) |
| Walking and cycling infrastructure | -2.32 (0.423) | 0.27 (0.921) | -9.09 (0.050) | 3.89 (0.400) | -6.05 (0.374) | -2.28 (0.468) | -14.72 (0.010) * |
| Aesthetics | -2.54 (0.317) | 2.55 (0.285) | -4.13 (0.311) | 6.11 (0.106) | -5.32 (0.397) | -0.87 (0.766) | -6.17 (0.207) |
| Crime safety | -5.29 (0.006) * | -3.80 (0.038) * | -1.85 (0.562) | -3.61 (0.212) | 5.73 (0.290) | -3.71 (0.103) | -5.05 (0.154) |
| Traffic safety | 1.99 (0.375) | 0.12 (0.953) | 2.47 (0.506) | -0.52 (0.870) | 7.23 (0.229) | 0.963 (0.712) | -1.94 (0.650) |
| Personal safety | 0.09 (0.972) | 1.07 (0.649) | -3.53 (0.424) | 0.11 (0.976) | -1.69 (0.797) | 2.75 (0.348) | -6.85 (0.204) |
| Stranger danger | 2.84 (0.106) | -0.36 (0.825) | 3.33 (0.245) | 0.63 (0.788) | 1.85 (0.697) | -2.84 (0.221) | 0.78 (0.817) |
## Discussion
The current study investigated the associations between parental perceptions of home and neighborhood environment attributes and children’s accelerometer-measured MVPA in a sample of Ugandan children. We found few and inconsistent associations of the home and neighborhood environments with children’s MVPA, which replicated the patterns of evidence on this topic in LMICs [28]. To our knowledge, no study has been published on the relationship of the home and neighborhood environment on the PA behavior of children in Uganda.
This study’s main finding was that one of five home environment attributes and two of 14 neighborhood environment attributes were significantly associated with children’s MVPA in the full sample. However, two of the three significant associations were unexpected. At the home environment level, higher parental perception of play equipment at home was related to lower levels of children’s MVPA. Similarly, the International Study of Childhood Obesity, Lifestyle, and the Environment found an inverse association between higher perception of play equipment at home and children’s MVPA in Kenya (a low-income country), unlike the pattern in the HICs countries of Australia, Canada, and Finland, where more play equipment at home was related to more MVPA among children [41]. Perhaps, compared to children in HICs, the availability of play equipment at home may not be as important to children in LICs as having unrestricted access and time to use the equipment at home [42]. Social support is an important construct that can help explain children’s PA behavior at home [13, 15, 43], but consistent with previous LIC studies [6, 34, 41], we did not find associations between social support and children’s MVPA in the present study. Social support may be more important to children from HICs where children engage in more organized sports activities that require parents to pay for membership and transport their children to activity venues [41], unlike free play and transportation PA, which is predominant in LICs like Uganda [11].
At the neighborhood environment level, higher parental perception of safety from crime was unexpectedly related to lower levels of children’s MVPA, suggesting that higher parental perception of insecurity from crime leads to engagement in more MVPA among the children. The usually reported unexpected and inverse associations of crime safety with PA behaviors of children and adults in *Africa is* largely a reflection of environmental injustice, under conditions in which people have no choice but to walk for transportation to destinations and utilitarian purposes regardless of the pervasive insecurity from crime and traffic [44]. To address the unintended consequences of the circumstances in which children will have to be active despite the high perception of crime in the neighborhood, there is a recent call for “physical activity security” as an agenda for creating an enabling PA environment in LMICs and other highly inequitable settings [45].
The only correlation of children’s MVPA in the expected direction was the perception of higher residential density, replicating findings from international studies [6, 46]. High residential density neighborhoods have access to various destinations (friends’ homes, shops, or transit stops) and are well connected, providing more opportunities for children to walk, thus increasing PA [47, 48]. High residential density has also been associated with adolescent PA [49, 50] and adult PA [48], so this construct could be an important component of neighborhood walkability that is applicable across the age spectrum, even in Africa [51].
Sex-specific analyses provided inconsistent patterns of findings. Irrespective of school type, girls unexpectedly engaged in less MVPA when their parents perceived more access to various destinations, services, and transit stops. A review study also reported negative associations between access to various destinations and PA among children [25]. Girls attending public (low SES) schools engaged in less MVPA if their parents perceived more walking and cycling infrastructure, particularly sidewalks. Because public (low SES) schoolgirls are more likely to live in areas that are more compact with high residential density, this diminishes space for walking and cycling, reducing their PA [52]. For example, the description of the neighborhood environment characteristics in our study “S1 Table” revealed that most of the parents perceived the sidewalks to be indistinct from the road (vehicle traffic) and were not well maintained. Parents also perceived more informal walking or foot paths with no designated places to walk or cycle. These conditions of walking and cycling infrastructure may discourage girls from walking, thereby decreasing their PA.
Our findings raise questions about why girls participated in less MVPA despite parental perceptions of the presence of favorable neighborhood environment attributes. Cultural, social, and gender norms coupled with parental rules and concerns about unsupervised PA in the neighborhood may be critical factors. Cultural and social norms may initiate unfavorable gender stereotypes that limit girls’ social and behavioral expressions; for example, girls and boys may be offered distinct activities based on their gender, most commonly encouraging boys to play vigorously and girls to play quietly [53]. Literature also shows significant differences in independent mobility; boys experience far more freedom and spend more time in activity-enhancing environments than girls, particularly outdoors [46, 54, 55]. The time that children spend outdoors is consistently and positively correlated with their PA [56]. Other leisure time, particularly sedentary activities, such as engaging in screen-based activities (television, videogames, and computer use), may attract girls to stay at home rather than go out into the neighborhood [57]. On the other hand, the lack of association between parental-perceived neighborhood attributes and MVPA in boys may be due to the definition of the neighborhood in our study (the area surrounding the child’s home that they could walk to within 10 to 15 minutes). Since boys are reported to have a higher independent mobility, it is possible that environmental attributes outside the neighborhood may be associated with their MVPA [46]. In subsequent studies, it would be beneficial to examine the facilitators and barriers that girls encounter to achieve sufficient PA at home and in their neighborhoods.
## Strength and limitations
The primary strength of this study is the use of accelerometers to measure children’s MVPA in an understudied region. This study also provides preliminary evidence on the link between home and neighborhood environment attributes and PA behaviors of children in Uganda. However, the results of the current study should be interpreted with caution because of its cross-sectional study design, which limits the ability to make causal inferences. In addition, both the home and neighborhood environment characteristics were self-reported and, as such, may have been vulnerable to social desirability and bias. The low response rate and recruitment of participants from Kampala city limits the generalizability of the findings to other populations. The current study did not assess specific types of PA, such as leisure and transportation [58], and the home and neighborhood context of the MVPA outcomes [30, 34]. We also combined different items for the home and neighborhood environment variables into one scale, unlike other studies in which individual items were assessed [16, 33], which might have obscured the associations of the home and neighborhood environment with children’s MVPA. Previous studies have shown that the home environment is more likely to be correlated with sedentary behavior, which occurs predominantly at home, than with MVPA, which accumulates across the entire day in different settings [15, 59]. In addition, there was limited variability in parental perceptions of the home and neighborhood environments in the current study, which may have reduced the power to detect associations with children’s MVPA [60]. Therefore, the correlation of the home and neighborhood environment with children’s PA needs to be studied further with increased specificity and improved measures to improve the quality of evidence.
## Conclusion
Overall, very few home and neighborhood environment attributes were supported as correlates of MVPA among Ugandan children. The patterns of influence of the home and neighborhood environments on Ugandan children’s PA may be gender specific and different from those in the HICs. Future qualitative studies exploring the perceptions, facilitators, and barriers to PA participation among girls in LMICs are important since girls exhibit lower levels of PA than boys. More research is needed on correlations between the built environment and PA in Uganda, particularly using objective measures of the built environment and longitudinal cohort studies to better guide effective health promotion interventions and policies.
## References
1. Poitras VJ, Gray CE, Borghese MM, Carson V, Chaput JP, Janssen I. **Systematic review of the relationships between objectively measured physical activity and health indicators in school-aged children and youth**. *Appl Physiol Nutr Metab* (2016.0) **41** S197-239. DOI: 10.1139/apnm-2015-0663
2. Telama R, Yang X, Leskinen E, Kankaanpää A, Hirvensalo M, Tammelin T. **Tracking of physical activity from early childhood through youth into adulthood**. *Med Sci Sports Exerc* (2014.0) **46** 955-62. DOI: 10.1249/MSS.0000000000000181
3. Guthold R, Stevens GA, Riley LM, Bull FC. **Global trends in insufficient physical activity among adolescents: a pooled analysis of 298 population-based surveys with 1·6 million participants**. *Lancet Child Adolesc Health* (2020.0) **4** 23-35. DOI: 10.1016/S2352-4642(19)30323-2
4. Ding D, Lawson KD, Kolbe-Alexander TL, Finkelstein EA, Katzmarzyk PT, van Mechelen W. **The economic burden of physical inactivity: a global analysis of major non-communicable diseases**. *Lancet* (2016.0) **388** 1311-24. DOI: 10.1016/S0140-6736(16)30383-X
5. 5World Health Organization. Global action plan on physical activity 2018–2030: more active people for a healthier world. Geneva: World Health Organization; 2018.. *Global action plan on physical activity 2018–2030: more active people for a healthier world* (2018.0)
6. Bauman AE, Reis RS, Sallis JF, Wells JC, Loos RJF, Martin BW. **Correlates of physical activity: why are some people physically active and others not?**. *Lancet* (2012.0) **380** 258-71. DOI: 10.1016/S0140-6736(12)60735-1
7. Sallis F, Owen N, Fisher EB, Glanz K, Rimer BK, Vismanath K. *Health behaviour and health education: theory, research, and practice* (2008.0) 465-85
8. Farooq A, Martin A, Janssen X, Wilson MG, Gibson A-M, Hughes A. **Longitudinal changes in moderate-to-vigorous-intensity physical activity in children and adolescents: A systematic review and meta-analysis**. *Obes Rev* (2019.0) 1-15. PMID: 30230172
9. Schwarzfischer P, Gruszfeld D, Socha P, Luque V, Closa-Monasterolo R, Rousseaux D. **Longitudinal analysis of physical activity, sedentary behaviour, and anthropometric measures from ages 6 to 11 years**. *Int J Behav Nutr Phys Act* (2018.0) **15** 126. DOI: 10.1186/s12966-018-0756-3
10. Gomes TN, Katzmarzyk P, Hedeker D, Fogelholm M, Standage M, Onywera VO. **Correlates of compliance with recommended levels of physical activity in children**. *Sci. Rep* (2017.0) **7** 16507. DOI: 10.1038/s41598-017-16525-9
11. Muthuri SK, Wachira L-JM, Leblanc AG, Francis CE, Sampson M, Onywera VO. **Temporal trends and correlates of physical activity, sedentary behaviour, and physical fitness among school-aged children in sub-Saharan Africa: a systematic review**. *Int. J. Environ. Res. Public Health* (2014.0) **11** 3327-59. DOI: 10.3390/ijerph110303327
12. Brown HE, Atkin AJ, Panter J, Wong G, Chinapaw MJ, van Sluijs EM. **Family-based interventions to increase physical activity in children: a systematic review, meta-analysis, and realist synthesis**. *Obes Rev* (2016.0) **17** 345-360. DOI: 10.1111/obr.12362
13. Xu H, Wen LM, Rissel C. **Associations of parental influences with physical activity and screen time among young children: a systematic review**. (2015.0) **2015** 546925. DOI: 10.1155/2015/546925
14. Crawford D, Cleland V, Timperio A, Salmon J, Andrianopoulos N, Roberts R. **The longitudinal influence of home and neighbourhood environments on children’s body mass index and physical activity over 5 years: the CLAN study**. *Int J Obes (Lond)* (2010.0) **34** 1177-87. DOI: 10.1038/ijo.2010.57
15. Maitland C, Stratton G, Foster S, Braham R, Rosenberg M. **A place for play? The influence of the home physical environment on children’s physical activity and sedentary behaviour**. (2013.0) **10** 99. DOI: 10.1186/1479-5868-10-99
16. Liu Y, Zhang Y, Chen S, Zhang J, Guo Z, Chen P. **Associations between parental support for physical activity and moderate-to-vigorous physical activity among Chinese school children: A cross-sectional study**. *J Sport Health Sci* (2017.0) **6** 410-415. DOI: 10.1016/j.jshs.2017.09.008
17. Yao CA, Rhodes RE. **Parental correlates in child and adolescent physical activity: a meta-analysis**. *Int J Behav Nutr Phys Act* (2015.0) **12** 10. DOI: 10.1186/s12966-015-0163-y
18. Tandon P, Grow HM, Couch S, Glanz K, Sallis JF, Frank LD. **Physical and social home environment in relation to children’s overall and home-based physical activity and sedentary time**. *Prev Med* (2014.0) **66** 39-44. DOI: 10.1016/j.ypmed.2014.05.019
19. Noonan RJ, Boddy LM, Knowles ZR, Fairclough SJ. **Cross-sectional associations between high-deprivation home and neighbourhood environments, and health-related variables among Liverpool children**. (2016.0) **6** e008693. DOI: 10.1136/bmjopen-2015-008693
20. 20Lou, D. Sedentary behaviours, and youth: current trends and the impact on health. San Diego, CA: Active Living Research; 2014. www.activelivingresearch.org.
21. Sheldrick MP, Maitland C, Mackintosh KA, Rosenberg M, Griffith LJ, Fry R. **Associations between the Home Physical Environment and Children’s Home-Based Physical Activity and Sitting**. (2019.0) **16** 4178. DOI: 10.3390/ijerph16214178
22. Handy SL, Boarnet MG, Ewing R, Killingsworth RE. **How the built environment affects physical activity: views from urban planning**. (2002.0) **23** 64-73. DOI: 10.1016/s0749-3797(02)00475-0
23. Sallis JF, Floyd MF, Rodriguez DA, Saeden BE. **The role of built environments in physical activity, obesity, and CVD**. *Circulation* (2012.0) **125** 729-37. PMID: 22311885
24. Masoumi HE. **Associations of built environment and children’s physical activity: a narrative review**. *Rev Environ Health* (2017.0) **32** 315-331. DOI: 10.1515/reveh-2016-0046
25. Carlin A, Perchoux C, Puggina A, Aleksovka K, Buck C, Burns C. **A life course examination of the physical environmental determinants of physical activity behaviour: A "Determinants of Diet and Physical Activity" (DEDIPAC) umbrella systematic literature review**. *PLoS One* (2017.0) **12** e0182083. DOI: 10.1371/journal.pone.0182083
26. Smith M, Hosking J, Woodward A, Written K, MacMillan A, Field A. **Systematic literature review of built environment effects on physical activity and active transport—an update and new findings on health equity**. *Int J Behav Nutr Phys Act* (2017.0) **14** 158. DOI: 10.1186/s12966-017-0613-9
27. Day K.. **Physical Environment Correlates of Physical Activity in Developing Countries: A Review**. *J Phys Act Health* (2018.0) **15** 303-314. DOI: 10.1123/jpah.2017-0184
28. Elshahat S, O’Rorke M, Adlakha D. **Built environment correlates of physical activity in low- and middle- income countries: a systematic review**. *PLoS One* (2020.0) **15** e0230454. DOI: 10.1371/journal.pone.0230454
29. Jia P, Zou Y, Wu Z, Zhang D, Wu T, Smith M. **Street connectivity, physical activity, and childhood obesity: A systematic review and meta-analysis**. *Obes Rev* (2019.0). DOI: 10.1111/obr.12943
30. Tappe KA, Glanz K, Sallis JF, Zhou C, Saelens BE. **Children’s physical activity and parents’ perception of the neighbourhood environment: neighbourhood impact on kids study**. *Int J Behav Nutr Phys Act* (2013.0) **10** 39. DOI: 10.1186/1479-5868-10-39
31. Weinberg D, Stevens GWJM, Bucksch J, Inchley J, de Looze M. **Do country-level environmental factors explain cross-national variation in adolescent physical activity? A multilevel study in 29 European countries**. (2019.0) **19** 680. DOI: 10.1186/s12889-019-6908-9
32. Giles-Corti B, Vernez-Moudon A, Reis R, Turrell G, Dannenberg AL, Badland H. **City planning and population health: a global challenge**. *Lancet* (2016.0) **388** 2912-2924. DOI: 10.1016/S0140-6736(16)30066-6
33. Muthuri SK, Wachira LJ, Onywera VO, Tremblay MS. **Associations Between Parental Perceptions of the Neighbourhood Environment and Childhood Physical Activity: Results from ISCOLE-Kenya**. *J Phys Act Health* (2016.0) **13** 333-343. DOI: 10.1123/jpah.2014-0595
34. Uys M, Broyles ST, Draper CE, Hendericks S, Rae D, Naidoo N. **Perceived and objective neighbourhood support for outside of school physical activity in South African children**. *BMC Public Health* (2016.0) **16** 462. DOI: 10.1186/s12889-016-2860-0
35. 35Kampala Capital City Authority (KCCA): Strategic plan 2014/15–2018/19. Laying a foundation for Kampala city transformation. Retrieved from http://www.kcca.go.ug
36. 36Uganda Bureau of Statistics (UBOS). Uganda national household survey 2016/2017. Retrieved from https://www.ubos.org/wp-content/uploads/publications/03_20182016_UNHS_FINAL_REPORT.pdf
37. Oyeyemi AL, Kasoma SS, Onywera VO, Assah F, Adedoyin RA, Conway TL. **NEWS for Africa: adaptation and reliability of a built environment questionnaire for physical activity in seven African countries**. *Int J Behav Nutr Phys Act* (2016.0) **13** 33. DOI: 10.1186/s12966-016-0357-y
38. Sadeh A, Sharkey KM, Carskadon MA. **Activity-based sleep wake identification: An empirical test of methodological issues**. *Sleep* (1994.0) **17** 201-207. DOI: 10.1093/sleep/17.3.201
39. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG. **Calibration of two objective measures of physical activity for children**. *J Sports Sci* (2008.0) **26** 1557-65. DOI: 10.1080/02640410802334196
40. 40Joe L, Carlson J: Active Where? Individual Item Reliability Report.: Active Living Research; 2010. http://www.drjamessallis.sdsu.edu/measures.html
41. Harrington DM, Gillison F, Broyles ST, Chaput J-P, Fogelholm M, Hu G. **Household-level correlates of children’s physical activity levels in and across 12 countries**. *Obesity (Silver Spring)* (2016.0) **24** 2150-2157. DOI: 10.1002/oby.21618
42. Sirard JR, Laska MN, Patnode CD, Farbakhsh K, Lytle LA. **Adolescent physical activity and screen time: associations with the physical home environment**. *Int J Behav Nutr Phys Act* (2010.0) **7** 82. DOI: 10.1186/1479-5868-7-82
43. Verloigne M, Van Lippevelde W, Maes L, Brug J, De Bourdeaudhuij I. **Family- and school-based correlates of energy balance-related behaviours in 10-12-year-old children: a systematic review within the ENERGY (European Energy balance Research to prevent excessive weight Gain among Youth) project**. *Public Health Nutr* (2012.0) **15** 1380-1395. DOI: 10.1017/S1368980011003168
44. Oyeyemi AL, Kolbe-Alexander TL, Lambert EV, Siefken K, Ramirez A, Schulenkorf N. *Physical activity in low- and middle-income countries* (2020.0)
45. Lambert EV, Kolbe-Alexander T, Adlakha D, Oyeyemi A, Anokye NK, Goenka S. **Making the case for ’physical activity security’: the 2020 WHO guidelines on physical activity and sedentary behaviour from a Global South perspective**. *Br J Sports Med* (2020.0) **54** 1447-1448. DOI: 10.1136/bjsports-2020-103524
46. De Meester F, Van Dyck D, De Bourdeaudhuij I, Cardon G. **Parental perceived neighbourhood attributes: associations with active transport and physical activity among 10-12-year-old children and the mediating role of independent mobility**. *BMC Public Health* (2014.0) **14** 631. DOI: 10.1186/1471-2458-14-631
47. Ewing R, Cervero R. **Travel, and the built environment: a meta-analysis**. *J Am Plann Assoc* (2010.0) **76** 265-94
48. Sallis JF, Cerin E, Conway TL, Adams MA, Frank LD, Pratt M. **Physical activity in relation to urban environments in 14 cities worldwide: a cross-sectional study**. *Lancet* (2016.0) **387** 2207-17. DOI: 10.1016/S0140-6736(15)01284-2
49. van Loon J, Frank LD, Nettlefold L, Naylor PJ. **Youth physical activity and the neighbourhood environment: examining correlates and the role of neighbourhood definition**. *Soc Sci Med* (2014.0) **104** 107-15. DOI: 10.1016/j.socscimed.2013.12.013
50. Hinckson E, Cerin E, Mavoa S, Smith M, Badland H, Stewart T. **Associations of the perceived and objective neighbourhood environment with physical activity and sedentary time in New Zealand adolescents**. *Int J Behav Nutr Phys Act* (2017.0) **14** 145. DOI: 10.1186/s12966-017-0597-5
51. Oyeyemi AL, Conway TL, Adedoyin RA, Akinroye KK, Aryeetey R, Assah F. **Construct Validity of the Neighbourhood Environment Walkability Scale for Africa**. *Med Sci Sports Exerc* (2017.0) **49** 482-491. DOI: 10.1249/MSS.0000000000001131
52. Devarajan R, Prabhakaran D, Goenka S. **Built environment for physical activity-An urban barometer, surveillance, and monitoring**. *Obes Rev* (2020.0) **21** e12938. DOI: 10.1111/obr.12938
53. Oyeyemi AL, Ishaku CM, Deforche B, Oyeyemi AY, De Bourdeaudhuij I, Van Dyck D. **Perception of built environmental factors and physical activity among adolescents in Nigeria**. *Int J Behav Nutr Phys Act* (2014.0) **11** 56. DOI: 10.1186/1479-5868-11-56
54. Carver A, Timperio A, Crawford D. **Playing it safe: the influence of neighbourhood safety on children’s physical activity. A review**. *Health Place* (2008.0) **14** 217-227. DOI: 10.1016/j.healthplace.2007.06.004
55. Faulkner G, Mitra R, Buliung R, Fusco C, Stone M. **Children’s outdoor playtime, physical activity, and parental perceptions of the neighbourhood environment**. *International Journal of Play* (2015.0) **4** 84-97. DOI: 10.1080/21594937.2015.1017303
56. Kneeshaw-Price S, Saelens BE, Sallis JF, Glanz K, Frank LD, Kerr J. **Children’s objective physical activity by location: why the neighbourhood matters**. *Pediatr Exerc Sci* (2013.0) **25** 468-86. DOI: 10.1123/pes.25.3.468
57. Sterdt E, Liersch S, Walter U. **Correlates of physical activity of children and adolescents: A systematic review of reviews**. *Health Education Journal* (2014.0) **73** 72-89. DOI: 10.1177/0017896912469578
58. Giles-Corti B, Timperio A, Bull F, Pikora T. **Understanding physical activity environmental correlates: increased specificity for ecological models**. *Exerc Sport Sci Rev* (2005.0) **33** 175-181. DOI: 10.1097/00003677-200510000-00005
59. Prince SA, Butler GP, Rao DP, Thompson W. **Evidence synthesis—Where are children and adults physically active and sedentary?—a rapid review of location-based studies**. *Health Promot Chronic Dis Prev Can* (2019.0) **39** 67-103. DOI: 10.24095/hpcdp.39.3.01
60. Sallis JF, Prochaska JJ, Taylor WC. **A review of correlates of physical activity of children and adolescents**. *Med Sci Sports Exerc* (2000.0) **32** 963-975. DOI: 10.1097/00005768-200005000-00014
|
---
title: Physical activity engagement in Eldoret, Kenya, during COVID-19 pandemic
authors:
- Karani Magutah
- Grace Mbuthia
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021677
doi: 10.1371/journal.pgph.0000339
license: CC BY 4.0
---
# Physical activity engagement in Eldoret, Kenya, during COVID-19 pandemic
## Abstract
The World Health Organization (WHO) recommends that individuals of all ages participate in regular physical activity (PA) for optimal health and to support with the control of multiple non-communicable diseases. In Kenya however, involvement in PA across the general population is low and there is an increase in sedentary lifestyles in both rural and urban areas. An inverse relationship exists between socioeconomic status and involvement in PA. The novel COVID-19 ushered in associated control measures to limit the spread of the virus. These measures included staying at home, social distancing, and closure of physical spaces such as gyms, public parks, sports grounds, outdoor playing areas and schools. The impact was immediate, impacting patterns and routines of PA in Kenya. The primary aim of this study was to verify if COVID-19 affected PA prevalence and patterns amongst adults in Eldoret, Kenya. The secondary aim was to ascertain if the modification in behaviour is consistent amongst individuals from different socioeconomic backgrounds. We used a cross-sectional study to examine self-reported PA data amongst 404 participants. All participants were ≥18 years and resided in Eldoret, Kenya. Data were collected using a self-administered, structured questionnaire adapted from the WHO Global Physical Activity Questionnaire (WHO GPAQ). The characteristics of participants’ is summarized using descriptive statistics, and bivariate analyses for measures of associations of variables was done using Chi-squared and Fishers exact tests. Binary logistic regressions were performed to adjust for the various factors and report associations between variables. The p-value considered for significant differences was set at <0.05. Participants in this study had mean age of 30.2±9.8 years. Almost $90\%$ of the participants were not aware of the current WHO guidelines on PA, $9\%$ stopped PA engagement after COVID-19 was first reported in Kenya, and only $25\%$ continued regular PA. Less than half maintained PA intensity after the advent of COVID-19, with almost half reporting a drop. Males had a drop in time taken per PA session while females maintained session lengths after COVID-19 ($$p \leq 0.03$$). Males preferred gym-setup or mixed-type PA while females opted for indoor (home) aerobics before and after COVID-19 ($$p \leq 0.01$$, $$p \leq 0.02$$ respectively). Compared to males, females were less likely to achieve both vigorous- and moderate-intensity PA recommendations ($p \leq 0.01$ and $$p \leq 0.02$$ respectively). Zone of residence was associated with participation in aerobic PA ($$p \leq 0.04$$; $95\%$ CI = 0.02499–0.96086) and, similarly, level of education was associated with knowledge of WHO recommendations for PA ($$p \leq 0.01$$; $95\%$ CI = -1.7544 - -0.2070). A majority of the urban population of Eldoret, Kenya and especially those with lower level of education are unaware of WHO recommendations for PA, and $30\%$ of them have not engaged in any form of PA for many years. The majority that report involvement in PA do not achieve the WHO recommended threshold levels of PA. The results also indicated that COVID-19 has negatively affected intensity of PA, and that there has been an increase in time spent sitting/reclining amongst individuals in the higher socio-economic classes and specifically amongst females.
## Introduction
The World Health Organization (WHO) recommends regular physical activity (PA) across all age groups to prevent and manage non-communicable diseases and promote optimal health and wellness [1]. Physical activity data are plenty and a positive curvilinear relationship between involvement in PAs of various intensities and fitness has long been established [2]. Physical activity largely contributes to a reduction in premature mortality by controlling cardiovascular disease, various cancers, diabetes and also by promoting mental health [2, 3]. This in turn translates to better overall health for individuals. It is this knowledge that has informed the existing guidelines and recommendations and also the current pool of evidence suggesting that even different exercise regimes as one form of PA would achieve similar health benefits. According to the WHO, 150 minutes of moderate-intensity PA done weekly yields the same health benefit as 75 minutes of vigorous-intensity PA. Our recent studies found that bouts of moderate-intensity PAs that are as short as 7.5 minutes, if performed thrice daily so that their cumulative time equals the WHO recommendation, yields similar health benefits, presenting additional options of regimens of PA to choose from [1, 4, 5].
Despite these WHO recommendations or other proposed options, PA involvement in Kenya remains generally low. Sedentary lifestyles are on the rise, and high inactivity levels can be found in both rural and urban areas [6–9]. Sedentary lifestyle and physical inactivity is the fourth leading risk factor for mortality globally [1] and this is worrying because in Eldoret, Kenya, $82\%$ of elderly inhabitants do not participate in any known form of PA [10]. The start of the novel COVID-19 pandemic immediately led to control measures that entailed individuals having to social distance and stay at home. Due to the pre-existing low involvement levels of participation in outdoor PA, there was concern that the levels of PA would decline further [10]. There have been extensive debates and discussions on how to improve outdoor PA in the past, and there was now concern that even indoor home-based PAs may also be negatively impacted [1, 4, 5, 10–12]. While PA continued to be recommended during confinement associated with COVID-19 [13], the emerging data on how COVID-19 is affecting PA suggests that individuals are further reducing their participation in PA [14]. This is despite temporary recommendations suggesting home-based PAs during the COVID-19 confinement [15]. This drop has specifically been observed in older age groups and in individuals with non-communicable diseases such as diabetes [16, 17], and comes at a time when a rise in sedentary lifestyles and inactivity are associated with the increase in non-communicable diseases in Kenya and globally [6–9, 18–21]. We currently however do not have data on the effect of COVID-19 on PA in Kenya or from the eastern African region, and therefore with the absence of data are unable to paint an accurate picture of how Kenya compares globally in relation to PA changes during the pandemic.
Participation in PA has been low across all ages and populations locally and globally and while the quest to overcome this problem is unresolved [1, 4, 5, 10–12], there is a major concern that COVID-19 may reverse overall population fitness gains in general [22, 23]. Lower education and income levels have already been shown to be associated with lower involvement in PA in both males and females of all ages, and that an inverse relationship exists between social economic status and sedentary time, compounding the concern in our set up. Individuals in the lower socio economic strata have more sitting and screen-time compared to those of the higher socio economic statuses [24–28]. Local data is minimal however, and this paucity of data from lower socio-economic countries contribute the current recommendation of need for more work to adequately demonstrate how socioeconomic status impacts PA involvement [29].
Recently, new data showed that PA regimens involving shorter periods of activity that are repeated several times a day led to improved PA adherence in individuals from all socioeconomic statuses [4, 5, 30–33]. However, the start of COVID-19 mitigation measures such as physical isolation have introduced further barriers to participation in PA.
We are aware that stay-at-home-orders may negatively affect PA involvement, but we are unsure exactly how this has impacted our population’s participation in PA. Recent emerging studies have documented a decline in metabolic equivalent minutes of PA in United States of America and the United Kingdom [34–36]. Given the many adjustments in behavioural aspects of life during the pandemic, we envisage a similar drop in PA participation in Kenya. Currently, the Kenyan Ministry of Health (MOH) has instructed individuals over 58 years and those with underlying health conditions to work from home and to limit outdoor engagements. Earlier, the MOH had recommended a lockdown that entailed stay-at-home for most of the population [37].
In order to develop responsive health policies, it is important to conduct research to obtain data on the current PA participation rates one year since the first case of COVID-19 was identified in Kenya. A comprehensive set of data will explain how working / staying at home and limiting outdoor engagements have affected the general population and their participation in PA, and how this may affect lifestyle diseases for the population in the future. It is hoped that this study will provide timely evidence-based data that will facilitate interventions to mitigate a potential increase in non-communicable lifestyle diseases due to the COVID-19 measures aimed to mitigate the spread of the virus [22, 23, 34–36]. The results of this study will directly contribute to the global enquiry to develop common recommendations for PA during the pandemic that can be implemented as policy by the various stakeholders [38]. The primary aim of this study is to verify if COVID-19 has affected PA prevalence and patterns amongst adults in Eldoret, Kenya. The secondary aim is to verify if the modifications to PA is the same for the different sexes and if the modifications differ depending on individuals’ socio-economic backgrounds.
## Ethics statement
Ethical approval was granted by Moi Teaching and Referral Hospital / Moi University research ethics committee (MTRH/MU IREC), approval no. 0003800. All participants provided written informed consent for participation. Participants were assured of confidentiality and anonymity, and no identifiers were used throughout the study.
## Design
This was a cross-sectional study.
## Study population and site
Participants were adults (aged ≥18 years) residing in Eldoret town and its peri-urban area within a 10 kilometers radius from the central business district (CBD).
## Sample size
The prevalence of PA participation in *Eldoret is* $18\%$ [39]. This study is not only too old but was also conducted before the COVID-19 pandemic. Thus, we assumed PA involvement prevalence of 0.5, a $5\%$ level of precision, and a Z value of 1.96 corresponding to $95\%$ CI, yielding a sample size of 384 participants.
## Sampling procedure
Participants’ selection first entailed a random selection of the estates from where a systematic criterion was employed. For participating estates, we listed estates of Eldoret town that are within a radius of ten kilometers. From these estates, we used anecdotal allocation into 2 groups for the affluent and the less affluent estates based on the availability of basic infrastructure, housing cost and affordability. An individual’s income places them into certain socio-economic status largely collapsed into higher or lower economic status, and choice of residence is associated with this. Each of the two groups contributed half of the sample. A computerized random selection of four [4] estates from each category was done. We selected participants using systematic sampling based on information on the estimated number of total homes/households in the estate obtained from the village/estate elders. We determined sampling interval by first allocating the sample of 384 into the two categories equally. Thereafter, proportionate allocation based on total household numbers per estate was done, which determined the sampling interval for each of the estates by dividing the households number by the allocated sample for that estate. From the selected households, we sampled the first respondent that we encountered and although we had a replacement criterion if there was an ineligible household where respondents were below 18 years old, this never became necessary. This is summarized in Fig 1.
**Fig 1:** *A summary of participants’ recruitment.*
## Eligibility
The inclusion criteria was male and female individuals aged ≥18 years regardless of their current or previous PA history, residing within a 10-kilometer radius from Eldoret CBD, and without medical advice to keep off PAs.
## Data collection
Four trained research assistants collected data from 15th March to 14th April 2021 in private rooms or outdoors at participants’ homes. The collected data included bio-demographic characteristics and self-reported information regarding PA. Data were collected using a structured questionnaire adapted from the WHO Global Physical Activity Questionnaire (WHO GPAQ). The tool included any activity that raised heart rate during performance as classified using the WHO GPAQ generic showcards that were attached to it. Thus, PA was categorized into mild, moderate or vigorous-intensity. Participants also gave an estimate of the time they spent on each activity which yielded a cumulative weekly minutes of PA. The tool was designed to either be self-administered by participants who could read and write or to be interviewer-administered for illiterate participants.
## Data management and analysis
Stata version 13 (College Station, TX: StataCorp LP) was used for data entry and analysis. Univariate and bivariate analyses were performed. Continuous data such as age were summarized using means and their standard deviations while categorical data such as PA time-length were summarized using frequencies and percentages to depict participants’ characteristics based on their zones of residence described above. Bivariate analyses were based on the various PA-related variables studied per sex groups and age categorizations. These bivariate analyses for measures of associations were done using Chi-squared tests. Odds ratios and $95\%$ confidence intervals were reported when associations were significant for binary outcomes associated with PA. To report associated factors and adjust for the various confounders, logistic regressions at $95\%$ confidence interval were performed for effect of various independent variables on select binary outcomes. P values of <0.05 were considered statistically significant.
## Results
The mean age of the participants was 30.2±9.8 years (males 28.9±8.3 ($$n = 190$$); females 31.4±10.9 ($$n = 214$$)). One in every four participants had a minimum of secondary level of education and the majority were in the working class. The demographic characteristics are shown in Table 1.
**Table 1**
| Variable | Zone 0 (Lower income) Mean ± SD or n (%) | Zone 1 (Higher income) Mean ± SD or n (%) | Overall Mean ± SD or n (%) |
| --- | --- | --- | --- |
| Mean age (N = 404) | 30.6±10.3 (n = 202) | 29.9±9.3 (n = 202) | 30.2±9.8 (n = 404) |
| Mean age (N = 404) | [Males 29.0±8.5 (n = 94); Females 31.9±11.6 (n = 108)] | [Males 28.8±8.2 (n = 96); Females 30.8±10.2 (n = 106)] | [Males 28.9±8.3 (n = 190); Females 31.4±10.9 (n = 214)] |
| Education level (N = 400) | | | |
| | 4 (2) | 2 (1) | 6 (1.5) |
| Primary | 50 (25) | 13 (6.5) | 63 (15.75) |
| Secondary | 85 (42.5) | 58 (29) | 143 (35.75) |
| Tertiary | 61 (30.5) | 127 (63.5) | 188 (47) |
| Employment (N = 402) | | | |
| | 77 (38.3) | 83 (41.3) | 160 (39.8) |
| Self Employed | 87 (43.3) | 63 (31.3) | 150 (37.3) |
| Formally Employed | 37 (18.4) | 55 (27.4) | 92 (22.9) |
| Work entails physical exertion (N = 402) | | | |
| Yes | 77 (38.7) | 48 (24.9) | 125 (31.9) |
| No | 122 (61.3) | 145 (75.1) | 267 (68.1) |
## Physical activity awareness and patterns
Almost 9 in 10 participants were unaware of the current WHO guidelines on PA and the majority ($52.9\%$) of these respondents came from the lower-income zones. Nine percent of the participants reported that they had stopped exercising after COVID-19 was first reported in Kenya, and most of these individuals lived in lower-economic zones/estates. Based on reported session duration and number of days of PA per week, only $25\%$ of participants achieved the recommended threshold of weekly minutes of PAs, a majority of whom came from the lower-economic estates. Physical activity awareness and reported PA history are shown in Table 2 while patterns of PA are shown in Table 3.
Bivariate analyses showed that while individuals from lower-income estates were more likely to participate alone in PA ($$p \leq 0.01$$), in organized gym sessions (p = <0.001) and preferred outdoor running ($$p \leq 0.02$$), those from higher-income estates did PA either alone or with others, preferred indoor PA at home or in the gym, and performed a mixture of PAs which included park/estate running both before and after COVID-19. Those from high socio-economic estates preferred gym aerobics and indoors/home aerobics while those from lower socio-economic estates preferred walking. Before COVID-19, the lower socio-economic estate individuals preferred doing morning PAs as opposed to the evening as preferred by the higher socio-economic estates’ participants. Table 4 summarizes variables with significant differences between zones.
**Table 4**
| Unnamed: 0 | Pearson χ2 | P value |
| --- | --- | --- |
| Participation in hard physical labour | 8.62 | <0.001 |
| Normalcy of exercise (alone, group, combinations) | 14.2 | 0.01 |
| Preferred exercises location (gym/home/park/estate/mix) before COVID | 20.6 | <0.001 |
| Preferred exercises location (gym/home/park/estate/mix) currently | 16.6 | <0.001 |
| Preferred exercises (running/ball games/walking/gymnastics) currently | 9.57 | 0.02 |
| Preferred time-of-day for a workout before COVID (morning) | 8.58 | 0.04 |
Males were twice more likely to be aware of the WHO recommendations for PA compared to females (OR = 2.09 ($95\%$ CI 1.05–4.30); $$p \leq 0.02$$). Females tended to participate in PA alone or in combination with a family member whereas males preferred to exercise in groups or at the gym (p = <0.001). Females also preferred mild-to-moderate intensity PAs while males preferred vigorous-intensity (P = <0.001). The odds of achieving the recommendations for vigorous-intensity PA were 5.21 ($95\%$ CI 2.35–12.7) times higher in males than females. The odds of achieving moderate-intensity PA recommendations were also higher in males than females (OR 1.6 ($95\%$ CI 1.06–2.46). Before COVID-19, males had longer sessions of PA ($$p \leq 0.01$$) and this sex difference was lost after COVID-19. Males had a drop in time taken per PA session while females maintained session lengths after COVID-19 ($$p \leq 0.03$$). Males also preferred gym-setup or mixed-type PAs while females opted for indoor (home) aerobics before and after COVID-19 ($$p \leq 0.01$$, and $$p \leq 0.02$$ respectively). Further, both before and after COVID-19, males preferred running and ball game PAs as opposed to walking and indoor aerobics for females ($p \leq 0.001$ for both). Males also preferred planned (morning or evening) PAs while females participated in PA only as time became available both before and after COVID-19 ($$p \leq 0.02$$ and $$p \leq 0.03$$ respectively). Table 5 summarizes associations of sex and the various variables, with odds ratios provided for binary outcomes.
**Table 5**
| Unnamed: 0 | Pearson χ2 | P value | Odds Ratio (binary outcomes) |
| --- | --- | --- | --- |
| Knowledge of WHO exercise recommendation | 5.15 | 0.02 | 2.09 (1.05–4.30) |
| Participation in hard physical labour | 0.33 | 0.57 | |
| Whether currently active in aerobic exercises | 0.06 | 0.80 | |
| Normalcy of exercise (alone, group, combinations) | 15.7 | <0.001 | |
| Current exercise intensity | 22.5 | <0.001 | |
| Whether exercise intensity changed since COVID-19 | 5.08 | 0.08 | |
| Whether meeting 150 minutes moderate intensity exercise | 5.64 | 0.02 | 1.61 (1.06–2.46) |
| Whether meeting 75 minutes vigorous intensity exercise | 21.2 | <0.001 | 5.21 (2.35–12.7) |
| Regularity of aerobic exercises (current) | 0.6 | 0.90 | |
| Regularity of aerobic exercises (before COVID-19) | 2.0 | 0.57 | |
| Exercise session length | 9.12 | 0.01 | |
| Exercise session length (before COVID-19) | 5.88 | 0.05 | |
| Change in exercise session length | 7.3 | 0.03 | |
| Exercise location (gym/home/park/estate/mix) before COVID-19 | 14.9 | 0.01 | |
| Current exercises location (gym/home/park/estate/mix) | 12.0 | 0.02 | |
| Pre-COVID exercise type (running/ball games/walking/gymnastics) | 50.7 | <0.001 | |
| Preferred exercises (run/ball games/walk/gymnastics) currently | 36.9 | <0.001 | |
| Preferred day of workout before COVID-19 | 1.94 | 0.38 | |
| Preferred day of workout currently | 5.23 | 0.07 | |
| Preferred time-of-day for workout before COVID-19 | 10.3 | 0.02 | |
| Preferred time-of-day for workout currently | 8.9 | 0.03 | |
| Change in time spent sitting/reclining since COVID-19 | 0.79 | 0.68 | |
To determine the associated factors and degree of such associations, a binary logistic regression model to predict factors that affected current participation in aerobic PA while controlling for other confounders showed that only zone of residence had statistically significant association ($$p \leq 0.04$$; $95\%$ CI = 0.02499–0.96086) where individuals from lower socio-economic zones had higher participation. Age, sex, knowledge of WHO recommendation for PA and level education all were not significantly associated with participation in PA. Similarly for knowledge of WHO recommendations for PA, only level of education was associated ($$p \leq 0.01$$; $95\%$ CI = -1.7544 - -0.2070) such that individuals with higher education levels were more likely to be aware of existing recommendations compared to those with lower education. Variables such as zone of residence, age, and sex were not associated with knowledge of WHO recommendations for PA (all $p \leq 0.05$).
## Discussion
To the best of our knowledge, this is the first study to report on how COVID-19 has affected PA amongst individuals in Kenya. The current study found that 9 in 10 participants were unaware of the current WHO guidelines that recommend 150 or 75 weekly minutes of moderate- or vigorous-intensity PA, respectively [1]. Bio-demographic characteristics of age and sex had no statistical association with this knowledge and only level of education was statistically associated where individuals with higher education were more likely to be aware of existing recommendations for PA. Except for level of knowledge where the current study shows an association, our findings replicate what has already been shown elsewhere [40]. Our study was carried out in a region of Kenya renown for producing world-class athletes and this could probably explain why individuals with higher level of education may be aware of the recommendation for PA as they probably read about and follow on performance of such elite athletes, a feat individuals with less education may not study through or relate.
One in three participants had not engaged in any PA for years, which mirrors WHO reports that 25–$33\%$ of individuals worldwide do not engage in PAs at all [41]. A total of $68\%$ of participants considered themselves aerobically active, however only $30\%$ of these individuals actually achieved the recommended threshold of weekly minutes of PA. This translated to an overall of $25\%$ prevalence in attainment of PA recommendations. Amongst the PA achievers, the majority came from the lower-economic estates, tended to be males, and were performing moderate-intensity PAs. Zone of residence was statistically associated with participation in PA where individuals from lower-economic zones were more likely to be involved when compared to those from the higher-economic zones. The findings are consistent with data showing that lower, rather than higher income, is positively associated with higher PA involvement [41]. While our current work only examined individuals from one town in Kenya with differing socio-economic status, the findings do however mirror other studies comparing higher- and lower-income countries which show that lower-income individuals tend to be more active than their higher income counterparts [41]. There is however an ongoing debate in the literature with dissenting studies refuting this association [42].
While it has been shown that it is still possible for some to attain PA recommendations during the COVID-19 pandemic [43], $9\%$ of participants in our current study stopped exercising after COVID-19 was first reported in Kenya, and the majority of these participants came from lower economic zones/estates. Less than half of the participants maintained previous PA intensity after the advent of COVID-19, with almost half reporting a drop in activity. It should be noted that there was no change in preferred day-of-the-week for PA, PA type, session length or place of PA for those who maintained their activity levels during the pandemic. The time spent sitting increased for the majority of the participants, with only one-third maintaining pre-COVID-19 sitting time and only $8\%$ of participants reducing their sitting time, echoing studies elsewhere [44, 45]. Reviews of recent studies have highlighted that COVID-19 may have negatively impacted PA involvement as per the results of our study [46, 47]. Our study further added that the higher socio-economic strata was more adversely affected during COVID-19 as it pertain to PA levels, although there was an overall drop in PA levels across both higher and lower socioeconomic classes. Only $25\%$ of participants, who mostly hailed from lower economic estates, reported exercising regularly. These individuals from the lower-socioeconomic estates who participated in PA reported doing PA while alone, in organized outdoor-exercises (park or estate), and preferred to do walks. This differed from those from higher socio-economic status who preferred various combinations for whom to exercise with, and who, further, preferred PAs done indoors either at home or at a gym (or mix), and preferred aerobic activities both before and after COVID-19. It appears COVID-19 did not change preferences of where to do PA or with whom for all participants. The start of the novel COVID-19 did however affect the preferred time of day for PA. Before COVID-19, participants from higher socio-economic estates preferred evening PAs and those in the lower socio-economic estates tending to morning hours. However, after COVID-19, this association was lost. This could be associated with COVID-19 mitigation measures that have reduced opportunities for PAs due to the suspension of outdoor engagement prospects. We still however have insufficient data to fully demonstrate the entire impact of the pandemic on PA [44, 47–49].
Females were twice less likely than males to be aware of the WHO recommendations for PA, and, were more likely to participate alone in PA or in combination with a family member compared to group/gym sessions for males. Males were however more likely to have vigorous-intensity PAs compared to mild-to-moderate intensities for females and were 5 times more likely to achieve WHO recommendations for vigorous-intensity PA. They were also 1.6 times more likely to achieve moderate-intensity PA recommendations compared to their female counterparts. There however was no significant association between sex and awareness of WHO recommendations for PA or participation in the same. Previous research has shown that males reduced vigorous-intensity PA while females increased moderate-intensity PA after COVID-19 [45]. Our current work however, adds that males and females maintained their PA intensity preferences even with reported overall decline in PA involvement after COVID-19. Before COVID-19, being male was associated with longer average sessions of PA, but this ceased with the start of the pandemic resulting in a significant drop in time taken per PA session. This differed from females who managed to maintain their pre-COVID-19 session lengths. Recent research has shown that the pandemic has negatively impacted the types and intensities of PA with which individuals engage, and daily sitting time has increased [45]. Except for exercise intensity, there has not been any segregation based on sex and age that we are aware of [44, 45]. Our study attempts to segregate by sex, with additional variables. Concerning the venue for PAs, both males and females maintained their preference for gym set-up. Additionally, males seemed to maintain mixed-type PA, running and ball games while females maintained indoor (home) activities, walking and indoor aerobics. The current study also found that being male was associated with planned (either morning or evening) PA while being females was associated with PA only as time became available, without prior planning both before and after the start of COVID-19.
## Limitations
The cross-sectional design employed in the current study was unable to assess the actual effect of COVID-19 on PA participation. Our attempt to handle this may have introduced a recall bias that might have affected the data for the period before COVID-19. We attempted to reduce this by including only those questions we thought had minimal recall challenges, but we note that this may not have totally eliminated the bias.
## Conclusions
A majority of the urban population of Eldoret, Kenya and especially those with lower level of education were unaware of the WHO recommendations for PA and $30\%$ of them have not engaged in any PA for years. For those reporting participation in PA, the majority do not achieve the recommended WHO threshold levels. COVID-19 has reduced participation in PA and increased the time spent sitting/reclining especially for individuals in higher socio-economic class and for females.
## References
1. 1World Health Organization (2010) Global Recommendations on Physical Activity for Health.. *Global Recommendations on Physical Activity for Health* (2010.0)
2. Warburton DER, Bredin SSD. **Health benefits of physical activity: a systematic review of current systematic reviews**. *Curr Opin Cardiol* (2017.0) **32** 541-556. DOI: 10.1097/HCO.0000000000000437
3. Nuzum H, Stickel A, Corona M, Zeller M, Melrose RJ, Wilkins SS. **Potential Benefits of Physical Activity in MCI and Dementia.**. *Behav Neurol* (2020.0) **2020** 7807856. DOI: 10.1155/2020/7807856
4. Magutah K, Meiring R, Patel NB, Thairu K. **Effect of short and long moderate-intensity exercises in modifying cardiometabolic markers in sedentary Kenyans aged 50 years and above**. *BMJ Open Sport Exerc Med* (2018.0) **4** e000316. DOI: 10.1136/bmjsem-2017-000316
5. Magutah K, Patel NB, Thairu K. **Effect of moderate-intensity exercise bouts lasting <10 minutes on body composition in sedentary Kenyan adults aged >/ = 50 years**. *BMJ Open Sport Exerc Med* (2018.0) **4** e000403. DOI: 10.1136/bmjsem-2018-000403
6. Onywera VO, Adamo KB, Sheel AW, Waudo JN, Boit MK, Tremblay M. **Emerging evidence of the physical activity transition in Kenya**. *J Phys Act Health* (2012.0) **9** 554-562. DOI: 10.1123/jpah.9.4.554
7. Ojiambo RM, Easton C, Casajus JA, Konstabel K, Reilly JJ, Pitsiladis Y. **Effect of urbanization on objectively measured physical activity levels, sedentary time, and indices of adiposity in Kenyan adolescents**. *J Phys Act Health* (2012.0) **9** 115-123. DOI: 10.1123/jpah.9.1.115
8. Wachira LM, Muthuri SK, Ochola SA, Onywera VO, Tremblay MS. **Screen-based sedentary behaviour and adiposity among school children: Results from International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE)—Kenya**. *PLoS One* (2018.0) **13** e0199790. DOI: 10.1371/journal.pone.0199790
9. Ssewanyana D, Abubakar A, van Baar A, Mwangala PN, Newton CR. **Perspectives on Underlying Factors for Unhealthy Diet and Sedentary Lifestyle of Adolescents at a Kenyan Coastal Setting**. *Front Public Health* (2018.0) **6** 11. DOI: 10.3389/fpubh.2018.00011
10. Magutah K, Patel NB, Thairu K. **Majority of Elderly Sedentary Kenyans Show Unfavorable Body Composition and Cardio-Metabolic Fitness**. *J Aging Sci* (2016.0) **4** 160
11. Garber CE, Blissmer B, Deschenes MR, Franklin BA, Lamonte MJ, Lee IM. **American College of Sports Medicine position stand. Quantity and quality of exercise for developing and maintaining cardiorespiratory, musculoskeletal, and neuromotor fitness in apparently healthy adults: guidance for prescribing exercise**. *Med Sci Sports Exerc* (2011.0) **43** 1334-1359. DOI: 10.1249/MSS.0b013e318213fefb
12. Chodzko-Zajko WJ, Proctor DN, Fiatarone Singh MA, Minson CT, Nigg CR, Salem GJ. **American College of Sports Medicine position stand. Exercise and physical activity for older adults**. *Med Sci Sports Exerc* (2009.0) **41** 1510-1530. DOI: 10.1249/MSS.0b013e3181a0c95c
13. Polero P, Rebollo-Seco C, Adsuar JC, Pérez-Gómez J, Rojo-Ramos J, Manzano-Redondo F. **Physical Activity Recommendations during COVID-19: Narrative Review**. *Int J Environ Res Public Health* (2020.0) **18**. DOI: 10.3390/ijerph18010065
14. Nyenhuis SM, Greiwe J, Zeiger JS, Nanda A, Cooke A. **Exercise and Fitness in the Age of Social Distancing During the COVID-19 Pandemic**. *J Allergy Clin Immunol Pract* (2020.0) **8** 2152-2155. DOI: 10.1016/j.jaip.2020.04.039
15. Schwendinger F, Pocecco E. **Counteracting Physical Inactivity during the COVID-19 Pandemic: Evidence-Based Recommendations for Home-Based Exercise**. *Int J Environ Res Public Health* (2020.0) **17**. DOI: 10.3390/ijerph17113909
16. Ruiz-Roso MB, Knott-Torcal C, Matilla-Escalante DC, Garcimartín A, Sampedro-Nuñez MA, Dávalos A. **COVID-19 Lockdown and Changes of the Dietary Pattern and Physical Activity Habits in a Cohort of Patients with Type 2**. *Diabetes Mellitus. Nutrients* (2020.0) **12**
17. Ghosh A, Arora B, Gupta R, Anoop S, Misra A. **Effects of nationwide lockdown during COVID-19 epidemic on lifestyle and other medical issues of patients with type 2 diabetes in north India**. *Diabetes Metab Syndr* (2020.0) **14** 917-920. DOI: 10.1016/j.dsx.2020.05.044
18. BeLue R, Okoror TA, Iwelunmor J, Taylor KD, Degboe AN, Agyemang C. **An overview of cardiovascular risk factor burden in sub-Saharan African countries: a socio-cultural perspective**. *Global Health* (2009.0) **5** 10. DOI: 10.1186/1744-8603-5-10
19. Ikem I, Sumpio BE. **Cardiovascular disease: the new epidemic in sub-Saharan Africa**. *Vascular* (2011.0) **19** 301-307. DOI: 10.1258/vasc.2011.ra0049
20. van der Sande MA. **Cardiovascular disease in sub-Saharan Africa: a disaster waiting to happen**. *Neth J Med* (2003.0) **61** 32-36. PMID: 12735418
21. Hendriks ME, Wit FW, Roos MT, Brewster LM, Akande TM, de Beer IH. **Hypertension in sub-Saharan Africa: cross-sectional surveys in four rural and urban communities**. *PLoS One* (2012.0) **7** e32638. DOI: 10.1371/journal.pone.0032638
22. Flanagan EW, Beyl RA, Fearnbach SN, Altazan AD, Martin CK, Redman LM. **The Impact of COVID-19 Stay-At-Home Orders on Health Behaviors in Adults**. *Obesity (Silver Spring)* (2021.0) **29** 438-445. DOI: 10.1002/oby.23066
23. Almandoz JP, Xie L, Schellinger JN, Mathew MS, Gazda C, Ofori A. **Impact of COVID-19 stay-at-home orders on weight-related behaviours among patients with obesity**. *Clin Obes* (2020.0) **10** e12386. DOI: 10.1111/cob.12386
24. Zapata-Lamana R, Poblete-Valderrama F, Cigarroa I, Parra-Rizo MA. **The Practice of Vigorous Physical Activity Is Related to a Higher Educational Level and Income in Older Women**. *Int J Environ Res Public Health* (2021.0) **18**. DOI: 10.3390/ijerph182010815
25. Huikari S, Junttila H, Ala-Mursula L, Jämsä T, Korpelainen R, Miettunen J. **Leisure-time physical activity is associated with socio-economic status beyond income—Cross-sectional survey of the Northern Finland Birth Cohort 1966 study**. *Econ Hum Biol* (2021.0) **41** 100969. DOI: 10.1016/j.ehb.2020.100969
26. Nicolson G, Hayes C, Darker C. **Examining total and domain-specific sedentary behaviour using the socio-ecological model—a cross-sectional study of Irish adults**. *BMC Public Health* (2019.0) **19** 1155. DOI: 10.1186/s12889-019-7447-0
27. Puciato D, Rozpara M, Mynarski W, Oleśniewicz P, Markiewicz-Patkowska J, Dębska M. **Physical Activity of Working-Age People in View of Their Income Status**. *Biomed Res Int* (2018.0) **2018** 8298527. DOI: 10.1155/2018/8298527
28. Scholes S, Mindell JS. **Inequalities in participation and time spent in moderate-to-vigorous physical activity: a pooled analysis of the cross-sectional health surveys for England 2008, 2012, and 2016**. *BMC Public Health* (2020.0) **20** 361. DOI: 10.1186/s12889-020-08479-x
29. DiPietro L, Al-Ansari SS, Biddle SJH, Borodulin K, Bull FC, Buman MP. **Advancing the global physical activity agenda: recommendations for future research by the 2020 WHO physical activity and sedentary behavior guidelines development group**. *Int J Behav Nutr Phys Act* (2020.0) **17** 143. DOI: 10.1186/s12966-020-01042-2
30. Magutah K, Thairu K, Patel N. **Effect of short moderate intensity exercise bouts on cardiovascular function and maximal oxygen consumption in sedentary older adults**. *BMJ Open Sport Exerc Med* (2020.0) **6** e000672. DOI: 10.1136/bmjsem-2019-000672
31. Macfarlane DJ, Taylor LH, Cuddihy TF. **Very short intermittent vs continuous bouts of activity in sedentary adults**. *Prev Med* (2006.0) **43** 332-336. DOI: 10.1016/j.ypmed.2006.06.002
32. Miyashita M, Burns SF, Stensel DJ. **Accumulating short bouts of running reduces resting blood pressure in young normotensive/pre-hypertensive men**. *J Sports Sci* (2011.0) **29** 1473-1482. DOI: 10.1080/02640414.2011.593042
33. Quinn TJ, Klooster JR, Kenefick RW. **Two short, daily activity bouts vs. one long bout: are health and fitness improvements similar over twelve and twenty-four weeks?**. *J Strength Cond Res* (2006.0) **20** 130-135. DOI: 10.1519/R-16394.1
34. Fearnbach SN, Flanagan EW, Höchsmann C, Beyl RA, Altazan AD, Martin CK. **Factors Protecting against a Decline in Physical Activity during the COVID-19 Pandemic**. *Med Sci Sports Exerc* (2021.0). DOI: 10.1249/MSS.0000000000002602
35. McCarthy H, Potts HWW, Fisher A. **Physical Activity Behavior Before, During, and After COVID-19 Restrictions: Longitudinal Smartphone-Tracking Study of Adults in the United Kingdom**. *J Med Internet Res* (2021.0) **23** e23701. DOI: 10.2196/23701
36. Yang Y, Koenigstorfer J. **Determinants of physical activity maintenance during the Covid-19 pandemic: a focus on fitness apps**. *Transl Behav Med* (2020.0) **10** 835-842. DOI: 10.1093/tbm/ibaa086
37. 37Ministry of Health (2020) INTERIM GUIDELINES ON MANAGEMENT OFCOVID-19 IN KENYA.. *INTERIM GUIDELINES ON MANAGEMENT OFCOVID-19 IN KENYA* (2020.0)
38. Bentlage E, Ammar A, How D, Ahmed M, Trabelsi K, Chtourou H. **Practical Recommendations for Maintaining Active Lifestyle during the COVID-19 Pandemic: A Systematic Literature Review**. *Int J Environ Res Public Health* (2020.0) **17**. DOI: 10.3390/ijerph17176265
39. Nambakai JE, Kamau J, Amusa LO, Goon DT, Andanje M. **Factors influencing participation in physical exercise by the elderly in Eldoret West District, Kenya.**. *African Journalfor Physical, Health Education, Recreation and Dance (AJPHERD)* (2011.0) **17** 462-472
40. Wong MK, Cheng SYR, Chu TK, Lee CN, Liang J. **Hong Kong Chinese adults’ knowledge of exercise recommendations and attitudes towards exercise**. *BJGP Open 1: bjgpopen17X100929* (2017.0). DOI: 10.3399/bjgpopen17X100929
41. 41World Health Organization (2020) https://www.who.int/news-room/fact-sheets/detail/physical-activity.
42. Martins TCR, Pinho L, Brito M, Pena GDG, Silva RRV, Guimarães ALS. **Influence of socioeconomic status, age, body fat, and depressive symptoms on evel of physical activity in adults: a path analysis**. *Cien Saude Colet* (2020.0) **25** 3847-3855. DOI: 10.1590/1413-812320202510.24742018
43. Carvalho VO, Gois CO. **COVID-19 pandemic and home-based physical activity**. *J Allergy Clin Immunol Pract* (2020.0) **8** 2833-2834. DOI: 10.1016/j.jaip.2020.05.018
44. Ammar A, Brach M, Trabelsi K, Chtourou H, Boukhris O, Masmoudi L. **Effects of COVID-19 Home Confinement on Eating Behaviour and Physical Activity: Results of the ECLB-COVID19 International Online Survey**. *Nutrients* (2020.0) **12**. DOI: 10.3390/nu12061583
45. Castañeda-Babarro A, Arbillaga-Etxarri A, Gutiérrez-Santamaría B, Coca A. **Physical Activity Change during COVID-19 Confinement**. *Int J Environ Res Public Health* (2020.0) **17**. DOI: 10.3390/ijerph17186878
46. Freiberg A, Schubert M, Romero Starke K, Hegewald J, Seidler A. **A Rapid Review on the Influence of COVID-19 Lockdown and Quarantine Measures on Modifiable Cardiovascular Risk Factors in the General Population**. *Int J Environ Res Public Health* (2021.0) **18**. DOI: 10.3390/ijerph18168567
47. Clemente-Suárez VJ, Beltrán-Velasco AI, Ramos-Campo DJ, Mielgo-Ayuso J, Nikolaidis PA, Belando N. **Physical activity and COVID-19. The basis for an efficient intervention in times of COVID-19 pandemic**. *Physiol Behav* (2021.0) **244** 113667. DOI: 10.1016/j.physbeh.2021.113667
48. Hall G, Laddu DR, Phillips SA, Lavie CJ, Arena R. **A tale of two pandemics: How will COVID-19 and global trends in physical inactivity and sedentary behavior affect one another?**. *Prog Cardiovasc Dis* (2021.0) **64** 108-110. DOI: 10.1016/j.pcad.2020.04.005
49. Sonza A, Da Cunha de Sá-Caputo D, Bachur JA, Rodrigues de Araújo MDG, Valadares Trippo KVT, Ribeiro Nogueira da Gama D. **Brazil before and during COVID-19 pandemic: Impact on the practice and habits of physical exercise**. *Acta Biomed* (2020.0) **92** e2021027. DOI: 10.23750/abm.v92i1.10803
|
---
title: 'Assessing quality of care in maternity services in low and middle-income countries:
Development of a Maternity Patient Reported Outcome Measure'
authors:
- Fiona M. Dickinson
- Barbara Madaj
- Onesmus M. Muchemi
- Charles Ameh
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021686
doi: 10.1371/journal.pgph.0000062
license: CC BY 4.0
---
# Assessing quality of care in maternity services in low and middle-income countries: Development of a Maternity Patient Reported Outcome Measure
## Abstract
Globally, low and middle-income countries bear the greatest burden of maternal and newborn mortality. To reduce these high levels, the quality of care provided needs to be improved. This study aimed to develop a patient reported outcome measure for use in maternity services in low and middle-income countries, to facilitate improvements in quality of care. Semi-structured interviews and focus groups discussions were conducted with women who had recently given birth in selected healthcare facilities in Malawi and Kenya. Transcripts of these were analysed using a thematic approach and analytic codes applied. Draft outcomes were identified from the data, which were reviewed by a group of clinical experts and developed into a working copy of the Maternity Patient Reported Outcome Measure (MPROM). A further sample of new mothers were asked to evaluate the draft MPROM during cognitive debriefing interviews, and their views used to revise it to produce the final proposed measure. Eighty-three women were interviewed, and 44 women took part in 10 focus group discussions. An array of outcomes was identified from the data which were categorised under the domains of physical and psychological symptoms, social issues, and baby-related health outcomes. The draft outcomes were configured into 79 questions with answers provided using a five-point Likert scale. Minor revisions were made following cognitive debriefing interviews with nine women, to produce the final proposed MPROM. In conjunction with women from the target population and clinical experts, this study has developed what is believed to be the first condition-specific PROM suitable for assessing care quality in maternity services in low and middle-income countries. Following further validation studies, it is anticipated that this will be a useful tool in facilitating improvements in the quality of care provided to women giving birth in healthcare facilities in these settings.
## Maternal & newborn health globally
It is estimated that worldwide, nearly 300,000 women die each year giving birth [1], in addition to the approximately 4.5 million babies that are either stillborn or die within the first week of life [2, 3]. Of these, most occur in low and middle-income countries and, with good care, are preventable [4]. The targets for reducing maternal and newborn deaths included in the Millennium Development Goals (MDGs) were not achieved [5]. In order to achieve the new maternal health targets proposed in the Sustainable Development Goals (SDGs), improvements in the quality of care are essential [5].
## Quality of care
Previous policies targeting improvement in population health have largely focused on increasing the availability and access to healthcare services, with the assumption that they are of sufficient quality or that the quality will improve with availability. This, however, may not always be correct, and in some situations, patients avoid healthcare facilities with good justification [6].
Defining quality of care (QoC) has been highlighted by Donabedian [7] as a key challenge in addressing and improving it, and he questions whether it is the process of caring for patients or a goal or objective of the process. The framework developed by the WHO [8] divides quality of care into three aspects in line with the Donabedian model [7]: Structure ie. the health system; Process, combining the provision and experience of care as well as human and physical resources; and Outcome, which includes coverage of key practices, and people-centred and health outcomes. Combining the provision and receipt of care in one model highlights the relevance and importance of including the service user perspective in assessing healthcare.
Measuring care quality can facilitate the identification of instances of poor-quality care and assist in assessing the effectiveness of quality improvement activities. Various methods have been employed to measure QoC in maternal health, including direct observation, patient interviews and standards-based audits [9–11] but each comes with its own drawbacks, including the effect of conducting the research itself. Other methods such as reviewing medical records are largely reliant on the quality of the records, which in many resource-constrained settings, may be lacking [12]. An alternative measurement approach is the use of patient reported outcomes.
## Patient Reported Outcome Measures and their uses
Patient Reported Outcome Measures (PROMs) have been variously described but are essentially structured questionnaires that assess health or health-related quality of life outcomes recounted by the patient themselves. They are developed as a survey-type form, either paper-based or increasingly commonly, electronic. The data generated by PROMs has been used to standardise research outcomes, promote patient choice, allocate financial resources, and measure quality of care [13].
A 2019 literature review to identify existing PROMs suitable for assessing quality of care in maternity services, failed to locate any suitable tools [14]. Those that were available largely focussed on specific aspects of pregnancy or the postnatal period, such as hyperemesis gravidarum [15], gestational diabetes mellitus [16], obstetric and postpartum haemorrhage [17, 18] and postnatal depression [19]. The only PROM that explored pregnancy and childbirth more broadly was that developed by Symon et al [20], however this was an individualised tool, with outcomes specified by each woman. This lack of specificity was felt to be unsuitable for assessing the quality of care as the outcomes of interest at the time of completion might have little if anything to do with the care experienced whilst giving birth. Another limitation of the existing identified childbirth related PROMs was that they had all been developed in high-income countries (UK, USA, Netherlands, Poland, Australia and New Zealand), potentially limiting their applicability to LMICs, where $94\%$ of maternal deaths occur [21]. In addition, a search of grey literature identified the pregnancy and childbirth standard set [22] recommended by the International Consortium for Health Outcomes Measurement (ICHOM). This comprised of different measures (eg Edinburgh Postnatal Depression Scale), as well as administrative data (including mortality), a patient satisfaction scale and generic PROMs. These were described as patient centred and developed by clinicians and researchers in higher income countries, making the set inappropriate for our purposes.
Subsequently, the Obstetric quality of recovery score was published (ObsQoR-11) [23], a PROM designed to assess recovery from caesarean section, and based on the previous quality-of-recovery tool for general surgery. This was subsequently revised to form the ObsQoR-10, a slightly shorter version of the initial PROM and evaluated for use with women immediately following vaginal delivery [24]. Although these tools assessed aspects of recovery following childbirth, they were largely limited to the physical domain (items included pain, nausea, dizziness, mobilisation, and ability to feed baby without assistance), with the only psychological item relating to feeling ‘in control’. The time frame for response specified by the ObsQoR-11 and 10 was the previous 24 hours and the tool was administered on day 1 of the postnatal period. As is common with other tools described above, the ObsQoR-10 was developed and validated in high income countries (USA & UK). These factors would potentially limit the applicability of the ObsQoR-10 in a low-income setting, among postnatal women following discharge from hospital. The lack of tools suitable for assessing women’s global postpartum recovery was highlighted by Sultan et al [25] and Landau [26] and indicates the need for a PROM to assess more broadly, women’s health outcomes following childbirth.
A review was carried out to determine best practice when developing new PROMs. No single guideline or method was identified but a number of key aspects were recognised in order to ensure that the resulting PROM was relevant and acceptable to the patients it was addressing. These aspects included interviews with members of the target population to inform the outcome identification and establishing face validity through the use of cognitive debriefing methods. The review also indicated roles for clinicians in the PROM development process, including the identification of domains and refining outcomes classified through the patient interview process. The findings from the review formed the basis of the methodology of this study, which aimed to develop a condition-specific PROM to assess the quality of care provided to women and newborns using maternity services in LMICs.
PROMs on their own may have a small impact on improving QoC by virtue of the Hawthorne effect [27], with staff being aware that an assessment is being conducted and therefore modifying their behaviour accordingly. However, the greatest benefit from PROMs is likely to be seen when they are used in conjunction with quality improvement (QI) activities. An example of their use might be to administer a PROM to a number of facilities in order to identify those with the poorest outcomes. These could then be targeted with QI interventions and reassessed to determine the extent to which the intervention has achieved its goal. The aim of this study was to identify health outcomes which were important to and could be reported on by women who had recently given birth. These would then be developed into a PROM, the aggregated from which would be used as an indication of the quality of care provided in the respective healthcare facilities.
## Method
This study used a generic qualitative approach employing qualitative interviews and focus group discussions, to identify health outcomes important to women who had recently given birth, which were used to develop the final proposed maternity PROM. The study was divided into three phases–outcome identification, item generation, and pre-testing. This is in line with methodology recommended by the US Food & Drug Administration [28] (S1 Fig, MPROM development process).
## Phase 1: Outcome identification
Based on the initial review [14], a set of draft domains were drawn up and presented to a group of experienced clinicians working in maternity services in LMICs, the Clinician Review Group (CRG). The CRG validated the four primary domains of physical, psychological, social and baby outcomes, identified from the first literature review. A study by Sultan et al [29] proposed 13 domains to describe outpatient recovery following childbirth in the Unites States, which also broadly fit within the physical/psychological/social domains identified.
## Recruitment
Semi-structured topic guides were developed based on the literature review and identified domains, and face to face interviews and focus group discussions were carried out by FD, OM (Kenya) and a research assistant (Malawi), with women who had recently given birth in one of 19 purposively selected healthcare facilities in Malawi and Kenya. These are largely English speaking, low and lower-middle income countries in sub-Saharan Africa, with a wide range of socio-demographics and different ethnic and tribal groups.
As noted in the introduction, there is no guideline or method for developing new PROMs, however based on a review of PROM development literature by the study team, a target of 100 women per country was estimated to be sufficient. Participants were recruited from six-week postnatal/vaccination clinics in both government and charity run healthcare facilities, in both countries. This time frame allowed time for women to have recovered from the birth, whilst still being able to remember any health outcomes. The facilities were purposively selected to include those from rural and urban health centres and hospitals.
Participant selection was largely opportunistic with women attending a facility on the day of data collection and who met the inclusion criteria, being invited to take part. Inclusion criteria comprised having recently given birth to a live baby, attending a vaccination or postnatal care clinic in a target facility and being able to give informed consent. Initial recruitment targets were set for women who had experienced either uncomplicated vaginal, complicated vaginal, or Caesarean section births, in order to include as a broad range of potential outcomes as possible.
## Data collection and analysis
Data collection was conducted using a combination of semi-structured in-depth interviews (IDI) and focus group discussions (FGD) during September 2017 (Malawi) and January 2018 (Kenya). Topic guides focussed on the quality of care the women had received in the healthcare facilities and the health outcomes they had experienced. Interviews allowed individual women, particularly those who had experienced difficult or complicated births, to talk more openly about personal issues, in a one-to-one setting, without fear of their confidentiality being compromised, whilst FGDs enabled a dynamic discussion between women who had had similar births, drawing out the similarities and diversities of their experiences. The three data collectors were experienced midwives and researchers, two of whom were local to the included countries. Informed consent was obtained prior to inclusion in the study and interviews and FGDs were conducted in either English or the local language, depending on the preference of the participants. Interviews and FGDs were recorded with consent and stored securely. Recordings were transcribed verbatim, and where necessary, translated by professional transcribers. FGDs and interviews formed the primary source of data for the outcomes in the final proposed MPROM.
Data were analysed using an inductive thematic method [30] a flexible method of coding data, generated from the data itself. This facilitated analysis across the entire data set, to explore recurrent themes.
## Phase 2: Item generation
In order to develop the interview and FGD data into a working PROM, a list of potential outcomes was drawn up from the themes and transcripts from Phase 1. The CRG was asked to review the draft outcomes, in order to highlight any that they thought were overlapping or that needed expanding further, and identify any additional outcomes they thought were missing. The role of the CRG was to complement the contributions of the women, ensuring that the final MPROM would be suitable for assessing QoC in a clinical setting, but they were not privileged over the views expressed by the women.
Subsequently, based on the findings of the literature review examining methods of developing PROMs, the draft outcomes were expanded to form questions that could be answered by women, using the root: ‘Since the birth of your baby have you…’. This was followed by the outcome formed into a statement, such as ‘suffered from fever with shivering’. The baby related questions were constructed in a similar way using the root: ‘Since they were born, has your baby…’. Answer options were based on a five-point Likert scale, with two positive options, a neutral and two negative options. ‘ Not applicable’ was given as an additional choice where appropriate.
In addition to the outcome questions, demographic questions were added, relating to the method and location of giving birth, and two questions about their health generally, since the birth of the baby and on the day of answering the questions. At this point the question set was labelled as the ‘Draft MPROM’.
## Phase 3: Pre-testing
Phase 3 of the study entailed pilot testing the draft MPROM with a group of women. This was done using cognitive debriefing techniques [31] and helped to establish face validity of the MPROM. Cognitive debriefing involved administering the draft MPROM to a small number of women in Malawi and Kenya, recruited using the same criteria as the Phase 1 participants. They were asked to complete the draft questionnaire and to talk about why they completed the questions in the way they did, how easily they understood the questionnaire, and if there were any other issues which they thought should be added. Due to practical constraints, the draft MPROM was only available in English. The feedback from the cognitive debriefing interviews was reviewed, and any issues identified, were addressed.
## Ethical considerations
Ethical approval was obtained from: LSTM, who also sponsored the study (Research protocol 17–007); National Health Services Research Committee (Malawi) (Ref: $\frac{17}{05}$/1806); and Kenyatta National Hospital-University of Nairobi Ethics Research Committee (Kenya) (Ref: P$\frac{297}{06}$/2017). All women participating in the study were asked to give informed consent with study information sheets and consent forms being available in the local languages, as well as English. All data collected were anonymous and stored securely, separate from the consent forms. No incentives were given to women for taking part in the study, but they were given light refreshments during the data collection process, and a small contribution towards travel costs, to reduce any further impingement on their time. Women who had experienced a stillbirth or early neonatal death were excluded from the study, as there were no resources available to offer them appropriate support. All participants were reminded that participation was entirely voluntary and that they were free to withdraw at any point, for any reason, without penalty.
## Phase 1—Outcome identification
A total of 137 women, across Malawi and Kenya, took part in 83 interviews and 10 FGDs, in Phase 1 of the study. The women were aged between 18 and 45 years, with $52.6\%$ ($$n = 72$$) having had an uncomplicated vaginal birth, $36.5\%$ ($$n = 50$$) having a complicated vaginal birth and $11\%$ ($$n = 15$$) giving birth by Caesarean section. 36 of the women ($26.3\%$) of the women reported no pregnancies prior to most recent pregnancy, 84 women ($61.3\%$) reported 1–3 previous pregnancies, and 17 women ($12.4\%$) reported 4+ previous pregnancies. Complications associated with vaginal births included postpartum haemorrhage, second degree perineal tears or episiotomy, retained placenta, and neonatal sepsis. The babies were aged between 1 and 16 weeks at the time of the interviews. The median was 6 weeks, as this was the age at which they were scheduled to attend for routine postnatal care and vaccinations.
The women described the quality of care experienced in terms of their interactions with members of staff within the healthcare facilities, the availability of staff and the environment of the facility itself.
## Health outcomes
A range of themes were identified under the domains of Physical, Psychological, Social and Baby outcomes (S2 Fig, Development of themes per domain).
## Physical
Physical issues reported by the women included pain, blood loss, perineal trauma and incontinence, as well as issues relating to breasts and breastfeeding, and sexual intercourse. These symptoms were also frequently reported to affect other aspects of the women’s lives such as work, domestic chores, mental health, and their ability to care for the new baby and other family members. Pain was often associated with the normal process of pregnancy, childbirth, and the postnatal period, such as after-pains, but other sources of pain included perineal trauma, Caesarean section wounds, discomfort from engorged breasts or nipple trauma, headaches, and backache. Similarly, to pain, a certain amount of vaginal blood loss was to be expected following the birth of the baby, but when more severe, its reporting was sometimes indicative of serious complications.
## Psychological
The emotions and feelings expressed by the women ranged from happy that they and their baby had survived the birth experience to anxiety and stress about their current situations, particularly relating to finances and other family members. Some women also described what might be considered as signs of depression including self-isolation, being ‘tired of life’, crying for no obvious reason and having difficulty sleeping, although it was beyond the remit of this study to diagnose mental illness. There seemed to be little conscious awareness of the issue of mental health surrounding childbirth or the availability of treatment, with one woman commenting that “concerns do not have medication” (FGD, Malawi). The support that was available largely revolved around family, the local community or the church.
## Social
Although not necessarily an obvious health outcome, issues relating to women’s social well-being often seemed to impact or be impacted by, other aspects of the women’s health. Issues raised by the women related to housework, relationships with partners and other family members, work and finances, and other social activities. For many of the women, particularly in Malawi, daily chores and housework often involved heavy physical tasks, such as carrying water from the water source, washing clothes by hand, or collecting firewood. Their ability to do these and other domestic tasks was reportedly affected by their physical wellbeing following the birth. Psychological consequences from the birth sometimes affected their relationships with others and problems with relationships, particularly with partners and close family members, and were reported to affect the women’s mental wellbeing, particularly where financial hardship or extramarital affairs were involved.
## Baby
The health of the baby was reported by the mother and largely related to feeding and stomach problems or infections. There was, however, also evidence of some misunderstanding of health matters concerning the baby, including confusion between jaundice and yellow fever, and the misconception that umbilical hernias could be caused by air in the abdomen, resulting from the umbilical cord not being tied properly at birth.
In total, 26 key themes were generated from the combined Malawi and *Kenya data* and towards the end of the coding process there was a large degree repetition of code allocation and no new codes being generated.
## Phase 2—Item generation
Following analysis and collation of the interview and FGD transcripts from the two countries, 112 initial potential outcomes were attached to the themes. Following further review, and discussion with the CRG, these were reduced to a total of 79 draft outcomes across the four domains, through a process of combining themes where appropriate, such as “pain” which was included in several themes. We also removed draft outcomes that were difficult to define, such as “confused”, “crying” and “sleep” in relation to depression. Care was taken to ensure that the priorities of the women were preserved and where new outcomes were added, these were defined based on the interview and FGD transcripts. In total, 40 physical, 7 psychological, 17 social, and 15 baby related outcomes were included. Once formed into 79 items, these formed the draft MPROM used for pre-testing in Phase 3. Examples of the outcome and item generation process are included in S3 Fig, Examples of outcome and item generation process.
## Phase 3—Pre-testing
Nine women were asked to complete the draft MPROM using cognitive debriefing methods, four in Malawi and five in Kenya. Completion of the MPROM took approximately 20 minutes. Participants were recruited using the same criteria with the addition of being confident reading and understanding English, from two healthcare facilities used in Phase 1. The women were aged between 19 and 37, with a range of 2 to 12 weeks since the birth of their babies. Six had uncomplicated vaginal births, one had an assisted vaginal birth and two had Caesarean sections. Key issues identified from the cognitive debriefing related to comprehension of the questions, understanding and appropriateness of the answer options, and the format of the questionnaire. All the women were happy to complete the draft MPROM and felt that it would be acceptable to women who were attending postnatal clinics. No additional outcomes were suggested by the interviewees, although a few suggestions were made relating to how the PROM could be improved.
## Comprehension of questions
A few of the women encountered difficulty in understanding a small number of terms used in the PROM. Challenges in understanding the questions could be divided into two categories: linguistic misunderstanding and conceptual misunderstanding. Linguistic misunderstandings occurred where women did not understand specific words in English but when it was explained to them, they understood the concept and were able to discuss how they would have answered the question. Examples of this included the words ‘cope’, ‘stools’, and ‘abscess’. Conceptual misunderstandings occurred when women not only did not understand the word in English but when it was explained, had no experience or understanding of the concept itself. Two particular examples of this were ‘varicose veins’ and ‘assisted vaginal delivery’.
## Answer options
The answer options employed for the draft MPROM were based on a review of existing PROMs and gave the women a choice from five responses, ranging from strongly positive to strongly negative with a neutral option. For some questions an additional ‘Not applicable’ response was provided, where the question might not be relevant, such as bleeding from perineal trauma or relationships with other children. The ‘Not applicable’ option seemed to cause confusion among a few of the respondents, with some using it instead of a negative response, whilst others did not use it when it would have been appropriate. Participants suggested that clearer instructions detailing exactly what each response meant, might be helpful. Alternatively, the final version of the MPROM could use filter questions to avoid needing the ‘Not applicable’ option.
## Questionnaire format
Feedback from the women suggested that they would prefer the layout of the questionnaire to be altered, so that the instructions on how to answer the questions were closer to the questions themselves. They also indicated a preference for having a gap between responses, and using a visual analogue type scale for the two questions asking about their health generally. These could be easily integrated into the final proposed MPROM.
## Principal findings
Following extensive interviews and FGDs with women who had recently given birth in Malawi and Kenya, this study identified a range of health outcomes relevant to them under the domains of physical, psychological, social, and baby. These outcomes were reviewed by a group of clinical experts and then developed into a draft PROM for pre-testing. The cognitive debriefing interviews used for pre-testing found that the draft MPROM was acceptable to women, and that they would be willing to complete it whilst attending the postnatal clinic. Some suggestions for improvements were made by the women relating to a few of the terms used and the layout of the PROM.
For the purposes of the pre-testing, it was not feasible to translate the draft MPROM into the local languages of Malawi and Kenya, so it was necessary to only recruit women who were confident in the use of English.
## Strengths and limitations
The previously published systematic review [14] was not able to identify an existing PROMs suitable for assessing quality of hospital-based maternity care, making the MPROM the first of its kind. It is also the first PROM relating to pregnancy and childbirth, developed specifically for women on LMICs, a population who carry the greatest burden of maternal and newborn mortality, globally. A key strength of this study was the relatively large number of women from two LMICs, who were included in outcome identification data collection. These interviews and FGDs ensured as broad a range of experience and perspectives as possible, to maximise the applicability of the final PROM to the target population. As there is no existing ‘gold standard’ approach to developing PROMs, the study adopted the most widely accepted methods, based on the systematic review of published information (awaiting publication), from PROM development studies in other specialities, as well as the widely cited FDA recommendations [23].
Although the inclusion of women from two different countries contributed the rigour of the study, it also presented some challenges, particularly in relation to language. The two countries were chosen, partly due to the widespread use of English as an official language. However, for many of the women interviewed, this was not their native language and necessitated the use of research assistants, to conduct some of the interviews and FGDs, and professional translation of the transcripts. The women’s limited understanding of English also presented a few challenges during the cognitive debriefing interviews and highlights the need for the final MPROM to be translated and validated into the local language of the women completing it. The importance of the provision of local language versions of tools is similarly highlighted in the EORTC guidelines on PROM development [32]. Other limitations identified during the study were the lack of understanding by some of the women of particular health problems that they had not personally experienced, and potential confusion of the causes of medical conditions such as umbilical hernias underlining the importance of cognitive debriefing [25]. The two countries that were chosen for the development of the MPROM were deliberately selected as they represented both low and middle-income countries, with varied cultural and socio-economic populations, however, further validation studies would need to be conducted in other countries in which it was applied.
## Implications and future research
It is anticipated that once validated, the MPROM would be routinely administered to women attending six-week postnatal clinics in healthcare facilities, with the aggregated findings being used at facility or district regional level to identify hospitals that would benefit from quality improvement (QI) activities. It could also be used as part of a ‘before and after’ model to assess the efficacy of QI activities. It is anticipated that further engagement with the national and regional ministries of health and other stakeholders in each country, will facilitate the utilisation and enhance the ultimate benefits of the MPROM.
To achieve widespread use of the MPROM, further research will be required to develop a scoring system, with each domain being scored separately. This could potentially enable its use without the ‘baby’ questions, for women who have experienced stillbirths or neonatal loss, although further validation research would be required as these women were not included in the initial MPROM development. Additional data collection using the tool will also ensure its validity and demonstrate that it does measure what it is expected to.
## Conclusion
This study conducted interviews and FGDs in Malawi and Kenya, with women who had recently given birth, in order to identify relevant health outcomes that they may have experienced. Outcomes were grouped under four domains, physical, psychological, social and baby. Following consultation with a group of clinical experts, the outcomes were used to develop a draft version of the MPROM, which was pre-tested with small groups of women, using cognitive debriefing methods, to establish face validity. This resulted in the production of the proposed MPROM, believed to be the first of its kind, for use in assessing the quality of care provided in healthcare facilities in LMICs. Further research is intended to establish the psychometric validity of the tool and promote its use in Kenya and Malawi.
## References
1. 1United Nations Children’s Fund (UNICEF). Maternal mortality. https://data.unicef.org/topic/child-survival/maternal-mortality/
2. 2United Nations (UN). The Millennium Development Goals Report 2015. New York: UN; 2015. https://www.un.org/en/development/desa/publications/mdg-report-2015.html
3. 3WHO (2020a) The Global Health Observatory. https://www.who.int/data/gho
4. 4World Health Organization (WHO). National, regional, and worldwide estimates of stillbirth rates in 2009 with trends since 1995: Policy brief. 2011. https://www.who.int/reproductivehealth/publications/maternal_perinatal_health/rhr_11_03/en/
5. 5World Health Organization (WHO) Strategies for ending preventable maternal mortality (EPMM). Geneva, WHO. 2015.
6. Hanefeld J, Powell-Jackson T, Balabanova D. **Understanding and measuring quality of care: dealing with complexity**. (2017.0) **95** 368-374. DOI: 10.2471/BLT.16.179309
7. Donabedian A.. **Evaluating the quality of medical care**. (2005.0) **83** 691-719. DOI: 10.1111/j.1468-0009.2005.00397.x
8. Tuncalp O, Were WM, MacLennan C, Oladapo OT, Gulmezoglu AM, Bahl R. **Quality of care for pregnant women and newborns–the WHO vision**. (2015.0) **122** 1045-49. DOI: 10.1111/1471-0528.13451
9. Kongnyuy EJ, van den Broek N. **Criteria for clinical audit of women friendly care and providers’ perception in Malawi**. *BMC Pregnancy and Childbirth* (2008.0) **8**. DOI: 10.1186/1471-2393-8-28
10. Tuncalp O, Hindin JM, Adu-Bonsaffoh K, Adanu R. **Listening to women’s voices: The quality of care of women experiencing severe maternal morbidity, in Accra, Ghana**. *PLoS ONE* (2012.0) **7** e44536. DOI: 10.1371/journal.pone.0044536
11. Fisseha G, Berhane Y, Worku A, Terefe W. **Quality of the delivery services in health facilities in Northern Ethiopia**. *BMC Health Services Research* (2017.0) **17**. DOI: 10.1186/s12913-017-2125-3
12. King JJC, Das J, Kwan A, Daniels B, Powell-Jackson T, Makungu C. **How to do (or not do)… using the standardized patient method to measure clinical quality of care in LMIC health facilities**. (2019.0) **34** 625-634. DOI: 10.1093/heapol/czz078
13. Devlin NJ, Appleby J. (2010.0)
14. Dickinson F, McCauley M, Smith H, van den Broek N. **Patient reported outcomes measures for use in pregnancy and childbirth: A systematic review**. (2019.0) **19**. DOI: 10.1186/s12884-019-2318-3
15. Fletcher SJ, Waterman H, Nelson L, Carter LA, Dwyer L, Roberts C. **Holistic assessment of women with hyperemesis gravidarum: A randomised controlled trial**. (2015.0) **52** 1669-1677. DOI: 10.1016/j.ijnurstu.2015.06.007
16. Kopec JA, Ogonowski J, Rahman MM, Miazgowski T. **Patient-reported outcomes in women with gestational diabetes: A longitudinal study**. (2015.0) **22** 206-213. DOI: 10.1007/s12529-014-9428-0
17. Thompson JF, Roberts CL, Ellwood DA. **Emotional and physical health outcomes after significant primary post-partum haemorrhage (PPH): A multicentre cohort study**. (2011.0) **51** 365-71. DOI: 10.1111/j.1479-828X.2011.01317.x
18. de Visser SM, Kirchner CA, van der Velden BGJ, de Wit AC, Dijkman A, Huisjes AJM. **Major obstetric haemorrhage: Patients’ perspective on the quality of care**. (2018.0) **224** 146-152. DOI: 10.1016/j.ejogrb.2018.03.032
19. Yawn BP, Dietrich AJ, Wollan P, Bertram S, Graham D, Huff J. **TRIPPD: A practice-based network effectiveness study of postpartum depression screen and management**. (2012.0) **10** 320-9. DOI: 10.1370/afm.1418
20. Symon A., Downe S, Finlayson KW, Knapp R, Diggle P. **The feasibility and acceptability of using the Mother-Generated Index (MGI) as a patient reported outcomes measure in a randomised controlled trial of maternity care**. (2015.0) **15** 1004
21. 21World Health Organization (WHO). Maternal mortality. 2019. https://www.who.int/news-room/fact-sheets/detail/maternal-mortality
22. 22ICHOM Pregnancy and Childbirth Standard Set. https://connect.ichom.org/standard-sets/pregnancy-and-childbirth/
23. Ciechanowicz S, Setty T, Robson E, Sathasivam C, Chazapis M, Dick J. **Development and evaluation of an obstetric quality-of-recovery score (ObsQoR-11) after elective Caesarean delivery**. (2019.0) **122** 69-78. DOI: 10.1016/j.bja.2018.06.011
24. Sultan P, Kormendy F, Nishimura S, Charvalho B, Guo N, Papageorgiou C. **Comparison of spontaneous versus operative vaginal delivery using Obstetric Quality of Recover-10 (ObsQoR-10): An observational cohort study**. (2020.0) **63**. DOI: 10.1016/j.jclinane.2020.109781
25. Sultan P, Sharawi N, Blake L, Ando K, Sultan E, Aghaeepour N. **Use of patient-reported outcome measures to assess outpatient postpartum recovery: A systematic review**. (2021.0) **4**. DOI: 10.1001/jamanetworkopen.2021.11600
26. Landau R.. **Deconstructing current postpartum recover research–The need to contextualize patient-reported outcome measures**. (2021.0) **4** 27n
27. Green J., Thorogood N.. (2018.0)
28. 28US Food and Drug Administration (FDA). Guidance for industry Patient Reported Outcome Measures: Use in medical product development to support labelling claims. 2009. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/patient-reported-outcome-measures-use-medical-product-development-support-labeling-claims
29. Sultan P, Jensen SE, Taylor J, El-Sayed Y, Carmichael S, Cella D. **Proposed domains for assessing postpartum recovery: A concept elicitation study**. (2021.0). DOI: 10.1111/1471-0528.16937
30. Braun V, Clarke V. **Using thematic analysis in psychology**. (2006.0) **3** 77-101
31. Collins D.. **Pretesting survey instruments: An overview of cognitive methods**. (2003.0) **12** 229-238. DOI: 10.1023/a:1023254226592
32. 32Johnson C, Aaronson N, Blazeby JM, Bottomley A, Fayers P, Koller M, et al. EORTC Quality of Life Group guidelines for developing questionnaire modules. 2011. https://qol.eortc.org/manuals/
|
---
title: 'The burden of catastrophic and impoverishing health expenditure in Armenia:
An analysis of Integrated Living Conditions Surveys, 2014–2018'
authors:
- Jacob Kazungu
- Christina L. Meyer
- Kristine Gallagher Sargsyan
- Seemi Qaiser
- Adanna Chukwuma
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021688
doi: 10.1371/journal.pgph.0000494
license: CC BY 4.0
---
# The burden of catastrophic and impoverishing health expenditure in Armenia: An analysis of Integrated Living Conditions Surveys, 2014–2018
## Abstract
Armenia’s health spending is characterized by low public spending and high out-of-pocket expenditure (OOP), which not only poses a financial barrier to accessing healthcare for Armenians but can also impoverish them. We analyzed Armenia’s Integrated Living Conditions Surveys 2014–2018 data to assess the incidence and correlates of catastrophic health expenditure (CHE) and impoverishment. Households were considered to have incurred CHE if their annual OOP exceeded 40 percent of the per capita annual household non-food expenditure. We assessed impoverishment using the US$1.90 per person per-day international poverty line and the US$5.50 per person per-day upper-middle-income country poverty line. Logistic regression models were fitted to assess the correlates of CHE and impoverishment. We found that the incidence of CHE peaked in 2017 before declining in 2018. Impoverishment decreased until 2017 before rising in 2018. After adjusting for sociodemographic factors, households were more likely to incur CHE if the household head was older than 34 years, located in urban areas, had at least one disabled member, and had at least one member with hypertension. Households with at least one hypertensive member or who resided in urban areas were more likely to be impoverished due to OOP. Paid employment and high socioeconomic status were protective against both CHE and impoverishment from OOP. This detailed analysis offers a nuanced insight into the trends in Armenia’s financial risk protection against catastrophic and impoverishing health expenditures, and the groups predominantly affected. The incidence of CHE and impoverishment in Armenia remains high with a higher incidence among vulnerable groups, including those living with chronic disease, disability, and the unemployed. Armenia should consider different mechanisms such as subsidizing medication and hospitalization costs for the poorest to alleviate the burden of OOP.
## Introduction
As one of the key components of universal health coverage (UHC), financial risk protection in health aims to prevent households from exposure to financial hardship as a result of their direct health spending [1]. Since 2010, direct household out-of-pocket (OOP) payments still account for well over a third of low- and middle-income countries’ (LMICs) total health spending, contributing to global financial hardship and impoverishment [2–4]. Globally, 996 million individuals experienced catastrophic health expenditures (CHE) out-of-pocket health spending that exceeds ten percent of a household’s budget—while 435 million people were impoverished by OOP payments in 2017 [5].
Compared with other LMICs, Armenia has one of the highest ratios of OOP spending to total health expenditure, growing by 45 percent over a twelve-year period to reach 85 percent in 2019 [2]. As of 2018, Armenia’s public health spending as a proportion of the country’s gross domestic product (GDP) (1.2 percent) is much lower than other upper-middle income (UMI) countries (3.2 percent) and the average among countries in Europe and Central Asia (6.7 percent) [6]. Similarly, the country’s public health spending as a proportion of the Armenian government’s budget (5.2 percent) is lower than its European and Central Asian counterparts in 2018 (15.2 percent) [7]. The most recent examination of Armenia’s CHE found that 16 percent of Armenian households’ OOP payments exceeded 10 percent of annual household consumption in 2013 [3]. With an average annual 3.3 percent increase in the incidence of catastrophic OOP payments between 2010 and 2013, Armenia’s incidence of catastrophic health payments has grown faster than any other country in the world [3].
Armenia provides public spending for services via an explicitly defined basic benefits package (BBP). In practice, there are three packages. The basic BBP package available to the whole population consists of outpatient services including primary care, maternity services, and sanitary epidemiological services. However, 38 percent of the population has access to the more extensive BBP package with inpatient service coverage, including the poor, vulnerable, and special groups. There is also a special package for civil and military servants [8]. Those who qualified for the more generous coverage, in 2006,paid 45 percent less during health care visits and had 36 percent higher rates of using outpatient care than those with less generous BBP coverage [9]. As of 2018, approximately 18 percent of Armenian households have at least one member with access to the BBP, including government workers and select socially vulnerable individuals [10].
As a reflection of the above, 62 percent of Armenians have to pay entirely out-of-pocket for outpatient diagnostic care, medication, and most inpatient care [8, 9]. The rest of the population (recipients of the more generous BBP coverage)have also required copayments for the same services, challenging efforts to reduce the disease and economic burden from Armenia’s high prevalence of noncommunicable diseases [9, 11]. The cost of medication has historically been a major source of high OOP payments as the BBP pharmaceutical program coverage is limited, and pharmaceuticals are subject to a 20 percent value-added tax but not subject to other forms of pricing laws or regulations [8]. Moreover, because of the BBP’s limited budget for provider reimbursements, healthcare providers that administer care to BBP patients often subsidize their revenue by increasing formal and informal fees for non-BBP eligible patients, leading to higher OOP spending in Armenia [8]. As a result, Armenia’s OOP payments have been higher than other UMI countries as well as Armenia’s Europe and Central Asia regional counterparts. In a recent survey, nearly a third of Armenians who had not used health care services in the last 12 months reported their primary barrier to care was their inability to pay for services [12].
Given the high OOP cost of care for Armenians, the impetus for this study is to understand the factors contributing to the country’s financial barriers to health care access. Currently, there is no comprehensive analysis of the incidence of OOP payments for health care in Armenia and limited data on CHE and impoverishing health spending after 2013. While the global literature indicates that: household economic status, the incidence of hospitalization, the presence of an elderly or disabled household member in the family, and the presence of a family member with a chronic illness are common significant factors associated with household CHE [5, 13], there are additional drivers of catastrophic and impoverishing health spending that are context-specific [14–21]. Among some European and Central Asian countries, the high cost of medication, outpatient coverage gaps, and physician-induced demand are cited as contributors to high OOP [22, 23].
This analysis aims to examine trends in the current state of Armenia’s OOP expenditures and determine who is most at risk for CHEs or impoverishing health spending. It investigates sociodemographic factors, health status and service usage, and current benefit levels to understand their potential impact on catastrophic and impoverishing health expenditure in the Armenian context. By exploring the trends and inequalities in catastrophic and impoverishing health spending, this study both updates and adds nuance to the discussion around the current state of Armenia’s financial risk protection and those most impacted by the cost of healthcare.
## Data sources
We obtained data from Armenia’s Integrated Living Conditions Survey (ILCS) for five years from 2014 to 2018, within which the full complement of variables for the analysis was available and collected consistently [10]. However, the ILCS in Armenia has been conducted annually since 2001 by the National Statistical Service of the Republic of Armenia with support from the World Bank, United States Agency for International Development, and other donor organizations. The survey involves a two-stage stratified sampling approach where households are the ultimate sampling units. In this approach, the county is first stratified into sampling areas based on the location’s geographical region, urbanization, and population size. Once stratified, eight households are randomly selected from each sampling area. Since its introduction, Armenia’s ILCS has included a sample size of 5,184 households. However, the sample size was expanded to 7,872 between 2007 and 2011 following increased funding by the Millennium Challenge Account–Armenia (MCA-Armenia). The sample size reverted to 5,184 from 2012 except in 2017, where 7,776 households were included, because of budget availability. Ethical clearance for this study is deemed not to be required because it uses publicly available, de-identified data.
## Incidence of CHE and impoverishment
For each year between 2014 and 2018, we computed the incidence of CHE and impoverishment following widely applied methods [24, 25]. First, drawing on reported household spending, we calculated the total OOP expenditure incurred by patients while accessing care, summing individual expenditure at the household level. The OOP expenditure included spending on medicines, laboratory and radiology investigations, and other direct payments to the cashier or medical staff during inpatient or outpatient visits, including informal payments. However, the calculated OOP expenditure does not include the transportation costs to and from the health facilities, as this data is not available in the ILCS. We excluded any payments covered by health insurance. Cost variables for outpatient visits were based on a 30-day recall period. We annuitized these by multiplying the total OOP for outpatient visits by 12. However, inpatient costs were based on their last visit and a 12-month recall period. Therefore, we obtained total annual OOP costs for inpatient care by multiplying their respective inpatient costs in the last visit with their total number of visits in the last 12 months. We imputed inpatient costs for those with inpatient visits but missing inpatient costs by multiplying the median day costs per admission by the average length of stay for admission in their last visit and the total number of visits in the last 12 months. We imputed zero for the length of stay for those who did not have inpatient visits to include the length of stay variable in the model. Then, we adjusted the outpatient, inpatient and total healthcare costs to 2018 constant Armenian Drams (AMD) using Consumer Price *Index data* from the World Bank Database.
CHE was measured by estimating the catastrophic headcount. A catastrophic headcount was defined as the percentage of households whose OOP expenditure for health care was greater than an established threshold. Two thresholds were used to assess the incidence of CHE. A household was considered to have incurred CHE if their per capita annual expenditure on health care exceeded: 1) 10 percent of the per capita total annual household consumption expenditure, or 2) 40 percent of the per capita annual household non-food consumption expenditure [26]. The two thresholds were used for comparison as a result of their differences in denominators (total household expenditure versus non-food expenditure), the threshold levels as a percentage of the denominator, and given the fact that there is no consensus about which of the two is the better threshold for assessing CHE [26, 27].
We assessed the incidence of impoverishment in Armenia through several steps. First, we defined the poverty line. We adopted two poverty lines for comparison: 1) the US$1.90 per person per day international poverty line and 2) the US$5.50 per person per day recommended for upper-middle income countries as suggested by the World Bank. Second, we used the GDP deflator data for Armenia to convert the poverty line to 2014, 2015, 2016, 2017, and 2018 values. Third, we converted this international poverty line for each respective year by multiplying by its respective World Bank purchasing power parity conversion factor (LCU per international dollar) into Armenian Drams (AMD). Finally, we calculated the percentage of households pushed further into poverty due to health expenditure by examining changes in total per capita expenditure before and after the household incurred health spending (standardized by household size).
## Inequalities in the distribution of CHE and impoverishing health expenditure
We computed the inequalities in the incidence of CHE and impoverishment using well-established methodologies described by Wagstaff et al. [ 28]. We calculated the rich-poor difference, defined as the difference in the percentage of households incurring CHE or facing impoverishment between the households in the richest quintile (Q5) to the poorest (Q1) and ratios (Q5/Q1). Although these two measures of inequality are easier to calculate and interpret, they do not consider poor (Q2), middle (Q3), and rich (Q4). Consequently, we generated the concentration curves and Wagstaff’s concentration index for the CHE using both thresholds and impoverishing health expenditure for 2017 and 2018. The concentration curve plots indicated the cumulative percentage of CHE/impoverishment on the y-axis versus the cumulative population percentage ranked by wealth from poorest to richest [26]. When every household experiences the same share of CHE/impoverishment, the concentration curve transforms into a 45-degree line. CHE/impoverishment can be interpreted as concentrated among the rich when the curve lies below the 45-degree line and vice versa. The further the curve is from the 45-degree line, the higher the inequality. The concentration index is a summary measure obtained from the concentration curve and ranges from -1 to 1. A concentration index of 0 indicates equality in experiencing CHE/impoverishment, whereas the concentration index is a positive (negative) value when inequality is concentrated among the rich (poor) [29]. For this analysis, the inequalities in impoverishment were calculated using the US$5.50 per person per day poverty line only as it is the recommended poverty line for upper-middle countries like Armenia.
Fig 2 shows the concentration curves for CHE at both the 10 percent of total consumption and 40 percent of non-food consumption thresholds from 2014 to 2018. Fig 3 shows the concentration curves for impoverishment between 2014 and 2018 in Armenia. Table 4 also presents the incidence of CHE in the poorest quintile (Q1), richest quintile (Q5), absolute differences (Q5-Q1), rich-poor ratios (Q5/Q1), and concentration index (CIX). Overall, there is a higher incidence of both CHE and impoverishing health expenditure among the poorest quintile compared to the richest quintile in Armenia. Specifically, in absolute terms, the incidence of CHE peaked in 2017, where the poorest experienced a 2.88 percent and 5.34 percent increase in CHE incidence compared to the richest at the 10 percent total consumption and 40 percent non-food consumption thresholds, respectively. The incidence then slightly declined in 2018. Based on the 40 percent of non-food expenditure threshold for CHE, the inequalities in CHE were significantly higher among the poorest relative to all other quintiles except in 2014; (CIX = -0.051, 95 percent CI: -0.108 to 0.005, p-value = 0.074) in 2014, (CIX = -0.170, 95 percent CI:-0.234 to -0.105, p-value<0.001) in 2015, (CIX = -0.105, 95 percent CI: -0.167 to -0.042, p-value = 0.001) in 2016, (CIX = -0.154, 95 percent CI: -0.202 to -0.106, p-value<0.01) in 2017 and (CIX = -0.170, 95 percent CI: -0.240 to -0.099, p-value<0.001) in 2018. Impoverishing health expenditure was also characterized by a higher incidence among the poor relative to the other quintiles across all years examined but was only significantly higher in 2017 (Table 6).
**Fig 2:** *Concentration curve for CHE by threshold and year in Armenia—2014 to 2018 ILCS.* **Fig 3:** *Concentration Curve for impoverishment by year in Armenia– 2014 to 2018 ILCS.* TABLE_PLACEHOLDER:Table 6
## Correlates of CHE and impoverishing health expenditure
To assess the correlates of incurring CHE and impoverishment (using US$5.50 per person per day poverty line) from OOP expenditure on health care, we fit both bivariate and multivariable logistic regression models for each of the two outcomes: 1) whether a household incurred CHE at the 10 percent threshold (Yes/No) and 2) whether a household was pushed into poverty as a result of OOP expenditure on health (Yes/No). In each of these models, we incorporated sociodemographic factors including gender (Male/Female), age group (<25 years/25-34 years/45-54 years/55-64 years/65+ years), and level of education of the household head (None/Primary/Secondary/Tertiary), whether household has at least one member with hypertension (Yes/No), whether at least one household member has private health insurance (Yes/No), whether at least one household member belongs to a group considered socially vulnerable or eligible for the social package (disabled, pensioner, receives social benefits, military, children), whether at least one household member has access to coverage for vulnerable groups under the BBP (Yes/No), whether at least one household member had some paid work (Yes/No), household location (Urban/Rural), household size (1–2 members/3-4 members/5+ members), and the household’s socioeconomic status proxied by quintiles (poorest, poor, middle, rich, or richest quintile) of an asset index generated through principal component analysis. These independent variables were included in this analysis based on their relevance to the Armenian context (e.g., access to the BBP for vulnerable groups) or have previously been associated with the incidence of CHE and/or impoverishment in other studies [17–21, 24, 30]. Sampling weights were also applied to ensure that estimates were nationally generalizable while standard errors were clustered at the sampling area level. Additionally, we employed a complete case analysis where participants with missing data were excluded from the analysis. Overall, 98 percent of households had complete data on all variables included in the multivariable models.
Table 7 presents the unadjusted odds ratio (uOR), adjusted odds ratio (aOR), and p-values for the correlates of CHE health care in Armenia using the 2018 ILCS. The adjusted findings indicate that households headed by individuals older than 34 years were more likely to incur CHE than households headed by those under the age of 25 years. In addition, households with at least one person having hypertension were over five times (aOR = 5.20, 95 percent CI: 3.77–7.19, p-value<0.001) more likely to incur CHE compared to households where none of the members had hypertension. Furthermore, households with at least one member who belonged to a disabled group (aOR = 2.12, 95 percent CI: 1.56–2.87, p-value<0.001) or located in urban areas (aOR = 2.19, 95 percent CI: 1.03–4.65, p-value = 0.042) were significantly more likely to incur CHE. Besides, every additional household member aged >65 years was associated with a 47 percent (aOR = 1.47, 95 percent CI: 1.12–1.94, p-value<0.001) increase in the odds of incurring CHE. However, households with at least one member engaged in some paid work had a 45 percent (aOR = 0.55, 95 percent CI: 0.47–0.64, p-value<0.001) reduced odds of incurring CHE after controlling for other factors. Similarly, better off households had lower odds of incurring CHE compared to poor households (Table 7).
**Table 7**
| Unnamed: 0 | Catastrophic Health Expenditure (CHE) | Catastrophic Health Expenditure (CHE).1 | Catastrophic Health Expenditure (CHE).2 | Catastrophic Health Expenditure (CHE).3 |
| --- | --- | --- | --- | --- |
| | Un-adjusted odds ratio [95% CI] | p-value | Adjusted odds ratio [95% CI] | p-value |
| Gender of household head (Ref. Female) | Gender of household head (Ref. Female) | Gender of household head (Ref. Female) | Gender of household head (Ref. Female) | Gender of household head (Ref. Female) |
| Male | 0.74 [0.59–0.94] | 0.013 | 0.83 [0.58–1.18] | 0.296 |
| Age group of household head (Ref. <25) | Age group of household head (Ref. <25) | Age group of household head (Ref. <25) | Age group of household head (Ref. <25) | Age group of household head (Ref. <25) |
| 25–34 | 1.39 [0.59–3.27] | 0.449 | 1.98 [0.72–5.44] | 0.187 |
| 35–44 | 2.02 [1.49–2.74] | <0.001 | 2.87 [2.09–3.95] | <0.001 |
| 45–54 | 2.10 [1.45–3.03] | <0.001 | 2.58 [1.80–3.69] | <0.001 |
| 55–64 | 3.85 [3.05–4.86] | <0.001 | 3.83 [2.91–5.04] | <0.001 |
| 65+ | 5.75 [4.27–7.74] | <0.001 | 2.72 [1.76–4.22] | <0.001 |
| Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) |
| Married | 0.79 [0.64–0.98] | 0.03 | 1.25 [0.88–1.77] | 0.205 |
| Level of Education of household head (Ref. No education) | Level of Education of household head (Ref. No education) | Level of Education of household head (Ref. No education) | Level of Education of household head (Ref. No education) | Level of Education of household head (Ref. No education) |
| Primary | .01 [0.57–7.15] | 0.278 | 1.55 [0.43–5.56] | 0.503 |
| Secondary | 1.55 [0.35–6.86] | 0.566 | 1.44 [0.36–5.77] | 0.61 |
| Tertiary | 1.70 [0.41–7.09] | 0.464 | 1.58 [0.42–5.95] | 0.499 |
| Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) |
| Yes | 6.17 [4.91–7.74] | <0.001 | 5.20 [3.77–7.19] | <0.001 |
| Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) |
| Yes | 2.79 [2.30–3.38] | <0.001 | 2.12 [1.56–2.87] | <0.001 |
| Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) |
| Yes | 2.63 [2.23–3.10] | <0.001 | 1.22 [0.92–1.63] | 0.169 |
| Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) |
| Yes | 1.21 [0.93–1.59] | 0.162 | 0.74 [0.46–1.20] | 0.219 |
| Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) |
| Yes | 1.34 [0.95–1.88] | 0.091 | 0.93 [0.71–1.23] | 0.623 |
| Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) |
| Yes | 0.59 [0.32–1.11] | 0.103 | 0.74 [0.31–1.78] | 0.503 |
| Whether at least one member has access to BBP for vulnerable groups (Ref. None) | Whether at least one member has access to BBP for vulnerable groups (Ref. None) | Whether at least one member has access to BBP for vulnerable groups (Ref. None) | Whether at least one member has access to BBP for vulnerable groups (Ref. None) | Whether at least one member has access to BBP for vulnerable groups (Ref. None) |
| Yes | 1.49 [1.16–1.91] | 0.002 | 1.20 [0.96–1.50] | 0.114 |
| Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) |
| Yes | 0.77 [0.53–1.11] | 0.159 | 1.08 [0.69–1.67] | 0.746 |
| Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) |
| Yes | 0.33 [0.28–0.39] | <0.001 | 0.55 [0.47–0.64] | <0.001 |
| Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area |
| Urban | 1.84 [1.15–2.95] | 0.011 | 2.19 [1.03–4.65] | 0.042 |
| Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) |
| Average (3 to 4 members) | 0.47 [0.39–0.57] | <0.001 | 0.80 [0.61–1.07] | 0.133 |
| Bigger (5+) | 0.47 [0.34–0.64] | <0.001 | 0.61 [0.35–1.09] | 0.096 |
| Number of household members aged <18 years | Number of household members aged <18 years | Number of household members aged <18 years | Number of household members aged <18 years | Number of household members aged <18 years |
| | 0.74 [0.66–0.84] | <0.001 | 1.01 [0.87–1.17] | 0.935 |
| Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years |
| | 0.77 [0.71–0.83] | <0.001 | 1.03 [0.88–1.21] | 0.721 |
| Number of household members aged >65 years | Number of household members aged >65 years | Number of household members aged >65 years | Number of household members aged >65 years | Number of household members aged >65 years |
| | 1.75 [1.57–1.96] | <0.001 | 1.47 [1.12–1.94] | 0.006 |
| Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) |
| Poorer | 0.82 [0.63–1.08] | 0.157 | 0.60 [0.46–0.77] | <0.001 |
| Middle | 1.03 [0.78–1.37] | 0.843 | 0.70 [0.47–1.03] | 0.073 |
| Rich | 0.90 [0.60–1.35] | 0.608 | 0.62 [0.36–1.06] | 0.083 |
| Richest | 0.75 [0.49–1.14] | 0.187 | 0.52 [0.29–0.96] | 0.036 |
| N | | | 5078 | 5078 |
| R 2 | | | 0.1339 | 0.1339 |
Table 8 presents the uOR, aOR, and p-values for the correlates of impoverishing health care expenditures in Armenia using the 2018 ILCS.
**Table 8**
| Unnamed: 0 | Impoverishment | Impoverishment.1 | Impoverishment.2 | Impoverishment.3 |
| --- | --- | --- | --- | --- |
| | Un-adjusted odds ratio [95% CI] | p-value | Adjusted odds ratio [95% CI] | p-value |
| Gender of household head (Ref. Female) | Gender of household head (Ref. Female) | Gender of household head (Ref. Female) | Gender of household head (Ref. Female) | Gender of household head (Ref. Female) |
| Male | 1.15 [0.91–1.47] | 0.249 | 1.29 [0.90–1.85] | 0.163 |
| Age group of household head (Ref. <25) | Age group of household head (Ref. <25) | Age group of household head (Ref. <25) | Age group of household head (Ref. <25) | Age group of household head (Ref. <25) |
| 25–34 | 0.17 [0.06–0.50] | 0.001 | 0.47 [0.13–1.70] | 0.251 |
| 35–44 | 0.26 [0.13–0.52] | <0.001 | 0.63 [0.30–1.31] | 0.216 |
| 45–54 | 0.29 [0.16–0.55] | <0.001 | 0.80 [0.50–1.30] | 0.371 |
| 55–64 | 0.52 [0.39–0.69] | <0.001 | 1.18 [0.84–1.65] | 0.338 |
| 65+ | - | - | - | - |
| Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) | Current marital status of household head (Ref. Not married) |
| Married | 1.14 [0.95–1.35] | 0.151 | 1.44 [1.05–1.99] | 0.024 |
| Level of education of household head (Ref. No education) | Level of education of household head (Ref. No education) | Level of education of household head (Ref. No education) | Level of education of household head (Ref. No education) | Level of education of household head (Ref. No education) |
| Primary | 0.89 [0.11–7.41] | 0.913 | 0.52 [0.05–5.45] | 0.583 |
| Secondary | 0.81 [0.09–7.34] | 0.849 | 0.63 [0.05–7.22] | 0.710 |
| Tertiary | 0.75 [0.07–7.67] | 0.810 | 0.60 [0.05–7.79] | 0.694 |
| Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) | Whether at least one member has hypertension (Ref. None) |
| Yes | 3.45 [2.44–4.88] | <0.001 | 2.51 [1.56–4.03] | <0.001 |
| Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) | Whether at least one member is disabled (Ref. None) |
| Yes | 2.11 [1.63–2.73] | <0.001 | 1.01 [0.41–2.48] | 0.978 |
| Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) | Whether at least one member is a pensioner (Ref. None) |
| Yes | 3.51 [2.16–5.70] | <0.001 | 1.94 [1.28–2.93] | 0.002 |
| Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) | Whether at least one member receives social benefits (Ref. None) |
| Yes | 1.55 [0.92–2.60] | 0.099 | 1.81 [0.65–5.03] | 0.254 |
| Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) | Whether at least one member is in the military social group (Ref. None) |
| Yes | 2.37 [1.42–3.95] | 0.001 | 2.17 [1.23–3.81] | 0.007 |
| Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) | Whether at least one member is in the children’s social group (Ref. None) |
| Yes | 0.74 [0.32–1.71] | 0.488 | 0.85 [0.40–1.82] | 0.676 |
| Whether at least one member has access to BBP (Ref. None) | Whether at least one member has access to BBP (Ref. None) | Whether at least one member has access to BBP (Ref. None) | Whether at least one member has access to BBP (Ref. None) | Whether at least one member has access to BBP (Ref. None) |
| Yes | 1.52 [1.17–1.96] | 0.001 | 1.42 [0.86–2.33] | 0.167 |
| Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) | Whether at least one member has health insurance (Ref. None) |
| Yes | 0.65 [0.42–1.02] | 0.06 | 0.83 [0.49–1.44] | 0.514 |
| Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) | Whether at least one member has some paid work (Ref. None) |
| Yes | 0.41 [0.25–0.66] | <0.001 | 0.74 [0.53–1.02] | 0.068 |
| Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area |
| Urban | 1.61 [1.04–2.49] | 0.033 | 1.76 [1.10–2.83] | 0.019 |
| Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) | Household size (Ref. Small—1 to 2 members) |
| Average (3 to 4 members) | 0.49 [0.27–0.90] | 0.022 | 0.95 [0.47–1.92] | 0.881 |
| Bigger (5+) | 0.69 [0.53–0.91] | 0.008 | 1.00 [0.50–2.01] | 0.998 |
| Number of household members aged <18 years | Number of household members aged <18 years | Number of household members aged <18 years | Number of household members aged <18 years | Number of household members aged <18 years |
| | 0.87 [0.76–1.00] | 0.044 | 1.10 [0.92–1.32] | 0.291 |
| Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years | Number of household members aged 18 to 65 years |
| | 0.78 [0.68–0.91] | 0.001 | 0.89 [0.72–1.10] | 0.281 |
| Number of household members aged >65 years | Number of household members aged >65 years | Number of household members aged >65 years | Number of household members aged >65 years | Number of household members aged >65 years |
| | 2.11 [1.64–2.72] | <0.001 | 1.31 [0.95–1.80] | 0.105 |
| Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) | Household socioeconomic status (Ref. Poorest) |
| Poor | 1.10 [0.71–1.73] | 0.664 | 0.93 [0.64–1.35] | 0.72 |
| Middle | 1.58 [0.91–2.74] | 0.105 | 1.20 [0.83–1.75] | 0.334 |
| Rich | 0.79 [0.57–1.10] | 0.164 | 0.63 [0.44–0.90] | 0.011 |
| Richest | 0.94 [0.57–1.55] | 0.820 | 0.75 [0.44–1.30] | 0.309 |
| N | - | - | 5049 | 5049 |
| R2 | - | - | 0.0926 | 0.0926 |
Although the current marital status was not a significant factor in bivariate analyses, in the adjusted analysis, households with heads that were married had significantly 44 percent (aOR = 1.44, 95 percent CI: 1.05–1.99, p-value = 0.024) increased odds of being impoverished as a result of OOP payments. Additionally, households with at least one hypertensive member (aOR = 2.51, 95 percent CI: 1.56–4.03, p-value<0.001), at least one member belonging to the pensioner group (aOR = 1.94, 95 percent CI: 1.28–2.93, p-value = 0.002), at least one member belonging to the military social group (aOR = 2.17, 95 percent CI: 1.23–3.81, p-value = 0.007), or was located in urban areas (aOR = 1.76, 95 percent CI: 1.10–2.83, p-value = 0.019) were significantly more likely to be impoverished compared to households without hypertension, a member belonging to a pensioner or military social groups, or located in rural areas. Again, households with high socioeconomic status were protected from becoming impoverished because of OOP payments for health compared to households with a low socioeconomic status (Table 8).
## Inclusivity in global research
Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in S1 Checklist.
## Results
Table 1 presents the sociodemographic characteristics of the households surveyed in Armenia’s 2014–2018 ILCS. Overall, the majority of surveyed households lived in urban areas; had elderly heads of household (65+ years); had at least one member with paid work; and were comprised of 3–4 household members. The proportion of female household heads in the study sample ranged from 31.4 percent in 2016 to 33.6 percent in 2017. Nearly all heads of surveyed households had a secondary education, half of whom received a tertiary education. Fewer than one out of ten surveyed households had one household member with hypertension. The proportion of households with a disabled member decreased from over 17 percent in 2014 to 14 percent in 2018. The poorest households made up a slightly larger share of surveyed households, ranging between 22.2 percent in 2017 and 24.4 percent in 2016.
**Table 1**
| Unnamed: 0 | 2014 | 2014.1 | 2015 | 2015.1 | 2016 | 2016.1 | 2017 | 2017.1 | 2018 | 2018.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | N | % | N | % | N | % | N | % | N | % |
| Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head | Gender of household head |
| Female | 1635 | 31.54 | 1706 | 32.91 | 1625 | 31.35 | 2610 | 33.56 | 1703 | 32.85 |
| Male | 3549 | 68.46 | 3478 | 67.09 | 3559 | 68.65 | 5166 | 66.44 | 3481 | 67.15 |
| Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head | Age group of household head |
| <25 | 37 | 0.71 | 31 | 0.6 | 28 | 0.54 | 52 | 0.67 | 30 | 0.58 |
| 25–34 | 292 | 5.63 | 293 | 5.65 | 267 | 5.15 | 414 | 5.32 | 272 | 5.25 |
| 35–44 | 522 | 10.07 | 570 | 11 | 615 | 11.86 | 787 | 10.12 | 576 | 11.11 |
| 45–54 | 1115 | 21.51 | 1043 | 20.12 | 989 | 19.08 | 1384 | 17.8 | 893 | 17.23 |
| 55–64 | 1379 | 26.6 | 1497 | 28.88 | 1512 | 29.17 | 2143 | 27.56 | 1517 | 29.26 |
| 65+ | 1839 | 35.47 | 1750 | 33.76 | 1773 | 34.2 | 2996 | 38.53 | 1896 | 36.57 |
| Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head | Current marital status of household head |
| Not Married | 1948 | 37.59 | 1924 | 37.11 | 1889 | 36.44 | 2967 | 38.16 | 1932 | 37.27 |
| Married | 3234 | 62.41 | 3260 | 62.89 | 3295 | 63.56 | 4809 | 61.84 | 3252 | 62.73 |
| Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head | Level of Education of household head |
| No Education | 40 | 0.77 | 35 | 0.68 | 24 | 0.46 | 44 | 0.57 | 16 | 0.31 |
| Primary | 691 | 13.33 | 616 | 11.88 | 594 | 11.46 | 825 | 10.61 | 535 | 10.32 |
| Secondary | 2237 | 43.15 | 2342 | 45.18 | 2343 | 45.2 | 3486 | 44.83 | 2472 | 47.69 |
| Tertiary | 2216 | 42.75 | 2191 | 42.26 | 2223 | 42.88 | 3421 | 43.99 | 2161 | 41.69 |
| Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension | Whether at least one member has hypertension |
| | 4696 | 90.59 | 4651 | 89.72 | 4640 | 89.51 | 7138 | 91.8 | 4782 | 92.25 |
| Yes | 488 | 9.41 | 533 | 10.28 | 544 | 10.49 | 638 | 8.2 | 402 | 7.75 |
| Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled | Whether at least one member is disabled |
| | 4279 | 82.54 | 4290 | 82.75 | 4308 | 83.1 | 6473 | 83.24 | 4439 | 85.63 |
| Yes | 905 | 17.46 | 894 | 17.25 | 876 | 16.9 | 1303 | 16.76 | 745 | 14.37 |
| Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner | Whether at least one member is a pensioner |
| | 2585 | 49.86 | 2647 | 51.06 | 2577 | 49.71 | 3769 | 48.47 | 2556 | 49.31 |
| Yes | 2599 | 50.14 | 2537 | 48.94 | 2607 | 50.29 | 4007 | 51.53 | 2628 | 50.69 |
| Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits | Whether at least one member receives social benefits |
| | 4886 | 94.25 | 4899 | 94.5 | 4935 | 95.2 | 7218 | 92.82 | 4931 | 95.12 |
| Yes | 298 | 5.75 | 285 | 5.5 | 249 | 4.8 | 558 | 7.18 | 253 | 4.88 |
| Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group | Whether at least one member is in the military social group |
| | 5125 | 98.86 | 5105 | 98.48 | 5109 | 98.55 | 7640 | 98.25 | 5109 | 98.55 |
| Yes | 59 | 1.14 | 79 | 1.52 | 75 | 1.45 | 136 | 1.75 | 75 | 1.45 |
| Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group | Whether at least one member is in the children’s social group |
| | 4986 | 96.18 | 5012 | 96.68 | 5028 | 96.99 | 7603 | 97.78 | 5055 | 97.51 |
| Yes | 198 | 3.82 | 172 | 3.32 | 156 | 3.01 | 173 | 2.22 | 129 | 2.49 |
| Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups | Whether at least one member has access to the BBP for vulnerable groups |
| | 4274 | 82.45 | 4171 | 80.46 | 4151 | 80.07 | 5971 | 76.79 | 4238 | 81.75 |
| Yes | 910 | 17.55 | 1013 | 19.54 | 1033 | 19.93 | 1805 | 23.21 | 946 | 18.25 |
| Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance | Whether at least one member has health insurance |
| | 4572 | 88.19 | 4511 | 87.02 | 4363 | 84.16 | 6716 | 86.37 | 4519 | 87.17 |
| Yes | 612 | 11.81 | 673 | 12.98 | 821 | 15.84 | 1060 | 13.63 | 665 | 12.83 |
| Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work | Whether at least one member has some paid work |
| | 1286 | 24.81 | 1302 | 25.12 | 1327 | 25.6 | 1952 | 25.1 | 1233 | 23.78 |
| Yes | 3898 | 75.19 | 3882 | 74.88 | 3857 | 74.4 | 5824 | 74.9 | 3951 | 76.22 |
| Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area | Whether household is in an urban area |
| Rural | 3348 | 64.58 | 3348 | 64.58 | 3348 | 64.58 | 5184 | 66.67 | 1944 | 37.5 |
| Urban | 1836 | 35.42 | 1836 | 35.42 | 1836 | 35.42 | 2592 | 33.33 | 3240 | 62.5 |
| Household size | Household size | Household size | Household size | Household size | Household size | Household size | Household size | Household size | Household size | Household size |
| Small (1 to 2 members) | 1511 | 29.15 | 1607 | 31 | 1594 | 30.75 | 2580 | 33.18 | 1777 | 34.28 |
| Average (3 to 4 members) | 1776 | 34.26 | 1847 | 35.63 | 1844 | 35.57 | 2632 | 33.85 | 1829 | 35.28 |
| Bigger (5+ members) | 1897 | 36.59 | 1730 | 33.37 | 1746 | 33.68 | 2564 | 32.97 | 1578 | 30.44 |
| Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status | Household socioeconomic status |
| Poorest | 1227 | 23.67 | 1202 | 23.19 | 1264 | 24.38 | 1704 | 22.19 | 1222 | 24.06 |
| Poorer | 1131 | 21.82 | 1034 | 19.95 | 1012 | 19.52 | 1566 | 20.4 | 1056 | 20.8 |
| Middle | 1098 | 21.18 | 1177 | 22.7 | 1101 | 21.24 | 1681 | 21.89 | 1038 | 20.44 |
| Rich | 821 | 15.84 | 885 | 17.07 | 1110 | 21.41 | 1395 | 18.17 | 856 | 16.86 |
| Richest | 907 | 17.5 | 886 | 17.09 | 697 | 13.45 | 1332 | 17.35 | 906 | 17.84 |
Sampled households had varied access to governmental benefits or additional financial support. Whereas about half of surveyed households had at least one pensioner, fewer than one in ten sampled households had a member who received social benefits. By 2018, nearly one in five surveyed households had a member covered by the BBP for vulnerable groups, while just over a tenth of the surveyed households had a member with access to private health insurance. Less than two percent of surveyed households had a member in the military, while approximately three percent had a child who qualified as a part of the children’s group [8].
## Health care costs–OOP payments
Table 2 shows the average and median household OOP payments for outpatient, inpatient, and total health care costs from 2014 to 2018 in constant 2018 Armenian Drams (AMD). Within each year, the average and median OOP payments for outpatient costs were higher than those for inpatient care. Overall, the mean OOP payments per household increased significantly from AMD 118,992 (95 percent confidence intervals (CI): 110,436–127,548) in 2014 to AMD 137,360 (95 percent CI: 119,581–155,140) in 2016 (p-value<0.009) and then dropped to AMD 124,733 (95 percent CI: 115,049–134,418) in 2017 (p-value = 0.209) and AMD 114,613 (95 percent CI: 98,235–130,992) in 2018 (p-value = 0.823). Between 2014 and 2018, the mean total OOP payments by households declined by 3.68 percent, with a larger share of this contributed by OOP payments for inpatient care, which declined by 11.57 percent while outpatient care also declined by 3.19 percent.
**Table 2**
| Unnamed: 0 | Unnamed: 1 | Outpatient (AMD) | Inpatient (AMD) | Total health care costs (AMD) |
| --- | --- | --- | --- | --- |
| 2014 | N | 5184 | 5184 | 5184 |
| 2014 | Mean [95% CI] | 111,994 [103,611–120,378] | 6,998 [5,896–8100] | 118,992 [110,436–127,548] |
| 2014 | Median [IQR] | 19,057 [0–104,176] | 0 [0–0] | 22,938 [0–114,339] |
| 2015 | N | 5184 | 5184 | 5184 |
| 2015 | Mean [95% CI] | 117,780 [98,100–137,460] | 6,695 [5,672–7,719] | 124,475 [104,723–144,227] |
| 2015 | Median [IQR] | 25,474 [0–122,474] | 0 [0–0] | 33,068 [0–122,474] |
| 2016 | N | 5184 | 5184 | 5184 |
| 2016 | Mean [95% CI] | 128,201 [110,557–145,844] | 9,160 [7,739–10,580] | 137,360 [119,581–155,140] |
| 2016 | Median [IQR] | 26,914 [0–31,675] | 0 [0–0] | 124,217 [0–124,217] |
| 2017 | N | 7776 | 7776 | 7776 |
| 2017 | Mean [95% CI] | 116,619 [107,067–126,171] | 8,115 [7,161–9,068] | 124,733 [115,049–134,418] |
| 2017 | Median [IQR] | 27,065 [0–123,24] | 0 [0–0] | 30,756 [0–123,24] |
| 2018 | N | 5184 | 5184 | 5184 |
| 2018 | Mean [95% CI] | 108,425 [92,136–124,714] | 6,188 [5,211–7,165] | 114,613 [98,235–130,992] |
| 2018 | Median [IQR] | 22,200 [0–96,000] | 0 [0–0] | 24,000 [0–102,000] |
## Trends in the incidence of CHE and impoverishing health expenditure
Table 3 outlines CHE incidence at the 10 percent of total consumption and 40 percent of non-food expenditure thresholds. Similar trends in CHE are found upon examining both thresholds. At the 10 percent threshold, the incidence of CHE peaked in 2017, where 21.0 percent (95 percent CI: 20.02–22.02) of the population incurred CHE. On the other hand, the percentage of the population that incurred CHE is much lower each year when considering the 40 percent of non-food consumption threshold. For instance, the incidence of CHE was 7.93 percent (95 percent CI: 7.29–8.63) in 2017 using the 40 percent threshold. Irrespective of whichever threshold is used, over five percent of Armenians incurred CHE due to OOP payments for outpatient and inpatient care annually. Using the World Bank estimates of Armenia’s population in 2020, this translates to over 148,162 individuals [31]. Regardless of the threshold, fewer Armenian households incurred CHE in 2018 compared to 2014. For instance, 47,928 (representing 171,004 Armenians) and 289,131 (representing 1,031,592 Armenians) fewer households incurred CHE in 2018 compared to 2014, when the 10 percent and 40 percent thresholds are considered, respectively.
**Table 3**
| Unnamed: 0 | Threshold—10% of total consumption | Threshold—10% of total consumption.1 | Threshold—10% of total consumption.2 | Threshold—40% of non-food consumption | Threshold—40% of non-food consumption.1 | Threshold—40% of non-food consumption.2 |
| --- | --- | --- | --- | --- | --- | --- |
| | Percentage | 95% CI | 95% CI | Percentage | 95% CI | 95% CI |
| | Percentage | Lower | Upper | Percentage | Lower | Upper |
| 2014.0 | 19.86 | 18.65 | 21.13 | 8.50 | 7.67 | 9.42 |
| 2015.0 | 19.64 | 18.45 | 20.88 | 6.26 | 5.57 | 7.03 |
| 2016.0 | 19.50 | 18.32 | 20.73 | 6.77 | 6.06 | 7.56 |
| 2017.0 | 21.00 | 20.02 | 22.02 | 7.93 | 7.29 | 8.63 |
| 2018.0 | 18.71 | 17.55 | 19.93 | 5.53 | 4.87 | 6.27 |
Table 4 shows the results of the impoverishing effects of OOP payments using the US$1.90 per day international poverty line. Before incurring any health-related expenditure, 1.22 percent and 0.89 percent of Armenians lived below the international poverty line in 2014 and 2018, respectively. However, after spending on health, the poverty headcounts increased by 0.66 percent (95 percent CI: 0.45–0.97) [2014] and 0.70 percent (95 percent CI: 0.48–1.02) [2018], meaning that 19,300 and 20,662 Armenians were pushed into poverty in 2014 and 2018, respectively.
**Table 4**
| Unnamed: 0 | Gross OOP health care payments | Gross OOP health care payments.1 | Gross OOP health care payments.2 | Net OOP health care payments | Net OOP health care payments.1 | Net OOP health care payments.2 | Difference | Difference.1 | Difference.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Percentage | 95% CI | 95% CI | Percentage | 95% CI | 95% CI | Percentage | 95% CI | 95% CI |
| | Percentage | Lower | Upper | Percentage | Lower | Upper | Percentage | Lower | Upper |
| 2014.0 | 1.22 | 0.94 | 1.58 | 1.88 | 1.52 | 2.33 | 0.66 | 0.45 | 0.97 |
| 2015.0 | 1.00 | 0.75 | 1.33 | 1.33 | 1.04 | 1.69 | 0.33 | 0.21 | 0.53 |
| 2016.0 | 0.68 | 0.49 | 0.96 | 1.27 | 0.98 | 1.64 | 0.58 | 0.39 | 0.87 |
| 2017.0 | 0.46 | 0.34 | 0.64 | 0.84 | 0.66 | 1.07 | 0.37 | 0.26 | 0.55 |
| 2018.0 | 0.89 | 0.66 | 1.19 | 1.59 | 1.26 | 2.01 | 0.70 | 0.48 | 1.02 |
When applying the US$5.50 per day poverty line, there was a considerable shift in the proportion of households below the poverty line before and after spending on healthcare (Table 5). 38.32 percent and 31.61 percent of households were already below the poverty line before incurring any OOP expenditures for health in 2014 and 2018, respectively. After considering these expenditures, 4.28 percent and 4.77 percent more Armenian households were impoverished in 2014 and 2018, respectively.
**Table 5**
| Unnamed: 0 | Gross of OOP healthcare payments | Gross of OOP healthcare payments.1 | Gross of OOP healthcare payments.2 | Net of OOP healthcare payments | Net of OOP healthcare payments.1 | Net of OOP healthcare payments.2 | Difference | Difference.1 | Difference.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | Percentage | 95% CI | 95% CI | Percentage | 95% CI | 95% CI | Percentage | 95% CI | 95% CI |
| | Percentage | Lower | Upper | Percentage | Lower | Upper | Percentage | Lower | Upper |
| 2014.0 | 38.32 | 36.88 | 39.77 | 42.59 | 41.12 | 44.08 | 4.28 | 3.70 | 4.94 |
| 2015.0 | 34.92 | 33.52 | 36.34 | 39.52 | 38.07 | 40.99 | 4.60 | 4.01 | 5.28 |
| 2016.0 | 30.16 | 28.82 | 31.54 | 33.36 | 31.97 | 34.77 | 3.19 | 2.72 | 3.74 |
| 2017.0 | 30.46 | 29.36 | 31.58 | 36.00 | 34.85 | 37.17 | 5.54 | 5.00 | 6.13 |
| 2018.0 | 31.61 | 30.24 | 33.01 | 36.38 | 34.95 | 37.84 | 4.77 | 4.15 | 5.48 |
Fig 1 reports the incidence of impoverishing effect of OOP payments for health care in Armenia between 2014 and 2018. The percentage of Armenians pushed further into poverty after health care payments fluctuated between 2014 and 2018. On average, when we applied the US$1.90 per person per day international poverty line, 5.63 percent more Armenian households (translating to 166,138 Armenians) were pushed into poverty because of OOP payments for health care in 2018 compared to 2014.
**Fig 1:** *Trends in impoverishment in Armenia between 2014 and 2018.*
## Discussion
This study presents the most recent comprehensive update of catastrophic and impoverishing health expenditures in Armenia since 2013. Moreover, it is one of the first studies to examine the correlates of Armenia’s catastrophic and impoverishing health expenditures. Based on our analysis, Armenia’s OOP payments declined between 2014 and 2018 due to declining outpatient and inpatient costs. Although there was a general decrease in CHE incidence among Armenian households between 2014 and 2018, CHE dramatically increased between 2013 and 2014, and in 2017 –coinciding with a fall in per capita public health spending between 2016 and 2017—before dipping again in 2018 [32]. Similarly, impoverishing health spending fell overall between 2014 and 2018, despite a peak in 2017.
The variation in CHE and impoverishment between 2014–2018 can be attributed to various reasons. First, over time, there have been changes to the benefits package composition, such as services, tariffs, and qualifying groups [33, 34]. For example, the maximum age for children eligible for more generous coverage under the BBP has continued to change, from three years old in 2001 to eighteen in 2019 [34]. Moreover, the BBP’s service list differs over time as the list is primarily informed by recommendations from Armenia’s Minister of Health. Thus, variable coverage levels of services and populations may in part explain changes in OOP and the incidence of CHE and impoverishment. Second, Armenia’s wealth has increased over time. For instance, the Gross National Income (GNI) per capita (international US dollars) has increased by 26 percent, from US$ 10,490 in 2014 to US$ 13,230 in 2018 [35].
The increases in wealth translated to growth in household consumption. Monthly adult consumption increased overall between 2014 and 2018 from 47,622 AMD to 48,575 AMD–an annual increase of 24 percent [36]. The rise in consumption expenditure and parallel falls in OOP, translated to reductions in catastrophic health spending. In addition, the proportion of the population living below the UMI poverty line of $5.50 a day decreased between 2014 and 2016 by seven percent before trending upwards in 2017 (increase of four percent) [37]. Despite the overall falls in poverty and rise in household consumption in the study period, 2017 was marked by a decline in public health spending that shifted the burden of health care to households and led to a rise in impoverishing and catastrophic health spending. The improvement in poverty rate and increasing consumption over the studied period strengthens the argument that OOP payments are the main driver of increasing CHE. In other words, despite the increasing prosperity enjoyed by Armenians, it was insufficient in protecting households against the impoverishing effects of CHE. Despite efforts by the government to target vulnerable groups with generous coverage, the incidence of catastrophic and impoverishing health expenditures was disproportionally concentrated among Armenia’s more vulnerable socioeconomic groups between 2014 and 2018. Examining the impoverishing health expenditure concentration curves and index values demonstrates that the poorest Armenians are much more likely to be pushed into poverty from health spending than any other wealth quintiles. These disparities have also grown over time. While the wealthiest were less likely to experience CHE and impoverishment in 2018 compared to 2014, Armenia’s poorest were more likely to experience CHE and impoverishment in 2018 than in 2014 (i.e., more people in the lowest quintile experienced CHE and impoverishment in 2018 compared to 2014). The concentration of financial barriers to health care access among vulnerable groups suggests a need to increase or better target public spending for health benefits. Given ongoing efforts to target health benefits, these inequalities may also highlight the importance of universality in designing health benefits, which may improve equity and reduce inefficiencies arising from the cost of implementing targeting mechanisms. Our findings on the concentration of CHE and impoverishment among the poor are similar to those reported by other studies [3, 18, 38, 39].
Besides a household’s level of wealth, a significant predictor of CHE was whether a household member had hypertension, a common risk factor for chronic noncommunicable diseases. After accounting for all other factors, households with a hypertensive member were more than five times as likely to incur a CHE. Households with elderly over the age of 65 were also 47 percent more likely to experience a CHE than households without any elderly members, even if they could receive some form of social benefits. These findings are not just reflective of global trends but were expected as elderly and individuals with chronic conditions often tend to have more facility visits than young and healthy individuals [5]. Our findings point to the importance of strengthening financial protection for people living with NCDs in Armenia, which is essential to prevent the development of complications that are relatively more expensive to manage, and productivity losses due to premature death and preventable disability. Similarly, households with a disabled member had more than twice the odds of incurring CHE compared to households without any disabled members, regardless of their benefits or insurance status. Interestingly, in contrast to other countries’ experiences [40–42], urban residents had a greater odds of incurring CHE than their rural counterparts controlling for all other factors, including participation in social benefit programs. This may be explained by the fact that urban residents tend to bypass primary health care providers for expensive specialist care in the urban polyclinic model in which the scope of care of specialists and family physicians overlaps [43].
A critical protective factor against CHE was whether a household had at least one member with paid work. Regardless of one’s level of wealth or other characteristics, households with at least one member with paid work had 45 percent reduced odds of incurring CHE compared to households without paid work. The presence of paid work—informal or formal—may be a proxy for a higher ability to pay for health services, buttressing the fact that the poorest households continue to incur a disadvantage in terms of access to care.
As for the impoverishing impact of health spending, urban households, households with members belonging to the military social group, households with at least one pensioner, and households with a hypertensive patient had higher odds of IHE. Although households with a disabled member were more likely to face impoverishment, these odds were not statistically different from their non-disabled counterparts. Surprisingly, access to social benefits or Armenia’s BBP did not appear significantly impact the odds of experiencing impoverishing health expenditures, indicating Armenia’s benefits package does not offer a sufficient level of support to meaningfully protect Armenians from impoverishing health spending. Hence, there is a need to re-examine the depth and service scope of coverage in the BBP in Armenia to ensure that it confers adequate financial risk protection to all Armenians.
Armenia has made strides to expand services covered under the BBP, including providing primary health services for the general population as well as inpatient services for the poor and other vulnerable groups. However, with Armenia’s growing burden of chronic noncommunicable diseases and the high cost of outpatient and diagnostic treatment for most, the country’s current health system and BBP cannot address the growing disparities in catastrophic and impoverishing health spending. Similar to other European and Central Asian countries, outpatient medications are a key driver of OOP payments in Armenia [44, 45]. Without pharmaceutical pricing regulation, the VAT tax on medication, and a negative perception of generic drugs among patients and providers, Armenia’s reliance on OOP payments to finance expensive outpatient treatment puts many households at financial risk [45].
Increasing the government’s financing for health would allow the government to strategically finance more health services with prepaid public resources [46]. For example, in a study of Eastern European countries, Estonia not only has the second highest public spending on health and second lowest levels of OOP payment but also offers a generous package that covers most hospital care with cost-sharing elements for pharmaceuticals, dental care, and therapeutic appliances [47, 48]. Several considerations may inform the options for financing an expanded benefits package in Armenia, including the population age structure, formal employment rates, the strength of tax administrative mechanisms, and the broader fiscal policies [5, 9, 31].
The study has several key limitations. The analysis of Armenia’s ILCS demonstrates correlations between factors and outcomes and does not identify any causal relationships between specific factors and the levels of health expenditures. The analysis cannot directly ascertain the individual circumstances that lead to differing household healthcare spending and garner further study. Similarly, while the authors conducted a literature review of common correlates of OOP payments, CHE, and IHE, the study may be at risk of omitted variable bias if there are other context-specific variables that we did not account for in our analysis. Given the survey’s focus on Armenia, the results of our analysis may not apply to other contexts outside of Armenia.
This study did not include direct non-medical costs, such as transportation, due to data unavailability, which have proven to increase the incidence of CHE and IHE in other settings [18, 39]. Lastly, although examining healthcare expenditures precludes an analysis of those who may not seek care, analyzing the burden of catastrophic and impoverishing health expenditures provides insight into how cost affects healthcare use broadly amongst Armenians, which may also have implications for those who do not seek out healthcare. Future studies should include and explore these factors.
## Conclusion
This study offers a detailed examination of the burden of catastrophic and impoverishing health spending in Armenia to date. As the analysis draws from a nationally representative survey of Armenia, the results are generalizable throughout the population, providing a detailed picture of what Armenians are likely to experience on average. Investigating the association between socioeconomic factors and healthcare spending trends in Armenia provides greater insight into the progress Armenia has made on healthcare spending and which groups need more robust support to reduce the incidence and intensity of CHE and IHE. Considering nearly a fifth of Armenian households experienced a CHE in 2018, Armenia’s efforts to expand UHC and offer patients greater financial risk protection may require the country to undertake reforms to increase prepaid pooled financing to finance an expanded benefits package for the whole population.
## References
1. Grépin KA, Irwin BR, Sas Trakinsky B. **On the Measurement of Financial Protection: An Assessment of the Usefulness of the Catastrophic Health Expenditure Indicator to Monitor Progress Towards Universal Health Coverage.**. *Health Syst Reform* (2020.0) **6** e1744988. DOI: 10.1080/23288604.2020.1744988
2. 2World Health Organization Global Health Expenditure Database. Out-of-pocket (OOPS) as% of Current Health Expenditure (CHE), Year-2018 [Internet]. 2021. Available from: https://data.worldbank.org/indicator/SH.XPD.OOPC.CH.ZS
3. Wagstaff A, Flores G, Hsu J, Smitz MF, Chepynoga K, Buisman LR. **Progress on catastrophic health spending in 133 countries: a retrospective observational study.**. *Lancet Glob Health.* (2018.0) **6** e169-79. DOI: 10.1016/S2214-109X(17)30429-1
4. 4World Health Organization, International Bank for Reconstruction and Development. Global monitoring report on financial protection in health 2019 [Internet].
Geneva: World Health Organization; 2020. Available from: https://apps.who.int/iris/handle/10665/331748. *International Bank for Reconstruction and Development. Global monitoring report on financial protection in health 2019 [Internet].* (2020.0)
5. 5World Health Organization, World Bank. Global Monitoring Report on Financial Protection in Health 2021 [Internet].
Washington, DC: World Bank; 2021 Dec [cited 2022 Apr 18]. Available from: https://openknowledge.worldbank.org/handle/10986/36723. *Global Monitoring Report on Financial Protection in Health 2021 [Internet].* (2021.0)
6. 6World Health Organization Global Health Expenditure Database. Domestic general government health expenditure (% of GDP) [Internet]. 2021. Available from: http://apps.who.int/nha/database
7. 7World Health Organization Global Health Expenditure Database. Domestic general government health expenditure (% of general government expenditure) [Internet]. 2021. Available from: http://apps.who.int/nha/database
8. Lavado R, Hayrapetyan S, Kharazyan S. *Expansion of the Benefits Package: The Experience of Armenia.* (2018.0)
9. Fraser N, Chukwuma A, Koshkakaryan M, Yengibaryan L, Hou X, Wilkinson T. *Reforming the Basic Benefits Package in Armenia: Modeling Insights from the Health Interventions Prioritization Tool. [Internet]* (2021.0)
10. 10National Statistical Service of the Republic of Armenia, World Bank. Armenia Integrated Living Conditions Survey (ILCS) 2018, Ref. ARM_2018_ILCS_v01_M [Internet]. 2020. Available from: https://microdata.worldbank.org/index.php/catalog/3617
11. Farrington J, Kontsevaya A, Fediaev D, Grafton D, Khachatryan H, Schmitt A. *Prevention and control of noncommunicable diseases in Armenia. The case for investment [Internet]* (2019.0)
12. Harutyunyan T, Hayrumyan V. **Public opinion about the health care system in Armenia: findings from a cross-sectional telephone survey.**. *BMC Health Serv Res.* (2020.0) **20** 1005. DOI: 10.1186/s12913-020-05863-6
13. Azzani M, Roslani AC, Su TT. **Determinants of Household Catastrophic Health Expenditure: A Systematic Review.**. *Malays J Med Sci* (2019.0) **26** 15-43. DOI: 10.21315/mjms2019.26.1.3
14. Ravangard R, Jalali FS, Bayati M, Palmer AJ, Jafari A, Bastani P. **Household catastrophic health expenditure and its effective factors: a case of Iran.**. *Cost Eff Resour Alloc.* (2021.0) **19** 59. DOI: 10.1186/s12962-021-00315-2
15. Obse AG, Ataguba JE. **Assessing medical impoverishment and associated factors in health care in Ethiopia.**. *BMC Int Health Hum Rights.* (2020.0) **20** 7. DOI: 10.1186/s12914-020-00227-x
16. Mulaga AN, Kamndaya MS, Masangwi SJ. **Examining the incidence of catastrophic health expenditures and its determinants using multilevel logistic regression in Malawi.**. *PLOS ONE.* (2021.0) **16** e0248752. DOI: 10.1371/journal.pone.0248752
17. Li Y, Wu Q, Xu L, Legge D, Hao Y, Gao L. **Factors affecting catastrophic health expenditure and impoverishment from medical expenses in China: policy implications of universal health insurance**. *Bull World Health Organ* (2012.0) **90** 664-71. DOI: 10.2471/BLT.12.102178
18. Barasa EW, Maina T, Ravishankar N. **Assessing the impoverishing effects, and factors associated with the incidence of catastrophic health care payments in Kenya.**. *Int J Equity Health.* (2017.0) **16** 31. DOI: 10.1186/s12939-017-0526-x
19. Choi JW, Kim TH, Jang SI, Jang SY, Kim WR, Park EC. **Catastrophic health expenditure according to employment status in South Korea: a population-based panel study**. *BMJ Open* (2016.0) **6** e011747. DOI: 10.1136/bmjopen-2016-011747
20. Galárraga O, Sosa-Rubí SG, Salinas-Rodríguez A, Sesma-Vázquez S. **Health insurance for the poor: impact on catastrophic and out-of-pocket health expenditures in Mexico.**. *Eur J Health Econ.* (2010.0) **11** 437-47. DOI: 10.1007/s10198-009-0180-3
21. Aregbeshola BS, Khan SM. **Determinants Of Impoverishment Due To Out Of Pocket Payments In Nigeria.**. *J Ayub Med Coll Abbottabad JAMC.* (2017.0) **29** 194-9. PMID: 28718230
22. Arenliu Qosaj F, Froeschl G, Berisha M, Bellaqa B, Holle R. **Catastrophic expenditures and impoverishment due to out-of-pocket health payments in Kosovo.**. *Cost Eff Resour Alloc.* (2018.0) **16** 26. DOI: 10.1186/s12962-018-0111-1
23. Goginashvili K, Nadareishvili M, Habicht T. *Can people afford to pay for health care? New evidence on financial protection in Georgia.* (2021.0)
24. Xu K, Evans DB, Kawabata K, Zeramdini R, Klavus J, Murray CJ. **Household catastrophic health expenditure: a multicountry analysis**. *The Lancet* (2003.0) **362** 111-7. DOI: 10.1016/S0140-6736(03)13861-5
25. Wagstaff A, Doorslaer E van. **Catastrophe and impoverishment in paying for health care: with applications to Vietnam 1993–1998.**. *Health Econ* (2003.0) **12** 921-33. DOI: 10.1002/hec.776
26. O’Donnell O, Van Doorslaer E, Wagstaff A, Lindelow M. *Analyzing health equity using household survey data: a guide to techniques and their implementation.* (2008.0)
27. Cylus J, Thomson S, Evetovits T. **Catastrophic health spending in Europe: equity and policy implications of different calculation methods**. *Bull World Health Organ* (2018.0) **96** 599-609. DOI: 10.2471/BLT.18.209031
28. Wagstaff A.. **The concentration index of a binary outcome revisited.**. *Health Econ.* (2011.0) **20** 1155-60. DOI: 10.1002/hec.1752
29. Wagstaff A, van Doorslaer E, Paci P. **On the measurement of horizontal inequity in the delivery of health care.**. *J Health Econ.* (1991.0) **10** 169-205. DOI: 10.1016/0167-6296(91)90003-6
30. Xu Y, Gao J, Zhou Z, Xue Q, Yang J, Luo H. **Measurement and explanation of socioeconomic inequality in catastrophic health care expenditure: evidence from the rural areas of Shaanxi Province.**. *BMC Health Serv Res.* (2015.0) **15** 256. DOI: 10.1186/s12913-015-0892-2
31. 31World Bank Data Bank. Population, total—Armenia Year-2020 (SP.POP.TOTL) [Internet]. 2021. Available from: https://data.worldbank.org/indicator/SP.POP.TOTL?locations=AM
32. 32World Health Organization. Global Health Expenditures Database [Internet]. online; 2020. Available from: https://apps.who.int/nha/database
33. 33World Health Organization. Regional Office for Europe, Policies EO on HS and, Kutzin J, Cashin C, Jakab M. Implementing health financing reform: lessons from countries in transition [Internet]. World Health Organization. Regional Office for Europe; 2010 [cited 2022 Apr 23]. 411 p. Available from: https://apps.who.int/iris/handle/10665/326420
34. Chukwuma A, Messen B, Lylozian H, Gong E, Ghazaryan E. *Strategic Purchasing for Better Health in Armenia [Internet]* (2020.0)
35. 35World Bank Data Bank. GNI per capita, PPP (current international $)—Armenia (NY.GNP.PCAP.PP.CD) [Internet]. 2022. Available from: https://data.worldbank.org/indicator/NY.GNP.PCAP.PP.CD?end=2018&locations=AM&start=2014
36. 36Part 1: Armenia—Poverty snapshot over 2008–2017 [Internet]. Yerevan, Armenia: Statistical Committee Republic of Armenia; 2018 [cited 2022 Aug 22]. Available from: https://armstat.am/file/article/poverty_2018_english_2.pdf
37. 37Poverty headcount ratio at $5.50 a day (2011 PPP) (% of population)—Armenia | Data [Internet]. [cited 2022 Aug 22]. Available from: https://data.worldbank.org/indicator/SI.POV.UMIC?locations=AM
38. Piroozi B, Mohamadi-Bolbanabad A, Moradi G, Safari H, Ghafoori S, Zarezade Y. **Incidence and Intensity of Catastrophic Health-care Expenditure for Type 2 Diabetes Mellitus Care in Iran: Determinants and Inequality.**. *Diabetes Metab Syndr Obes Targets Ther* (2020.0) **13** 2865-76. DOI: 10.2147/DMSO.S263571
39. Mutyambizi C, Pavlova M, Hongoro C, Booysen F, Groot W. **Incidence, socio-economic inequalities and determinants of catastrophic health expenditure and impoverishment for diabetes care in South Africa: a study at two public hospitals in Tshwane.**. *Int J Equity Health* (2019.0) **18** 73. DOI: 10.1186/s12939-019-0977-3
40. zhi Fu X, wei Sun Q, qing Sun C, Xu F, jian He J. **Urban-rural differences in catastrophic health expenditure among households with chronic non-communicable disease patients: evidence from China family panel studies.**. *BMC Public Health.* (2021.0) **21** 874. DOI: 10.1186/s12889-021-10887-6
41. Iamshchikova M, Mogilevskii R, Onah MN. **Trends in out of pocket payments and catastrophic health expenditure in the Kyrgyz Republic post “Manas Taalimi” and “Den Sooluk” health reforms, 2012–2018.**. *Int J Equity Health* (2021.0) **20** 30. DOI: 10.1186/s12939-020-01358-2
42. 42World Bank Group. Tracking universal health coverage: 2017 global monitoring report (English) [Internet]. Washington D.C.; Available from: http://documents.worldbank.org/curated/en/640121513095868125/Tracking-universal-health-coverage-2017-global-monitoring-report
43. 43Armenia—Achievements and Challenges in Improving Health Care Utilization: A MultiProject Evaluation of the World Bank Support to the Health System Modernization (2004–2016) [Internet].
Washington, D.C.: World Bank Group; 2019
Mar [cited 2022 Aug 18]. (Independent Evaluation Group, Project Performance Assessment Report 134584). Available from: https://documents1.worldbank.org/curated/en/725551557425652702/pdf/Armenia-Achievements-and-Challenges-in-Improving-Health-Care-Utilization-A-Multiproject-Evaluation-of-the-World-Bank-Support-to-the-Health-System-Modernization-2004-2016.pdf. *Armenia—Achievements and Challenges in Improving Health Care Utilization: A MultiProject Evaluation of the World Bank Support to the Health System Modernization (2004–2016) [Internet].* (2019.0)
44. Thomson S, Cylus J, Evetovits T, Srakar A. *Can people afford to pay for health care? new evidence on financial protection in Europe: regional report [Internet]* (2019.0)
45. 45World Health Organization Regional Office for Europe. Pharmaceutical pricing and reimbursement systems in Eastern Europe and Central Asia (2020) [Internet]. Copenhagen, Denmark; 2020 [cited 2021 Oct 26]. Available from: https://www.euro.who.int/en/health-topics/Health-systems/health-technologies-and-medicines/publications/2020/pharmaceutical-pricing-and-reimbursement-systems-in-eastern-europe-and-central-asia-2020
46. Wagstaff A, Eozenou P, Smitz M. **Out-of-Pocket Expenditures on Health: A Global Stocktake.**. *World Bank Res Obs.* (2020.0) **35** 123-57
47. Tambor M, Klich J, Domagała A. **Financing Healthcare in Central and Eastern European Countries: How Far Are We from Universal Health Coverage?**. *Int J Environ Res Public Health* (2021.0) **18** 1382. DOI: 10.3390/ijerph18041382
48. 48OECD/European Observatory on Health Systems and Policies. Estonia Country Health Profile 2021. Brussels: OECD Publishing; 2021. (State of Health in the EU).. *Estonia Country Health Profile 2021* (2021.0)
|
---
title: 'Multimorbidity and health-related quality of life among patients attending
chronic outpatient medical care in Bahir Dar, Northwest Ethiopia: The application
of partial proportional odds model'
authors:
- Fantu Abebe Eyowas
- Marguerite Schneider
- Shitaye Alemu Balcha
- Sanghamitra Pati
- Fentie Ambaw Getahun
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021695
doi: 10.1371/journal.pgph.0001176
license: CC BY 4.0
---
# Multimorbidity and health-related quality of life among patients attending chronic outpatient medical care in Bahir Dar, Northwest Ethiopia: The application of partial proportional odds model
## Abstract
### Background
Multimorbidity, the presence of two or more chronic non-communicable diseases (NCDs) in a given person affects all aspects of people’s lives. Poor quality of life (QoL) is one of the major consequences of living with multimorbidity. Although healthcare should support multimorbid individuals to achieve a better quality of life, little is known about the effect of multimorbidity on the QoL of patients living with chronic conditions. This study aimed to determine the influence of multimorbidity on QoL among clients attending chronic outpatient medical care in Bahir Dar, Northwest Ethiopia.
### Methodology
A multi-centered facility-based study was conducted among 1440 participants aged 40+ years. Two complementary methods were employed to collect sociodemographic and disease related data. We used the short form (SF-12 V2) instrument to measure quality of life (QoL). The data were analyzed by STATA V.16, and a multivariate partial proportional odds model was fitted to identify covariates associated with quality of life. Statistical significance was considered at p-value <0.05.
### Principal findings
Multimorbidity was identified in $54.8\%$ ($95\%$ CI = $52.2\%$-$57.4\%$) of the sample. A significant proportion ($33.5\%$) of the study participants had poor QoL and a quarter ($25.8\%$) of them had moderate QoL. Advanced age, obesity and living with multimorbidity were the factors associated with poor QoL. Conversely, perceived social support and satisfaction with care were the variables positively associated with better QoL.
### Conclusion
The magnitude of multimorbidity in this study was high and individuals living with multimorbidity had a relatively poorer QoL than those without multimorbidity. Care of people with chronic multiple conditions has to be oriented to the realities of multimorbidity burden and its implication on QoL. It is also imperative to replicate the methods we employed to measure and analyze QoL data in this study for facilitating comparison and further development of the approaches.
## Introduction
The increasing demographic and social changes with ageing populations are leading to rapid epidemiological transitions, including the rise of chronic non-communicable diseases (NCDs) and multimorbidity [1].
Multimorbidity refers to the presence of two or more coexisting long-term conditions, being related or not in a given person [2]. Despite the inconsistency in the methodologies employed to define and measure multimorbidity [3], the burden of multimorbidity is shown to be growing globally [4, 5], in LMICs [6, 7] and Ethiopia [8, 9]. The prevalence of multimorbidity is projected to double by 2035, and the majority of the people surviving beyond 65 years will have four or more chronic conditions [10, 11].
Advanced age [4, 12–15], socioeconomic deprivation [15], obesity [16], female sex [17, 18], physical inactivity [19], use of tobacco and alcohol [20, 21] and psychosocial factors such as poor social networks and external locus of control [22, 23] were the most common factors associated with multimorbidity in the global literature.
Multimorbidity affects all aspects of patients’ lives. Poor quality of life (QoL) along with disability, functional decline and high health care costs are major consequences of living with multimorbidity [19, 24, 25]. The effect of multimorbidity on QoL was reported to be high among middle-aged, elderly and female population groups, and individuals with comorbid mental illnesses [17].
Living with multimorbidity is beyond the mere sum of individual chronic conditions [26]. The specific disease clusters that an individual is living with would have a different effect on their physical and psychological functioning [27]. For instance, the health-related QoL has always been lower among people living with multimorbidity compared to those without multimorbidity [28].
Lifelong presence of multimorbidity is also posing a significant challenge for the health system [29]. Individuals living with different types and combinations of NCDs may have different needs and priorities [30]. Nevertheless, too little attention is paid to what matters to people living with multiple health problems [31, 32].
Moreover, the current model of care and guidelines being developed at a time when single disease frameworks were predominant, tend to focus on diseases in isolation rather than the needs and circumstances of the person with complexity care needs as a whole [33, 34]. As a result, people living with multiple conditions will be in contact with multiple health professionals who may not communicate each other and with inadequate data flow across the healthcare system, leading to a fragmented, uncoordinated and siloed patient management [2, 35]. Further, the rapid emergence of infections such as COVID-19 fuels the complexity and adds a huge burden to the health system with worsening outcomes for patients with preexisting chronic diseases and multimorbidity [36, 37].
Several valuable studies investigated the relationship between QoL and multimorbidity [38, 39]. However, most of them were conducted in high-income countries and the tools employed to measure QoL among people living with multimorbidity have not been consistent [40]. Some studies in high-income countries used Euro QoL (EQ-5D-5l) [41, 42], while others used either WHOQOL brief [40] or SF-36 [40, 43] or SF-12 tools [40, 44–46]. Although the use of all of these tools has been widespread, the short form (SF-12) version is an efficient algorism to reproduce SF-36 tool to measure health related quality of life [47].
The observed variations in the existing literature also included the methods of data analyses used [48]. The way the data have been generated is particularly important for analyzing quality of life assessments scores [49]. Health related QoL is often measured by Likert-type scales and the scores are treated as if they were continuous and normally distributed, which often is not the case [50]. Scholars in the field have noted that analyzing ordinal data as if they were a metric one (continuous) can systematically lead to distorted effect-size estimates, inflated errors rates and inaccurate parameter estimates [51, 52].
Neither are the methods used for binary data adequate to fully take account of the properties of ordered outcomes such as QoL [48, 53]. Hence, a more sensitive and comprehensive model is required. Evidence suggests that the ordinal regression models are superior to the methods commonly used to analyze data of an ordered nature [54, 55]. The ordinal models provide better theoretical interpretation and numerical inference than the metric (linear) models for ordered outcomes [56, 57].
The ordinal regression model provides unbiased estimates when the data meet the proportional odds (PPO) assumption [53, 56]. The PPO assumption implies that all observations have a common variance on the underlying continuum, and the coefficients that describe the relationship between, say, the lowest versus all higher categories of the response variable are the same except in the cut-off points [49, 55, 58] However, it is often difficult to find data for which a proportional odds model is a plausible description, and evidence suggests that the assumptions of the ordered logit (proportional odds) model are frequently violated [54]. When the given data violates the parallel regression assumption, a more realistic approach, the partial proportional odds (PPO) model would be suitable [54]. This model is robust in revealing unobserved heterogeneity in the group and identify correlates contributing to negative health outcomes, including impaired QoL [48, 53]. The primary reason for the formulation of the partial proportional odds models is to relax the stringent assumption of constant odds ratio over all the cut-points for a given covariate [55].
Supporting people living with long-term conditions to maintain a good quality of life is one of the key challenges facing the healthcare and social care systems today [25]. Studies have suggested that the management of patients with multimorbidity should take into account the impact of multimorbidity on a person’s quality of life and the person’s own priorities [59, 60]. However, little is known about the effect of multimorbidity on health-related QoL in Ethiopia. If health systems are to meet the needs and priorities of individuals living with multimorbidity, we need to adequately measure the magnitude and impact of multimorbidity on QoL among patient population with chronic conditions.
The present study aimed to understand the influence of multimorbidity on QoL among individuals attending chronic outpatient medical care in Bahir Dar, Northwest Ethiopia.
## Materials and methods
This facility-based study was conducted in eight health facilities providing chronic NCDs care in Bahir Dar City, Ethiopia. This study presents the result of one part of an ongoing research project in the region. All methods were carried out in accordance with relevant guidelines and regulations. The detail of the methods employed in this study has been published elsewhere [61].
## Design
This multi-center, facility-based cross-sectional study was conducted in public and private health facilities rendering health services in Bahir Dar City, Ethiopia. The city is the capital of the Amhara regional state, the second most populous region in the country with a population of about 31 million.
## Study setting and population
This study was conducted in five hospitals (three public and two private) and three private higher/specialty clinics in the city. These facilities also serve as referral center for primary care facilities surrounding the regional capital. Chronic NCDs care and management in these facilities is supposed to be provided in a relatively uniform fashion using the national NCDs treatment guideline [62]. However, the nature of patients visiting these facilities may vary and there remains a concern on the quality and affordability of NCDs care in public hospitals and private health facilities, respectively.
Only facilities which were providing chronic NCDs care by medical doctors (general practitioners or specialist physicians) for at least a duration of one year prior to the data collection period were considered. Older adults (40 years or more) diagnosed with at least one NCD and on chronic diseases follow up care for at least six months at the time of data collection were recruited into the study. However, pregnant women and individuals who were too ill to be interviewed and admitted patients were excluded.
## Sample size
Key issues considered to estimate the sample size required were the nature of the dependent and predictor variables and the anticipated data analysis techniques. The input values: α (type I error = 0.05), power (1-β = 90), confidence level ($95\%$) and the estimated non-response and attrition during follow-up ($20\%$) remain constant. The sample size yielded by the general linear multivariate model with Gaussian errors (GLIMMPSE) sample size and power calculator(32–34) formula was chosen owing to its adequacy to answer all the study objectives compared to other techniques. Based on the given assumptions and the approach we used, the calculated sample size required became 600. As the nature of participants is likely to be different by the type of facility (public or private) where they receive care, we employed stratification to ensure fair representation in the sample for important sub-groups analysis. Hence, a design effect of 2 was considered to avoid the possible loss of sample during stratification. Adding $20\%$ to the possible loss to follow-up (considering the longitudinal study) and nonresponse, the sample size needed was calculated to be 1440. About half ($$n = 728$$, $50.84\%$) of the participants were enrolled from public facilities.
## Sampling technique
A stratified random sampling method was employed for recruiting eight eligible facilities and a corresponding number of participants. The sample size from each facility was determined based on the notion of probability proportional to size (PPS) using the pool of chronic NCD patients (≥ 40yrs) registered for chronic follow-up over the year preceding our assessment (January—December 2020) in each participating facility. Health facilities and eligible clients were randomly selected for the study.
## Definition and measurement of dependent variable (HRQoL)
HRQoL (stated as QoL in this study) is defined as individuals’ perception of their position in life in the context of physical, psychological and social functioning and well-being [63]. QoL was measured using the interviewer-administered short form (SF-12 V2) assessment tool [64, 65], which is derived from the SF-36 QoL assessment tool [47].
The SF-12 tool is extensively validated and widely used generic tool for measuring QoL in multimorbidity across different contexts, including Sub-Saharan Africa [46, 66, 67]. The tool was translated and pilot tested according to the study protocol we published [61]. The tool measures eight health aspects, namely physical functioning (PF), role limitations due to physical health problems (RP), bodily pain (BP), general health perceptions (GH), vitality (VT), social functioning (SF), role limitations due to emotional problems (RE), and mental health (psychological distress and psychological well-being) (MH). Two summary measures are derived from the SF-12: physical health (Physical Component Summary-PCS) and mental health (Mental Component Summary -MCS). However, owing to the possibility of correlation (lack of uni-dimensionality) between the PCS and MCS scores, some studies criticized the use of these scoring algorithms and recommended raw sum scores instead [47, 52]. The use of a single raw sum score enables a consistent assessment of the impact of multimorbidity and how this varies across a given population [68]. Thus, we applied this approach for analyzing the QoL data.
First, we reverse coded the scores for items 1, 9 and 10 and computed the raw total. The overall scores were scaled from 0 to 100, with 0 representing worst health [69]. Although popularly used in previous studies, the notion of fitting linear regression models to summarize categorical data such as the QoL data has been questioned [49, 52]. The linear regression models may potentially lose important variability in the data particularly when the QoL data is collected by Liker-type scales such as the SF-12 tool [48, 53]. Recent advances in the field recommend the interpretation of QoL rather as a categorical (group continuous) variable than as a metric variable [49]. Studies suggest that ordinal regression models (OLR) are superior to other method for analyzing ordinal data, including health-related QoL data [49, 50]. Hence, we ranked the scaled QoL scores into three ordered and non-overlapping categories as poor QoL (a scaled value <75), moderate QoL (scaled value from 75–89.9) and high QoL (scaled value from 90–100) [53]and fitted into the OLR and partial proportional odds (PPO) models.
## Measurement of independent variables
Independent variables including socio-demographic characteristics [age, gender, education, marital status, residence and occupation] were assessed using validated tools. Data to calculate body mass index (BMI) and waist to hip circumference were directly measured from patients according to the approaches described in our study protocol published elsewhere [61].
Social networking and support systems were assessed through face-to-face interview using the Oslo Scale [70]. The tool was translated and pilot tested among 29 patients attending chronic care in health facilities which were not selected for the actual study. A scale ranging from 3–8 was interpreted as poor social support, 9–11 moderate social support and 12–14 strong social support) [70].
A wealth Index at a household level was generated from a combination of material assets and housing characteristics [71]. The Wealth index was scored using principal component analysis (PCA) technique. The score was classified into quintiles, for urban and rural residents separately., Quintile 1 represents the poorest and quintile 5 the wealthiest [72]. It was collapsed into three classes as low, middle and high income to ease the analysis and interpretation.
Multimorbidity was operationalized as the co-occurrence of two or more of the chronic NCDs, including hypertension, diabetes, heart diseases (heart failure, angina and heart attack), stroke, bronchial asthma, chronic obstructive pulmonary diseases (COPD), depression, cancer, musculoskeletal disorders (arthritis, chronic back pain and osteoporosis), thyroid disorders (hyperthyroidism and multinodular goiter), chronic kidney disease, gastrointestinal disorders (chronic liver, gall bladder and gastric diseases) and Parkinson’s disease (PD). The list of NCDs identified for the study was determined based on a review study [6] and preliminary and pilot studies conducted prior to the main study. Information on these chronic conditions was obtained from interview and review medical records using standardized tools [61].
Functional status (limitation) was assessed using the WHO’S Disability Assessment tool (WHODAS 2.0) [73]. The tool is a five-point scale (none = 1, mild = 2, moderate = 3, severe = 4, extreme/cannot do = 5). The results of score on the 12 items were summed up and categorized into three (with a score ≤50 as no limitation, 51–75 moderate limitation and 76–100 severe limitation). The 12 items WHODAS 2.0 has been validated and used in Ethiopia [74].
## Data collection tools and procedures
As mentioned above, the data were collected mainly from two different sources: interview and review of medical records. The combined questionnaire to collect the data was translated to Amharic (local language) and pilot tested for cross-cultural adaptability based on standard protocols [75, 76]. The data were collected by the Kobo Toolbox software [77]. Patients were interviewed and assessed following their regular consultation visit. Data were primarily collected by ten graduate nurses recruited from institutions outside the study facilities. Moreover, physicians and nurses working in the chronic care unit were facilitated in the data collection process.
To ensure good data quality, data collectors and supervisors were provided with a two-day training detailing the study, including obtaining written consent, conducting face-to-face interviews, performing physical measurement, medical record review and navigating through the questionnaires in the Kobo toolbox platform preloaded into their smart phones. The data collection process was monitored by trained supervisors, and the principal investigator. The data sent to the Kobo toolbox server were checked daily for completeness, accuracy and clarity. Feedback and coaching support were given for the data collectors on the quality and completeness of the information they submitted daily. Once the data were sent to the server, they were deleted from the data collection devices, and deletion of any data remained from smart phones and the questionnaire was made by the PI immediately after completion of the data collection.
## Data analysis
The data from the Kobo toolbox server were downloaded into excel spreadsheet and exported to SPSS V. 21 for cleaning and were analyzed using STATA V. 16. Descriptive statistics were computed to describe the sociodemographic characteristics of participants. The number of individual chronic conditions and multimorbidity status was determined by combining the data from patient interviews and medical record reviews.
In addition, the proportion of individuals falling into each of the QoL category was calculated. QoL as an ordered outcome was categorized as low, moderate and high, and coded as 0, 1 and 2, respectively while fitting into the ordinal logistic regression model. The association between each explanatory variable and QoL was assessed separately and model fitness was checked using the proportional odds (test for parallel regression) assumption [54]. The proportional odds (PO) assumption is said to be satisfied when we fail to reject the null hypothesis (a p-value of >0.05 in the Brant test) in the ordinal logistic regression model [54, 55].
For variables which fail to satisfy the PO assumptions, the partial proportional odds (PPO) model is more appropriate [48, 53], as the OLR model cannot fit the data well [55]. The partial proportional odds (PPO) model bridge the gap between ordered and non-ordered modeling frameworks [56, 78]. While the ordinal logistic regression model is restrictive and assumes that the effect of independent variables remain the same (fixed) for all levels of the dependent variables, the PPO allows the independent variables to take into account the individual differences in their effect on the dependent variables[55, 79]. Compared to the OLR model, the PPO has performed well in studies that compared different analytical models fitted for QoL data [49, 53]. Hence, we fitted the PPO (gologit2, autofit lrforce and gologit2, autofit lrforce gamma commands) model for determining covariates associated with QoL and to clearly identify the variables which violate the assumptions.
The independent variables fitted into the PPO model include residence, sex, age, marital status, education, BMI, social support, SES, multimorbidity, self-rated functional capacity and satisfaction with care. Independent variables having more than two categories were collapsed into two categories while fitting the PPO model [54]. The association between QoL and independent variables was assessed by fitting univariate and multivariate odds ratio (OR) with $95\%$ confidence intervals and p-values are reported for each of the independent variable analyzed. Variables having a p-value <0.2 were fitted into multivariable PPO models to predict the adjusted effect of the independent variables on QoL. Before running the multivariable analysis, multi-collinearity between independent variables was checked using the Variance Inflation Factor (VIF) and variables were not strongly correlated (the highest value was 1.05). To make the interpretation more straightforward, we expressed the effects in terms of odds ratio rather than as regression coefficients [53]. In all cases, a p-value < 0.05 was taken as a statistically significant relationship.
We have conducted the univariate and multivariate analyses using the svyset, svy and gsvy prefixes to account for the potential difference of the sample drawn from the two strata (public and private health facilities) [80].
Looking at the Brant test of significance, most of the explanatory variables satisfied the proportional odds assumption. However, two independent variables (social network scale and satisfaction with care) violated the assumption of parallel lines regression (p-value ≤0.05), warranting the application of multivariate partial proportional odds model. Considering the effect of stratification, we also applied syv prefix before the ologit command, and output shows changes in the parameter estimates, brant test values and level of significance in the univariate analysis.
As stated above, the nature of the independent variables necessitated fitting of the partial proportional odds (PPO) model. The partial PPO model allows variables that meet the assumption to be modeled with the proportional odds assumption, whilst allowing others to have odds ratios that vary for the different categories that are compared. Only the variables with a p-value <0.2 in the univariate (using svy syntax) ordinal logistic regression analysis were fitted into the multivariate partial proportional odds model using gsvy (glogit2) command accounting for the stratified sampling.
Fitting the partial proportional odds assumption requires that the independent variables to have only two categories. Accordingly, except for age and social support score, we coded independent variables as a binary [0, 1] response category, where higher values were coded as “1” and low values were given “0” and treated as a base category. Therefore, sex was coded as male [0] and female [1], SES as low [0] and middle or high [1], BMI as ≤24.99 [0] and ≥25 [1], multimorbidity was scored as no [0] and yes [1], functioning was scored as severe limitation [0] and no or mild limitation [1] and satisfaction was scored as not satisfied [0] and satisfied [1]. Whereas, age and social support scales were treated as continuous independent variables.
Without adjustment and weighting, application of a stratified sampling method could affect parameter estimates of a given sample [56, 80]. Hence, we applied a more stringent model of analysis using the gsvy: gologit2 auto lrforce command.
In the final multivariate ordinal logistic regression model, the Wald test of parallel-lines assumption became significant (Chi-square = 31.25, p-value <0.001) indicating the need for fitting a less restrictive model, the partial proportional odds model through applying gologit2 and gsvy commands. The authors observed that using the prefix gsvy (to account for sampling stratification) in the multivariate partial proportional odds model changed the parameter estimates and level of significance compared to the baseline output of the default gologit2 auto-fit model (without gsvy prefix).
The outcome variable, QoL (Y) is categorized into three (poor, moderate and high), so the model produced two panels. The first panel contrasts category 1 (poor QoL) with category 2 (moderate QoL) and 3 (high QoL) and the second panel contrasts category 1 and 2 with category 3. An odds ratio value greater than 1 (positive coefficient) on the explanatory variable indicates that it is more likely that the respondent will be in a higher category of Y than the current one (increasing in the explanatory variable led to better levels of QoL); whereas, an odds value below 1 (negative coefficient) indicates the likelihood of being in the current or a lower category.
Since social support and satisfaction with care violated the proportional odds assumption, the odds ratios for this variable were allowed to vary between panels (AOR1 ≠ AOR2). AOR1 stands for panel one (low versus moderate or high QoL), while AOR2 refers to the second panel (low/ moderate versus high QoL). However, for the independent variables which met the parallel regression assumption (Brant test value ≥0.05), the odds ratio would be the same (AOR1 = AOR2) for the two panels.
## Ethics approval and consent to participate in the study
As this is a part of an ongoing PhD study, permission to conduct the study was obtained from the Institutional Review Board (IRB) of the College of Medicine and Health Sciences, Bahir Dar University with a protocol number $\frac{003}{2021.}$ Study participants were enrolled after giving verbal consent to participate in the study. The adequacy of oral consent was approved by the IRB and the consent was documented on participants’ information sheet. Permission was obtained from the health facilities involved in the study. Moreover, confidentiality of the data obtained from the study participants and medical records have been strictly maintained through anonymizing identities and applying pertinent legal and ethical protocols.
## Characteristics of the study participants
Complete data were obtained from 1432 individuals giving rise to a response rate of $99.4\%$. Females constitute a slightly higher ($51\%$) percentage in terms of sex distribution. The mean (±SD) age of the participants was 56.4 (±11.8) years. Individuals aged 45–54 years and 55–64 years accounted almost equally ($27.9\%$) for the age distribution and those aged 65+ had a $26.9\%$ share from the total sample (Table 1).
**Table 1**
| Variables | Frequency | Percentage |
| --- | --- | --- |
| Age group | | |
| ≤44Yrs | 247.0 | 17.3 |
| 45-54Yrs | 399.0 | 27.9 |
| 55-64Yrs | 400.0 | 27.9 |
| 65+Yrs | 386.0 | 26.9 |
| Sex | | |
| Male | 702.0 | 49.0 |
| Female | 730.0 | 51.0 |
| Marital status | | |
| Currently married | 1081.0 | 75.5 |
| Single* | 351.0 | 24.5 |
| Education | | |
| No formal education | 780.0 | 54.5 |
| Primary education (Grade 1–8) | 166.0 | 11.6 |
| Secondary (9–12) | 171.0 | 11.9 |
| College level and above | 315.0 | 22.0 |
| Residence | | |
| Urban | 1007.0 | 70.3 |
| Rural | 425.0 | 29.7 |
| Occupation | | |
| Housewife | 329.0 | 23.0 |
| Employed (government and private) | 328.0 | 22.9 |
| Farmer | 288.0 | 20.1 |
| Trader | 207.0 | 14.5 |
| Retired | 141.0 | 9.8 |
| Unemployed | 139.0 | 9.7 |
| Wealth Index (SES) | | |
| Poorest | 269.0 | 18.8 |
| Poorer | 334.0 | 23.3 |
| Middle | 267.0 | 18.6 |
| Rich | 252.0 | 17.6 |
| Richest | 310.0 | 21.6 |
The majority of participants ($75.5\%$) were married at the time of data collection. Looking into the education level of the respondents, a little more than half ($54.5\%$) of them did not attend any formal education. Urban residents accounted the largest ($70.3\%$) proportion, and housewives ($23\%$) and employed individuals ($22.9\%$) represent the largest proportion in the occupation category. The highest proportion of SES were those with low SES ($37.4\%$) (Table 1).
## Body mass index (BMI)
For body mass index (BMI), the highest percentage ($53.3\%$) were those with normal BMI, with about one third ($32\%$) of the participants being either overweight or obese (Fig 1).
**Fig 1:** *Proportion of individuals in different level of nutritional status based on the BMI indices.*
## Psychosocial Characteristics
The mean of social support scale was 10.2 and a standard deviation (SD) of ± 2.17 scores. Just over half ($50.7\%$) of the participants reported that they have moderate social support, and about one third ($28\%$) reported strong social support, while the remaining $21\%$ reported that they have a poor social support.
## Magnitude of NCDs and number of chronic NCDs identified per person
The magnitude of each of the chronic conditions considered in this study is shown in Fig 2. The number of NCDs identified per person ranged from one to four (mean = 1.74, SD = 0.78). Hypertension was the most frequently reported NCD ($63.5\%$), followed by diabetes ($42.5\%$) and heart diseases ($25.6\%$).
**Fig 2:** *List of NCDs studied and their magnitude among participants attending chronic outpatient NCDs care, Bahir Dar, Ethiopia (N = 1432).*
## Magnitude of multimorbidity
More than half $54.8\%$ (CI = $52.2\%$, $57.4\%$) of the study participants had multimorbidity, with $39.6\%$ having two chronic NCDs and $15.2\%$ three or more chronic NCDs (Fig 3).
**Fig 3:** *Patterns of NCDs morbidity among individuals attending chronic NCDs care in Bahir Dar, Ethiopia (N = 1432).*
The most prevalent NCDs greatly contributed to shaping the patterns of multimorbidity in this study. For example, hypertension co-existed with diabetes and heart diseases in $38.2\%$ and $19.0\%$ of the participants, respectively. Similarly, co-occurrence of diabetes was observed among individuals with heart diseases, depression and other types of reported chronic conditions. Hypertension remained the most frequently reported NCD ($87.2\%$) among individuals living with three or more NCDs in our study. Diabetes was reported by $51\%$ of those who had three or more chronic NCDs and heart diseases were reported by $39\%$ of the participants from this group (Table 2).
**Table 2**
| Single morbidity | Single morbidity.1 | Common pairs of NCDS | Common pairs of NCDS.1 | Common Triples of NCDs | Common Triples of NCDs.1 |
| --- | --- | --- | --- | --- | --- |
| Chronic NCD | Frequency (%) | Combination | Frequency (%) | Combination | Frequency (%) |
| Hypertension alone | 245 (37.9) | Hypertension +Diabetes | 217 (38.2) | Hypertension +Diabetes+ heart diseases | 19 (8.7) |
| Diabetes alone | 225 (34.8) | Hypertension + Heart diseases | 108 (19.0) | Hypertension +Diabetes + depression | 18 (8.3) |
| Heart diseases alone | 120 (18.5) | Hypertension + stroke | 38 (6.7) | Hypertension +Diabetes + other NCDs | 49 (22.5) |
| All other forms of single NCDsa | 57 (8.8) | Hypertension +Musculoskeletal diseases | 23 (4.0) | Hypertension +heart diseases + other NCDs | 43 (19.7) |
| | | Hypertension + Asthma | 21 (3.7) | Hypertension + Diabetes + heart diseases + other NCDs | 12 (5.5) |
| | | Hypertension + Chronic Renal diseases | 21 (3.7) | Hypertension + Diabetes + two other NCDs | 13 (6.0) |
| | | Hypertension + Depression | 18 (3.2) | Hypertension + two or other NCDs | 36 (16.5) |
| | | Hypertension + other chronic diseases | 25 (4.4) | Diabetes + two or more other NCDs | 13 (6.0) |
| | | Diabetes + Depression | 8 (1.4) | Heart diseases + two or more other NCDs | 11 (5.0) |
| | | Diabetes + heart disease | 6 (1.0) | Triple or quadruple of all other NCDs | 4 (1.8) |
| | | Diabetes + other chronic NCDs | 25 (4.4) | | |
| | | Heart disease +Depression | 16 (2.8) | | |
| | | Heart diseases + other chronic diseases | 27 (4.8) | | |
| | | Comorbidity of other NCDs | 14 (2.5) | | |
A third ($33.5\%$) of the study participants had poor quality of life and about a quarter of them had moderate QoL (Fig 4).
**Fig 4:** *Number of individuals classified in different categories of health-related QoL.*
Individuals with multimorbidity had a relatively poor QoL than those without multimorbidity ($62\%$ vs. $38\%$). Similarly, a higher proportion of individuals with severe functional limitations had poor QoL compared to those without severe limitations (Fig 5).
**Fig 5:** *Graphic presentation of the relationship between QoL, functioning and multimorbidity.*
Table 3 compares the output between the default univariate ordinal logistic regression analysis (ologit) model and the model adjusted for stratification (using the svy: ologit command). Although there are difference in the parameter estimates between the two models, age, BMI, functioning, satisfaction with care and presence of multimorbidity remain significantly associated with QoL in the model accounting for stratification. However; sex, SES and perceived social support lost their significance in the model accounting for a stratified sampling (Table 3).
**Table 3**
| Variables | QoL category | QoL category.1 | QoL category.2 | P-value | P-value: svy output(accounting for stratification) | Brant test |
| --- | --- | --- | --- | --- | --- | --- |
| Variables | Poor QoL | Moderate QoL | High QoL | P-value | P-value: svy output(accounting for stratification) | Brant test |
| Residence | | | | | | |
| Urban | 272 | 273 | 462 | base | | |
| Rural | 208 | 97 | 120 | 0.474 | | |
| Sex | | | | | | |
| Male | 207 | 174 | 321 | base | base | <0.564 |
| Female | 273 | 196 | 262 | <0.0.001 | 0.072 | <0.564 |
| Age | | | | | | |
| Mean | 60.0 | 56.0 | 53.6 | <0.001 | 0.037* | 0.377 |
| SD | 12.61 | 11.30 | 10.62 | <0.001 | 0.037* | 0.377 |
| Education | | | | | | |
| Below primary | 360 | 191 | 229 | 0.557 | | |
| Primary and above | 120 | 179 | 353 | base | | |
| Marital status | | | | | | |
| Married | 312 | 284 | 486 | base | | |
| Single | 168 | 87 | 96 | <0.648 | | |
| BMI | | | | | | |
| ≤24.99 | 250 | 209 | 304 | base | base | 0.665 |
| ≥25 | 115 | 41 | 55 | <0.023 | 0.020* | 0.665 |
| SES | | | | | | |
| Low | 234 | 128 | 174 | base | base | <0.365 |
| Middle or high | 133 | 109 | 163 | 0.003 | 0.173 | <0.365 |
| Social supportscale | | | | | | |
| Mean | 9.3 | 10.6 | 10.7 | <0.001 | 0.092 | <0.001** |
| SD | 2.17 | 1.70 | 2.19 | <0.001 | 0.092 | <0.001** |
| Overall functioning | | | | | | |
| Limited/weak capacity | 258 | 29 | 5 | base | base | <0.140 |
| Strong capacity | 222 | 341 | 577 | <0.001 | 0.006* | <0.140 |
| Care satisfaction | | | | | | |
| not satisfied | 108 | 43 | 32 | base | base | <0.001** |
| Satisfied | 372 | 327 | 550 | 0.019* | 0.018* | <0.001** |
| Multimorbidity | | | | | | |
| No | 182 | 170 | 295 | base | Base | 0.478 |
| Yes | 298 | 200 | 287 | <0.001* | 0.011* | 0.478 |
## Multivariable partial proportional odds analysis
In the final model, we entered all the variables that had a P-value <0.2 in the univariate model adjusted for stratification. As indicated above, social support score and satisfaction with care did not satisfy the parallel lines regression assumptions, necessitating a fitting of the partial proportional odds (PPO) model. Therefore, for the two variables, the odds ratios were allowed to vary (AOR1 ≠ AOR2) in the two panels.
Whereas age, sex, SES, BMI, multimorbidity and perceived functional status met the parallel regression assumption (Brant test value ≥0.05). Hence, the odds ratio would be the same (AOR1 = AOR2) for the two panels. In the final model, AOR1 stands for panel one (low versus moderate or high QoL), while AOR2 refers to the second panel (low/ moderate versus high QoL).
Looking into the final model (Table 4), statistically significant differences were observed in terms of the effect of most of the explanatory variables on QoL, adjusting for all the covariates. However, self-reported functioning lost its significance in the final model (specified by gsvy: gologit2, Y x1, x2, x3… auto lrforce command). While social support score became significant in the final model (Table 4).
**Table 4**
| Explanatory Variables | Outcome variables (panels) | Outcome variables (panels).1 | Outcome variables (panels).2 | Outcome variables (panels).3 | Outcome variables (panels).4 |
| --- | --- | --- | --- | --- | --- |
| Explanatory Variables | Panel One (1 Vs. 2 and 3) | Panel One (1 Vs. 2 and 3) | Panel One (1 Vs. 2 and 3) | Panel Two (1 or 2 Vs 3) | Panel Two (1 or 2 Vs 3) |
| Explanatory Variables | AOR 1 (95%CI) | Coefficients constant (OR1 = OR2) | P-value | AOR2(95%CI) | P-value |
| Sex (Female vs. male[Ref]) | | 0.63(0.27, 1.46) | 0.141 | | |
| SES (high vs. low [ref]) | | 0.73(0.41, 1.29) | 0.138 | | |
| Self-rated functioning (Strong vs. weak [Ref]) | | 1.28(0.66, 3.16) | 0.233 | | |
| Satisfaction (satisfied vs. not satisfied [Ref]) | 0.93(0.57, 1.51) | | 0.588 | 2.63(1.83, 5.37) | 0.023 |
| Social support scale | 1.41(1.29, 1.56) | | 0.004 | 1.15(0.88,, 1.50) | <0.153 |
| Multimorbidity (yes vs.no [Ref]) | | 0.76(0.69,, 0.84) | 0.008 | | |
| Age in years | | 0.96(0.94, 0.99) | <0.032 | | |
| BMI (≥ 25 vs. <24.99 [Ref]) | | 0.80(0.66, 0.98) | 0.043 | | |
The final model indicates that the odds of being in the combined categories of moderate and high QoL versus poor QoL was 2.6 times higher for patients satisfied with care than individuals that are not satisfied, when holding the other variables constant [AOR1 = 2.6 ($95\%$ CI: 1.83, 5.37)]. However, the odds of being in the combined categories of moderate and low versus high QoL was lower by a factor of 0.07 for satisfied patients than patients which were not satisfied. It was not, however, statistically significant [AOR2 = 0.93($95\%$ CI: 0.57, 1.51)].
In this study, for a unit increase in the social support scale, the odds of being in the higher categories of QoL versus lower categories was 1.41 times greater [AOR1 = 1.41($95\%$ CI: 1.29, 1.56)], given the other variables held constant.
Meanwhile, the variables that did not violate the PPO assumption had a constant beta coefficient (AOR1 = AOR2) for each of the two QoL categories; hence a single odds ratio was reported. We found that, for every one-year increase in age of the participants, the odds of being in the poorer QoL category was increased by a factor of 0.04 [AOR1 = AOR2: 0. 96 ($95\%$ CI: 0.94, 0.99)].
Similarly, the odds of being in the higher categories of QoL was 0.24 times lower for individuals with multimorbidity than those people without multimorbidity [AOR1 = AOR2: 0.76 ($95\%$ CI: 0.69, 0.84)]. Moreover, individuals having a higher categories of BMI score were $20\%$ more likely to have a poorer QoL than individuals in the lower BMI category, holding the other variable remain constant [AOR1 = AOR2: 0. 80 ($95\%$ CI: 0.66, 0.98)] (Table 4).
## Discussion
Understanding the effect of multimorbidity on health related quality of life (QoL) is one of the top research priorities in the existing literature [20, 81]. A broad sample of health facilities where most of the people living with chronic NCDs receive their care and corresponding number of patients were randomly selected and enrolled to determine the magnitude of multimorbidity and its association with QoL in the study area. We employed a blend of methods (face-to-face interview and review of medical records) to better determine the presence of individual NCDs and their pairwise and triple combination among a broad sample of 1432 individuals (aged 40+) attending chronic medical care in hospitals and specialized health facilities in Bahir Dar city, Northwest Ethiopia.
The implication of our findings should be interpreted in light of the variations in the way QoL has been measured and analyzed globally. The method of analysis we employed- the partial proportional odds (PPO) is relatively new in the context of analyzing QoL data [53]. Compared to other methods, the PPO model is said to be a robust analytic method for categorical QoL data [48, 53].
In this study, the authors found that multimorbidity is common, affecting the majority ($55\%$) of the individuals receiving chronic outpatient medical care. The high burden of multimorbidity in the study area implies that individuals living with chronic conditions have already been facing the overwhelming consequences of multimorbidity. Previous studies have also reported the challenges of living with multimorbidity in the country [8]. Studies show the most common risk factors contributing for the increasing burden of multimorbidity are advanced age, obesity, physical inactivity, socioeconomic deprivation and use of tobacco and alcohol [21]. This implies, the majority of the risk factors for multimorbidity are modifiable [6].
It was found that a third of ($33.5\%$) of individuals living with chronic NCDs had poor QoL, with $62\%$ of these having multimorbidity. Several studies have shown that multimorbidity is a key factor contributing for poor QoL [38, 39, 43]. Although direct comparison may not be possible with most of the previous studies owing to methodological variations, the authors observed that patients with multimorbidity had significantly poorer quality of life compared to patients with single chronic conditions. Studies which utilized the PPO model to analyze QoL data have also reported consistent results corroborating the negative association between multimorbidity and QoL [53]. However, evidence shows that not only the mere sum of individuals conditions, but also the nature of disease clusters matter in relation to quality of life, functionality and survival [64].
Consistent with previous studies [17, 24], people with advanced age were shown to have reduced QoL in our study. Advanced age is known to impair molecular and cellular functions that leads to a gradual decline in the physiological reserves and capacity of the individuals [82]. The observed inverse relationship between advanced age and poor QoL may also be due to the mediated effect of multimorbidity as the probability of having multimorbidity was higher among the middle-aged and elderly in our study. Although it is expected that people in old age often face poor QoL due to physical disability, frailty and sensory impairment [29], earlier onset of multimorbidity and its effect on QoL was reported to be higher among young adults living in socioeconomically deprived areas [40].
Evidence shows a clear relationship between obesity and poor QoL. In this study, individuals having higher score of BMI had lower levels of QoL than their leaner counterparts. However, the association between higher scores of BMI and poor QoL might be due to the mediating effect of multimorbidity. It was observed in this study that obese individuals had higher odds of multimorbidity. On the other hand, individual having higher BMI scores could have increased risk of physical limitations [25, 83], which will contribute negatively to QoL [84]. The complex interplay between obesity, multimorbidity and poor QoL has also been well established globally [2].
Medical care alone cannot adequately improve QoL [25]. The presence of strong social support is helpful to improve patient’s adaptation to life and their QoL [85]. The authors found a positive and statistically significant association between perceived social support and QoL. However, some studies were inconsistent in reporting the effect of social support in modifying QoL among individuals with multimorbidity [86–88]. People with multimorbidity are generally less satisfied with the care they receive [2]. Ensuring satisfaction with care for people with multiple chronic health conditions is challenging because the notion of satisfaction is influenced by several actors, including caregivers, healthcare providers and the health system in general [25]. In this study, individuals satisfied with care were more likely than their counterparts to have higher odds of better QoL. This is consistent with previous findings [89]. However, other literature shows no difference between satisfaction with care and multimorbidity related QoL [35, 90].
Previous studies have shown that economically deprived people struggle to cope with everyday life activities and have a lower quality of life compared with more affluent patients with multimorbidity [15]. Further, multimorbidity was associated with a more significant reductions in QoL scores amongst participants living in the most deprived areas [40], signifying a coupling effect of poverty and multimorbidity on QoL. However, the authors did not find any statistically significant relationship between SES and QoL. This might be due to the difference in methodology or nature of study participants involved in our study.
## Implication for practice and research
The main goal of health care for the people living multiple chronic conditions is to help them achieve better QoL [25, 91]. Given that the magnitude of multimorbidity is high and that it poses a profound impact on QoL, the healthcare system needs to be guided by these findings in order to adequately respond to individual patient needs. Care for people with multimorbidity must be based on the needs and circumstances of the person as a whole rather than the different conditions a person happens to have [90]. The provision of patient-centered care in which all healthcare providers work together with patients to ensure coordination, consistency and continuity of care over time is essential [92]. This will in turn improve the wellbeing and survival of the people with multimorbidity in the study area.
The evidence base on the association between multimorbidity and QoL is growing, albeit slowly. However, the methodologies employed to study multimorbidity vary widely [3], and the methods applied to investigate the impact of multimorbidity on QoL have not been universally consistent [40]. We are aware of the possible limitation of comparing our results with studies that employed different tools and methods of analysis. Further research is needed on the application of ordinal regression and PPO models for analyzing QoL data and to identify the associated covariates. Understanding the longitudinal effect of individual NCDs, multimorbidity and disease severity on QoL would help fill the substantial gaps in our knowledge in this regard. It is also imperative to study the way health systems are organized to manage patients with multimorbidity, and to explore the perspective and lived experiences of individuals with multiple chronic conditions in Ethiopia.
## Strength and limitations of the study
Our study has the advantage of involving a broader range of health facilities rendering comprehensive care for the people with chronic NCDs. Guided by a published study protocol; this study employed three complementary methods to define the presence of chronic NCDs accurately. The authors utilized the widely accepted QoL measure, the SF-12V2 tool and analyzed the data using a relatively robust categorical data analytic method, the PPO model. The PPO model we fitted has taken into account variation in stratified sampling to analyze the QoL data and identify covariates associated with QoL in a relatively efficient, reliable and valid way. However, the findings of this facility-based study may not exactly represent the underlying epidemiology of multimorbidity and patterns of association between multimorbidity and QoL in the general population in region and beyond. It is also difficult to confirm that the observed association between the variables has a temporal relationship. Further, variables measured by Likert-type scales in general are subject to bias; therefore, care should be taken when interpreting studies using such scales. In addition, the lack of consistent methods to measure both multimorbidity and QoL globally makes our findings comparable to only some of the previous studies. However, the application of the PPO model makes our study more parsimonious than studies employing the traditional linear methods of analyzing QoL data in the field of multimorbidity research in the global context.
## Conclusion and Recommendations
The magnitude of multimorbidity in this study was high and the highest proportion of individuals with multimorbidity had poor QoL. The high multimorbidity estimate observed might be attributed to the fact that the study was conducted among health facilities where most of people living with chronic NCDs were attending care. Advanced age, living with multimorbidity and obesity were the variables negatively associated with QoL. In contrast, high-perceived social support and satisfaction with care were the variables associated with higher categories of QoL.
The literature on the relationship between multimorbidity and QoL is dominated by studies in high income countries. If health systems in LMICs are to meet the needs of the people with multimorbidity, it is essential to understand the full breadth of multimorbidity across the ages and its effect on individuals QoL, functioning and survival. Future studies may need to focus on understanding the epidemiology of multimorbidity and its effect on QoL and survival in the general population. Further studies areneeded to explore the longitudinal effect of multimorbidity on quality of life, functioning and survival, and to assess the way health services organized to meet the care needs of the people with multiple chronic conditions in the country. It is also imperative to replicate the methods that were employed to measure and analyze QoL data in this study in order to facilitate comparison and further development of the approaches.
## Patient and public involvement
No patient or the public were involved in the design, or conduct, or reporting, or dissemination plans of our research
## References
1. 1WHO. World health statistics 2016: monitoring health for the SDGs, sustainable development goals2016.
2. Mercer S, Salisbury C, Fortin M. *ABC of multimorbidity* (2014.0)
3. Ho IS-S, Azcoaga-Lorenzo A, Akbari A, Black C, Davies J, Hodgins P. **Examining variation in the measurement of multimorbidity in research: a systematic review of 566 studies**. *Lancet Public Health* (2021.0) **6** e587-97. DOI: 10.1016/S2468-2667(21)00107-9
4. Xu X, Mishra GD, Jones M. **Mapping the global research landscape and knowledge gaps on multimorbidity: a bibliometric study**. *Journal of global health* (2017.0) **7** 010414. DOI: 10.7189/jogh.07.010414
5. Zemedikun DT, Gray LJ, Khunti K, Davies MJ, Dhalwani NN. **Patterns of Multimorbidity in Middle-Aged and Older Adults: An Analysis of the UK Biobank Data**. *Mayo Clinic proceedings* (2018.0) **93** 857-66. DOI: 10.1016/j.mayocp.2018.02.012
6. Abebe F, Schneider M, Asrat B, Ambaw F. **Multimorbidity of chronic non-communicable diseases in low- and middle-income countries: A scoping review**. *Journal ofComorbidity* (2020.0) **10** 1-13. DOI: 10.1177/2235042X20961919
7. Asogwa OA, Boateng D, Marzà-Florensa A, Peters S, Levitt N, Olmen Jv. **Multimorbidity of non-communicable diseases in low-income and middleincome countries: a systematic review and meta-analysis**. *BMJ open* (2022.0) **12**
8. Bhagavathula AS, Gebreyohannes EA, Seid MA, Adane A, Brkic J, Fialová D. **Prevalence and Determinants of Multimorbidity, Polypharmacy, and Potentially Inappropriate Medication Use in the Older Outpatients: Findings from EuroAgeism H2020 ESR7 Project in Ethiopia**. *Pharmaceuticals* (2021.0) **14**
9. Eyowas FA, Schneider M, Alemu S, Pati S, Getahun FA. **Magnitude, pattern and correlates of multimorbidity among patients attending chronic outpatient medical care in Bahir Dar, northwest Ethiopia: The application of latent class analysis model**. *PLoS ONE* (2022.0) **17** e0267208. DOI: 10.1371/journal.pone.0267208
10. Calderon-Larranaga A, Vetrano DL, Ferrucci L, Mercer SW, Marengoni A, Onder G. **Multimorbidity and functional impairment: bidirectional interplay, synergistic effects and common pathways**. *Journal of internal medicine* (2018.0). DOI: 10.1111/joim.12843
11. KINGSTON A, ROBINSON L, BOOTH H, KNAPP M, JAGGER C. **Projections of multi-morbidity in the older population in England to 2035: estimates from the Population Ageing and Care Simulation (PACSim) model**. *Age and Ageing* (2018.0). DOI: 10.1093/ageing/afx201
12. Mounce LTA, Campbell JL, Henley WE, Tejerina Arreal MC, Porter I, Valderas JM. **Predicting Incident Multimorbidity**. *Annals of family medicine* (2018.0) **16** 322-9. DOI: 10.1370/afm.2271
13. Ornstein SM, Nietert PJ, Jenkins RG, Litvin CB. **The prevalence of chronic diseases and multimorbidity in primary care practice: a PPRNet report**. *Journal of the American Board of Family Medicine: JABFM* (2013.0) **26** 518-24. DOI: 10.3122/jabfm.2013.05.130012
14. Willadsen T, Jarbøl D, Reventlow S, Mercer S. **Olivarius NdF. Multimorbidity and mortality: A 15-year longitudinal registry-based nationwide Danish population study**. *Journal ofComorbidity* (2018.0) **8** 1-9
15. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. **Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study**. *The Lancet* (2012.0) **380** 37-43. DOI: 10.1016/s0140-6736(12)60240-2
16. Romano E, Ma R, Vancampfort D, Firth J, Felez-Nobrega M, Haro JM. **Multimorbidity and obesity in older adults from six low- and middle-income countries**. *Prev Med* (2021.0). DOI: 10.1016/j.ypmed.2021.106816
17. Violan C, Foguet-Boreu Q, Flores-Mateo G, Salisbury C, Blom J, Freitag M. **Prevalence, determinants and patterns of multimorbidity in primary care: a systematic review of observational studies**. *PLoS One* (2014.0) **9** e102149. DOI: 10.1371/journal.pone.0102149
18. Alimohammadian M, Majidi A, Yaseri M, Ahmadi B, Islami F, Derakhshan M. **Multimorbidity as an important issue among women: results of a gender difference investigation in a large population-based cross-sectional study**. *West Asia. BMJ open* (2017.0) **7** e013548. DOI: 10.1136/bmjopen-2016-013548
19. Skou ST, Mair FS, Fortin M, Guthrie B, Nunes BP, Miranda JJ. **Multimorbidity**. *Nature Reviews* (2022.0) **8**
20. 20AMS. Multimorbidity: a priority for global health research. 2018.
21. Freisling H, Viallon V, Lennon H, Bagnardi V, Ricci C, Butterworth AS. **Lifestyle factors and risk of multimorbidityof cancer and cardiometabolic diseases: amultinational cohort study**. *BMC Medicine* (2020.0) **18**. DOI: 10.1186/s12916-019-1474-7
22. Mvd Akker, Buntinx F, Metsemakers JFM, Roos S, Knottnerus JA. **Multimorbidity in General Practice: Prevalence, Incidence, and Determinants of Co-Occurring Chronic and Recurrent Diseases**. *J Clin Epidemiol* (1998.0) **51** 367-75. DOI: 10.1016/s0895-4356(97)00306-5
23. France EF, Wyke S, Gunn JM, Mair FS, McLean G, Mercer SW. **Multimorbidity in primary care: a systematic review of prospective cohort studies**. *The British journal of general practice: the journal of the Royal College of General Practitioners* (2012.0) **62** e297-307. DOI: 10.3399/bjgp12X636146
24. Marengoni A, Angleman S, Melis R, Mangialasche F, Karp A, Garmen A. **Aging with multimorbidity: a systematic review of the literature**. *Ageing Res Rev* (2011.0) **10** 430-9. DOI: 10.1016/j.arr.2011.03.003
25. 25Aiden H. Multimorbidity. Understanding the challenge. A report for the Richmond Group of Charities. 2018.
26. Harrison C, Henderson J, Miller G, Britt H. **The prevalence of diagnosed chronic conditions and multimorbidity in Australia: A method for estimating population prevalence from general practice patient encounter data**. *PLoS One* (2017.0) **12** e0172935. DOI: 10.1371/journal.pone.0172935
27. Hunter ML, Knuiman MW, Musk BAW, Hui J, Murray K, Beilby JP. **Prevalence and patterns of multimorbidity in Australian baby boomers: the Busselton healthy ageing study**. *BMC Public Health* (2021.0) **21**. DOI: 10.1186/s12889-021-11578-y
28. Hunger M, Thorand B, Schunk M, Doring A, Menn P, Peters A. **Multimorbidity and health-related quality of life in the older population: results from the German KORA-age study**. *Health and quality of life outcomes* (2011.0) **9** 53. DOI: 10.1186/1477-7525-9-53
29. 29NICE. Multimorbidity: clinical assessment and management: Multimorbidity: assessment, prioritisation and management of care for people with commonly occurring multimorbidity. NICE guideline NG56: National Institute for Health and Care Excellence; 2016.
30. Leijten FRM, Struckmann V, van Ginneken E, Czypionka T, Kraus M, Reiss M. **The SELFIE framework for integrated care for multi-morbidity: Development and description**. *Health policy (Amsterdam, Netherlands)* (2018.0) **122** 12-22. DOI: 10.1016/j.healthpol.2017.06.002
31. 31Charities TRGo. Just one thing after another’ Living with multiple conditions: A report from the Taskforce on Multiple Conditions. 2018.
32. Bayliss EA, Bonds DE, Boyd CM, Davis MM, Finke B, Fox MH. **Understanding the context of health for persons with multiple chronic conditions: moving from what is the matter to what matters**. *Annals of family medicine* (2014.0) **12** 260-9. DOI: 10.1370/afm.1643
33. Guthrie B, Payne K, Alderson P, McMurdo MET, Mercer SW. **Adapting clinical guidelines to take account of multimorbidity**. *BMJ (Clinical research ed)* (2012.0) **345**. DOI: 10.1136/bmj.e6341
34. Young CE, Boyle FM, Mutch AJ. **Are care plans suitable for the management of multiple conditions?**. *Journal of comorbidity* (2016.0) **6** 103-13. DOI: 10.15256/joc.2016.6.79
35. Salisbury C, Man MS, Bower P, Guthrie B, Chaplin K, Gaunt DM. **Management of multimorbidity using a patient-centred care model: a pragmatic cluster-randomised trial of the 3D approach**. *Lancet* (2018.0) **392** 41-50. DOI: 10.1016/S0140-6736(18)31308-4
36. Ailabouni NJ, Hilmer SN, Kalisch L, Braund R, Reeve E. **COVID-19 Pandemic: Considerations for Safe Medication Use in Older Adults with Multimorbidity and Polypharmacy**. *J Gerontol A Biol Sci Med Sci* (2020.0)
37. Guan W-j, Liang W-h, Zhao Y, Liang H-r, Chen Z-s, Li Y-m. **Comorbidity and its impact on 1590 patients with Covid-19 in China: A Nationwide Analysis**. *Eur Respir J* (2020.0). DOI: 10.1183/13993003.00547-2020
38. Fortin M, Dubois MF, Hudon C, Soubhi H, Almirall J. **Multimorbidity and quality of life: a closer look**. *Health and quality of life outcomes* (2007.0) **5** 52. DOI: 10.1186/1477-7525-5-52
39. Fortin M, Lapointe L, Hudon C, Vanasse A, Ntetu AL, Maltais D. **Multimorbidity and quality of life in primary care: a systematic review**. *Health and quality of life outcomes* (2004.0) **2** 51. DOI: 10.1186/1477-7525-2-51
40. Makovski TT, Schmitz S, Zeegers MP, Stranges S, Akker Mvd. **Multimorbidity and quality of life: systematic literature review and meta-analysis**. *Ageing Research Reviews* (2019.0). DOI: 10.1016/j.arr.2019.04.005
41. Bao X-Y, Xie Y-X, Zhang X-X, Peng X, Huang J-X, Du Q-F. **The association between multimorbidity and health-related quality of life: a crosssectional survey among community middle-aged and elderly residents in southern China**. *Health and quality of life outcomes* (2019.0) **17**
42. Bayliss M, Rendas-Baum R, White MK, Maruish M, Bjorner J, Tunis SL. **Health-related quality of life (HRQL) for individuals with self-reported chronic physical and/or mental health conditions: panel survey of an adult sample in the United States**. *Health and Quality of Life Outcomes* (2012.0) **10**. DOI: 10.1186/1477-7525-10-154
43. Fortin M, Bravo G, Hudon C, Lapointe L, Almirall J, Dubois MF. **Relationship between multimorbidity and health-related quality of life of patients in primary care**. *Quality of life research: an international journal of quality of life aspects of treatment, care and rehabilitation* (2006.0) **15** 83-91. DOI: 10.1007/s11136-005-8661-z
44. Lang C, Roessler M, Schmitt J, Bergmann A, Holthoff‑Detto V. **Health‑related quality of life in elderly, multimorbid individuals with and without depression and/or mild cognitive impairment using a telemonitoring application**. *Quality of Life Research (2021)* (2021.0) **30** 2829-41. DOI: 10.1007/s11136-021-02848-8
45. Millá-Perseguer M, Guadalajara-Olmeda N, Vivas-Consuelo D, Usó-Talamantes R. **Measurement of health-related quality by multimorbidity groups in primary health care**. *Health and quality of life outcomes* (2019.0) **17**. DOI: 10.1186/s12955-018-1063-z
46. Williams JS, Egede LE. **The Association Between Multimorbidity and Quality of Life, Health Status and Functional Disability**. *The American journal of the medical sciences* (2016.0) **352** 45-52. DOI: 10.1016/j.amjms.2016.03.004
47. Hagell P, Westergren A, Årestedt K. **Beware of the origin of numbers: Standard scoring of the SF-12 and SF-36 summary measures distorts measurement and score interpretations**. *Res Nurs Health* (2017.0) **40** 378-86. DOI: 10.1002/nur.21806
48. Lall R, Campbell MJ, Walters SJ, Morgan K. **A review of ordinal regression models applied on health-related quality of life assessments**. *StatisticalMethods in Medical Research* (2002.0) **11** 49-67. DOI: 10.1191/0962280202sm271ra
49. 49Lall R. The Application of Ordinal Regression Models in Quality of Life Scales used in Gerontology 2004.
50. Walters SJ, Campbell MJ, Lall R. **DESIGN AND ANALYSIS OF TRIALS WITH QUALITY OF LIFE AS AN OUTCOME: A PRACTICAL GUIDE**. *ournal of Biopharmaceutical Statistics* (2001.0) **11** 155-76. DOI: 10.1081/BIP-100107655
51. Liddell TM, Kruschke JK. **Analyzing ordinal data with metric models: What could possibly go wrong?**. *Journal of Experimental Social Psychology* (2018.0) **79** 328-48
52. McKenna SP, Heaney A. **Composite outcome measurement in clinical research: the triumph of illusion over reality?**. *Journal of medical economics* (2020.0) **23** 1196-204. DOI: 10.1080/13696998.2020.1797755
53. Abreu MNS, Siqueira AL, Cardoso CS, Caiaffa WT. *Ordinal logistic regression models: Application in quality of life studies* (2008.0) **Sup 4** S581-S91
54. Williams R. **Understanding and interpreting generalized ordered logit models**. *The Journal of Mathematical Sociology* (2016.0)
55. Peterson B, Frank E. **Harrell J. Partial Proportional Odds Models for Ordinal Response Variables**. *Journal of the Royal Statistical Society Series C (Applied Statistics)* (1990.0) **39** 205-17
56. Williams R.. **Generalized ordered logit/partial proportional odds models for ordinal dependent variables**. *The Stata Journal* (2006.0) **6** 58-82
57. Bürkner P-C, Vuorre M. **Ordinal Regression Models in Psychology: A Tutorial**. *Advances in Methods and Practices in Psychological Science* (2019.0) 1-25
58. Brant R.. **Assessing Proportionality in the Proportional Odds Model for Ordinal Logistic Regression**. *Biometrics* (1990.0) **46** 1171-8. PMID: 2085632
59. Austad B, Hetlevik I, Mjolstad BP, Helvik AS. **Applying clinical guidelines in general practice: a qualitative study of potential complications**. *BMC family practice* (2016.0) **17** 92. DOI: 10.1186/s12875-016-0490-3
60. Turner A, Mulla A, Booth A, Aldridge S, Stevens S, Begum M. **The international knowledge base for new care models relevant to primary care-led integrated models: a realist synthesis**. *HEALTH SERVICES AND DELIVERY RESEARCH* (2018.0) **6**. DOI: 10.3310/hsdr06250
61. Eyowas FA, Schneider M, Alemu S, Getahun FA. **Multimorbidity of chronic noncommunicable diseases: burden, care provision and outcomes over time among patients attending chronic outpatient medical care in Bahir Dar, Ethiopia—a mixed methods study protocol**. *BMJ-Open* (2021.0) **11** e051107. DOI: 10.1136/bmjopen-2021-051107
62. 62G/Michael M, Dagnaw W, Yadeta D, Feleke Y, Fantaye A, Kebede T, et al. Ethiopian National Guideline on Major NCDs 2016. 2016.
63. Skevington SM, Lotfy M, O’Connell KA. **The World Health Organization’s WHOQOL-BREF quality of life assessment: Psychometric properties and results of the international field trial A Report from the WHOQOL Group**. *Quality of Life Research* (2004.0) **13** 299-310. PMID: 15085902
64. Gonzalez-Chica DA, Hill CL, Gill TK, Hay P, Haag D, Stocks N. **Individual diseases or clustering of health conditions? Association between multiple chronic diseases and health-related quality of life in adults**. *Health and quality of life outcomes* (2017.0) **15** 244. DOI: 10.1186/s12955-017-0806-6
65. Carlozzi NE, Kratz AL, Downing NR, Goodnight S, Miner J, Migliore N. **Validity of the 12-item World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0) in individuals with Huntington disease (HD)**. *Quality of Life Research* (2015.0) **24** 1963-71. DOI: 10.1007/s11136-015-0930-x
66. WARE JEJ, KOSINSKI MM, KELLER SD. **A 12-Item Short-Form Health Survey: Construction of Scales and Preliminary Tests of Reliability and Validity**. *Ovid: WARE: Med Care, Volume 34(3)March 1996* (1996.0) **34** 220-33. DOI: 10.1097/00005650-199603000-00003
67. Ohrnberger J, Anselmi L, Fichera E, Sutton M. **Validation of the SF12 mental and physical health measure for the population from a low-income country in sub-Saharan Africa**. *Health and quality of life outcomes* (2020.0) **18**. DOI: 10.1186/s12955-020-01323-1
68. Lawson KD, Mercer SW, Wyke S, Grieve E, Guthrie B, Watt GC. **Double trouble: the impact of multimorbidity and deprivation on preference-weighted health related quality of life a cross sectional analysis of the Scottish Health Survey**. *International Journal for Equity in Health* (2013.0) **12**. DOI: 10.1186/1475-9276-12-67
69. Stubbs B, Vancampfort D, Veronese N, Kahl KG, Mitchell AJ, Lin PY. **Depression and physical health multimorbidity: primary data and country-wide meta-analysis of population data from 190 593 people across 43 low- and middle-income countries**. *Psychological medicine* (2017.0) **47** 2107-17. DOI: 10.1017/S0033291717000551
70. Kocalevent R-D, Berg L, Beutel ME, Hinz A, Zenger M, Härter M. **Social support in the general population: standardization of the Oslo social support scale (OSSS-3)**. *BMC Psychology* (2018.0) **6**
71. 71FAO. Wealth Index mapping in the Horn of Africa. Animal Production and Health Working Paper. No. 4. Rome. 2011.
72. Chakraborty NM, Fry K, Behl R, Longfielda K. **Simplified Asset Indices to Measure Wealth and Equity in Health Programs: A Reliability and Validity Analysis Using Survey Data From 16 Countries**. *Global Health: Science and Practice* (2016.0) **4**. DOI: 10.9745/GHSP-D-15-00384
73. Üstün TB, Chatterji S, Kostanjsek N, Rehm J, Kennedy C, Epping-Jordan J. **Developing the World Health Organization Disability Assessment Schedule 2.0**. *Bulletin of the World Health Organization* (2010.0) **88** 815-23. DOI: 10.2471/BLT.09.067231
74. Habtamu K, Alem A, Medhin G, Fekadu A, Dewey M, Prince M. **Validation of the World Health Organization Disability Assessment Schedule in people with severe mental disorders in rural Ethiopia**. *Health and Quality of Life Outcomes* (2017.0) **15**. DOI: 10.1186/s12955-017-0647-3
75. 75WHO. Process of translation and adaptation of instruments. 2014.
76. Hall DA, Domingo SZ, Hamdache LZ, Manchaiah V, Thammaiah S, Evans C. **A good practice guide for translating and adapting hearingr elated questionnaires for different languages and cultures**. *International Journal of Audiology* (2018.0) **57** 161-75. DOI: 10.1080/14992027.2017.1393565
77. 77OCHA. Manual Kobo Toolbox. https://www.kobotoolbox.org/: Office for the Coordination of Humanitarian Affairs (OCHA) in West and Central Africa; 2019.
78. Sasidharan L, Menéndez M. **Partial proportional odds model—An alternate choice for analyzing pedestrian crash injury severities**. *Accident Analysis and Prevention* (2014.0) **72** 330-40. DOI: 10.1016/j.aap.2014.07.025
79. Fullerton AS, Xu J. **The proportional odds with partial proportionality constraints model for ordinal response variables**. *Social Science Research* (2012.0) **41** 182-98. DOI: 10.1016/j.ssresearch.2011.09.003
80. 80Williams R. Analyzing Complex Survey Data: Some key issues to be aware of. 2021.
81. Tisminetzky M, Bayliss EA, Magaziner JS, Allore HG, Anzuoni K, Boyd CM. **Research Priorities to Advance the Health and Health Care of Older Adults with Multiple Chronic Conditions**. *J Am Geriatr Soc* (2017.0) **65** 1549-53. DOI: 10.1111/jgs.14943
82. 82WHO. World report on ageing and health. 2015.
83. Ryan A, Wallace E, O’Hara P, Smith SM. **Multimorbidity and functional decline in community-dwelling adults: a systematic review**. *Health and quality of life outcomes* (2015.0) **13** 168. DOI: 10.1186/s12955-015-0355-9
84. Rivera-Almaraz A, Manrique-Espinoza B, Ávila-Funes JA, Chatterji S, Naidoo N, Kowal P. **Disability, quality of life and all-cause mortality in older Mexican adults: association with multimorbidity and frailty**. *BMC Geriatrics* (2018.0) **18**
85. Garin N, Olaya B, Moneta MV, Miret M, Lobo A, Ayuso-Mateos JL. **Impact of multimorbidity on disability and quality of life in the Spanish older population**. *PLoS One* (2014.0) **9** e111498. DOI: 10.1371/journal.pone.0111498
86. Vogel I, Miksch A, Goetz K, Ose D, Szecsenyi J, Freund T. **The impact of perceived social support and sense of coherence on health-related quality of life in multimorbid primary care patients**. *Chronic Illness* (2012.0) **8** 296-307. DOI: 10.1177/1742395312445935
87. Demirer I, Bethge M, Spyra K, Karbach U, Pfaffe H. **Does social support mediate the effect of multimorbidity on mental wellbeing in the German working population? A longitudinal mediation analysis using structural equation modelling**. *SSM—population health* (2021.0) **13**. DOI: 10.1016/j.ssmph.2021.100744
88. Schäfer I, Hansen H, Kaduszkiewicz H, Bickel H, Fuchs A, Gensichen J. **Health behaviour, social support, socio-economic status and the 5-year progression of multimorbidity: Results from the MultiCare Cohort Study**. *Journal of Comorbidity* (2019.0) **9** 1-11. DOI: 10.1177/2235042X19883560
89. Kuipers SJ, Cramm JM, Nieboer AP. **The importance of patient-centered care and co-creation of care for satisfaction with care and physical and social well-being of patients with multi-morbidity in the primary care setting**. *BMC health services research* (2019.0) **19**. DOI: 10.1186/s12913-018-3818-y
90. Smith SM, Wallace E, O’Dowd T, Fortin M. **Interventions for improving outcomes in patients with multimorbidity in primary care and community settings**. *The Cochrane database of systematic reviews* (2016.0) **3** CD006560. DOI: 10.1002/14651858.CD006560.pub3
91. Smith SM, Wallace E, Salisbury C, Sasseville M, Bayliss E, Fortin M. **A Core Outcome Set for Multimorbidity Research (COSmm)**. *Annals of family medicine* (2018.0) **16** 132-8. DOI: 10.1370/afm.2178
92. Valderas JM, Gangannagaripalli J, Nolte E, Boyd CM, Roland M, Jones AS-SE. *Quality of care assessment for people with multimorbidity, scoping review* (2019.0)
|
---
title: 'Knowledge, attitude, and practice of Bangladeshi urban slum dwellers towards
COVID-19 transmission-prevention: A cross-sectional study'
authors:
- Md. Zahid Hasan
- A. M. Rumayan Hasan
- Md. Golam Rabbani
- Mohammad Abdus Selim
- Shehrin Shaila Mahmood
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021697
doi: 10.1371/journal.pgph.0001017
license: CC BY 4.0
---
# Knowledge, attitude, and practice of Bangladeshi urban slum dwellers towards COVID-19 transmission-prevention: A cross-sectional study
## Abstract
The first COVID-19 case in Bangladesh was detected on March 8, 2020. Since then, efforts are being made across the country to raise awareness among the population for preventing the spread of this virus. We aimed to examine the urban slum dwellers’ knowledge, attitude, and practice (KAP) towards COVID-19 transmission-prevention. A phone-based cross-sectional survey was conducted in five slums of Dhaka City. Total 476 adult slum dwellers were interviewed between October 31 to December 1, 2020 using a pre-tested questionnaire. During an interview, information was collected on participants’ demographic characteristics and KAP items towards COVID-19. We used quartiles for categorization of knowledge and practice score where the first quartile represents poor, the second and third quartiles represent average while the fourth quartile represents good. Attitude score was standardized using z-score and identified as positive and negative attitude. Multiple linear regression models were used separately to identify the socioeconomic predictors of the KAP scores. The results showed that $25\%$ of the respondents had good knowledge and $25\%$ had poor knowledge, $48\%$ had a positive attitude and $52\%$ had a negative attitude, and $21\%$ maintained good practice and $33\%$ maintained poor practice towards COVID-19 transmission-prevention. About $75\%$ respondents relied on television for COVID-19 related information. Regression results showed that knowledge and attitude scores were significantly higher if respondents had primary or secondary and above level of education compared to the uneducated group. Female respondents maintained significantly good practice compared to their male counterparts (β = 6.841; $p \leq 0.01$). This study has found that one third of the studied slum dwellers maintained poor practice and one fourth had poor knowledge towards COVID-19 transmission-prevention. As KAP domains are significantly correlated, efforts are needed to raise awareness of COVID-19 particularly targeting individuals with average and lower knowledge to improve attitude and practice for the prevention of COVID-19.
## Introduction
The coronavirus disease 2019 (COVID-19), a highly infectious disease, was first detected in Wuhan, China, at the end of 2019 [1]. The disease has spread rapidly across many countries and due to its terrible consequences, in March 2020, the World Health Organization (WHO) declared COVID-19 as a global pandemic [2]. In Bangladesh, the first COVID-19 case was detected on March 8, 2020, and the first person died of COVID-19 on March 18, 2020 [3]. By December 2021, Bangladesh reported more than 1,580,872 positive cases with 28,047 deaths from COVID-19 [4].
Being a highly infectious disease, COVID-19 can be transmitted in multiple ways such as being in close proximity to an infected person or being in an environment where droplets are generated from coughs, sneezes, or exhalation of an infected person; or by touching a contaminated surface, among others [5]. Though vaccines are now available, the risk of infection will remain until a large proportion of the population is vaccinated across the globe [6]. Therefore, controlling the spread of the virus with a behaviour change intervention related to transmission prevention is still considered as the most effective measure to protect people from the disease in the absence of pharmaceutical interventions [7]. Such interventions can guide how best to promote adherence to individuals’ personal protective behaviours (individual behaviours aimed at protecting oneself and others). However, different enacted personal protective behaviours require different types of interventions guided by various behaviour change principles [8]. The capability, opportunity, motivation, and behaviour (COM-B) model can be implemented to change the behaviours needed to limit COVID-19 transmission e.g., cough or sneeze etiquette, using face masks, keeping physical distance. According to the COM-B model a behaviour-change occurs when capability and opportunity are present and when the person is more motivated to enact that behaviour than any other [9]. The existence of a behavioural immune system in humans to control epidemic or pandemic is evident in literature [10]. For example, behaviour change during A/H1N1 influenza pandemic in 2009 [11, 12] and during Zika outbreak [13] reduced the transmission of virus. Following the evidences of behaviour theories and public health principles of infectious disease control, several measurement methods have been adopted and practices globally to control COVID-19 [14–17]. Globally, countries have adopted various controlling measures to fight against the spread of the virus, notably, frequent and proper handwashing, maintaining social distance, imposing lockdown to limit mobility of people, ensuring isolation or quarantine of the infected and suspected persons [5, 18, 19]. Studies showed that knowledge about COVID-19 and prevention measures to be taken, and attitudinal dispositions are significantly associated with appropriate infection prevention practice, which has the potential to play a significant role in the prevention and control of community transmission of COVID-19 [20]. The government of Bangladesh, within a few days of identifying the COVID-19 case, declared a nationwide lockdown for a few months and recommended its people to maintain the standard guidelines to prevent COVID-19 [21–23].
Around $38\%$ (62.5 million) of the total population of this country live in urban areas and about half of them live in slums [24, 25]. In Dhaka City, the capital of Bangladesh, there are more than 5,000 slums which are densely inhabited by an estimated four million people [26]. Almost three-fourths of the slum families live in a single room with shared toilets, bathrooms, shared water sources, poor access to electricity and communication system [27–29]. In addition, they are mainly involved in informal professions with limited earnings. Due to the lack of education, economic vulnerabilities, and living conditions, this population is highly susceptible to COVID-19 infection. The evidence of studies showed that the sociodemographic and economic characteristics of individuals are significantly associated with the level of KAP against the spread of COVID-19 [30–32]. Although some assessments on KAP towards COVID-19 have been conducted in Bangladesh [20, 30, 31, 33], studies focusing on the urban slum dwellers are very limited [30]. Therefore, rapid assessment of the KAP of slum population towards COVID-19 transmission-prevention is essential in, filling the existing knowledge gap and informing policymakers to design and implement required timely public health interventions in slum areas. Thus, the current study aims to assess the KAP of the slum dwellers towards COVID-19 transmission-prevention.
## Ethics statement
This study was approved by the Research Review Committee and the Ethics Review Committee of International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) (Protocol# PR-20092). Informed verbal consent was obtained from all the respondents before their interview and audio calls were recorded with permission from respondents. The respondents were assured about confidentiality and anonymity of their information.
## Study design
This study was a component of a rapid assessment of the health system impact of COVID-19 among the urban slum population in Bangladesh [34]. This was a cross-sectional telephone-based exploratory study conducted among 476 adult individuals (>18 years) of 476 households from the selected slums during 31 October to 1 December 2020.
## Study setting and population
The study was based on the sampling frame of the existing Urban Health and Demographic Surveillance System (UHDSS) of icddr,b that covers 31,577 households in five slums of the Dhaka North City Corporation, Dhaka South City Corporations, and Gazipur City Corporation. Adult male or female household members constituted the study population.
## Sample design and procedure
We assumed that $50\%$ of the adult slum population have proper KAP towards COVID-19. Considering this proportion ($$p \leq 0.5$$), with $95\%$ confidence level and $5\%$ precision level, an estimated 384 consenting individuals/households were required to be enrolled in the study. Assuming a $10\%$ non-response rate and 1.2 design effect for selecting respondents from different slums, 512 individuals/households were selected by multi-stage sampling technique. To capture the level of KAP, an adult informed member was selected from each household as respondent.
## Data collection instruments and measures
The study instrument was developed by the study team in consultation with experts from different fields to check its relevance and make necessary changes according to the study requirement. The development of instrument was grounded on theory of how knowledge and attitudes of individuals influence healthy behaviour and motivate them in taking actions towards prevention of infection [35–37]. We also reviewed published articles related to the assessment of KAP towards COVID-19 in similar settings, and the general guidelines recommended by the Bangladesh government and WHO [5, 38–40].
Initially, the tools were developed in English and translated into Bengali, contextualized for urban slum people of Bangladesh, pretested to ensure that the respondents understood all the KAP questions / items for control and transmission prevention of COVID-19, and then finalized the questionnaire for interview [41–49]. Both the initial and translated version of the questionnaire were rigorously reviewed by the Institutional Review Board (IRB) of icddr,b. The questionnaire included several parts, including sociodemographic condition, access to COVID-19 related information, and KAP towards COVID-19 preventive and curative measures.
There were 31 knowledge questions in five sub-domains, namely, about the disease [5], prevention [10], risk groups [6], symptoms [7], and access to COVID-19 related information [3]; 10 questions to understand attitude and 16 questions to evaluate prevention practices. There was a total of 57 questions related to KAP and the overall Cronbach α of the questions regarding the final data was 0.86, which indicates a satisfactory level of internal consistency of the items [50]. The knowledge and attitude questions had three levels with “Yes”, “No”, and “Don’t know”; whereas the preventive practice was only responded as “Yes” or “No”. Information on KAP level was assessed using the scoring method. Each positive response for knowledge and attitude was scored as “1” and a negative response was scored as “0” [51, 52].
## Data collection
We collected data by direct phone calls to the mobile number of the selected household as it was not feasible to conduct a face-to-face survey during the pandemic due to the risk of infection and restricted mobility. The UHDSS surveillance workers, during their routine data collection, took verbal consent from the households whether they would be interested to take part in this study or not and whether they would like to share their mobile numbers or not. Only the households those agreed to share their mobile phone numbers to take part in this study were used as a sampling frame for this survey. Six trained data collectors conducted interviews with the respondents over the phone. Respondents received BDT 200 (US$2.37) using a mobile financial service to compensate their time for participating in the study.
## Data analysis
Both descriptive and inferential statistical analyses were performed with the Two-tailed t-test for means of variables of two independent samples and the One-way ANOVA for variables of more than two independent samples to test the hypothesis of a statistical difference for each dimension of KAP scores with demographic and socioeconomic categories, separately. Statistical relationship between ‘Knowledge-Attitude’, ‘Knowledge-Practice’, and ‘Attitude-Practice’ were assessed using the pearson’s test of correlation. Principal component analysis was used to generate asset scores based on the household’s ownership of land and holdings of durable assets e.g., Television, Refrigerator, Mobile phone [53]. Then the asset scores were divided into five quintiles where the first quintile referred to the poorest households and the fifth quintile referred to the richest households. As the number of questions / items were not equal in three domains of the KAP, we developed weighted scores for each of the domains on the scale of 100. We used quartiles for categorization of the weighted scores where the first quartile represented ‘poor’, the second and third quartiles represented ‘average/moderate’ while the fourth quartile represented ‘good’ levels. However, for categorizing the sub-domains of knowledge, we used the score range for overall knowledge for three categories (good, moderate, poor) derived using the quartiles. The attitude score was standardized using z-score and using this z-score we estimated the proportion of individuals with positive and negative attitude.
Three multiple linear regression models were applied to identify the socioeconomic determinants of KAP scores, separately. In these models, KAP level were treated as dependent variables and household headship status, age, gender, education, occupation, marital status, household size, regular earning status, and wealth status were treated as explanatory variables. The multiple linear regression models were specified as follows: Yi=β0+β1X1i+β2X2i+β3X3i+⋯+εi……….. [1] Where, *Yi is* the knowledge, attitude, or practice scores of i-th individual; β0 is a constant; X1, X2, X3,… denote control variables e.g., age, gender, marital status, family size, earning status, asset quintiles; β1, β2, β3,… represent the coefficients, and εi is the random error term of the model. We analyzed the data using STATA version 16 [54].
## Demographic and socioeconomic characteristics
The results in the study showed that a total of 476 respondents from 476 households who participated were heads ($65\%$) of households with $82\%$ being adults (18–45 years of age), $18\%$ were older than 45 years. The proportion of male respondents was higher ($58\%$) than the female respondents. Regarding the educational qualification, $30\%$ of the respondents had a primary level of education, $28\%$ had a secondary level education, $9\%$ had higher secondary and above level education, and $26\%$ had no education. More than $50\%$ of the respondents did not have regular income and about $54\%$ did not work in the last 30 days before the survey (see Table 1).
**Table 1**
| Characteristics | n | % |
| --- | --- | --- |
| Relationship with household head | | |
| Household head | 309.0 | 64.9 |
| Others | 167.0 | 35.1 |
| Age in years | | |
| ≤30 | 188.0 | 39.5 |
| 31–45 | 204.0 | 42.9 |
| 46–60 | 67.0 | 14.1 |
| >60 | 17.0 | 3.6 |
| Sex | | |
| Male | 277.0 | 58.2 |
| Female | 199.0 | 41.8 |
| Education Level | | |
| No education | 125.0 | 26.3 |
| Primary | 145.0 | 30.5 |
| Secondary and above | 206.0 | 43.3 |
| Occupation | | |
| Housewife | 110.0 | 23.1 |
| Driver | 72.0 | 15.1 |
| Service | 70.0 | 14.7 |
| Business | 63.0 | 13.2 |
| Day labor | 40.0 | 8.4 |
| Unemployed/students | 32.0 | 6.7 |
| retired person | 27.0 | 5.7 |
| Garment’s worker | 22.0 | 4.6 |
| Technical labor/ Electrician | 20.0 | 4.2 |
| Housemaid | 13.0 | 2.7 |
| Others (e.g., beggar) | 7.0 | 1.5 |
| Marital Status | | |
| Married | 405.0 | 85.1 |
| Unmarried | 53.0 | 11.1 |
| Others | 18.0 | 3.8 |
| Family size | | |
| < = 3 members | 124.0 | 26.1 |
| > = 4 to < = 6 members | 314.0 | 66.0 |
| >6 members | 38.0 | 8.0 |
| Regular earning person | | |
| Yes | 222.0 | 46.6 |
| No | 254.0 | 53.4 |
| Worked in last 30 days | | |
| No working days | 255.0 | 53.6 |
| ≤ 10 | 14.0 | 2.9 |
| 11–20 | 36.0 | 7.6 |
| > 20 | 171.0 | 35.9 |
| Income in last 30 days in BDT (respondents) | | |
| No income | 259.0 | 54.4 |
| < = 8000 | 67.0 | 14.1 |
| >8000 to < = 14000 | 86.0 | 18.1 |
| >1400 to < = 20000 | 49.0 | 10.3 |
| >20000 | 15.0 | 3.2 |
| Asset quintiles | | |
| Poorest | 126.0 | 26.5 |
| 2nd | 66.0 | 13.9 |
| 3rd | 95.0 | 20.0 |
| 4th | 94.0 | 19.8 |
| Richest | 95.0 | 20.0 |
## Response to the items for KAP towards COVID-19
Out of the five sub-domains of the knowledge domain, in the knowledge about disease sub-domain, $99\%$ of the respondents reported that they heard about COVID-19 and only $70\%$ reported that COVID-19 was a curable disease. Under the prevention sub-domain, $94\%$ of the respondents reported that cleaning office, home and regularly used equipment with Sanitizer / soap / detergent can reduce the risk of COVID-19. Approximately $65\%$ of the respondents correctly responded that the COVID-19 patient recovers within 28 days. A majority of the respondents ($87\%$) knew that elderly people are more at risk of COVID-19 and $62\%$ believed that children are at risk of COVID-19. Regarding symptoms related knowledge, the highest percentage of respondents reported fever ($83\%$) and the lowest percentage of respondents reported odour-lessness ($46\%$) as the symptoms of COVID-19. About $51\%$ of the respondents mentioned that they received messages about COVID-19 prevention and treatment, and $20\%$ knew that they had access to a quarantine facility nearby their home. In the attitude domain, a majority of the respondents reported that they were aware that an infected person can increase the risk of spreading COVID-19 by roaming freely ($96\%$), infection can be prevented through proper precautions ($95\%$), practicing hygiene ($94\%$), raising awareness ($88\%$), and staying at home ($86\%$). More than half of the respondents incorrectly answered that bathing in hot water could prevent the disease ($64\%$) and that “COVID-19 is the curse of Allah” ($66\%$), while $46\%$ believed that death was the fate of COVID-19.
For the items related to practice, a majority of the respondents reported that they washed their hands in the last 24 hours with soap or sanitizer ($98\%$), wore a mask ($94\%$), maintained social distance ($78\%$) when went outside, and regularly washed their hands for 20–30 seconds ($80\%$). About $80\%$ said they did not eat half-boiled foods, did not spend time with friends ($75\%$), and did not touch mobile phone before washing hands ($73\%$). About $61\%$ of respondents mentioned that they shared bathrooms and toilets with other families and $50\%$ shared food/ water pot (see Table 2).
**Table 2**
| SL | Questions | Yes | No | Don’t know |
| --- | --- | --- | --- | --- |
| SL | Questions | n (%) | n (%) | n (%) |
| Knowledge towards COVOD-19 | Knowledge towards COVOD-19 | Knowledge towards COVOD-19 | Knowledge towards COVOD-19 | Knowledge towards COVOD-19 |
| | Knowledge about disease | | | |
| 1 | Have you heard about COVID-19? | 472(99.16) | 4(0.84) | 0 |
| 2 | Is COVID-19 a contagious disease? | 356(74.79) | 88(18.49) | 32(6.72) |
| 3 | Is COVID-19 a viral disease? | 416(87.39) | 29(6.09) | 31(6.51) |
| 4 | COVID-19 is a curable disease? | 333(69.96) | 90(18.91) | 53(11.13) |
| 5 | COVID-19 is a deadly disease? | 417(87.61) | 42(8.82) | 17(3.57) |
| | Knowledge about prevention | | | |
| 6 | Does COVID-19 transmit through air? | 349(73.32) | 74(15.55) | 53(11.13) |
| 7 | Do you know that symptoms of COVID-19 become visible within 3–14 days of infection? | 352(73.95) | 69(14.5) | 55(11.55) |
| 8 | Do you know that generally a COVID-19 patient gets recovered within 28 days? | 311(65.34) | 61(12.82) | 104(21.85) |
| 9 | Can a person be infected with COVID-19 after recovered from it? | 316(66.39) | 70(14.71) | 90(18.91) |
| 10 | Treatment of COVID-19 can be done at home? | 350(73.53) | 89(18.7) | 37(7.77) |
| 11 | Can COVID-19 be prevented through using face mask? | 381(80.04) | 73(15.34) | 22(4.62) |
| 12 | Without proper disinfecting, usage of cloth mask can increase the risk of infection? | 440(92.44) | 25(5.25) | 11(2.31) |
| 12 | Can frequent hand washing prevent COVID-19? | 416(87.39) | 46(9.66) | 14(2.94) |
| 13 | Maintaining 1 meter or 3 feet distance from infected person is helpful for preventing COVID-19? | 412(86.55) | 47(9.87) | 17(3.57) |
| 14 | Cleaning office, home and regular usage things with Sanitizer/soap/detergent can reduce the risk of COVID-19? | 448(94.12) | 18(3.78) | 10(2.1) |
| | Knowledge about high-risk group | | | |
| 15 | Elderly people are more at risk of COVID-19? | 416(87.39) | 39(8.19) | 21(4.41) |
| 16 | Smokers are at risk of COVID-19? | 372(78.15) | 54(11.34) | 50(10.5) |
| 17 | Rich people are at risk of COVID-19? | 302(63.45) | 111(23.32) | 63(13.24) |
| 18 | Children are at risk of COVID-19? | 297(62.39) | 148(31.09) | 31(6.51) |
| 19 | Pregnant women are at risk of COVID-19? | 305(64.08) | 100(21.01) | 71(14.92) |
| 20 | COVID-19 can be serious to the patients with diabetes and heart disease? | 403(84.66) | 33(6.93) | 40(8.4) |
| | Knowledge about symptoms | | | |
| 21 | Whether the followings are symptoms of COVID-19 | | | |
| 22 | Fever | 394(82.77) | 59(12.39) | 23(4.83) |
| 23 | Dry cough | 376(78.99) | 62(13.03) | 38(7.98) |
| 24 | Difficulty in breathing | 390(81.93) | 57(11.97) | 29(6.09) |
| 25 | Sore throat | 384(80.67) | 59(12.39) | 33(6.93) |
| 26 | Odor lessness | 218(45.8) | 122(25.63) | 136(28.57) |
| 27 | Diarrhea | 236(49.58) | 154(32.35) | 86(18.07) |
| 28 | Weakness | 247(51.89) | 143(30.04) | 86(18.07) |
| | Knowledge about access to information | | | |
| 29 | Do you have access to quarantine facility nearby your house/workplace? | 94(19.75) | 347(72.9) | 35(7.35) |
| 30 | Did you receive any message on COVID-19 prevention and treatment? | 244(51.26) | 232(48.74) | 0 |
| 31 | Do you know any hotline number to contact in case of any symptoms of you or others? | 234(49.16) | 242(50.84) | 0 |
| Attitudes towards COVID-19 | Attitudes towards COVID-19 | | | |
| 1 | Do you think, COVID-19 can be prevented? | 419(88.03) | 24(5.04) | 33(6.93) |
| 2 | Does bathing in hot water reduce risk of COVID-19? | 305(64.08) | 60(12.61) | 111(23.32) |
| 3 | Do you think that dwelling place, sitting place, crowded place or public transport handle (bus, train, taxi, auto rickshaw or rickshaw) are the major sources of COVID-19 infection? | 420(88.24) | 28(5.88) | 28(5.88) |
| 4 | Staying at home, COVID-19 can be prevented | 409(85.92) | 41(8.61) | 26(5.46) |
| 5 | Infected person can increase the risk of spreading COVID-19 by roaming freely? | 456(95.8) | 8(1.68) | 12(2.52) |
| 6 | Practicing hygiene can reduce the risk of COVID-19? | 446(93.7) | 19(3.99) | 11(2.31) |
| 7 | COVID-19 can be prevented through proper precautions? | 453(95.17) | 17(3.57) | 6(1.26) |
| 8 | Awareness is sufficient for COVID-19 prevention? | 420(88.24) | 39(8.19) | 17(3.57) |
| 9 | COVID-19 is the curse of ALLAH | 313(65.76) | 93(19.54) | 70(14.71) |
| 10 | Death is the fate of COVID-19? | 220(46.22) | 215(45.17) | 41(8.61) |
| Practice towards COVID-19 | Practice towards COVID-19 | | | |
| 1 | Do you sneeze into your elbow? | 289(60.71) | 187(39.29) | - |
| 2 | Do you often touch your mouth, eye, and nose? | 259(54.41) | 217(45.59) | - |
| 3 | Do you often eat half boiled fish, meat, egg, or vegetables? | 96(20.17) | 380(79.83) | - |
| 4 | Do you clean your house with detergent/disinfectant regularly? | 372(78.15) | 104(21.85) | - |
| 5 | Do you disinfect your mobile phone with sanitizer regularly? | 256(53.78) | 220(46.22) | - |
| 6 | Do you drink tea/coffee at street stall? | 193(40.55) | 283(59.45) | - |
| 7 | Do you usually spend time with your friends? | 118(24.79) | 358(75.21) | - |
| 8 | Do you touch your clothes, money bag, key ring, plate, glass, laptop, earphone, mobile before washing your hands? | 128(26.89) | 348(73.11) | - |
| 9 | Do you use public transport? | 295(61.97) | 181(38.03) | - |
| 10 | Do you maintain social distance when you go outside? | 369(77.52) | 107(22.48) | - |
| 11 | Do you wash your hands for 20–30 second regularly? | 381(80.04) | 95(19.96) | - |
| 12 | In the last 24 hours, did you wash your hand using soap or sanitizer? | 468(98.32) | 8(1.68) | - |
| 13 | Do you wear mask when you go outside? | 448(94.12) | 28(5.88) | - |
| 14 | Do you share food/water pot with other family members? | 237(49.79) | 239(50.21) | - |
| 15 | Do you share bathroom with other families? | 290(60.92) | 186(39.08) | - |
| 16 | Do you share toilets with other families? | 290(60.92) | 186(39.08) | - |
## Average weighted scores for KAP and sources of knowledge about COVID-19
Average knowledge score was 67.1 out of 100 with a standard deviation of 15.9. Overall, about $25.2\%$ of the respondents had good knowledge (fourth quartile) towards COVID-19, $49.8\%$ had moderate knowledge (second and third quartile) and the rest had poor knowledge (first quartile). Considering the sub-domains of knowledge, about $80\%$ of the respondents had good knowledge about the COVID-19 disease, about $70\%$ about prevention, $27\%$ about risk group, $45\%$ about symptoms, and $7\%$ about access to information. The average attitude score was 73.6 out of 100 with a standard deviation of 14.4. Overall, $48\%$ of the respondents had positive attitude, $52\%$ had negative attitude towards COVID-19. The average practice score was 65.1 out of 100 with a standard deviation of 16.6. Overall, about $21\%$ of the respondents maintained good practice, followed by moderate practice ($46\%$), and poor practice ($33\%$) towards COVID-19 (see Table 3).
**Table 3**
| Variable | No of questions | Weighted scale | Weighted mean ± SD | 95% CI | Level (%) | Level (%).1 | Level (%).2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Variable | | | | | Good | Moderate | Poor |
| Overall knowledge domain | 31.0 | 100.0 | 67.1 ± 15.9 | (65.6–68.5) | 25.2 | 49.8 | 25.0 |
| Knowledge about disease | 5.0 | 20.0 | 16.8 ± 4.0 | (16.4–17.1) | 79.8 | 13.5 | 6.7 |
| Knowledge about prevention | 10.0 | 20.0 | 15.9 ± 4.1 | (15.5–16.2) | 69.8 | 18.7 | 11.6 |
| Knowledge about risk group | 6.0 | 20.0 | 12.9 ± 3.4 | (12.6–13.3) | 26.9 | 43.1 | 30.0 |
| Knowledge about symptoms | 7.0 | 20.0 | 13.5 ± 6.5 | (12.9–14.1) | 45.2 | 17.0 | 37.8 |
| Knowledge about access to information | 3.0 | 20.0 | 8.0 ± 5.9 | (7.5–8.5) | 6.7 | 31.5 | 61.8 |
| Practice domain | 16.0 | 100.0 | 65.1 ± 16.6 | (63.6–66.6) | 21.2 | 46.2 | 32.6 |
| | | | | | Positive | Negative | Negative |
| Attitude domain | 10.0 | 100.0 | 73.6 ± 14.4 | (72.3–74.9) | 47.7 | 52.3 | 52.3 |
We found that television was the main source for three-quarters of the respondents ($75\%$) to get knowledge about COVID-19 followed by neighbours ($6.5\%$) and social media ($5.5\%$) (Fig 1).
**Fig 1:** *Sources of information on COVID-19 (in percentage).*
## Correlation among KAP scores
From the Pearson’s correlation coefficient test between knowledge-attitude, knowledge-practice and attitude-practice, we found that there was a statistically significant positive correlation of knowledge with attitude, knowledge with practice, and attitude with practice (see Table 4).
**Table 4**
| Variable | Correlation coefficient | p-value* |
| --- | --- | --- |
| Knowledge-Attitude | 0.4181 | 0.0 |
| Knowledge-Practice | 0.3122 | 0.0 |
| Attitude-Practice | 0.1226 | 0.007 |
## KAP scores across demographic and socioeconomic characteristics
Overall, KAP scores were higher if the respondents were members of a household compared to the respondents who were the head of a household. The attitude and practice score were significantly different by the status of the household head. On average, female respondents had a higher KAP score compared to the males, and there was a statistically significant difference in attitude and practice score between male and female respondents. We observed that KAP scores towards COVID-19 increased with the level of education. Differences in knowledge and attitude scores by educational level were statistically significant. KAP score also significantly varied across the types of occupation. A higher knowledge and attitude score were found among the unmarried respondents, in contrast, a higher practice score was found among the married respondents and the difference in knowledge by marital status was statistically significant. The average of KAP scores were higher among the non-earning respondents compared to the earning respondents and the difference between them was statistically significant. Similarly, the average KAP scores were higher among the person who did not work or worked up to 10 days in the last 30 days and the differences across the categories in knowledge and attitude score were statistically significant. We found that both knowledge and practice scores were higher among respondents who had an income of more than BDT 20,000 in the last 30 days. However, the attitude score was higher for the respondents who did not have an income. There was a statistically significant difference in attitude score among the respondents across the income group. Furthermore, we found that the respondents from richest quintile had higher KAP scores and the variation of scores of all three domains were statistically significant across the asset quintiles (see Table 5).
**Table 5**
| Description | n | Knowledge Score | p- value | Attitude Score | p- value.1 | Practice Score | p- value.2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Description | n | (Mean ± SD) | | (Mean ± SD) | | (Mean ± SD) | |
| Characteristics | | | | | | | |
| Relationship with HH head a) | | | | | | | |
| Household head | 309 | 66.1± 16.3 | 0.065 | 72.2± 15.4 | 0.005 | 63.9± 16.8 | 0.028 |
| Others | 167 | 68.9± 15.0 | | 76± 11.9 | | 67.4± 16 | |
| Age group b) | | | | | | | |
| ≤30 | 188 | 68.7± 15.2 | 0.098 | 76.2± 12 | 0.008 | 65± 15.7 | 0.348 |
| 31–45 | 204 | 66.7± 15.4 | | 72.5± 14.5 | | 66.4± 16.7 | |
| 46–60 | 67 | 65± 16.9 | | 70.4± 17.2 | | 62.5± 17.6 | |
| >60 | 17 | 60.4± 22.4 | | 70± 20.3 | | 62.5± 21.2 | |
| Sex a) | | | | | | | |
| Male | 277 | 66.5± 16.6 | 0.373 | 72.4± 15.1 | 0.038 | 62.3± 16.3 | 0.038 |
| Female | 199 | 67.8± 14.9 | | 75.2± 13.1 | | 69.1± 16.2 | |
| Education Level b) | | | | | | | |
| No education | 125 | 59.5± 17.8 | 0.001 | 68.6± 16.4 | 0.001 | 63.3± 17.1 | 0.352 |
| Primary | 145 | 68.1± 14.1 | | 74.6± 13 | | 65.9± 15.9 | |
| Secondary and above | 206 | 70.9± 14.3 | | 75.9± 13.2 | | 65.7± 16.8 | |
| Occupation b) | | | | | | | |
| Housewife | 70 | 67.5± 19.3 | 0.037 | 74.1± 13 | 0.005 | 65.3± 17.1 | 0.001 |
| Driver | 63 | 67.9± 13.6 | | 74.6± 11.6 | | 68.1± 14.6 | |
| Service | 110 | 67.1± 15 | | 76.2± 11.3 | | 69.4± 16.5 | |
| Business | 13 | 65.2± 8.3 | | 66.9± 14.4 | | 73.1± 10.9 | |
| Unemployed/students | 72 | 63.3± 16.7 | | 68.1± 18.3 | | 58.6± 15.1 | |
| Day labor | 40 | 63.4± 17.3 | | 69.5± 17.4 | | 57.7± 16.9 | |
| Retired person | 20 | 67.0± 15.8 | | 78.0± 13.6 | | 66.6± 19.6 | |
| Technical labor | 27 | 75.3± 10.1 | | 76.7± 10.7 | | 70.4± 15.5 | |
| Garment worker | 22 | 73.0± 8.7 | | 75.9± 11.4 | | 66.2± 14.8 | |
| Housemaid | 32 | 68.8± 18.1 | | 74.7± 18.5 | | 60.2± 18.2 | |
| Others (e.g., beggar) | 7 | 59.0± 19.4 | | 72.9± 12.5 | | 61.6± 9.1 | |
| Marital Status b) | | | | | | | |
| Married | 405 | 66.3± 15.9 | 0.025 | 73.2± 14.5 | 0.234 | 65.4± 16.3 | 0.388 |
| Unmarried | 53 | 72.6± 14.4 | | 75.1± 10.7 | | 64.5± 16.9 | |
| Others | 18 | 66.9± 17.8 | | 78.3± 20.1 | | 60.1± 22.3 | |
| Family size b) | | | | | | | |
| < = 3 members | 124 | 66.3± 15.2 | 0.733 | 74.4± 15.2 | 0.768 | 64.5± 17.4 | 0.835 |
| > = 4 to < = 6 members | 314 | 67.2± 16.5 | | 73.2± 14.4 | | 65.3± 16.2 | |
| >6 members | 38 | 68.5± 13.4 | | 73.7± 10.5 | | 66.1± 17.3 | |
| Regular earning person b) | | | | | | | |
| Yes | 222 | 65.6± 16.4 | 0.054 | 71.4± 15.6 | 0.002 | 63.6± 16.3 | 0.058 |
| No | 254 | 68.4± 15.4 | | 75.4± 12.9 | | 66.5± 16.8 | |
| Worked in last 30 days b) | | | | | | | |
| No working days | 255 | 68.3± 15.4 | 0.039 | 75.5± 12.9 | 0.001 | 66.3± 17 | 0.354 |
| ≤ 10 | 14 | 68.1± 15.7 | | 77.1± 9.1 | | 60.7± 21.3 | |
| 11 to 20 | 36 | 60.4± 18.1 | | 66.7± 18.2 | | 63.4± 14.1 | |
| > 20 | 171 | 66.5± 15.9 | | 71.9± 15.3 | | 64.1± 16.1 | |
| Income in last 30 days b) (BDT) | | | | | | | |
| No income | 259 | 68.1± 15.4 | 0.15 | 75.3± 13.1 | 0.045 | 66.5± 17 | 0.081 |
| < = 8000 | 67 | 66.9± 14.7 | | 73.4± 13.7 | | 64± 14.9 | |
| >8000 to < = 14000 | 86 | 63.8± 17.8 | | 71± 15.8 | | 63.8± 16.3 | |
| >14000-< = 20000 | 49 | 65.7± 17 | | 69.8± 17.6 | | 60.2± 15.8 | |
| >20000 | 15 | 72.3± 11.9 | | 72± 15.2 | | 70± 18.3 | |
| Asset quintiles b) | | | | | | | |
| Poorest | 126 | 63.4± 17.5 | 0.001 | 72.1± 14.7 | 0.008 | 62.2± 17.7 | 0.004 |
| 2nd | 66 | 61.9± 17.6 | | 70.5± 15.1 | | 65.7± 15.6 | |
| 3rd | 95 | 66.2± 15.6 | | 72.2± 14.7 | | 64.7± 17.6 | |
| 4th | 94 | 68.3± 14.3 | | 75.2± 14.3 | | 63.7± 16.2 | |
| Richest | 95 | 75± 10.5 | | 77.5± 12.1 | | 70.5± 14 | |
## Determinants of KAP towards COVID-19
The crude and adjusted association of the dependent variables i.e., knowledge, attitude, and practice scores with independent socioeconomic characteristics of the respondents i.e., age, sex, education level, occupation, asset quantiles, etc. are presented in Table 6. In the crude association, we found that age, education level, occupation, marital status, and asset quintiles were significantly associated with the knowledge score. However, in the adjusted association, compared to the group without education, the knowledge score was significantly higher among the respondents who had primary (β = 8.554; $p \leq 0.001$) and secondary and above (β = 11.431; $p \leq 0.001$) education level. The knowledge score was significantly higher among the respondents from the fourth (β = 4.463; $p \leq 0.05$) and the richest (β = 3.585; $p \leq 0.001$) quintiles compared to the respondents from the poorest quintile.
**Table 6**
| Characteristics | Unnamed: 1 | Model 1: Knowledge | Model 1: Knowledge.1 | Model 2: Attitude | Model 2: Attitude.1 | Model 3: Practice | Model 3: Practice.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Characteristics | n | Crude coef. (95% CI) | Adjusted coef. (95% CI) | Crude coef. (95% CI) | Adjusted coef. (95% CI) | Crude coef. (95% CI) | Adjusted coef. (95% CI) |
| Household head status | | | | | | | |
| Head | 309 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| Members | 167 | 2.821 (-0.17,5.82) | -0.659 (-5.29,3.97) | 3.815** (1.12,6.51) | 1.362 (-2.91,5.64) | 3.487* (0.37,6.61) | -2.992 (-7.96,1.97) |
| Age group | | | | | | | |
| ≤30 | 188 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| 31–45 | 204 | -2.005 (-5.15,1.14) | -1.281 (-4.88,2.31) | -3.670* (-6.50,-0.84) | -4.984** (-8.30,-1.67) | 1.431 (-1.87,4.73) | 2.970 (-0.88,6.82) |
| 46–60 | 67 | -3.710 (-8.14,0.72) | -1.941 (-6.93,3.05) | -5.722** (-9.70,-1.74) | -6.822** (-11.42,-2.22) | -2.460 (-7.10,2.18) | -0.285 (-5.63,5.06) |
| >60 | 17 | -8.371* (-16.26,-0.48) | -7.947 (-16.42,0.52) | -6.170 (-13.25,0.91) | -9.409* (-17.23,-1.59) | -2.460 (-10.72,5.80) | 0.485 (-8.59,9.56) |
| Sex | | | | | | | |
| Male | 277 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| Female | 199 | 1.317 (-1.59,4.22) | 0.349 (-4.37,5.07) | 2.757* (0.14,5.37) | -0.842 (-5.20,3.51) | 6.853*** (3.88,9.82) | 6.841** (1.78,11.90) |
| Education Level | | | | | | | |
| No education | 125 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| Primary | 145 | 8.554*** (4.90,12.21) | 7.611*** (3.89,11.33) | 5.992*** (2.62,9.36) | 4.562** (1.13,8.00) | 2.605 (-1.38,6.59) | 1.707 (-2.28,5.69) |
| Secondary and above | 206 | 11.431*** (8.04,14.82) | 8.903*** (5.12,12.69) | 7.362*** (4.23,10.49) | 4.204* (0.71,7.70) | 2.416 (-1.28,6.12) | 0.019 (-4.04,4.08) |
| Occupation | | | | | | | |
| Housewife | 110 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| Driver | 72 | 0.376 (-5.00,5.75) | -0.365 (-5.75,5.02) | 0.460 (-4.36,5.28) | 0.381 (-4.59,5.35) | 2.788 (-2.71,8.29) | 3.046 (-2.72,8.82) |
| Service | 70 | -0.456 (-5.19,4.28) | -5.064 (-11.68,1.56) | 2.039 (-2.21,6.28) | -4.345 (-10.45,1.77) | 4.107 (-0.73,8.95) | 0.119 (-6.98,7.21) |
| Business | 63 | -2.344 (-11.69,7.00) | 1.159 (-8.52,10.84) | -7.220 (-15.61,1.17) | -5.999 (-14.93,2.93) | 7.809 (-1.75,17.37) | 5.206 (-5.17,15.58) |
| Day labor | 40 | -4.279 (-9.47,0.92) | -2.676 (-8.04,2.68) | -6.087* (-10.75,-1.43) | -6.111* (-11.06,-1.16) | -6.674* (-11.99,-1.36) | -4.854 (-10.60,0.89) |
| Unemployed/students | 32 | -4.147 (-10.28,1.99) | -1.235 (-7.49,5.02) | -4.643 (-10.15,0.86) | -4.434 (-10.20,1.34) | -7.612* (-13.89,-1.34) | -5.877 (-12.58,0.82) |
| Retired person | 27 | -0.554 (-8.40,7.29) | -2.225 (-10.09,5.64) | 3.857 (-3.18,10.90) | 2.423 (-4.83,9.68) | 1.295 (-6.73,9.32) | 2.281 (-6.14,10.71) |
| Garment workers | 22 | 7.708* (0.70,14.72) | -2.159 (-11.34,7.02) | 2.524 (-3.77,8.81) | -4.423 (-12.89,4.05) | 5.103 (-2.07,12.27) | 5.465 (-4.37,15.30) |
| Technical labor | 20 | 5.433 (-2.13,13.00) | 6.786 (-0.54,14.11) | 1.766 (-5.02,8.55) | 1.240 (-5.52,8.00) | 0.925 (-6.81,8.66) | 2.172 (-5.68,10.02) |
| Housemaid | 13 | 1.296 (-5.31,7.90) | -3.141 (-10.42,4.13) | 0.545 (-5.38,6.47) | -3.100 (-9.81,3.61) | -5.112 (-11.87,1.64) | -5.732 (-13.53,2.07) |
| Others (e.g., beggar) | 7 | -8.559 (-20.83,3.71) | -4.202 (-16.25,7.84) | -1.286 (-12.29,9.72) | 1.006 (-10.11,12.12) | -3.661 (-16.21,8.89) | -5.152 (-18.06,7.76) |
| Marital status | | | | | | | |
| Married | 405 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| Unmarried | 53 | 6.268** (1.73,10.81) | 1.074 (-5.12,7.27) | 1.934 (-2.18,6.05) | -3.828 (-9.55,1.89) | -0.943 (-5.71,3.82) | -1.372 (-8.01,5.27) |
| Others (divorce/widow) | 18 | 0.610 (-6.88,8.10) | 3.610 (-4.16,11.38) | 5.173 (-1.62,11.96) | 9.326* (2.16,16.49) | -5.378 (-13.24,2.48) | -10.579* (-18.90,-2.26) |
| Family size | | | | | | | |
| < = 3 members | 124 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| > 4 to < = 6 members | 214 | 0.866 (-2.45,4.19) | 1.874 (-1.48,5.23) | -1.106 (-4.10,1.89) | 0.079 (-3.02,3.18) | 0.821 (-2.65,4.29) | 0.678 (-2.92,4.28) |
| > 6 members | 38 | 2.231 (-3.57,8.03) | 3.131 (-2.64,8.90) | -0.671 (-5.91,4.57) | 0.416 (-4.91,5.74) | 1.653 (-4.41,7.71) | 1.697 (-4.48,7.88) |
| Regular earning | | | | | | | |
| Yes | 222 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| No | 254 | 2.815 (-0.05,5.68) | 5.007* (1.10,8.91) | 3.992** (1.42,6.56) | 4.895** (1.29,8.50) | 2.888 (-0.10,5.88) | 1.900 (-2.28,6.08) |
| Asset quintiles | | | | | | | |
| Poorest | 126 | Ref. | Ref. | Ref. | Ref. | Ref. | Ref. |
| Second | 66 | -1.473 (-6.05,3.10) | -1.073 (-5.59,3.44) | -1.609 (-5.85,2.63) | -2.667 (-6.83,1.50) | 3.517 (-1.38,8.42) | 3.236 (-1.60,8.07) |
| Third | 95 | 2.822 (-1.27,6.91) | 2.204 (-1.83,6.24) | 0.147 (-3.65,3.94) | -1.006 (-4.73,2.72) | 2.469 (-1.91,6.85) | 1.012 (-3.31,5.34) |
| Fourth | 94 | 4.860* (0.76,8.96) | 4.436* (0.35,8.52) | 3.149 (-0.66,6.96) | 2.979 (-0.79,6.75) | 1.494 (-2.90,5.89) | 0.676 (-3.70,5.06) |
| Richest | 95 | 11.608*** (7.52,15.70) | 10.217*** (6.05,14.39) | 5.410** (1.62,9.20) | 4.066* (0.22,7.91) | 8.324*** (3.94,12.71) | 6.501** (2.03,10.97) |
| Constant | | | 56.326*** (49.91,62.74) | | 72.916*** (66.99,78.84) | | 59.035*** (52.16,65.91) |
| N | | | 476 | | 476 | | 476 |
| p< | | | 0.000 | | 0.000 | | 0.000 |
| R-squared | | | 0.177 | | 0.140 | | 0.133 |
We found that attitude score was significantly decreasing with the increase in respondents’ age. Attitude score was significantly higher among the respondents who had primary level (β = 4.563; $p \leq 0.01$) and secondary and above level of education (β = 4.204; $p \leq 0.05$) compared to the respondents having no-education. In terms of occupation, the attitude score was significantly lower among the day labourer compared to the housewives. Furthermore, attitude score was significantly higher among the respondents from the richest quintile (β = 4.066, $p \leq 0.05$) compared to the respondents from poorest quintile.
While examining the association of practice score with the characteristics of the participants, we found that the female respondents had significantly higher practice scores compared to the male counterparts (β = 6.841; $p \leq 0.01$). Divorced/ widowed respondents had a significantly lower practice score compared to those married (β = -10.579; $p \leq 0.05$). The practice score had a positive association with the asset quintile; respondents from the richest quintile had significantly higher practice score by 6.5 points compared to the respondents from the poorest quintile.
## Discussion
The study sought to assess the KAP of selected slum dwellers in Dhaka city, Bangladesh towards COVID-19 and found that a small proportion of the sample surveyed ($25.2\%$) demonstrated good knowledge regarding the infection, almost half of them ($48\%$) expressed a positive attitude about controlling and preventing this disease, and like good knowledge level, a small proportion of the sample surveyed ($21\%$) had favourable practices towards COVID-19 transmission control and prevention.
Slum dwellers in Bangladesh reside in congested spaces, while having a very poor surrounding environment, thus, these people are more susceptible to COVID-19 infection [55, 56]. A cross-sectional study in Bangladesh reported that the seroprevalence rate of COVID-19 was the highest ($64\%$) in slum areas, whereas $38\%$ in urban areas and $29\%$ in rural areas [57]. The current assessment, therefore, is useful to inform the policymakers about the level and the dimensions of KAP and the associated determinants to design targeted interventions.
A recent study conducted in slum areas reported that the overall correct rate of knowledge of COVID-19 was $36\%$, which was lower compared to our study ($67\%$). However, the correct attitude and practice rate reported in that study was higher compared to the current study ($88\%$ vs $74\%$) and ($82\%$ vs $65\%$), respectively [30]. The possible reason for this difference could be the difference in data collection period and choices of items for the different dimensions of KAP. Ferdous et al. [ 2020] conducted a study on general population that reported $48.3\%$ of the respondents had more accurate knowledge, $62.3\%$ had positive attitude, and $55.2\%$ had good practice. Compared to the general population, lower proportion of slum population had good knowledge ($48.3\%$ vs $25.2\%$), good attitude level ($62.3\%$ vs $48\%$) and good practice ($55.2\%$ vs $21\%$) [58]. As mentioned earlier that the slum people live in a disadvantaged environment have lower access to information which was also reflected in the findings of our study (only $6.7\%$ had good access to information). Moreover, they live in a congested place for which they might have limited scope of maintaining good practice towards COVID-19 without having a strong motivation. Another study on KAP towards COVID-19 conducted among Bangladeshi youth reported that a higher proportion of the respondents had good knowledge ($61.2\%$ vs $25.2\%$), positive attitude ($79\%$ vs $48\%$), and good practices ($52\%$ vs $21\%$) compared to our study [59]. In our study, it was also evident that KAP score was higher among the youth respondents compared to the elderly. It is noteworthy that in all three studies mentioned above, the reported attitude score was higher compared to the knowledge and practice score, which is similar to our findings. This might be attributable to the fact that the attitude reflects the respondents believe and perception which is different from their practice and the existing level of knowledge.
Our study also revealed that $75\%$ of the slum people depended on television to gather COVID-19 related information, which is remarkably higher than other sources, for example, internet, social media. This finding is similar to a study that was conducted to assess the KAP on COVID-19 in the slum area of Dhaka, Bangladesh [30]. Another study conducted to evaluate the KAP in all the administrative districts of Bangladesh also reported a similar finding on people’s access to information [60]. The lack of appropriate campaign, limited internet access, smartphone possession, and low access to health information could be the reasons for having such lower proportion of slum dwellers with good knowledge about the COVID-19. Therefore, interventions such as health education programs incorporating mass media with a variety of advertisements can be influential in the dissemination of information about COVID-19.
Two important characteristics e.g., education and asset quintile were significantly associated with the knowledge and attitude of the respondents. We found that higher education was significantly associated with having a higher knowledge score towards COVID-19. A similar association of knowledge related to COVID-19 with higher education were observed in other studies in Bangladesh [30] and Iran [39]. Educated individuals might have good access to information and are aware of the potential impact of COVID-19 from a variety of sources for example newspaper. As a result, they have more knowledge on COVID-19 compared to the uneducated individuals. This finding indicated the importance of educational intervention to improve knowledge of COVID-19 among the slum dwellers. Previous studies have also reported that educational interventions had a significant impact on improving knowledge level [30, 58, 61]. We found that knowledge and attitude were significantly correlated. Comparatively higher educated respondents also had a positive attitude towards COVID-19. A similar association of attitude with education level was observed in other recent local and international studies [30, 58, 62].
Among slum people, those who belonged to the richer quintile had a significantly higher level of KAP compared to the poorest quintile. The economic ability of respondents enables them to access information about COVID-19 which might be a significant factor behind this phenomenon.
About half of the respondents ($50\%$) believed that death is the ultimate fate of COVID-19. This may indicate the limited access to accurate and timely information among slum population [60]. Only one in five persons maintained good practice against the COVID-19 spreading in slum areas. Another KAP study in Bangladesh revealed a similar result that most of the respondents had poor practice towards COVID-19 [59]. The higher prevalence of poor practice may be the result of existing dwelling conditions in the slum areas. About $60\%$ of the respondents stated that they shared their toilets and bathrooms with other households. Furthermore, due to their socioeconomic status, they had to move outside frequently for earning their livelihoods. We found that female respondents had a significantly higher practice score compared to the male. In the context of Bangladesh, the male is the main wage earner in a household. Therefore, they have to go outside for work and travel by local transport, which ultimately exposes them to COVID-19.
## Limitations of the study
This study had some important limitations. Firstly, the respondents for this study included only the households with mobile phone users. Data were collected over mobile phone calls, which is a new process in Bangladesh. A majority of the consented respondents may include those who were more concerned about the COVID-19 emergency, which might include response bias. Therefore, the results may not be generalizable to other population who are not mobile phone users. Secondly, this is a cross-sectional study; thus, causal inferences cannot be drawn between the significant socioeconomic characteristics and the KAP level. Thirdly, as the study was a component of a rapid assessment, the questions related to KAP towards COVID-19 transmission control and prevention were adopted and contextualized for Bangladesh from published literature. Health-seeking paradigm of the health belief model could have been applied to validate the study instrument which would have strengthened the study findings [36]. However, this is one of the few studies that assess the KAP of urban slum dwellers in Bangladesh towards COVID-19 transmission prevention.
## Conclusions
This study provides a comprehensive assessment of KAP levels with respect to COVID-19 from the urban slum dwellers in Bangladesh. Overall, of the slum dwellers included in this study, one-fourth demonstrated poor knowledge, almost half had a good attitude level, and one-third maintained poor practice towards COVID-19. However, they had good knowledge about the disease and its prevention but had lack of knowledge about symptoms of the disease, risk population group, and access to COVID-19 related information. Despite one-third of the respondents having poor practice level, a majority of the respondents used a mask, washed their hands for 20–30 seconds, and used soap or sanitizer. Valuable insights on demographic characteristics associated with KAP among the slum population can help policymakers in designing health education programs, awareness raising campaigns, and behavioural change communication interventions. These programs can be designed in consultation with political, religious, and other influential community-based groups. Importance should be given to the groups who have lower KAP scores, such as individuals aged more than 60 years, uneducated, driver, day labourer, unemployed, people who were regular earning person of a household and belong to the poor socioeconomic group. Expanding living space is not possible in slums; thus, prioritizing household hygiene may enable them to maintain practices to the extent possible. Further studies, with more rigorously developed and validated KAP instruments can be conducted to obtain a more robust estimate of KAP towards COVID-19 transmission-prevention and compare the findings with the current study.
## References
1. 1World Health Organization. Novel Coronavirus (2019-nCoV): Situation Report-1. Data as reported by: 20 January 2020. Geneva; 2020.
2. 2World Health Organization. COVID 19 Public Health Emergency of International Concern (PHEIC). Global research and innovation forum: towards a research roadmap. Geneva; 2020.
3. 3World Health Organization. Bangladesh COVID-19 Situation Report #10 1. Dhaka; 2020.
4. 4Worldometer. COVID-19 CORONAVIRUS PANDEMIC: Reported Cases and Deaths by Country or Territory. 2021.
5. **Modes of transmission of virus causing COVID-19: implications for IPC precaution recommendations**. *N Engl J Med* (2020.0) **382** 1564-1567. PMID: 32182409
6. Wouters OJ, Shadlen KC, Salcher-Konrad M, Pollard AJ, Larson HJ, Teerawattananon Y. **Challenges in ensuring global access to COVID-19 vaccines: production, affordability, allocation, and deployment**. *Lancet (London, England)* (2021.0) **6736** 1-12. DOI: 10.1016/S0140-6736(21)00306-8
7. West R, Michie S, Rubin GJ, Amlôt R. **Applying principles of behaviour change to reduce SARS-CoV-2 transmission**. *Nat Hum Behav* (2020.0) **4** 451-459. DOI: 10.1038/s41562-020-0887-9
8. Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W. **The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions**. *Ann Behav Med a Publ Soc Behav Med* (2013.0) **46** 81-95. DOI: 10.1007/s12160-013-9486-6
9. Michie S, van Stralen MM, West R. **The behaviour change wheel: A new method for characterising and designing behaviour change interventions**. *Implement Sci* (2011.0) **6** 42. DOI: 10.1186/1748-5908-6-42
10. Schaller M.. **The behavioural immune system and the psychology of human sociality**. *Philos Trans R Soc B Biol Sci* (2011.0) **366** 3418-3426. DOI: 10.1098/rstb.2011.0029
11. Jones JH, Salathé M. **Early assessment of anxiety and behavioral response to novel swine-origin influenza a(H1N1)**. *PLoS One* (2009.0) **4** 2-9. DOI: 10.1371/journal.pone.0008032
12. Rubin GJ, Amlôt R, Page L, Wessely S. **Public perceptions, anxiety, and behaviour change in relation to the swine flu outbreak: Cross sectional telephone survey**. *BMJ* (2009.0) **339** 156. DOI: 10.1136/bmj.b2651
13. Pinchoff J, Serino A, Merritt AP, Hunter G, Silva M, Parikh P. **Evidence-based process for prioritizing positive behaviors for promotion: Zika prevention in latin America and the caribbean and applicability to future health emergency responses**. *Glob Heal Sci Pract* (2019.0) **7** 404-417. DOI: 10.9745/GHSP-D-19-00188
14. Yan QL, Tang SY, Xiao YN. **Impact of individual behaviour change on the spread of emerging infectious diseases**. *Stat Med* (2018.0) **37** 948-969. DOI: 10.1002/sim.7548
15. Weston D, Ip A, Amlôt R. **Examining the application of behaviour change theories in the context of infectious disease outbreaks and emergency response: A review of reviews**. *BMC Public Health* (2020.0) **20** 1-19. DOI: 10.1186/s12889-020-09519-2
16. Weston D, Hauck K, Amlôt R. **Infection prevention behaviour and infectious disease modelling: A review of the literature and recommendations for the future**. *BMC Public Health* (2018.0) **18** 1-16. DOI: 10.1186/s12889-018-5223-1
17. Verelst F, Willem L, Beutels P. **Behavioural change models for infectious disease transmission: A systematic review (2010–2015)**. *J R Soc Interface* (2016.0) **13**. DOI: 10.1098/rsif.2016.0820
18. 18WHO Bangladesh. COVID-19 Situation Report No. #6 07. 2020.
19. 19DGHS. National Guidelines on Clinical Management of Coronavirus Disease 2019 (COVID-19). Dhaka, Bangladesh; 2020.
20. Ejeh FE, Saidu AS, Owoicho S, Maurice NA, Jauro S, Madukaji L. **Knowledge, attitude, and practice among healthcare workers towards COVID-19 outbreak in Nigeria**. *Heliyon* (2020.0) **6** e05557. DOI: 10.1016/j.heliyon.2020.e05557
21. 21The Daily Star. Coronavirus outbreak: Govt orders closure of public, private offices from March 26 to April 4. Mar 2020.
22. 22The Daily Star. Coronavirus outbreak: shutdown won’t be extended after May 30. May 2020.
23. 23Dhaka Tribune. Restriction on public movement extended till August 31. Aug 2020.
24. 24The World Bank Group. Population living in slums (% of urban population)—Bangladesh | Data. 2020 [cited 3 Jan 2022]. https://data.worldbank.org/indicator/EN.POP.SLUM.UR.ZS?locations=BD
25. 25The World Bank Group. Urban population (% of total population) | Data. 2020 [cited 3 Jan 2022]. https://data.worldbank.org/indicator/SP.URB.TOTL.IN.ZS
26. 26UNICEF. Children in cities | UNICEF. 2020 [cited 3 Jan 2022]. https://www.unicef.org/bangladesh/en/children-cities
27. 27icddrb. Slum Health in Bangladesh: Coping with Ill Health in Urban Slums-Health Seeking and Healthcare Expenditure. Dhaka, Bangladesh; 2019.
28. Islam T, Kibria MG. **Challenges to the prevention of COVID-19 spread in slums of Bangladesh**. *Journal of public health (Oxford, England)* (2020.0) 637-638. DOI: 10.1093/pubmed/fdaa088
29. 29The Daily Star. Urban slums of Bangladesh. 2009. https://www.thedailystar.net/news-detail-93293
30. Islam S, Emran GI, Rahman E, Banik R. **Knowledge, attitudes and practices associated with the COVID-19 among slum dwellers resided in Dhaka City: a Bangladeshi interview-based survey**. *J Public Health (Oxf)* (2020.0) **43** 13-25. DOI: 10.1093/pubmed/fdaa182
31. Rabbani MG, Akter O, Hasan MZ, Samad N, Mahmood SS, Joarder T. **Knowledge, Attitude and Practice towards COVID-19 among people in Bangladesh during the pandemic: a cross-sectional study**. *medRxiv* (2020.0). DOI: 10.1101/2020.09.22.20198275
32. Taye GM, Bose L, Beressa TB, Tefera GM, Mosisa B, Dinsa H. **COVID-19 knowledge, attitudes, and prevention practices among people with hypertension and diabetes mellitus attending public health facilities in Ambo, Ethiopia**. *Infect Drug Resist* (2020.0) **13** 4203-4214. DOI: 10.2147/IDR.S283999
33. Rahman SMM, Akter A, Mostari KF, Ferdousi S, Ummon IJ, Naafi SM. **Assessment of knowledge, attitudes and practices towards prevention of coronavirus disease (COVID-19) among Bangladeshi population**. *Bangladesh Med Res Counc Bull* (2020.0) **46** 73-82. DOI: 10.3329/bmrcb.v46i2.49015
34. Mahmood SS, Hasan Z, Hasan AMR, Rabbani G, Begum F, Bin Yousuf T. **Health system impact of COVID- 19 on urban slum population of Bangladesh: a mixed- method rapid assessment study**. *BMJ Open* (2022.0). DOI: 10.1136/bmjopen-2021-057402
35. Xu N, Zhang Y, Zhang X, Zhang G, Guo Z, Zhao N. **Knowledge, Attitudes, and Practices of Urban Residents Toward COVID-19 in Shaanxi During the Post-lockdown Period**. *Front public Heal* (2021.0) **9** 659797. DOI: 10.3389/fpubh.2021.659797
36. Opatola Kikelomo O, Olanrewaju MF, Atulomah NO. **Knowledge, Attitude, Per‑ception of Covid‑19 Prevention Practices Among Residents in Selected Local Government Areas in Lagos State Nigeria**. *Afr J Biol Med Res* (2021.0) **4** 17-38
37. Karimy M, Bastami F, Sharifat R, Heydarabadi AB, Hatamzadeh N, Pakpour AH. **Factors related to preventive COVID-19 behaviors using health belief model among general population: a cross-sectional study in Iran**. *BMC Public Health* (2021.0) **21** 1934. DOI: 10.1186/s12889-021-11983-3
38. Zhong BL, Luo W, Li HM, Zhang QQ, Liu XG, Li WT. **Knowledge, attitudes, and practices towards COVID-19 among chinese residents during the rapid rise period of the COVID-19 outbreak: A quick online cross-sectional survey**. *Int J Biol Sci* (2020.0) **16** 1745-1752. DOI: 10.7150/ijbs.45221
39. Erfani A, Shahriarirad R, Ranjbar K, Mirahmadizadeh A, Moghadami M. **Knowledge, attitude and practice toward the novel coronavirus (COVID-19) outbreak- A population-based survey in Iran**. *Bull World Health Organ* (2020.0) 2-3. DOI: 10.2471/BLT.20.251561
40. 40WHO. Coronavirus disease (COVID-19): How is it transmitted? 2019.
41. Rosenstock IM. **The Health Belief Model and Preventive Health Behavior**. *Heal Educ Behav* (1977.0) **2** 354-386. DOI: 10.1177/109019817400200405
42. Janz NK, Becker MH. **The Health Belief Model: A Decade Later**. *Heal Educ Behav* (1984.0) **11** 1-47. DOI: 10.1177/109019818401100101
43. 43Karen Glanz, Barbara K. Rimer KV. Health Behavior adn Health Education. Jossey Bass. 2008.
44. 44Khaton SE. Effect of Preventive Program about Reproductive Tract Infections on Knowledge, Beliefs and Practices among Rural Women Based on Health Belief Model. 2020.
45. Jeihooni AK, Kashfi SH, Bahmandost M, Harsini PA. **Promoting preventive behaviors of nosocomial infections in nurses: The effect of an educational program based on health belief model**. *Investig y Educ en Enferm* (2018.0) **36**. DOI: 10.17533/udea.iee.v36n1e09
46. Li ZT, Yang SS, Zhang XX, Fisher EB, Tian BC, Sun XY. **Complex relation among Health Belief Model components in TB prevention and care**. *Public Health* (2015.0) **129** 907-915. DOI: 10.1016/j.puhe.2015.04.008
47. Erkin Ö, Özsoy S. **Validity and Reliability of Health Belief Model**. *Acad Res Int* (2012.0) **2** 31-40
48. Buldeo P, Gilbert L. **Exploring the Health Belief Model and first-year students responses to HIV/AIDS and VCT at a South African university**. *African J AIDS Res* (2015.0) **14** 209-218. DOI: 10.2989/16085906.2015.1052527
49. Sim SW, Moey KSP, Tan NC. **The use of facemasks to prevent respiratory infection: A literature review in the context of the Health Belief Model**. *Singapore Med J* (2014.0) **55** 160-167. DOI: 10.11622/smedj.2014037
50. Taber KS. **The Use of Cronbach’s Alpha When Developing and Reporting Research Instruments in Science Education**. *Res Sci Educ* (2018.0) **48** 1273-1296. DOI: 10.1007/s11165-016-9602-2
51. Akalu Y, Ayelign B, Molla MD. **Knowledge, Attitude and Practice Towards COVID-19 Among Chronic Disease Patients at Addis Zemen Hospital, Northwest Ethiopia**. *Infect Drug Resist* (2020.0) **13** 1949-1960. DOI: 10.2147/IDR.S258736
52. Seid MA, Hussen MS. **Knowledge and attitude towards antimicrobial resistance among final year undergraduate paramedical students at University of Gondar, Ethiopia**. *BMC Infect Dis* (2018.0) **18** 312. DOI: 10.1186/s12879-018-3199-1
53. Vyas S, Kumaranayake L. **Constructing socio-economic status indices: how to use principal components analysis**. *Heal Policy Plan* (2006.0) **21** 459-468. DOI: 10.1093/heapol/czl029
54. 54StataCorp. Stata Statistical Software: Release 16. College Station, TX: StataCorp LP.; 2019.
55. Friesen J, Pelz PF. **COVID-19 and Slums: A Pandemic Highlights Gaps in Knowledge About Urban Poverty**. *JMIR public Heal Surveill* (2020.0) **6** e19578-e19578. DOI: 10.2196/19578
56. Sahasranaman A, Jensen HJ. **Spread of COVID-19 in urban neighbourhoods and slums of the developing world**. *J R Soc Interface* (2021.0) **18** 20200599. DOI: 10.1098/rsif.2020.0599
57. Bhuiyan TR, Akhtar M, Akter A, Khaton F, Rahman SIA, Ferdous J. **Seroprevalence of SARS-CoV-2 antibodies in Bangladesh related to novel coronavirus infection**. *IJID Reg* (2022.0) **2** 198-203. DOI: 10.1016/j.ijregi.2022.01.013
58. Ferdous MZ, Islam MS, Sikder MT, Mosaddek ASM, Zegarra-Valdivia JA, Gozal D. **Knowledge, attitude, and practice regarding COVID-19 outbreak in Bangladesh: An onlinebased cross-sectional study**. *PLoS One* (2020.0) **15** 1-17. DOI: 10.1371/journal.pone.0239254
59. Banik R, Rahman M, Sikder MT, Rahman QM, Pranta MUR. **Knowledge, attitudes, and practices related to the COVID-19 pandemic among Bangladeshi youth: a web-based cross-sectional analysis**. *J Public Heal* (2021.0). DOI: 10.1007/s10389-020-01432-7
60. Barua Z, Barua S, Aktar S, Kabir N, Li M. **Effects of misinformation on COVID-19 individual responses and recommendations for resilience of disastrous consequences of misinformation**. *Prog Disaster Sci* (2020.0) **8** 100119. DOI: 10.1016/j.pdisas.2020.100119
61. Wong CL, Chen J, Chow KM, Law BMH, Chan DNS, So WKW. **Knowledge, attitudes and practices towards COVID-19 amongst ethnic minorities in Hong Kong**. *Int J Environ Res Public Health* (2020.0) **17** 1-13. DOI: 10.3390/ijerph17217878
62. Azlan AA, Hamzah MR, Sern TJ, Ayub SH, Mohamad E. **Public knowledge, attitudes and practices towards COVID-19: A cross-sectional study in Malaysia**. *PLoS One* (2020.0) **15** 1-15. DOI: 10.1371/journal.pone.0233668
|
---
title: Knowledge, attitude, and practice of Bangladeshi residents during COVID-19
pandemic
authors:
- Mili Saha
- Goutam Saha
- Mynul Islam
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021720
doi: 10.1371/journal.pgph.0000407
license: CC BY 4.0
---
# Knowledge, attitude, and practice of Bangladeshi residents during COVID-19 pandemic
## Abstract
Bangladeshi government has adopted some special steps to control the quick spread of the COVID-19 pandemic situation. However, the residents’ knowledge, attitudes, and practices towards the disease directly impact the success of the controlling measures taken by the state. This article explores knowledge (K) about preventions, attitude (A) to the disease, and practices (P) of preventing the COVID-19 infection risks of different age groups residing in Bangladesh. Quantitative data were collected online using a KAP questionnaire from 932 participants. Also, statistical t and F tests have been used and analyzed and p-value, $95\%$ Confidence Interval, Odd Ratio (OR), KAP scores, and multiple logistic regression analysis, are presented in this research. Results show the population is generally aware of the symptoms and social distancing. They are concerned about re-spreading and positive about staying home. The most significant findings of the study reveal that the old age group (age 50 or over) is the most alert group, male population are the most vulnerable with less care, people living outside Dhaka take less care and fewer preventive measures against the deadly virus, the young age group (age 18–25) is most optimistic while the female respondent group is best prepared among all the participants.
## 1. Introduction
Bangladesh detected her first COVID-19 infected patients on 8 March 2020, which began with three in number. However, it has increased to 2456 on day forty-one, which was the sixth week and indicates the fourth phase of infection [1] called ‘Sustained Human-to-Human Transmission’ [2]. The total number of deaths has reached 100 including ten deaths and around 300 new infections every day, which indicates the rapid rise period. The government has locked down the highly affected areas and a general leave has been declared to keep all the private and public schools, businesses, and service farms closed. Despite all these steps taken by the state, COVID-19 is still a rising infectious health issue and should be prevented perceptively by individuals. This is more about personal preclusion than authority care to reduce the infection rate. Knowledge, attitudes, and practices towards infectious diseases usually involve some panic emotions among the population, promoting complications in preventing the disease and spread [3]. Hence, understanding the public awareness level and readiness to combat the outbreak of COVID-19 in *Bangladesh is* crucial during the rapid rise period.
KAP refers to knowledge (what is known), attitude (what is thought), and practices (what is done). A KAP survey of any community about a particular topic serves as an educational diagnosis of that population through examining what people know, what they believe, and how they behave [4–6]. KAP is a very common tool used in health-seeking research [7]. This KAP study explores what Bangladeshi residents know about COVID-19 symptoms and prevention, how they view the socio-cultural effects of the disease, and what practices they use to prevent the infection. Understanding the KAP of the residents might enable the government authorities and other agencies to introduce and implement “a more efficient process of awareness creation programs” which will appropriately and effectively address the interventional needs of the community [8].
We performed this study to better understand the residents’ perceptions and beliefs, which might identify any knowledge gaps, negative attitudes, and factors influencing actions. No studies report about this community’s KAP of COVID-19 and the information essential to decide about the protective steps and select the program priorities. This research could be significantly useful for the local administration to reintegrate more impactful policies and successful measures to control infections and improve safety during and after the lock-down period. Bangladesh is almost at the peak of infection and the people have been continuing to fight against Coronavirus disease. To ensure that they win, “people’s adherence to the control measures are essential, which is largely affected by their knowledge, attitudes, and practices (KAP)” [9, 10].
People are, overall, willing to maintain social distance and quarantine which can slow down the infection, although they have high anxiety about possible infection, which can be reduced through increasing awareness and addressing mental health issues [11]. Most of the participants are young at the age of the Covid-19 studies and ‘younger individuals are more likely to be asymptomatic when infected and could be unaware they are putting others at risk’ [12]. On the other hand, the chances of infection and the severity of illness are much direr with aged people [13]. Hence, hasty lifting of lock-down can promote a secondary peak while lifting lock-down gradually can flatten it [14]. Studying individual awareness can itself motivate people to practice the preventive measures discussed in the research alerting them in turn and know about the practices of a large population to avoid mass contamination.
Inadequate understanding of the common people about the nature and effects of a viral disease can result in improper practices and less adherence to hygiene rules, delayed treatment, rapid spread, and fatal consequences of the infection. Hence, improving the knowledge, awareness, and perceptions of all population is crucial [15]. Studies analyzing the current level of awareness, attitudes, and practices towards COVID 19 among Nepali, Saudi Arabian, Ethiopian, Pakistani, Nigerian, and Indian [16–22] residents emphasize ensuring people’s willingness and active participation to minimize the pandemic effects. Bangladeshi male population, young generations, and rural participants have less knowledge, pessimistic attitudes, and malpractices towards the disease. Whereas, the female, elderly, and urban population have better knowledge and positive practices [23]. Some recent works on COVID-19 using the KAP model are also performed by many researchers [24–26].
## 2. Research design and methodology
The current research was conducted in Bangladesh and an online survey was performed using one of the popular Google tools called Google Forms. The link to the questionnaire was saved for future use. We provided ethical clearance and ensured the participants that the given personal information and opinion will be kept confidential. The research questions include three aspects of the pandemic spreading rapidly in the country:
## Participants
Most of these participants have a minimum educational background which is a Higher Secondary School Certificate level and use any one of the online platforms, such as Facebook, WhatsApp, Gmail. Our goal was to reach a bigger audience including a group of participants to elicit maximum responses.
## Ethics statement
Our survey study is performed following the authentic research guidance and protocol of the University of Dhaka, Dhaka-1000, Bangladesh and Jagannath University, Dhaka-1100, Bangladesh. This research survey is approved by the Chairman, Department of Mathematics, University of Dhaka, who gave us verbal consent to continue the present research. We also assured each of the participants that all the personal or institutional information gathered in this study will be used for academic publication purposes and their identities will never be exposed as it is stated in the ethical protocol. The consent was received from the participants through Google form and only those who agreed participated in this research survey.
## Tools
The survey questionnaire consists of two sections, including demographics and KAP inquiries. At the beginning, social demographic variables, such as age, gender, and place of residence (Dhaka vs. other districts in Bangladesh) were sought. Following the clinical and community management of COVID-19 guidelines by the Institute of Epidemiology, Disease Control and Research (IEDCR), Bangladesh, we prepared a COVID-19 awareness questionnaire including 14 questions (S1 Questionnaire). Five of the questions are about clinical arrangements (K1-K5), three of them are about the possible spreading medium (A1-A3), and six of them involve prevention and control (P1-P6) of the virus. The questions are answered on a three-item Likert scale consisting of yes, no, not sure options. We assigned 3 points for the first option, 2 points for the second option, and 1 for every third. The total knowledge score ranged from 1 to 30, which denotes better knowledge with higher scores. The Cronbach’s alpha coefficient was 0.79 indicating acceptable internal consistency of the KAP questionnaire.
## Data collection
The primary data on the participants’ attitudes for the online cross-sectional study were collected from 17 April 2020 to 20 April 2020. In this research, 565 male and 367 female in a total of 932 participants responded to the online survey. Among all, 574 participants belong to the age group ranging from 18 to 25 years, 349 of them belong to the age group ranging from 26 to 49 years, and only 9 participants are in the age group of 50 years and above. Also, 679 participants identify themselves as single and the remaining 253 of them are married. Besides, 514 respondents live in the Dhaka division, the most infected region in the country, and 418 participants live outside the Dhaka division.
Details of the above-mentioned various data types are also presented in the following: For calculating sample size, the proportion of the population having adequate knowledge about COVID-19 was considered as the indicator variable. The expected sample size was calculated as at least 576 where the z score for $95\%$ confidence interval was 1.9, the prevalence of adequate knowledge was assumed 0.5, the margin of error was 0.05 and the design effect was 1.5 for sampling variation. We provided the participants with the link using different online platforms including Facebook, Messenger, Google talk, email, and requested them to share the link with their friends and relatives on any social media so they can respond as well. In addition, we requested the participants’ demographic information comprising gender identity, marital status, and place of residence. Therefore, participants were included in the study through social network communities using snowball sampling. All the data were available in the following link: https://docs.google.com/spreadsheets/d/1Qndd80zdSltt_9W1x9uKsieUCASreqyYh8yFEfpjYkk/edit?usp=sharing.
## Data analysis
Later the data have been examined and categorized by the percentage of the agreement, disagreement, and personal preferences in each question. Also, a statistical analysis using statistical software SPSS with a $5\%$ significance level was done and a two-sided T-test are considered for statistical analysis of the data. In addition, statistical F test is also used and p-value, $95\%$ Confidence Interval (CI), Odd Ratio (OR), KAP scores, and multiple logistic regression analysis are also presented.
## 3.1 KAP scores
The participants’ knowledge, attitude, and practice score are shown using descriptive statistics in Table 1. The KAP scores are converted to a quantitative scale and samples are independent across the demographic variables. An independent sample t-test and ANOVA have been applied to understand whether the average participants’ knowledge, attitude, and practice towards COVID 19 scores significantly vary across gender, location, and age group. The normality assumption for implementing ANOVA has been checked through a q-q plot for each group. Thus, the comparison of knowledge, attitude, and practice score for demographic variables such as gender, location, and age group are shown in Table 2. The average knowledge score found for the participants was 11.15 (Table 1). The average knowledge scores of the male participants are higher than the female participants average. However, this difference is not statistically significant (p-value > 0.05). The average attitude towards the COVID-19 situation score was 6.43 (Table 1). Participants’ attitude scores range between 3 to 9 with SD 1.34. The analysis also demonstrates that the average attitude score does not vary significantly across the gender, location, and age group of the participants. This result also coincides with the results obtained in Maheshwari et al. [ 20]. The average practice score was 6.43 (Table 1). Participants’ practice scores range between 7 to 17 with SD 1.75. Participants’ practice scores vary significantly across gender, location, and age group of the participants. Male participants have a greater tendency to practice the preventions compared to the female ones for the safety against possible COVID-19 infections. And, the differences among the practice scores of the male and female participants are statistically significant at a $1\%$ level. The Covid-19 prevention practicing tendency is higher among elder age groups. Besides, the difference in practice scores among different age groups is statistically important at a $1\%$ level of significance. Observation shows Dhaka divisional participants have slightly higher practice scores than the participants residing outside this division with considerable differences at a $5\%$ level of significance. The majority of the participants ($$n = 652$$, $69.7\%$) have adequate knowledge. Following the independent sample t-test and ANOVA, it can be stated that the average knowledge score does not vary considerably across the location and age group of the participants at a $5\%$ level of significance. Similar kinds of results are found by Maheshwari et al. [ 20].
## Knowledge
K1: In response to the first survey question, $78.76\%$ of respondents report complete awareness about the varied symptoms of COVID-19 infection and common flu while $14.80\%$ are still unsure. And, $6.44\%$ are not aware of the differences at all, which can be either fatal or stressful for them. The findings reveal that female respondents and old age groups (50 years and above), and the population living in greater Dhaka are more aware of the differences between the symptoms of COVID-19 and other common flues than their counterparts who are most confused about the COVID-19 symptoms too.
K2: Also, the majority of people ($59.76\%$) are unsure about any proposed or available medicines that can be effective to treat the infected persons. And, $24.36\%$ of them do not rely on any medicine for treating the disease while $15.88\%$ believe if they are infected, these medicines can help. This confirms the other results demonstrating a particular number of people perceive the severity through media focus, which might be wrong in reality. More male respondents, young age groups (18–25), and people residing outside Dhaka believe in the medicines presented in various media than the female, elder, and Dhaka resident participants while the most hesitant population is the old aged group.
K3: A great majority of the respondents ($77.68\%$) maintain social distancing by being 1.5 meters away from other people while $8.91\%$ of them keep less distance and the rest $13.41\%$ are unsure about the practice, which defines the overall careless movement of the community. Again, the female, elderly, and non-Dhaka residents groups are found to maintain social distance more than their male, young, and Dhaka-residents counterparts who are most confused about maintaining the required social distance to avoid infection.
K4: Findings suggest that a considerable major group of $64.48\%$ of respondents believe getting closer to the infected persons is the worst potential source of COVID-19 transmission in Bangladesh while secondary contact through infected person’s used items is the next possible risk of contamination. And, a very small group of $5.48\%$ of participants are concerned with air transmission, which indicates this group’s extreme awareness about the virus’s existence in the environment. Also, it is found that 18–25 years and 26–49 years old age groups have comparable beliefs about the local sources of contamination, although the female participants tend to believe more in air transmission and infection through touching things used by infected persons than the male counterparts. The old age group completely ignores the possibility of air transmission.
K5: A smaller majority of the respondents ($42.51\%$) prefer to take all preventive measures than testing ($54.81\%$) if they discover COVID-19 symptoms in them. The worst risk is still $2.67\%$ of respondents plan to hide the infection news from others, which can be detrimental for their surroundings. Both male and female groups are equally interested in trying preventive measures, testing, and hiding about infections. The old age group is most interested in preventive measures; non-Dhaka residents are mostly aware of testing; Dhaka-residents are prone to hide infections most, although the old age group is not at all interested in hiding.
A summary of the percentage results for the knowledge section is presented in Table 3.
**Table 3**
| Type | Respondent Types | A (%) | B (%) | C (%) |
| --- | --- | --- | --- | --- |
| | Gender | | | |
| K1 | Male | 75.22 | 9.03 | 15.75 |
| K2 | Male | 19.47 | 24.25 | 56.28 |
| K3 | Male | 74.87 | 10.27 | 14.87 |
| K4 | Male | 4.25 | 66.90 | 28.85 |
| K5 | Male | 42.22 | 54.89 | 2.89 |
| K1 | Female | 84.2 | 2.45 | 13.35 |
| K2 | Female | 10.35 | 24.52 | 65.12 |
| K3 | Female | 82.02 | 6.81 | 11.17 |
| K4 | Female | 7.36 | 60.76 | 31.88 |
| K5 | Female | 43.10 | 54.88 | 2.02 |
| | Age groups | Age groups | Age groups | Age groups |
| K1 | 18–25 | 76.48 | 14.14 | 9.38 |
| K2 | 18–25 | 16.55 | 22.82 | 60.63 |
| K3 | 18–25 | 77.53 | 8.19 | 14.29 |
| K4 | 18–25 | 5.75 | 65.51 | 28.75 |
| K5 | 18–25 | 57.97 | 39.34 | 2.69 |
| K1 | 26–49 | 81.95 | 14.05 | 4.0 |
| K2 | 26–49 | 15.19 | 27.22 | 57.59 |
| K3 | 26–49 | 77.65 | 10.03 | 12.32 |
| K4 | 26–49 | 5.16 | 62.75 | 32.09 |
| K5 | 26–49 | 47.88 | 49.42 | 2.70 |
| K1 | 50 and over | 100 | 0 | 0 |
| K2 | 50 and over | 0 | 11.11 | 88.89 |
| K3 | 50 and over | 88.89 | 11.11 | 0 |
| K4 | 50 and over | 0 | 66.67 | 33.33 |
| K5 | 50 and over | 66.67 | 33.33 | 0.0 |
| | Place of residence | Place of residence | Place of residence | Place of residence |
| K1 | Dhaka division | 82.30 | 5.05 | 12.65 |
| K2 | Dhaka division | 13.82 | 24.90 | 61.28 |
| K3 | Dhaka division | 78.99 | 8.56 | 12.45 |
| K4 | Dhaka division | 6.23 | 63.62 | 30.16 |
| K5 | Dhaka division | 43.98 | 52.58 | 3.44 |
| K1 | Outside Dhaka division | 74.40 | 8.13 | 17.46 |
| K2 | Outside Dhaka division | 18.42 | 23.68 | 57.90 |
| K3 | Outside Dhaka division | 76.08 | 9.33 | 14.59 |
| K4 | Outside Dhaka division | 4.55 | 65.55 | 29.90 |
| K5 | Outside Dhaka division | 40.76 | 57.48 | 1.76 |
## Attitude
A1: The research participants are generally anxious about the negative social attitudes, negligence, and rudeness showed towards the COVID-19 patients in Bangladesh. Data shows $48.06\%$ of the respondents are frustrated, $40.24\%$ of them are afraid and $11.70\%$ are angry to see such insolence and humiliation including abundance, refusal, and rejection to treatment, and avoidance by family and society. The groups have similar reactions to the negative social attitudes, rudeness, and negligence shown to the Corona patients in Bangladesh, although the old age group is more angry and frustrated, who are least scared with this attitude.
A2: A great majority of the respondents ($49.25\%$) are found to be concerned that the virus will re-spread as soon as the lock-down is relaxed. A very minor group ($13.84\%$) of the respondents believe the situation will become normal while $36.91\%$ of them are unsure about the choices. The female participants are more anxious about COVID-19 re-spreading after the lock-down is lifted and are less optimistic about normalizing the situation than the male respondents. Also, the old age group is least concerned about re-spreading, least hopeful about normalizing, and most confused about choosing any of the relevant options.
A3: Frequently disinfecting things and washing hands are the second priority to the participants ($29.08\%$) while keeping a social distance is the highest priority to the $47.42\%$ of respondents, which the World Health Organization emphasizes. Although wearing safety masks and gloves is the least preferred measure after the lock-down period, this is the most common practice at this rapid rise period. However, male participants outperform the female respondent group regarding the post-lock-down awareness. Data shows the male participants group is more likely to wear safety masks and gloves in addition to maintaining social distance while the female respondents are prone to washing hands and disinfecting things more frequently than their male counterparts. Also, young participants groups and non-Dhaka residents outperform the counter old aged and Dhaka-resident groups with more awareness about the upcoming risks.
A summary of the percentage results for the attitude section is presented in Table 4.
**Table 4**
| Type | Respondent Types | A (%) | B (%) | C (%) |
| --- | --- | --- | --- | --- |
| | Gender | | | |
| A1 | Male | 10.97 | 48.85 | 40.18 |
| A2 | Male | 48.67 | 15.22 | 36.11 |
| A3 | Male | 26.73 | 49.38 | 23.89 |
| A1 | Female | 12.81 | 46.87 | 40.33 |
| A2 | Female | 50.14 | 11.72 | 38.15 |
| A3 | Female | 22.89 | 44.41 | 32.70 |
| | Age groups | Age groups | Age groups | Age groups |
| A1 | 18–25 | 11.15 | 46.34 | 42.51 |
| A2 | 18–25 | 46.69 | 15.33 | 37.98 |
| A3 | 18–25 | 23.69 | 50.70 | 25.61 |
| A1 | 26–49 | 12.32 | 50.43 | 37.25 |
| A2 | 26–49 | 53.87 | 11.75 | 34.38 |
| A3 | 26–49 | 23.21 | 42.12 | 34.67 |
| A1 | 50 and over | 22.22 | 66.67 | 11.11 |
| A2 | 50 and over | 33.33 | 0.0 | 66.67 |
| A3 | 50 and over | 22.22 | 44.44 | 33.33 |
| | Place of residence | Place of residence | Place of residence | Place of residence |
| A1 | Dhaka division | 13.04 | 48.05 | 38.91 |
| A2 | Dhaka division | 50.39 | 14.20 | 35.41 |
| A3 | Dhaka division | 23.35 | 43.19 | 33.46 |
| A1 | Outside Dhaka division | 10.05 | 48.09 | 41.87 |
| A2 | Outside Dhaka division | 47.85 | 13.40 | 38.76 |
| A3 | Outside Dhaka division | 23.68 | 52.63 | 23.68 |
## Practice
P1: Despite being remarkably alert about keeping social distance and avoid close contacts with the community members, the majority of the research population ($81.65\%$) are relaxed about eating animal protein including eggs, meat, or fish. Whereas, $7.84\%$ of participants are seriously aware about such transmission through animal bodies and they have been avoiding eggs and milk too. Also, another group consisting of $10.51\%$ respondents avoid eating meats and fish. Therefore, the majority of people are at great risk in case the local animal transmission begins. Surprisingly old age groups are least concerned about consuming animal proteins to prevent COVID-19 infection. Next comes the female participant group who avoid less animal protein than the male population. And, Dhaka residents avoid eating animal protein more than the residents living outside Dhaka. Gender and ages significantly influence the participants’ practices about protecting themselves from COVID-19 infection and treatment of the disease.
P2: Since $52.36\%$ of participants prefer to stay home and $34.23\%$ of the respondents go out once a week, it seems those $13.41\%$ of participants who go for daily shopping are at a greater risk. As usual, male respondents, mid-young age (26–49 years) group, and non-Dhaka resident group go out of the home more frequently than the female, early-young, Dhaka-resident, and old age groups. However, female and old age groups mostly stay home.
P3: Another positive preventative measure that a great majority of the population ($67.17\%$) avails is always disinfecting things and taking a bath right after returning home. However, $27.04\%$ of respondents take this measure less frequently and $5.79\%$ of the population do it seldom, which puts the risk of contaminating the home environment unintentionally and reduces the chance of escaping infection. Female, mid-young age, and Dhaka-residents groups are more careful about taking bath and disinfecting things after coming back from the outside than the counter male, early-young, non-Dhaka residents, and old groups, while the male group is the least and old age group is the most aware regarding the precaution.
P4: The majority of the participants ($51.06\%$) only avoid cold food items and drinks as prevention from the infection, although $44.56\%$ of them are more cautious about drinking warm water to stop the virus living inside the body and $4.38\%$ only take steam as they think it can help them to prevent the virus. It is found that more female, mid-young aged, and non-Dhaka resident participants avoid cold food and drinks. A greater number of male, old-aged and Dhaka-resident participants drink warm water. More male, old-aged and non-Dhaka resident respondents take steam to prevent COVID-19 infection than their counter groups. Although female groups are quite aware of avoiding cold foods and drinks, they are the least responsive group to drinking warm water and taking steam.
P5: The critical aspect of this research findings include that a considerable number of participants ($43.13\%$) rely on home care and nearly an equal $41.63\%$ of those who prefer to shift to hospital initially after being detected as COVID-19 positive. This indicated the local populations’ reliance on the state policies and physicians’ availability for treating COVID-19 patients. However, $15.24\%$ of the population is indecisive regarding the matter, making it chaos and delaying the cure of the infected individuals. The mid young age groups are most prone to stay home when they are infected while early young age groups prefer shifting to the hospital for Covid-19 treatment. And, old age groups are the most indecisive which is risky too. Also, Dhaka residents rely more on home care and the people residing outside Dhaka prefer to move to the hospital than relying on homecare. However, gender shows no considerable correlation to choosing COVID-19 treatment facilities among the respondents.
P6: Regarding social services, such as raising social awareness and community care, $77.86\%$ of participants mainly use the online platform and social media to ensure social distancing while $15.16\%$ try to make people aware through practicing the health rules, and $6.98\%$ report about demonstrating the safety tasks to instruct people how to save from infecting. Responses regarding raising mass awareness during the pandemic period demonstrate the groups’ almost equal share of work using any particular mode of demonstration.
A summary of the percentage results for the practice section is presented in Table 5.
**Table 5**
| Type | Respondent Types | A (%) | B (%) | C (%) |
| --- | --- | --- | --- | --- |
| | Gender | Gender | Gender | Gender |
| P1 | Male | 10.44 | 11.33 | 78.23 |
| P2 | Male | 19.82 | 43.54 | 36.64 |
| P3 | Male | 57.70 | 34.51 | 7.79 |
| P4 | Male | 49 | 45.7 | 5.3 |
| P5 | Male | 42.65 | 41.59 | 15.75 |
| P6 | Male | 14.39 | 7.19 | 78.42 |
| P1 | Female | 3.81 | 9.26 | 86.92 |
| P2 | Female | 3.54 | 19.89 | 76.57 |
| P3 | Female | 81.74 | 15.53 | 2.72 |
| P4 | Female | 54.15 | 42.86 | 2.99 |
| P5 | Female | 43.87 | 41.69 | 14.44 |
| P6 | Female | 16.34 | 6.65 | 77.0 |
| | Age groups | Age groups | Age groups | Age groups |
| P1 | 18–25 | 8.36 | 10.80 | 80.84 |
| P2 | 18–25 | 11.50 | 31.01 | 57.49 |
| P3 | 18–25 | 64.11 | 29.44 | 6.45 |
| P4 | 18–25 | 50.93 | 44.93 | 4.14 |
| P5 | 18–25 | 31.71 | 51.39 | 16.90 |
| P6 | 18–25 | 13.36 | 6.15 | 80.49 |
| P1 | 26–49 | 7.16 | 10.32 | 82.52 |
| P2 | 26–49 | 16.62 | 39.83 | 43.55 |
| P3 | 26–49 | 72.49 | 22.64 | 4.87 |
| P4 | 26–49 | 52.45 | 43.02 | 4.53 |
| P5 | 26–49 | 62.18 | 26.07 | 11.75 |
| P6 | 26–49 | 17.94 | 8.53 | 73.53 |
| P1 | 50 and over | 0 | 0 | 100 |
| P2 | 50 and over | 11.11 | 22.22 | 66.67 |
| P3 | 50 and over | 55.56 | 44.44 | 0 |
| P4 | 50 and over | 0 | 83.33 | 16.67 |
| P5 | 50 and over | 33.33 | 22.22 | 44.44 |
| P6 | 50 and over | 25.0 | 0.0 | 75.0 |
| | Place of residence | Place of residence | Place of residence | Place of residence |
| P1 | Dhaka division | 8.37 | 10.51 | 81.13 |
| P2 | Dhaka division | 9.34 | 35.21 | 55.45 |
| P3 | Dhaka division | 74.51 | 20.82 | 4.67 |
| P4 | Dhaka division | 48.29 | 47.80 | 3.9 |
| P5 | Dhaka division | 46.50 | 37.35 | 16.15 |
| P6 | Dhaka division | 15.08 | 6.35 | 78.57 |
| P1 | Outside Dhaka division | 7.18 | 10.52 | 82.30 |
| P2 | Outside Dhaka division | 18.42 | 33.01 | 48.57 |
| P3 | Outside Dhaka division | 58.13 | 34.69 | 7.18 |
| P4 | Outside Dhaka division | 54.36 | 40.70 | 4.94 |
| P5 | Outside Dhaka division | 39.0 | 46.89 | 14.11 |
| P6 | Outside Dhaka division | 15.25 | 7.75 | 77.0 |
The unusual situations and unexpected circumstances created by the coronavirus pandemic have undoubtedly changed the livelihoods, attitudes, and priorities at the local and global levels. Bangladeshi residents are no exceptions. Since the coronavirus outbreak, people have been loaded with massive information about different aspects of the virus infections, affects, and cure during the last few months. Some might be exaggerated and unnecessary, while some other essential information might be missing. These include symptoms, precautions, preventions, infection types, treatment, recovery, risks, deaths, etc. Although a vast majority of the local and global population are well informed about the essential aspects, the high-risk population including youth groups, males, and marginalized communities need to be educated and aware [21]. Hence, more awareness and preparation programs should be introduced regarding COVID-19 for the people residing in remote and underdeveloped areas where mass communication through online technology is limited. People need to continue to strengthen KAP towards COVID-19 to win the current and future battles against the disease. Policymakers need to put more emphasis on informing and awakening the less educated, low-income, and male young [17]. Besides, the government needs to design educational sessions for less knowledgeable people to enhance their knowledge [19].
Generally, as the findings show, educated people are supportive of the adopted measures to combat and control COVID-19 spread. Although, low-income people have no way out to maintain all such measures. Almost $80\%$ of citizens understand and enforce the necessity of undergoing complete isolation, disinfection, and prevention. However, people including day laborers and front liners might expect the authorities to provide protective masks and other staff for all citizens, and local governments have scopes to consider such an initiative. As the current studies expose, 10–$20\%$ population is a bit careless and aware of the symptoms, social distancing, outing, and protections in the different parts of the world, especially the young man and peripheral groups have some resistances to extreme caution. Although the possibility of individual contraction has not been reduced, some citizens have a low level of concern. Many adult men do not properly maintain cleaning and disinfecting but move around the need to be aware of maintaining protection. They adopt far fewer safeguard measures practically than women, so these populations should communicate further regarding COVID-19 risk management. Young adults need effective health education campaigns enhancing and encouraging knowledge, a positive mindset, and essential preventions of COVID-19 even more [27].
Overall, age is significantly correlated to inadequate knowledge, inappropriate practices, and poor perceptions about COVID-19 in the United Arab Emirates [15]. Especially, male and youth groups need to be empowered through acknowledging their preventive practices against COVID-19 and the sense of responsibilities in Bangladesh [23]. However, the current findings contradict Hayat’s [18] conclusion that increasing age is associated with poor knowledge and hence, poor practices. Also, gender has noticeable effects on positive attitudes and good practices, such as avoiding crowded places in Saudi Arabia, Pakistan, and India, and the Philippines [17, 19–21]. Although female participants are also more aware regarding many aspects, including cleaning, disinfecting, and further outbreak, they are less active in self-care and rarely avoid animal meat or fish which might be unsafe. Rural residency, low educational status, and poor income significantly correlate to insufficient knowledge and bad practices in Pakistan and India [19, 20]. The youth age group is more optimistic while the female respondent group is prepared most of all participants.
## 3.3 Regression analysis
Multiple logistic regressions are performed and shown in Table 6 to find out the adjusted effect of the potential risk factors on adequate knowledge, positive attitude, and good practice on COVID-19. To perform logistic regression, mean KAP score is considered as cutoff score to define sufficient knowledge, positive attitudes, and good practices. In multiple logistic regressions, the covariates are taken on the basis of having a significant association or relation with the outcome variables. According to the independent sample t-test, significant relations between gender, location, and age group with knowledge, attitude, and practice score have been found. Thus, these variables are included in logistic regression analysis. The odds of having adequate knowledge do not vary significantly across the gender and location of the participants, as shown in Table 2. Moreover, Table 2 shows the participants of the mid-age group 26–49 are more likely (1.41 times) of possessing adequate knowledge than the participants of the young age group 18–25. The odds of having adequate knowledge for the participants of age 50 or above is 3 times higher than the participants of age group 18–25, although this difference is not statistically significant. A high percentage of the participants ($$n = 690$$, $73.7\%$) showed positive attitudes towards COVID-19 disease. Multiple logistic regressions reveal that there is no association between positive attitudes and participant characteristics variables such as gender, age group, and location (Table 6). A large proportion of the participants ($$n = 646$$, $69\%$) had good practice towards COVID-19. A significant difference between male and female participants in terms of good practice at a $5\%$ significance level. Male participants are $43\%$ more likely to practice prevention than the female participant’s group. Participants with age group 26–49 have a 1.50 times higher chance of performing good practices towards COVID-19 than the participants with age group 18–25. The odds of doing good practices for the participants of age 50 or above is 3.48 times higher than the participants of age group 18–25, although this difference is not statistically significant. Participants in the Dhaka division are more intended to do good practice towards COVID-19 than that of the participants outside the Dhaka division. Therefore, it is recommended that women, young age groups, and people residing outside Dhaka should pay more attention to practice prevention against COVID-19.
**Table 6**
| Variables | Sufficient Knowledge | Sufficient Knowledge.1 | Positive Attitudes | Positive Attitudes.1 | Good Practices | Good Practices.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Variables | n (%) | OR adjusted (95% CI) | n (%) | OR adjusted (95% CI) | n (%) | OR adjusted (95% CI) |
| Gender | | | | | | |
| Woman | 250 (68.1) | Ref. | 269 (73.3) | Ref. | 242 (65.9) | Ref. |
| Man | 420 (71.2) | 1.19 (0.89–1.59) | 421 (74.5) | 1.05 (0.78–1.42) | 404 (71.5) | 1.43* (1.06–1.90) |
| Age-group | | | | | | |
| 18–25 | 385 (67.1) | Ref. | 428 (74.6) | Ref. | 377 (65.7) | Ref. |
| 26–49 | 259 (74.2) | 1.41* (1.04–1.90) | 255 (73.1) | 0.93 (0.69–1.27) | 261 (74.8) | 1.50** (1.11–2.02) |
| 50+ | 8 (88.9) | 3.74 (0.46–30.25) | 7 (77.8) | 1.22 (0.25–5.98) | 8 (88.9) | 3.48 (0.43–28.19) |
| Location | | | | | | |
| Outside Dhaka | 286 (68.4) | Ref. | 313 (74.9) | Ref. | 268 (64.1) | Ref. |
| Dhaka | 366 (71.2) | 1.11 (0.84–1.49) | 377 (73.3) | 0.94 (0.69–1.26) | 378 (73.5) | 1.56** (1.17–2.08) |
## 4. Conclusion
Findings show Bangladeshi women and old age groups are generally socially and psychologically connected, although physically distanced. And thus, they are preventing the risks of infection. However, young age groups are comparatively less aware of the symptoms, social distancing, and frequent outings, which indicates the greater risks involved with the male population and those living in divisions other than Dhaka. The most significant findings of the study reveal that the old age group is the most alert group, male respondents are the most vulnerable with less care. People living outside Dhaka have less knowledge and fewer preventive measures against the deadly virus. The young age group is more optimistic while the woman respondent group is prepared for most of all participants. Hence, more awareness and preparation programs should be introduced regarding COVID-19 for the people residing in remote and peripheral areas where mass communication through online technology is limited. The overall findings demonstrate that 16–$20\%$ of respondents comprising mainly the young man groups are at high risk of infection. Also, the analysis of the survey results confirms a correlation between the knowledge and practice of COVID-19 protection in Bangladesh.
## Recommendations
Recommendations are presented in the following:
## Limitation of the study
Some limitations of this research are presented below:
## References
1. Moral S.. **Bangladesh at the 4th stage of coronavirus infection**. *Daily Prothom Alo* (2020.0)
2. 2World Health Organization. Current WHO phase of pandemic alert for Pandemic (H1N1). 2009. Available from: https://www.who.int/csr/disease/swineflu/phase/en/.. *Current WHO phase of pandemic alert for Pandemic (H1N1)* (2009.0)
3. Zhong BL, Luo W, Li HM, Zhang QQ, Liu XG, Li WT. **Knowledge, attitudes, and practices towards COVID-19 among Chinese residents during the rapid rise period of the COVID-19 outbreak: a quick online cross-sectional survey**. *International journal of biological sciences* (2020.0) **16** 1745-1752. DOI: 10.7150/ijbs.45221
4. Good B. *Medicine, rationality, and experience: an anthropological perspective* (1994.0)
5. Ribeaux P, Poppleton SE. *Psychology and Work: an introduction* (1978.0)
6. Tannahill A.. **Beyond evidence-to ethics: a decision-making framework for health promotion, public health, and health improvement**. *Health Promot. Int* (2008.0) **23** 380-390. DOI: 10.1093/heapro/dan032
7. 7World Health Organization. Advocacy, communication and social mobilization for TB control: a guide to developing knowledge, attitude and practice surveys. 2008. Available form: http://whqlibdoc.who.int/publications/2008/9789241596176_eng.pdf.. *Advocacy, communication and social mobilization for TB control: a guide to developing knowledge, attitude and practice surveys* (2008.0)
8. Kaliyaperumal K.. **Diabetic Retinopathy Project Guideline for Conducting a Knowledge, Attitude, and Practice (KAP) Study**. *Community Ophthalmology* (2004.0) **4** 7-9
9. Ajilore K, Atakiti I, Onyenankey K. **College students’ knowledge, attitudes, and adherence to public service announcements on Ebola in Nigeria: Suggestions for improving future Ebola prevention education programs**. *Health Education Journal* (2017.0) **76** 648-660. DOI: 10.1177/0017896917710969
10. Tachfouti N, Slama K, Berraho M, Nejjari C. **The impact of knowledge and attitudes on adherence to tuberculosis treatment: a case-control study in a Moroccan region**. *Pan. Afr. Med. J.* (2012.0) **12** 52. PMID: 22937192
11. Roy D, Tripathy S, Kara SK, Sharmaa N, Vermaa SK, Kaushalb V. **Study of knowledge, attitude, anxiety & perceived mental healthcare need in Indian population during COVID-19 pandemic**. *Asian Journal of Psychiatry* (2020.0) **51** 102083. DOI: 10.1016/j.ajp.2020.102083
12. Dong Y, Mo X, Hu Y, Xin Q, Jiang F, Jiang Z. **Epidemiological characteristics of 2143 pediatric patients with 2019 coronavirus disease in China**. *Pediatrics* (2020.0) **58** 712-713. DOI: 10.1542/peds.2020-0702
13. Zhou F, Yu T, Du R, Fan G, Liu Y, Liu Z. **Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study**. *Lancet* (2020.0) **395** 1054-1062. DOI: 10.1016/S0140-6736(20)30566-3
14. Prem K, Liu Y, Russell TW, Kucharski AJ, Eggo RM, Davies N. **The effect of control strategies to reduce social mixing on outcomes of the COVID-19 epidemic in Wuhan, China: a modeling study**. *Lancet Public Health* (2020.0) **20** 30073-6. DOI: 10.1016/S2468-2667
15. Bhagavathula AS, Aldhaleei WA, Rahmani J, Mahabad MA, Bandari DK. **Knowledge and Perceptions of COVID-19 Among Health Care Workers: Cross-Sectional Study**. *JMIR Public Health Surveill* (2020.0) **6** 1-9. DOI: 10.2196/19160
16. Asraf H, Garima T, Singh BM, Ram R, Tripti RP. **Knowledge, attitudes, and practices towards COVID-19 among Nepalese Residents: A quick online cross-sectional survey**. *Asian Journal of Medical Sciences* (2020.0) **11** 6-11. DOI: 10.3126/ajms.v11i3.28485
17. Akalu Y, Ayelign B, Molla MD. **Knowledge, Attitude and Practice Towards COVID-19 Among Chronic Disease Patients at Addis Zemen Hospital. Northwest Ethiopia**. *Infection and Drug Resistance* (2020.0) **13** 1949-1960. DOI: 10.2147/IDR.S258736
18. Hayat K, Rosenthal M, Xu S, Arshed M, Li P, Zhai P. **View of Pakistani Residents toward Coronavirus Disease (COVID-19) during a Rapid Outbreak: A Rapid Online Survey**. *International Journal of Environmental Research and Public Health* (2020.0) **17** 1-10. DOI: 10.3390/ijerph17103347
19. Isah MB, Abdulsalam M, Bello A, Ibrahim MI, Usman A, Nasir A. **Coronavirus Disease 2019 (COVID-19): Knowledge, attitudes, practices (KAP) and misconceptions in the general population of Katsina State, Nigeria**. *medRxiv preprint* (2020.0). DOI: 10.1101/2020.06.11.20127936
20. Maheshwari S, Gupta PK, Sinha R, Rawat P. **Knowledge, attitude, and practice towards coronavirus disease 2019 (COVID-19) among medical students: A cross-sectional study**. *Journal of Acute Disease* (2020.0) **9** 100-104. DOI: 10.4103/2221-6189.283886
21. Lau LL, Hung N, Go DJ, Ferma J, Choi M, Dodd W. **Knowledge, attitudes, and practices of COVID-19 among income-poor households in the Philippines: A cross-sectional study**. *Journal of Global Health* (2020.0) **10** 1-11. DOI: 10.7189/jogh.10.011007
22. Zhang M, Zhou M, Tang F, Wang Y, Nie H, Zhang L. **Knowledge, attitude, and practice regarding COVID-19 among healthcare workers in Henan, China**. *Journal of Hospital Infection* (2020.0) **105** 183-187. DOI: 10.1016/j.jhin.2020.04.012
23. Banik R, Rahman M, Sikder T, Rahman QM, Pranta MUR. **Investigating knowledge, attitudes, and practices related to COVID-19 outbreak among Bangladeshi young adults: A web-based cross-sectional analysis**. *Journal of Public Health* (2021.0). DOI: 10.21203/rs.3.rs-37946/v1
24. Bates BR, Botero AV, Grijalva MJ. **Knowledge, attitudes, and practices towards COVID-19 among Colombians during the outbreak: an online cross-sectional survey**. *Journal of Communication in Healthcare* (2020.0) **13** 262-270. DOI: 10.1080/17538068.2020.1842843
25. Yousaf MA, Noreen M, Saleem T, Yousaf I. **A Cross-Sectional Survey of Knowledge, Attitude, and Practices (KAP) Toward Pandemic COVID-19 Among the General Population of Jammu and Kashmir, India**. *Social Work in Public Health* (2020.0) **3** 569-578. DOI: 10.1080/19371918.2020.1806983
26. Jiang M, Feng L, Wang W, Gong Y, Ming W-K, Hayat K. **Knowledge, attitudes, and practices towards influenza among Chinese adults during the epidemic of COVID-19: A cross-sectional online survey**. *Human Vaccines & Immunotherapeutics* (2020.0) **17** 1412-1419. DOI: 10.1080/21645515.2020.1812312
27. Al-Hanawi MK, Angawi K, Alshareef N, Qattan A, Helmy HZ, Abudawood Y. **Knowledge, Attitude and Practice toward COVID-19 among the public in the Kingdom of Saudi Arabia: A cross-sectional study**. *Frontiers in public health* (2020.0) **8** 217. DOI: 10.3389/fpubh.2020.00217
28. 28Rapid Assessment KAP (Knowledge Attitude Practice)—Covid 19 Response, Public opinion polling in Bosnia and Herzegovina, Sarajevo. 2020.
|
---
title: Changes in apparent consumption of staple food in Mexico associated with the
gradual implementation of the NAFTA
authors:
- Néstor A. Sánchez-Ortiz
- Mishel Unar-Munguía
- Sergio Bautista-Arredondo
- Teresa Shamah-Levy
- M. Arantxa Colchero
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021749
doi: 10.1371/journal.pgph.0001144
license: CC BY 4.0
---
# Changes in apparent consumption of staple food in Mexico associated with the gradual implementation of the NAFTA
## Abstract
In 1994, the United States, Canada, and Mexico signed the North American Free Trade Agreement (NAFTA) to remove trade barriers and facilitate cross-border trade in goods and services. Worldwide, trade agreements, urbanization and economic development have shaped significant changes in dietary habits. This study aims to evaluate the association between the gradual implementation of NAFTA and changes in apparent consumption of staple foods in Mexico. We analyzed national apparent consumption of animal- and vegetable-source foods, using data from the Food and Agriculture Organization of the United Nations (FAO) from 1970 to 2018. Association between NAFTA and apparent consumption was estimated using interrupted time series analysis (ITSA) with synthetic controls and included two inflection points based on the implementation of NAFTA: 1994, when the agreement began, and 2008 when it was fully implemented. As a result, comparing Mexico with the synthetic control, we found a significant decrease in apparent consumption of pulses, -3.22 and -1.92 kcal/capita/day in the post-1994 and post-2008 periods, respectively. The vegetable-source foods showed an increase of 5.79 kcal/capita/day after 2008. The trends of apparent consumption of animal-source foods, eggs, and milk had significant increases after 1994 and 2008. The apparent consumption of meat increased only after 2008. The implementation of NAFTA was associated with an increase in apparent consumption of food from animal-source and a decrease in consumption of pulses. After 2008, an increase in apparent consumption of vegetable-source foods was observed.
## Introduction
Globally, urbanization and economic development have been accompanied by significant changes in dietary patterns [1]. Traditional diets, such as the Mexican diet, based on a limited number of staple foods, have changed to an increased proportion of animal-source foods (meat, milk, and eggs), vegetable oils, and sugars. This process is known as "the nutrition transition” [2–4], and it has been associated to changes in employment, urbanization; macroeconomic causes such as economic reforms; and the introduction of new technologies (for food production and processing) [4–6]. Among the most important economic causes of the "nutrition transition” is the global integration of markets through free trade agreements, which remove all or some trade barriers [7,8]. Trade liberalization affects the availability and the retail prices by lowering or eliminating tariffs of imports, facilitating foreign investment in the food industry, processing, retailing and advertising, and encouraging the growth of transnational food companies [9]. Trade agreements promote monoculture agriculture (e.g., sugar, corn, soybean) to produce ultra-processed products [10,11]. These changes affect local food consumption patterns [1,12,13].
Removing trade barriers and increasing investment incentives facilitate transnational food corporations to buy small companies, products, and services across borders [4]. The result is that a single company takes over the production, distribution, and sales of a particular food, promoting changes in the food system that encourage the consumption of processed foods [14].
Some studies have analyzed the relationship between free trade agreements and changes in food consumption. In Thailand, the integration of the Association of Southeast Asian Nations (ASEAN) Free Trade Area has been associated with an increase in snack food consumption [1]. In Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua, the implementation of the Central America-Dominican Republic Free Trade Agreement (CAFTA-DR) was followed by an increase in the consumption of meat and processed foods [12,15]. In Vietnam and Peru, the implementation of regional trade agreements yielded increased consumption of sweetened beverages [16,17].
The North American Free Trade Agreement (NAFTA) was signed in 1994, which aimed to remove trade barriers and facilitate cross-border trade in goods and services between the United States, Canada, and Mexico, over the following 15-years [18,19]. NAFTA successfully increased investment opportunities in the territories of the parties involved [19]. In the food industry alone, investment in Mexico from the United States increased from $2.3 to $8.7 billion between 1993 and 2007 [18]. However, during the NAFTA implementation period, Mexico experienced an increase in type two diabetes [20], overweight and obesity [19].
The World Trade Organization claims that health protection is a significant concern, and therefore trade agreements must seek a balance between "trade and health" [21]. There is evidence that NAFTA has promoted unhealthy foods such as processed meat or sweeteners [22–24], but to our knowledge, the relationship between the gradual implementation of NAFTA and changes in apparent consumption of other staple foods has not yet been explored. To isolate the potential effect of NAFTA, we evaluated changes in apparent consumption of staple foods divided by animal- and vegetable-source food between Mexico and a synthetic control (constructed with information from a pool of lower-middle and upper-middle income countries) before and after two inflection periods: 1994, when the agreement came into force, and 2008, when NAFTA was fully implemented.
## Data sources
We analyzed apparent consumption data from the Food Balance Sheets (FBS) of the Food and Agriculture Organization of the United Nations (FAO), for Mexico and a group of 55 lower-middle and upper-middle-income countries between 1970 and 2018. Apparent consumption is defined as the average supply of food and nutrients, expressed in grams or calories, by individual or grouped food items, available for consumption, and divided by the country’s population in the middle of a given year [25]. The FBS estimates apparent national consumption of the main staple food groups and used nutritional factors to convert kilograms of staple food groups into calories [26]. We divided annual calories by 365 days to determine daily calories by food group.
For this study, available information on eighteen food groups, were classified into two major categories: animal and vegetable, depending on their origin. Animal foods includes meat, offal, animal fats, eggs, milk, fish, and seafood. We also defined the group of animal-source food products as the sum of meat, offal, eggs, and milk. Vegetable foods includes starchy roots, cereals, sugar plants, pulses, tree nuts, vegetable oils, vegetables, and fruits. In this group, we included two additional food categories: vegetable-source food, which encompasses all vegetable-origin foods, and fruits and vegetables, which include only pulses, vegetables, and fruits.
## Interrupted time series analysis with synthetic controls
In experimental designs where treatment and control groups are randomly assigned, comparability between them at baseline in basic characteristics is guaranteed. In the absence of such a design, we rely on statistical tests to check for comparability. When the baseline magnitude and trends in the outcomes of interest (kcal/capita/day for each food group) between a control and treatment group are not statistically different in the pre-intervention period, we expect that in the absence of a shock, such as NAFTA, the parallel trends would continue. Therefore, the difference in the outcomes between control and treatment groups after implementation can be attributed to the intervention. If the trends are not comparable, for instance, if the pre-intervention trend is steeper for the treatment group, the post-intervention difference would be upward biased.
We used interrupted time series analysis (ITSA) with synthetic controls to estimate the association between NAFTA and changes in apparent staple food consumption in Mexico [27–29]. ITSA is a method for assessing the effectiveness of large-scale interventions at the aggregate level [27] when there is no control group from an experimental design and the intervention is expected to change the trend of outcome variables [30]. The ITSA assesses the comparability of the study groups with two parameters: the intercept and the average change for the pre-intervention period. The addition of a synthetic control group is roughly equivalent to including treated and nontreated units, that were similar in observed characteristics prior to the intervention. Therefore, the addition of a control group strengthens the ITSA design [27].
The synthetic controls method predicts outcomes from a hypothetical control group based on a weighted average of preselected covariates from a group of nontreated units (donor pool) [31]. We included a list of lower-middle and upper-middle-income countries for which complete information on all variables was available. We excluded all countries that had bilateral trade agreements with the United States or Canada during the period analyzed. We obtained this information from World Bank databases [32].
The weights given to each donor unit in the construction of synthetic control are determined by the similarity between the observed characteristics of the countries in the donor pool and Mexico. Observations from countries that match better receive a higher weight, with values between 0 and 1 [33]. To estimate the effect of NAFTA, the synthetic control model predicts trends in apparent consumption of staple foods in a synthetic country that is comparable to Mexico but does not implement NAFTA. We chose the following variables to predict staple food consumption: a) apparent food consumption (in calories/capita/day) and supply of fat and protein expressed in gr/capita/day provided by the modeled food group; b) gross domestic product, life expectancy at birth, Co2 emissions per capita (as a proxy for industrial and other economic activities), the proportion of the rural population, and primary school enrollment. School enrollment is a rate for the total number of students enrolled in primary education, calculated as a percentage of the total population of official primary school age. This indicator may be higher than $100\%$ due to the inclusion of students older or younger than the official age [32].
## Association between the gradual implementation of NAFTA with changes in apparent consumption of staple foods
Studies have used this method to estimate the effect of NAFTA on the consumption of sweeteners, therefore we adapted the methodology to estimate changes in apparent consumption of staple foods [22,24]. To evaluate the association between the gradual implementation of NAFTA and changes in apparent consumption of staple foods, we modeled three splines: before the NAFTA implementation from 1970 to 1993, from the beginning of the agreement in 1994 to 2008, when it was fully implemented, and after 2008. We estimated an ITSA model as follows: Yt=β0+β1Tt+β2−3Xi,t+β4−5Xi,tTt+β6D+β7DTt+β8−9DXi,t+β10−11DXtTt+Ɛt Where Yt represents the apparent consumption of food groups (outcome variable) at time t, *Tt is* a count variable representing the years from 1970 to 2018, Xi,t are two dummy variables (pre-intervention period = 0 and intervention = 1) for each spline, and D is a dummy variable indicating the intervention or synthetic control group. The coefficient β0 denotes the intercept for the control group, β1 represents the pre-intervention trend for the control group, β2–3 are the change in the outcome level for the control group in each spline after the intervention, and β4–5 shows the change in the outcome trend in each spline for the control group.
A strength of ITSA models with treatment and control units is that we can assess comparability between study groups in the covariates. The parameters β6 and β7 indicate whether the treatment and control units are balanced in the outcome variable’s level and trend during the pre-intervention period. In the context of an experimental design, we would not expect to find differences [27]. It also tests for differences in level between the intervention and control units in each intervention splines points (1994 and 2008), representing the outcome level immediately after the interventions happens (β8–9). Finally, it tests for differences in the trends between the intervention and control groups in the post-interventions period, compared with the pre-intervention period, which is similar to a difference in difference estimation (β10–11). The Ɛt represents the error term.
## ITSA with synthetic control
*We* generated 18 synthetic control units, one for each food group. All controls were balanced with Mexico regarding the variables defined in the methods section (S1 Table shows the weighted values of each donor unit for the synthetic control and S2 Table presents the predictor variables for Mexico and the synthetic control). Table 1 shows the mean values of the outcome and predicting variables for staple food consumption for Mexico and the synthetic control. Once the country selection was completed, fifty-five units were available to form the donor pool used to construct the synthetic control.
**Table 1**
| Unnamed: 0 | Mexico (n = 25 years) | Mexico (n = 25 years).1 | Donor Pool (n = 1350 years) | Donor Pool (n = 1350 years).1 |
| --- | --- | --- | --- | --- |
| Variable | Mean | (SD) | Mean | (SD) |
| Outcomes (Kcal/capita/day) | | | | |
| Vegetable-source food | 2430.21 | (141.94) | 2050.20 | (344.06) |
| Fruits and vegetables | 268.46 | (38.30) | 181.13 | (91.29) |
| Cereals | 1415.50 | (41.23) | 1129.23 | (385.44) |
| Oilseeds | 22.50 | (6.11) | 78.70 | (125.46) |
| Starchy Roots | 23.58 | (2.28) | 191.01 | (243.63) |
| Sugar and Sweeteners | 424.46 | (43.18) | 236.47 | (136.77) |
| Fruits | 94.92 | (9.72) | 95.48 | (73.95) |
| Vegetables | 26.04 | (5.86) | 29.42 | (22.70) |
| Pulses | 147.50 | (35.34) | 55.26 | (45.44) |
| Vegetable Oils | 211.46 | (53.31) | 167.63 | (100.35) |
| Nuts | 4.63 | (0.92) | 5.67 | (10.04) |
| Animal-source food | 385.38 | (51.71) | 255.84 | (173.56) |
| Meat | 194.88 | (34.76) | 138.34 | (118.54) |
| Eggs | 30.21 | (7.99) | 12.96 | (11.24) |
| Milk | 148.63 | (16.95) | 98.47 | (80.44) |
| Fish and seafood | 14.38 | (5.75) | 28.08 | (28.28) |
| Offal | 11.67 | (2.58) | 6.51 | (5.33) |
| Animal Fats | 52.25 | (17.36) | 47.57 | (48.74) |
| Predictors | | | | |
| Food supply (kcal/capita/day) | 2882.13 | (201.78) | 2367.42 | (409.98) |
| Protein supply quantity (g/capita/day) | 76.26 | (6.14) | 61.09 | (14.78) |
| Fat supply quantity (g/capita/day) | 73.60 | (10.48) | 56.46 | (22.48) |
| *Gross domestic product (million dollars) | $169,000.00 | ($113,000.00) | $28,600.00 | ($58,700.00) |
| life expectancy at birth (years) | 67.00 | (3.31) | 60.12 | (8.31) |
| Co2 emissions (metric tons per capita) | 3.46 | (0.65) | 1.56 | (1.91) |
| School enrollment, primary (%) | 114.32 | (5.77) | 95.76 | (23.99) |
| Rural population (%) | 33.40 | (4.20) | 63.00 | (17.98) |
Table 2 shows the coefficients for the intercept and pre-intervention trend to see if Mexico and the control groups are comparable. The ITSA model test for differences in the intercept (difference in the apparent consumption of the staple food group at baseline -1970-) between Mexico and the synthetic control; and, the pre-intervention trend that is the average change per year in the apparent consumption of a staple food before 1994 in Mexico compared to the synthetic control. After assessing comparability, only eight of the eighteen groups showed parallel trends in the pre-intervention period so were kept for subsequent analyzes. Selected groups were: vegetable-source foods, fruits and vegetables, pulses, nuts, animal-source foods, meat, eggs, and milk (Table 2). Fig 1 shows the graphical representation of the observed and model-predicted apparent consumption trends in Mexico and the synthetic control.
**Fig 1:** *Observed and predicted values from the ITSA regression model before and after NAFTA implementation, comparing Mexico and the synthetic control (kcal/capita/day by food group).A) Vegetable-source food; B) Fruits and vegetables; C) Pulses D) Nuts; E) Animal-source food; F) Meat; G) Eggs; H) Milk. The control group was estimated by synthetic controls from a pool of lower-middle and upper-middle-income countries, adjusted for supply of fat and protein (gr/capita/day) provided by the modeled food group, gross domestic product, life expectancy at birth, Co2 emissions per capita, the proportion of the rural population, and primary school enrollment.* TABLE_PLACEHOLDER:Table 2
## Staple food consumption
Table 3 shows NAFTA’s effect estimations on staple food consumption in Mexico. Post-1994 trend shows no statistically significant changes in apparent consumption of vegetable-source food. However, the post-2008 trend shows a significant increase of 5.05 kcal/capita/day ($$p \leq 0.02$$). The synthetic control’s slope coefficient was negative but not statistically significant; the difference between the slopes was 5.79 kcal/capita/day ($$p \leq 0.02$$), representing a significant increase in the apparent consumption of vegetable-source foods in Mexico after NAFTA was fully implemented.
**Table 3**
| Unnamed: 0 | Unnamed: 1 | 1994-initial implementation | 1994-initial implementation.1 | 1994-initial implementation.2 | 2008- full implementation | 2008- full implementation.1 | 2008- full implementation.2 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | Coefficient | p-value | 95% CI | Coefficient | p-value | 95% CI |
| Vegetable-source food | Mexico | -0.19 | 0.880 | [-2.70, 2.31] | 5.05 | 0.020 | [0.84, 9.26] |
| Vegetable-source food | Synthetic control | 0.64 | 0.440 | [-1.01, 2.28] | -0.74 | 0.510 | [-2.95, 1.48] |
| Vegetable-source food | Difference | -0.83 | 0.580 | [-3.82, 2.17] | 5.79 | 0.020 | [1.04, 10.55] |
| Fruits and vegetables | Mexico | 0.71 | 0.270 | [-0.57, 1.99] | 1.58 | 0.160 | [-0.66, 3.82] |
| Fruits and vegetables | Synthetic control | 1.50 | 0.030 | [0.14, 2.86] | 2.03 | 0.000 | [3.71, 7.83] |
| Fruits and vegetables | Difference | -0.79 | 0.400 | [-2.66, 1.08] | -0.45 | 0.710 | [-2.84, 1.95] |
| Pulses | Mexico | -1.70 | 0.020 | [-3.12, -0.28] | -0.80 | 0.140 | [-1.86, 0.26] |
| Pulses | Synthetic control | 1.52 | 0.000 | [0.54, 2.49] | 1.12 | 0.030 | [0.12, 2.12] |
| Pulses | Difference | -3.22 | 0.000 | [-4.94, -1.50] | -1.92 | 0.010 | [-3.38, -0.46] |
| Nuts | Mexico | 1.16 | 0.000 | [0.88, 1.43] | -0.44 | 0.050 | [-0.88, 0.01] |
| Nuts | Synthetic control | 0.14 | 0.010 | [0.04, 0.24] | 0.08 | 0.040 | [0.00, 0.16] |
| Nuts | Difference | 1.01 | 0.000 | [0.72, 1.31] | -0.52 | 0.020 | [-0.97, -0.07] |
| Animal-source food | Mexico | 11.41 | 0.000 | [9.34, 13.48] | 8.01 | 0.000 | [6.30, 9.71] |
| Animal-source food | Synthetic control | 4.90 | 0.000 | [3.19, 6.61] | -1.29 | 0.370 | [-4.11, 1.54] |
| Animal-source food | Difference | 6.51 | 0.000 | [3.82, 9.20] | 9.30 | 0.000 | [6.00, 12.60] |
| Meat | Mexico | 6.81 | 0.000 | [5.73, 7.89] | 6.17 | 0.000 | [4.85, 7.49] |
| Meat | Synthetic control | 5.59 | 0.000 | [4.46, 6.73] | 4.68 | 0.000 | [4.01, 5.34] |
| Meat | Difference | 1.22 | 0.130 | [-0.35, 2.79] | 1.50 | 0.050 | [0.02, 2.97] |
| Eggs | Mexico | 1.87 | 0.000 | [1.53, 2.21] | 1.02 | 0.000 | [0.78, 1.26] |
| Eggs | Synthetic control | 0.92 | 0.000 | [0.67, 1.17] | 0.23 | 0.040 | [0.02, 0.45] |
| Eggs | Difference | 0.95 | 0.000 | [0.53, 1.37] | 0.78 | 0.000 | [0.46, 1.11] |
| Milk | Mexico | 2.64 | 0.000 | [1.85, 3.44] | 0.75 | 0.000 | [0.40, 1.11] |
| Milk | Synthetic control | -1.45 | 0.050 | [-2.92, 0.03] | -1.46 | 0.000 | [-1.91, -1.02] |
| Milk | Difference | 4.09 | 0.000 | (2.42, 5.76) | 2.22 | 0.000 | (1.65, 2.79) |
No changes in trends were seen in the fruit and vegetable groups in Mexico. Regarding the synthetic control, we saw increases of 1.5 kcal/capita/day in 1994 and 2.03 kcal/capita/day in 2008. However, no differences were found between treatment and control groups. Regarding pulses, after 1994, Mexico’s estimated trend was -1.70 while the control’s trend was 1.52 kcal/capita/day ($p \leq 0.05$). The direction of the coefficients remained the same after 2008. However, it was significant only for the control, and the difference was -1.92 kcal/capita/day ($$p \leq 0.01$$), indicating that Mexico experienced a decline in apparent consumption of pulses after the full release of NAFTA.
During the first phase of NAFTA (1994–2007), both Mexico and the synthetic control experienced an increase in apparent consumption of nuts, with values of 1.16 ($$p \leq 0.00$$) and 0.14 ($$p \leq 0.01$$) kcal/capita/day. The difference between trends implies an increase of 1.01 ($$p \leq 0.00$$) kcal/capita/day due to NAFTA. In the second phase after full implementation of NAFTA (2008–2018), the difference in the apparent consumption trends of nuts reflected a significant reduction, with a value of -0.52 kcal/capita/day with $$p \leq 0.02.$$
Trends in apparent consumption of animal-source foods during the post-1994 period show a significant increase in Mexico, with values of 11.40 kcal/capita/day ($$p \leq 0.00$$). The synthetic control showed a significant increase of 4.90 kcal/capita/day ($$p \leq 0.00$$) during the same period. The difference was 6.51 kcal/capita/day ($$p \leq 0.00$$). The difference in the second phase was 9.30 kcal/capita/day ($$p \leq 0.00$$) comparing Mexico with control group, therefore attributable to the NAFTA implementation.
Apparent consumption of meat shows a significant increase in both treatment and control. The post-1994 consumption trends were 6.81 and 5.59 kcal/capita/day ($p \leq 0.05$), and the post-2008 trends were 6.17 and 4.68 kcal/capita/day ($p \leq 0.05$) for Mexico and the synthetic control, respectively. The 1.50 kcal/capita/day ($$p \leq 0.05$$) difference in apparent consumption between study groups was significant in the second phase (2008–2018). There were significant increases in eggs consumption in Mexico; 1.86 and 1.02 kcal/capita/day ($p \leq 0.05$) in the 1994 and 2008 post-intervention periods. However, the synthetic control also increased by 0.92 kcal/capita/day and 0.23 kcal/capita/day ($p \leq 0.05$) in post-1994 and post-2008 periods respectively. Differences between groups was 0.94 and 0.78 kcal/capita/day ($p \leq 0.05$) for both pos intervention periods, higher consumption in Mexico.
Finally, for milk, the trends of apparent consumption in Mexico were positive in the post-1994 and 2008 periods, with values of 2.64 and 0.75 kcal/capita/day ($p \leq 0.05$). On the other hand, the synthetic control reported negative values. As a result, the difference in the trends of apparent consumption reflected values of 4.09 and 2.22 kcal/capita/day ($p \leq 0.05$), representing a significant increase in the apparent consumption of milk in Mexico due to NAFTA.
## Discussion
We estimated changes of apparent staple food consumption (kcal/capita/day) associated with the gradual implementation of NAFTA in Mexico. To approximate the causal effect of NAFTA on the outcome of interest, we compared Mexico with a synthetic control using a multiple group ITSA model.
Our results show that when comparing Mexico’s staple food consumption trends and a synthetic control (not exposed to a trade agreement), there was a significant reduction in the apparent consumption of pulses after initial implementation of NAFTA in 1994 and after full implementation in 2008. The vegetable-source foods showed a significant increase after 2008. Nuts showed a positive difference after 1994, but a negative one after 2008. On the other hand, the apparent consumption trends of animal-source foods, eggs, and milk showed significant increases between 1994 and 2008, and the meat group showed only an increase after 2008.
The relationship between the implementation of NAFTA and the increase in consumption of animal-source food is consistent with evidence from other regional trade agreements. Countries participating in the CAFTA-DR agreement significantly increased meat imports from the US: percentage changes between 2005 and 2016 were 2,$153\%$ for cattle, 1,$135\%$ for pigs, $460\%$ for poultry, and $488\%$ for processed meat products [34]. Between 1990 and 2007, trade liberalizations were also associated with increased imports of livestock products and poultry in Samoa and sheep meat in Fiji [35].
In 2019, an analysis based on the National Household Income and Expenditure Survey (ENIGH for its acronyms in Spanish) examined trends in food consumption based on the degree of processing during the period 1984–2016 [36]. Although this study did not focus on the effects of trade liberalization, the authors adjusted trends by household and macroeconomic variables, and the results show an increase in consumption of food of animal origin and a decrease in consumption of pulses. These results are consistent with the trends we observed.
In the absence of an experimental design, we cannot fully isolate NAFTA’s impact on food consumption from other factors such as economic growth or technological development [9]. However, the inclusion of a synthetic control with balanced observable characteristics such as previous trends, gross domestic product, the proportion of the rural population, or CO2 emissions provides a good approximation.
The data reported by FBS represent only the average of stocks available in a country but are not a direct indicator of individual final food consumption [26]. The data matrix does not consider correction factors for domestic food waste; especially in low- and middle-income countries, domestic food waste can be substantial [37] under this assumption, the amounts of apparent consumption could be overestimated for vegetables and underestimates for pulses and nuts [38]. However, this potential bias is present in both treatment and control, and it does not affect the validity of the estimated differences. Another limitation is that the FBS data does not include information on the type of preparation or the degree of food processing at consumption. Therefore, it is not possible to show associations between food consumption and negative health outcomes (cardiovascular disease, diabetes, or obesity) at individual level.
The results of this study should be interpreted with caution. The ideal strategy to evaluate a program is an experimental design where treatment and control groups are randomly assigned which are not plausible for policies implemented at the national level such as NAFTA. We acknowledge the limitations of using a pool of countries as control groups for comparison. The construction of a reliable synthetic control depends on the quality of the data and finding a good set of donor units can be challenging. In this study, we used data from the World Bank and FAO to create the synthetic control. Although the donor pool was made up of 55 countries, the construction of each synthetic control only considered between 3 and 7 countries that were not statistically different from Mexico in the variables used to construct this pool. In addition, using ITSA models allowed to test basic characteristics of an adequate control group: no statistical difference in the magnitude and trend pre-NAFTA between the synthetic control and Mexico. We acknowledge that even if we excluded countries with no formal trade agreements with the United States or Canada, donor units may not be completely free of any transnational trade.
Despite the limitations mentioned above, FBS estimates of apparent consumption of staple foods provide a reliable overview of the nutrition situation in Mexico [25]. Moreover, the ITSA model with multiple groups and symmetrically distributed observations over time (30 years before and 30 years after the intervention) is considered one of the most robust quasi-experimental designs for assessing the impact of large-scale interventions [30].
## Policy implications
This study provides estimates of the impact of NAFTA on apparent staple food consumption in Mexico. Ours is one of the first studies to examine the relationship between trade liberalization and changes in the trend of staple food consumption using a quasi-experimental design.
NAFTA affected food of animal- and vegetable-source consumption differently. Compared with the synthetic control, apparent consumption of animal-source foods increased, consumption of pulses decreased, while fruits and vegetables remained stable. Animal foods are a rich source of proteins with high biological value, necessary for adequate growth and human development, which is particularly important in low- and middle-income countries [39]. However, high consumption of animal-source food also negatively affects health and the environment (e.g., greenhouse gas emissions and water pollution) [40]. Shifting to a balanced diet with higher protein intake from plant-based foods such as whole grains, pulses, nuts, seeds, fresh fruits, and vegetables represents a significant step forward in improving human health and environmental sustainability [41]. Many of the resources used to produce livestock feed could be used more efficiently by growing plant foods for human consumption [42].
Presenting the results at an aggregate level does not allow us to observe the differential impact of NAFTA on different populations in Mexico (depending on the socioeconomic level or geographic area). Future studies should assess the impact of NAFTA on population groups that are particularly vulnerable to unhealthy dietary changes, such as people of low socioeconomic status [4]. This will help inform the development and evaluation of future strategies to protect population nutrition and health. There is a lack of information on fiscal strategies explicitly aimed at reducing meat consumption. However, it has been observed that specific fiscal policies, such as taxes on saturated fat, can reduce the consumption of highly processed meat products [43]. There is a need to generate evidence on the structural determinants of meat and derivative consumption and the possible strategies to promote an optimal intake in terms of quality and quantity. A key premise of free trade is that the removal of barriers leads to increased welfare and a more efficient allocation of resources [11]. However, there is evidence that trade agreements can generate negative externalities in social, environmental, food and therefore health aspects [21,44,45], so it is important to include policy measures that strengthen local healthy food production and processing, to improve the environment and human wellbeing.
## Conclusion
Our research suggests that the implementation of NAFTA had a differential impact on the apparent consumption of staple foods when comparing Mexico and a synthetic control. NAFTA is associated with an increase in apparent consumption of animal-source food and a decrease in consumption of pulses, while there are no significant differences in consumption of fruits and vegetables. Given the environmental and health impact of high consumption of animal-source foods, it is essential to promote policies at the macro level aimed at creating a healthier food system and, at the individual level, to equip the population with tools to make more informed and sustainable food choices. Although for this study we excluded countries with trade agreements with the US or Canada to create a pool of units as controls not exposed to the program, future studies should compare countries with trade agreements similar to Mexico to analyze if changes in consumption or dietary patterns were similar.
## References
1. Hawkes C.. **Uneven dietary development: Linking the policies and processes of globalization with the nutrition transition, obesity and diet-related chronic diseases**. *Global Health* (2006.0) **2** 1-18. DOI: 10.1186/1744-8603-2-4
2. Popkin BM. **Urbanization, Lifestyle Changes and the Nutrition Transition**. *World development* (1999.0) **27** 1905-16
3. Popkin BM G-LP. **The nutrition transition: worldwide obesity dynamics and their determinants**. *Int J Obes Relat Metab Disord* (2004.0) **28** 2-9. PMID: 14710166
4. Rayner G, Hawkes C, Lang T, Bello W. **Trade liberalization and the diet transition: a public health response**. *Health Promot Int* (2006.0) **21 Suppl 1** 67-74. DOI: 10.1093/heapro/dal053
5. Popkin BM. **The nutrition transition and its health implications in lower-income countries**. *Public Health Nutr* (1998.0) **1** 5-21. DOI: 10.1079/phn19980004
6. Smith RD, Lee K, Drager N. **Trade and health: an agenda for action**. *Lancet* (2009.0) **373** 768-73. DOI: 10.1016/S0140-6736(08)61780-8
7. Barlow P, McKee M, Basu S, Stuckler D. **The health impact of trade and investment agreements: A quantitative systematic review and network co-citation analysis**. *Global Health* (2017.0) **13** 1-9. DOI: 10.1186/s12992-017-0240-x
8. Dür A, Baccini L, Elsig M. **The design of international trade agreements: Introducing a new dataset**. *Rev Int Organ* (2014.0) **9** 353-75. DOI: 10.1007/s11558-013-9179-8
9. Popkin BM. **The nutrition transition and its health implications in lower-income countries**. *Public Health Nutrition.* (1997.0) **I**
10. Fardet A, Rock E. **Ultra-processed foods and food system sustainability: What are the links?**. *Sustain* (2020.0) **12**. DOI: 10.3390/SU12156280
11. Grübler J, Stöllinger R, Tondl G. **Wanted! Free Trade Agreements in the Service of Environmental and Climate Protection**. *wiiw Research Report* (2021.0)
12. Hawkes C, Thow AM. **Implications of the Central America-Dominican Republic-Free Trade Agreement for the nutrition transition in Central America**. *Rev Panam Salud Pública* (2008.0) **24** 345-60. DOI: 10.1590/s1020-49892008001100007
13. Hawkes C.. **The role of foreign direct investment in the nutrition transition**. *Public Heal Nutr* (2004.0) **8** 357-365
14. Martinez SW. *Vertical Coordination of Marketing Systems: Lessons From the Poultry, Egg, and Pork Industries* (2002.0) 1-45
15. Thow AM, Hawkes C. **The implications of trade liberalization for diet and health: A case study from Central America**. *Global Health* (2009.0) **5** 1-15. DOI: 10.1186/1744-8603-5-5
16. Schram A, Labonte R, Baker P, Friel S, Reeves A, Stuckler D. **The role of trade and investment liberalization in the sugar-sweetened carbonated beverages market: A natural experiment contrasting Vietnam and the Philippines**. *Global Health* (2015.0) **11** 1-13. DOI: 10.1186/s12992-015-0127-7
17. Baker P, Friel S, Schram A, Labonte R. **Trade and investment liberalization, food systems change and highly processed food consumption: A natural experiment contrasting the soft-drink markets of Peru and Bolivia**. *Global Health* (2016.0) **12** 1-13. DOI: 10.1186/s12992-016-0161-0
18. Siegel AD. *Wash* (2016.0) **50**
19. Villareal MA, Fergusson IF. *The North American Free Trade Agreement (NAFTA)* (2017.0)
20. Villalpando S, Shamah-Levy T, Rojas R, Aguilar-Salinas CA. **Trends for type 2 diabetes and other cardiovascular risk factors in Mexico from 1993–2006**. *Salud Publica Mex* (2010.0) **52**. DOI: 10.1590/s0036-36342010000700011
21. Barlow P, Stuckler D. **obalization and health policy space: Introducing the WTOhealth dataset of trade challenges to national health regulations at World Trade Organization, 1995–2016**. *Soc Sci Med* (2021.0) **275** 113807. DOI: 10.1016/j.socscimed.2021.113807
22. Barlow P, McKee M, Basu S, Stuckler D. **Impact of the North American free trade agreement on high-fructose corn syrup supply in Canada: A natural experiment using synthetic control methods**. *Cmaj* (2017.0) **189** E881-7. DOI: 10.1503/cmaj.161152
23. Krishnapillai S.. **Impact of NAFTA on the preference for meat consumption in USA: An inverse demand system approach**. *Int J Econ Financ Issues* (2012.0) **2** 79-84
24. Unar-Munguía M, Monterubio Flores E, Colchero MA. **Apparent consumption of caloric sweeteners increased after the implementation of NAFTA in Mexico**. *Food Policy* (2019.0) **84** 103-10. DOI: 10.1016/j.foodpol.2019.03.004
25. 25FAO. Food balance sheets: A handbook. 2001.. *Food balance sheets: A handbook* (2001.0)
26. 26FAOSTAT. [cited 2021 Aug 1]. Available from: http://www.fao.org/faostat/en/#home.
27. Linden A.. **Combining synthetic controls and interrupted time series analysis to improve causal inference in program evaluation**. *J Eval Clin Pract.* (2018.0) **24** 447-53. DOI: 10.1111/jep.12882
28. Huitema BE, McKean JW. **Design specification issues in time-series intervention models**. *Educ Psychol Meas* (2000.0) **60** 38-58. DOI: 10.1177/00131640021970358
29. Linden A.. **Conducting interrupted time-series analysis for single- and multiple-group comparisons**. *Stata J* (2010.0) **10** 288-308
30. Linden A, Adams JL. **Applying a propensity score-based weighting model to interrupted time series data: Improving causal inference in programme evaluation**. *J Eval Clin Pract* (2010.0) **17** 1231-8. DOI: 10.1111/j.1365-2753.2010.01504.x
31. Bouttell J, Craig P, Lewsey J, Robinson M, Popham F. **Synthetic control methodology as a tool for evaluating population-level health interventions**. *J Epidemiol Community Health* (2018.0) **72** 673-8. DOI: 10.1136/jech-2017-210106
32. 32World Bank. World Bank Data. 2020 [cited 2021 Aug 1]. Available from: https://datos.bancomundial.org/pais.. *World Bank Data* (2020.0)
33. Abadie A, Diamond A, Hainmueller AJ. **Synthetic control methods for comparative case studies: Estimating the effect of California’s Tobacco control program**. *J Am Stat Assoc* (2010.0) **105** 493-505. DOI: 10.1198/jasa.2009.ap08746
34. Werner M, Isa Contreras P, Mui Y, Stokes-Ramos H. **International trade and the neoliberal diet in Central America and the Dominican Republic: Bringing social inequality to the center of analysis**. *Soc Sci Med* (2019.0) **239** 112516. DOI: 10.1016/j.socscimed.2019.112516
35. Thow AM, Heywood P, Schultz J, Quested C, Jan S, Colagiuri S. **Trade and the nutrition transition: Strengthening policy for health in the pacific**. *Ecol Food Nutr* (2011.0) **50** 18-42. DOI: 10.1080/03670244.2010.524104
36. Marrón-Ponce JA, Tolentino-Mayo L, Hernández-F M, Batis C. **Trends in ultra-processed food purchases from 1984 to 2016 in Mexican households**. *Nutrients* (2019.0) **11** 1-15. DOI: 10.3390/nu11010045
37. Buzby Jean C., Wells Hodan F.. **The Estimated Amount, Value, and Calories of Postharvest Food Losses at the Retail and Consumer Levels in the United States**. *Econ Res Serv* (2014.0) **EIB-121** 1-30. DOI: 10.1016/j.foodpol.2010.10.010\nhttp://www.nrdc.org/food/files/wasted-food-IP.pdf?mkt_tok=3RkMMJWWfF9wsRonuqjPZKXonjHpfsX56+woXaS1lMI/0ER3fOvrPUfGjI4ATMphI/qLAzICFpZo2FFUH+GbbIFU8g==\n
38. Del Gobbo LC, Khatibzadeh S, Imamura F, Micha R, Shi P, Smith M. **Assessing global dietary habits: A comparison of national estimates from the FAO and the global dietary database**. *Am J Clin Nutr* (2015.0) **101** 1038-46. DOI: 10.3945/ajcn.114.087403
39. Daphna KD, Allen LH. **The importance of milk and other animal-source foods for children in low-income countries**. *Food Nutr Bull* (2011.0) **32**
40. Nardone A, Ronchi B, Lacetera N, Ranieri MS, Bernabucci U. **Effects of climate changes on animal production and sustainability of livestock systems**. *Livest Sci* (2010.0) **130** 57-69. DOI: 10.1016/j.livsci.2010.02.011
41. Willett W, Rockström J, Loken B, Springmann M, Lang T, Vermeulen S. **Food in the Anthropocene: the EAT–Lancet Commission on healthy diets from sustainable food systems**. *Lancet* (2019.0) **393** 447-92. DOI: 10.1016/S0140-6736(18)31788-4
42. Shepon A, Eshel G, Noor E MR. **The opportunity cost of animal based diets exceeds all food losses**. *Proc Natl Acad Sci U S A* (2018.0) **10** 3804-9. DOI: 10.1073/pnas.1713820115
43. Jensen JD, Smed S. **The Danish tax on saturated fat—Short run effects on consumption, substitution patterns and consumer prices of fats**. *Food Policy* (2013.0) **42** 18-31. DOI: 10.1016/j.foodpol.2013.06.004
44. Cherniwchan J.. **Trade liberalization and the environment: Evidence from NAFTA and U.S. manufacturing**. *J Int Econ* (2017.0) **105** 130-49. DOI: 10.1016/j.jinteco.2017.01.005
45. Mahrinasari MS, Haseeb M, Ammar J. **Is trade liberalization a hazard to sustainable environment? Fresh insight from asean countries**. *Polish J Manag Stud* (2019.0) **19** 249-59. DOI: 10.17512/pjms.2019.19.1.19
|
---
title: Investigating the role of climate-related disasters in the relationship between
food insecurity and mental health for youth aged 15–24 in 142 countries
authors:
- Isobel Sharpe
- Colleen M. Davison
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021754
doi: 10.1371/journal.pgph.0000560
license: CC BY 4.0
---
# Investigating the role of climate-related disasters in the relationship between food insecurity and mental health for youth aged 15–24 in 142 countries
## Abstract
Food insecurity (FI) represents a major global health challenge. Because climate-related disasters are a determinant of both FI and poor mental health, we investigated whether the severity of these disasters intensifies the relationship between FI and youth mental health. Data on FI and mental health came from the Gallup World Poll, a nationally representative survey of individuals in 142 countries, which included 28,292 youth aged 15–24. Data on climate-related disasters came from the International Disaster Database, a country-level record of disasters. Multilevel negative binomial regression was used to calculate relative risk (RR) of poor mental health. Youth with moderate or severe FI were significantly more likely to report poor mental health experiences compared to those with none/mild FI (moderate: RR 1.37, $95\%$ confidence interval (CI) 1.32–1.41; severe: 1.60, $95\%$ CI 1.54–1.66). We also observed a weak yet significant interaction effect ($p \leq 0.0001$), which suggested that the country-level relationship between FI and poor mental health is slightly stronger at greater disaster severity. While further research is needed to improve our understanding of these complex relationships, these findings suggest that mental health should be considered when undertaking national climate change actions and that additional FI-related supports may work to improve youth mental health.
## Introduction
Food insecurity (FI), or the lack of access to sufficient, safe, and nutritious food [1], represents a major global health challenge [2]. FI is associated not only with various negative physical health outcomes, including reduced nutrient intake, chronic disease, and premature death [3–7], but also with negative mental health outcomes [8, 9]. Individuals may experience feelings of stress, anxiety, depression, shame, and alienation due to FI itself or their FI status relative to others [9–11]. Conversely, poor mental health may contribute to FI, for example through inability to generate steady income or to cope with challenges [12–14]. Complex, and likely bidirectional, associations exist between FI and mental health.
Youth, ages 15 to 24 years, are vulnerable with respect to FI and mental health. Youth often brings important life changes such as starting an occupation or changes in relationships and living situations, presenting potential opportunities for worsened FI status [15, 16]. This is especially true at the upper end of this age range, when young adults often become more independent from their parents [17, 18]. Many youth also struggle with mental health concerns. In 2019, approximately 86 million youth aged 15–19 were living with a mental disorder [19]. Furthermore, suicide was the fourth leading cause of death among this age group [20]. Poor mental health has the potential to affect psychological development, learning, and relationship building in youth and throughout the life course. Relatively few studies have assessed the relationship between FI and mental health among youth, particularly at the global level and using well-tested measures of FI [21, 22].
Climate change, largely driven by human greenhouse gas emissions, represents one of the most serious current threats to human health and wellbeing [23]. One particularly concerning aspect of climate change is the increase in frequency and intensity of climate-related natural disasters such as floods, storms, heatwaves, and droughts [23, 24]. Climate-related disasters are a key contextual determinant of both FI and mental health. The 2020 Global Report on Food Crises identified climate-related disaster events as the second-most common driver of acute FI, affecting 34 million people across 25 countries [25]. Climate-related disasters contribute to FI through unexpected disruptions to crop yields and to physical and economic access to food. Climate-related disasters also harm mental health through their connection to psychological trauma for populations facing these events [26–28]. The effects of climate-related disasters fall disproportionately on those populations most vulnerable [29], working to intensify existing inequities such as those related to both FI and mental health. Youth is often a time of developmental and sociodemographic change, meaning that the cascading effects of climate-related disasters, such as disruptions to school and work, economic instability, and displacement [30], may create vulnerabilities within this population.
The potential three-way relationship among climate-related disasters, FI, and mental health has not been well studied. To our knowledge, there has just been a single study with this focus. Using a nationally representative Australian sample, Friel and colleagues [31] measured the association between FI and mental health at varying levels of drought. The authors found support for the idea that drought exposures modify the association between FI and mental health, but emphasised that these relationships were complex and challenging to quantify. Notably, the study tested three binary indicators of FI and therefore may not have been accurate in capturing the overall FI construct in comparison to an internationally validated multi-component scale [31]. Another methodological issue, common within this area of literature, was that the study explored the effects of a single disaster event on a relatively small population [31, 32]. In summary, more work is needed to better understand the relationships among climate-related disasters, FI, and mental health, particularly at a global scale using validated measures.
Thus, the main objective of our study was to determine whether climate-related disasters modify the relationship between FI and poor mental health among youth globally. We were particularly interested in understanding whether the severity of climate-related disasters intensifies the FI-mental health relationship.
## Data sources
The proposed objective was addressed using a cross-sectional design with data from two secondary sources. The first data source was the 2017 cycle of the Gallup World Poll (GWP), a nationally representative survey of individuals aged 15 and older from 148 countries [33]. In each country, a representative sample of ~1000 individuals was collected using one of two techniques. Telephone surveys were used in those countries with at least $80\%$ telephone coverage or where telephone surveys were customary. To sample individuals by phone, either random digit dialling or selection from a nationally representative phone number list was used. In the remainder of countries, face-to-face surveys were conducted. Individuals were selected through a stratified multi-stage cluster sampling process. In the first stage, clusters of households were stratified by population size and/or geography and 100–125 clusters were selected. In the second stage, 8–10 households were selected from each of the clusters using random route procedures. In the third stage, one respondent was randomly selected from each household. Further details on the GWP survey methodology are available elsewhere [34]. To address our specific research objective, we sampled ‘youth’ between ages 15–24 based on the United Nations (UN) definition [35]. For the purposes of our study, the GWP provided individual-level data on the exposure, outcome, and potential confounders of interest. Gallup obtained informed consent from all participants and the governing bodies of each country approved the survey protocols.
The second data source was the International Disaster Database (EM-DAT), a public global database of natural and technological disasters and their impacts developed by the Centre for Research on the Epidemiology of Disasters [36]. All disasters included in the database met at least one of the following criteria: 1) at least 10 people were reported killed, 2) at least 100 people were reported affected, 3) a state of emergency was declared, or 4) a call for international assistance occurred. For the current study, EM-DAT provided country-level data on the effect modifier of interest: climate-related disaster severity. Based on the GWP 2017 survey year, we chose to capture climate-related disaster severity data from 2015–2017. This three-year observation window was selected in an attempt to capture the time period through which climate-related disasters may impact individuals’ FI and/or mental health in the GWP data year 2017 [37–40]. While some disaster-related effects are very immediate (e.g., local food sources destroyed, disaster-related trauma), others are more gradual (e.g., long-term disruptions to crop yield, lasting mental health problems).
## Food insecurity
FI was measured using the GWP’s Food Insecurity Experiences Scale (FIES) [41]. The FIES is an 8-item questionnaire that measures individual-level FI over the previous 12 months (S1 Text). The FIES is a valid and reliable psychometric measure [42–44]. Its Rasch reliability falls between 0.70–0.80 for $79\%$ of countries included in the GWP [42]. Further, strong correlations between the FIES and related measures, such as prevalence of undernourishment (0.79) and child malnutrition (0.60), suggest its validity [43]. For the current analysis, responses to all eight FIES questions were coded as ‘yes’ = 1 and ‘no’ = 0. The responses were then summed and categorised into the following levels of FI: none or mild (score 0–3), moderate (score 4–6), or severe (score 7–8). Scores were only calculated for those who provided a valid response to each question; those with missing data or who responded ‘don’t know’ or ‘refused’ to any of the eight items were counted as missing.
## Poor mental health
Poor mental health was measured using the GWP’s Daily Experience Index (DEI) (S1 Text). The DEI measures level of wellbeing through the existence or absence of positive and negative feelings [34]. At the country level, the DEI has a Cronbach’s alpha value of 0.72 [34], indicating a satisfactory level of internal consistency [45]. In the present analysis, the DEI was calculated by coding all ten individual items where an answer reflecting negative emotion received a score of 1 and all other answers received a score of 0. We summed the responses to all 10 items, resulting in a count score for poor mental health ranging from 0 to 10 (where a higher score represents worse mental health). Following Gallup’s methodology [34], we did not calculate the DEI score if more than two individual items were missing (‘don’t know’ or ‘refused’).
## Climate-related disasters
Data on climate-related disasters came from the EM-DAT. Climate-related disasters were defined as those with meteorological, hydrological, or climatological origins, as described by EM-DAT’s disaster classification system [36]: extreme temperatures, fog, storms, floods, landslides, wave actions, droughts, glacial lake outbursts, and wildfires.
EM-DAT reports disasters at the country level. For the present analysis, we were interested in identifying those countries whose populations may be most seriously affected by climate-related disasters and comparing them with countries whose populations are less seriously affected. Thus, we categorised each country as ‘high’ or ‘low’ according to their level of climate-related disaster severity and assigned the severity variable to all individuals within that country. Disaster severity was based on the total number of people killed per 1,000,000 population in a given country over the time period of interest. This was, therefore, a country-level sum of those presumed dead as a result of all climate-related disaster events that occurred in that country between 2015–2017. To isolate those countries most seriously affected by climate-related disasters, we categorised the top $10\%$ of countries as high severity and the bottom $90\%$ as low severity. We chose the 90th percentile as a cut-off value due to the exponential distribution of this variable and the obvious divide between countries with many climate-related disasters and those with far fewer at about the 90th percentile mark (S1 Text).
Missing values for number of deaths were set to 0 under the assumption that disaster events associated with high mortality were less likely to be unreported than those with few or no deaths [46]. Data from the World Bank [47] were used to adjust for country population size. As per Ward and Shively [48], we used population data from 2014 (one year before the time period of interest) to account for the fact that a country’s population size may be affected by the disasters themselves.
## Potential confounders
The following variables were selected as potential confounders based on existing literature [8, 22, 49–52] and their availability in the GWP dataset: the respondent’s age (years) and gender (male, female), level of urbanicity of the respondent’s home (rural or farm, small town or village, suburb of a large city, large city), number of children in household <15 years of age, marital status of the respondent (single/never married, married, separated, divorced, widowed, domestic partner), highest level of education completed by the respondent (elementary or less, secondary, tertiary), employment status (full time for an employer, full time self-employed, part time and not want to work full time, part time and want to work full time, unemployed, out of the workforce) and annual household income (international dollars). The household income variable was log transformed due to its diminishing returns on mental wellbeing [8]. Due to their small number, all ‘don’t know’ or ‘refused’ responses for these variables were counted as missing.
## Statistical analysis
The final study sample included 142 countries. Six countries–Brazil, Maldives, Mauritania, Moldova, Turkmenistan, and Vietnam–were removed due to large amounts of missing data for the variables of interest. An additional 194 observations were removed due to missing outcome data. The remaining number of missing observations was proportionally small (<$5\%$) and treated as missing at random. To accommodate for the differences in sampling design between those countries using face-to-face surveys (multi-stage stratified cluster sampling) and those using telephone surveys (stratified random sampling), all strata with a single primary sampling unit were pooled.
Descriptive statistics were presented for the study sample, accounting for the appropriate stratification, clustering, and sampling weights. Further, poststratification weights were adopted to improve the global representativeness of the sample at the country level [8]. All descriptive statistics were presented as mean (standard error of the mean; SEM) for continuous variables and as frequency (%) for categorical variables.
Multilevel negative binomial regression models with robust standard error estimates were used to generate a relative risk (RR) and corresponding $95\%$ confidence interval (CI) for the poor mental health outcome. The negative binomial model was used due to overdispersion in the Poisson model. All models were unweighted and fitted using SAS PROC GLIMMIX, where we included a random intercept to account for added variation at the country level. We first generated bivariate models to examine each predictor. The backwards selection technique was then used to generate a parsimonious adjusted model, with the likelihood ratio test used to assess changes in model fit. The first adjusted model tested FI as the main predictor. To better understand how the relationship between FI and poor mental health varied around the world, we stratified our findings by country income level, UN sub-region, and individual country. Country-specific results were presented on a world map using ArcGIS software by Esri. The second adjusted model evaluated the addition of an interaction between FI and climate-related disaster severity (high vs. low). Besides the two main models, we performed sensitivity analyses to test various conceptualisations of the climate-related disaster severity variable.
The regression analyses were ≥$90\%$ powered to detect a RR of 1.5 for outcomes ranging in prevalence from $20\%$-$60\%$. Level of significance was set at $p \leq 0.05$ unless otherwise specified. All statistical analyses were conducted using SAS 9.4 (SAS Inc., Cary, North Carolina, USA).
## Results
The final study sample from the 2017 GWP survey consisted of 28,292 youth from 142 countries (S1 Text). The characteristics of the study sample are presented in Table 1. Based on FIES scores, $77.8\%$ of youth reported none or mild FI, $11.2\%$ reported moderate FI, and $11.0\%$ reported severe FI. The mean score for poor mental health was 2.29 (SEM 0.06) out of 10. S1 Text presents a histogram of poor mental health scores, showing that the majority of youth scored low (≤2) on this measure. During the three-year period leading up to the GWP survey year (2015–2017) a total of 828 climate-related disasters occurred within the 142 countries of interest, resulting in 27,508 reported deaths (S1 Text). Of those disasters, $47.7\%$ were floods, $33.2\%$ were storms, $6.6\%$ were landslides, $4.5\%$ were wildfires, $4.1\%$ were droughts, and $3.9\%$ were extreme temperatures. The countries with the highest total number of people killed per 1,000,000 population over the 2015–2017 period were Sierra Leone (158.5 per 1,000,000), Haiti (61.1 per 1,000,000), France (49.9 per 1,000,000), Belgium (36.8 per 1,00,000), and Sri Lanka (26.1 per 1,000,000).
**Table 1**
| Characteristic | Weighted n | % |
| --- | --- | --- |
| Food insecurity (FIES score) | | |
| None or mild (0–3) | 21014 | 77.8 |
| Moderate (4–6) | 3028 | 11.2 |
| Severe (7–8) | 2974 | 11.0 |
| Missing | 1422 | |
| Gender | | |
| Male | 14665 | 51.7 |
| Female | 13709 | 48.3 |
| Urbanicity | | |
| Rural | 11071 | 39.1 |
| Small town | 9562 | 33.8 |
| Suburban | 2325 | 8.2 |
| Urban | 5353 | 18.9 |
| Missing | 63 | |
| Marital status | | |
| Single | 21669 | 76.5 |
| Married | 6073 | 21.4 |
| Separated | 60 | 0.2 |
| Divorced | 100 | 0.4 |
| Widowed | 51 | 0.2 |
| Domestic partner | 366 | 1.3 |
| Missing | 56 | |
| Education (highest level completed) | | |
| Elementary or less | 9271 | 32.7 |
| Secondary | 17242 | 60.9 |
| Tertiary | 1818 | 6.4 |
| Missing | 43 | |
| Employment | | |
| Full-time (employer) | 6340 | 22.3 |
| Full-time (self-employed) | 2846 | 10.0 |
| Part-time (seeking full-time) | 2455 | 8.7 |
| Part-time (not seeking full-time) | 1201 | 4.2 |
| Unemployed | 2140 | 7.5 |
| Out of workforce | 13393 | 47.2 |
| | Mean | SEM |
| Age (years) | 19.53 | 0.08 |
| Number of children in household aged <15 years | 1.09 | 0.03 |
| Missing (weighted n) | 30 | |
| Poor mental health (DEI score) | 2.29 | 0.06 |
| Annual household income (international $, thousands) | 17.64 | 2.80 |
| Log annual household income (international $) | 8.90 | 0.05 |
Table 2 shows the results of the multilevel negative binomial regression models examining the association between FI and poor mental health with country as a random effect. Compared to the bivariate models, most of the fully adjusted RR estimates from the multivariate model were slightly attenuated but remained consistent in their relative size and direction. Notably, FI produced the strongest effect of all predictors in the multivariate model. Youth with moderate or severe FI were significantly more likely to report experiences of poor mental health compared to those with none or mild FI (moderate FI: RR 1.37, $95\%$ CI 1.32–1.41; severe FI: 1.60, $95\%$ CI 1.54–1.66). In addition, experiences of poor mental health were significantly predicted by older age (RR 1.02, $95\%$ CI 1.02–1.03), greater number of children in the household (RR 1.01, $95\%$ CI 1.00–1.01), and lower annual household income (RR 0.96, $95\%$ CI 0.95–0.97). Those who were educated at an elementary level or lower were significantly more likely to report poor mental health experiences compared to those with a tertiary level education (RR 1.06, $95\%$ CI 1.00–1.13). Youth who were widowed (RR 1.34, $95\%$ CI 1.13–1.58), separated (RR 1.14, $95\%$ CI 1.02–1.28), or married (RR 1.06, $95\%$ CI 1.01–1.10) were significantly more likely to report experiences of poor mental health compared to those who were single. Further, those working part-time (seeking full-time hours: RR 0.94, $95\%$ CI 0.90–0.98; not seeking full-time hours: RR 0.95, $95\%$ CI 0.91–0.99) or who were out of the workforce (RR 0.87, $95\%$ CI 0.83–0.91) were significantly less likely to report poor mental health experiences compared to those working full-time hours.
**Table 2**
| Unnamed: 0 | Bivariate Models | Bivariate Models.1 | Multivariate Model | Multivariate Model.1 |
| --- | --- | --- | --- | --- |
| Factor | RR (95% CI) | p | RR (95% CI) | p |
| Food insecurity | | | | |
| None or mild | 1 (ref) | | 1 (ref) | |
| Moderate | 1.43 (1.39–1.48) | <0.0001 | 1.37 (1.32–1.41) | <0.0001 |
| Severe | 1.70 (1.63–1.77) | <0.0001 | 1.60 (1.54–1.66) | <0.0001 |
| Gender | | | | |
| Male | 1 (ref) | | 1 (ref) | |
| Female | 1.03 (1.00–1.06) | 0.0270 | 1.03 (1.00–1.05) | 0.0690 |
| Urbanicity | | | | |
| Urban | 1 (ref) | | 1 (ref) | |
| Suburban | 1.00 (0.96–1.04) | 0.9522 | 0.99 (0.95–1.03) | 0.6755 |
| Small village | 1.01 (0.98–1.04) | 0.3482 | 0.97 (0.95–1.00) | 0.0527 |
| Rural | 1.08 (1.04–1.13) | <0.0001 | 1.01 (0.97–1.05) | 0.7062 |
| Marital status | | | | |
| Single | 1 (ref) | | 1 (ref) | |
| Domestic partner | 1.14 (1.08–1.21) | <0.0001 | 1.03 (0.98–1.08) | 0.2743 |
| Married | 1.17 (1.12–1.22) | <0.0001 | 1.06 (1.01–1.10) | 0.0092 |
| Separated | 1.38 (1.24–1.53) | <0.0001 | 1.14 (1.02–1.28) | 0.0177 |
| Divorced | 1.27 (1.13–1.42) | <0.0001 | 1.13 (0.99–1.29) | 0.0621 |
| Widowed | 1.49 (1.28–1.73) | <0.0001 | 1.34 (1.13–1.58) | 0.0005 |
| Education | | | | |
| Tertiary | 1 (ref) | | 1 (ref) | |
| Secondary | 1.01 (0.97–1.06) | 0.6235 | 1.03 (0.98–1.09) | 0.1779 |
| Elementary or less | 1.09 (1.03–1.15) | 0.0042 | 1.06 (1.00–1.13) | 0.0423 |
| Employment | | | | |
| Full-time (employer) | 1 (ref) | | 1 (ref) | |
| Full-time (self-employed) | 0.99 (0.94–1.04) | 0.6305 | 0.97 (0.93–1.02) | 0.2670 |
| Part-time (seeking full-time) | 0.94 (0.90–0.99) | 0.0127 | 0.94 (0.90–0.98) | 0.0062 |
| Part-time (not seeking full-time) | 0.93 (0.89–0.97) | 0.0008 | 0.95 (0.91–0.99) | 0.0197 |
| Unemployed | 1.06 (1.01–1.11) | 0.0100 | 1.01 (0.97–1.06) | 0.5327 |
| Out of workforce | 0.83 (0.80–0.87) | <0.0001 | 0.87 (0.83–0.91) | <0.0001 |
| Number of children in household <15 years | 1.01 (1.01–1.02) | <0.0001 | 1.01 (1.00–1.01) | 0.0035 |
| Age | 1.03 (1.03–1.04) | <0.0001 | 1.02 (1.02–1.03) | <0.0001 |
| Log annual household income | 0.92 (0.91–0.93) | <0.0001 | 0.96 (0.95–0.97) | <0.0001 |
| Disaster severitya | | | | |
| Low | 1 (ref) | | 1 (ref) | |
| High | 1.13 (0.99–1.30) | 0.0801 | 1.05 (0.93–1.19) | 0.4217 |
We performed a series of stratified analyses to determine whether the relationship between FI and poor mental health varied across different areas of the world. First, we stratified the multivariate model by country income level (S1 Text). We observed a consistent positive association between FI and poor mental health among youth across low-, middle-, and high-income levels, with RR values ranging from 1.31–1.45 for moderate FI and from 1.53–1.74 for severe FI (vs. none or mild FI). Second, we stratified the model by UN sub-region (Fig 1). In all 19 sub-regions, youth with moderate and/or severe FI were significantly more likely to report poor mental health experiences compared to those with none or mild FI. Third, we stratified the model by individual country (Fig 2). The strongest positive associations between FI and poor mental health among youth were largely concentrated in the Western countries, including Canada, the United States, Australia, Germany, the United Kingdom, and much of Eastern Europe.
**Fig 1:** *Forest plot showing the association between food insecurity (Food Insecurity Experiences Scale) and poor mental health (Daily Experience Index score) among n = 28,292 youth from the 2017 Gallup World Poll survey, by United Nations sub-region.Notes: There are 19 UN sub-regions (https://unstats.un.org/unsd/methodology/m49/overview/). Estimates were adjusted for gender, urbanicity, marital status, education, employment, number of children <15 in household, age, log annual household income, disaster severity (fixed effects), and country (random effect). Abbreviations: United Nations (UN), relative risk (RR), confidence interval (CI), food insecurity (FI).* **Fig 2:** *Heat map depicting the strength of association between food insecurity (Food Insecurity Experiences Scale) and poor mental health (Daily Experience Index score) in n = 137 countries.Notes: Presents relative risk estimates for moderate or severe FI vs. none or mild FI. Estimates were adjusted for gender, urbanicity, marital status, education, employment, number of children <15 in household, age, and log annual household income (fixed effects). Five of the 142 countries in the Gallup World Poll sample were not included in the map (the regression models for Hungary, Latvia, North Macedonia, and Slovenia did not converge, Kosovo was not available in the map template). Base map available at: https://www.arcgis.com/home/item.html?id=ee8678f599f64ec0a8ffbfd5c429c896 [53].*
To determine whether climate-related disasters modify the relationship between FI and mental health among youth, we added a FI*disaster severity interaction term to the multivariate model from Table 2. This term was not statistically significant (interaction $$p \leq 0.5362$$; S1 Text) and model fit did not improve according to the likelihood ratio test. Fig 3 (left) provides a graphical representation of the interaction, showing that while poor mental health increased with greater FI, this effect did not differ by level of disaster severity.
**Fig 3:** *Interaction plots showing the marginal effect of food insecurity (Food Insecurity Experiences Scale) and disaster severity on poor mental health (Daily Experience Index score) among n = 28,292 youth from the 2017 Gallup World Poll survey.Left: disaster severity represented as a dichotomous variablea. Right: disaster severity represented as a continuous variable with 95% confidence intervalsb. Notes: Adjusted for gender, urbanicity, marital status, education, employment, number of children <15 in household, age, log annual household income, disaster severity (fixed effects), and country (random effect). aDisaster severity was measured at the country level. High disaster severity was defined as those countries in the top 10% of total number of climate-related disaster deaths per 1,000,000 population between the years of 2015–2017. Low disaster severity was defined as those countries in the bottom 90% of total number of climate-related disaster deaths per 1,000,000 population between the years of 2015–2017. bDisaster severity was measured at the country level, defined as the total number of climate-related disaster deaths per 1,000,000 population between the years of 2015–2017 (continuous).*
## Sensitivity analyses
We performed three sensitivity analyses to address some of the limitations associated with our conceptualisation of the climate-related disaster severity variable. First, we tested disaster severity as a continuous measure of disaster deaths, as opposed to its initial representation as a binary (high vs. low) variable. In doing so, we observed a very weak yet significant effect of the FI*disaster severity interaction on poor mental health (interaction $p \leq 0.0001$). Table 3 displays the results of this model, showing that the interaction was significant for severe FI ($$p \leq 0.0063$$) but not moderate FI ($$p \leq 0.1817$$) when compared to none or mild FI. These findings are also reflected in the interaction plot (Fig 3, right), which shows a very small increase in poor mental health with increasing disaster severity for the severe FI group (thus a more steeply sloped line) compared to none or mild FI group. Second, we tested disaster severity as a continuous frequency measure, capturing the total number of climate-related disasters per 1,000,000 km2 country surface area. In this case, the FI*disaster severity interaction was not statistically significant (interaction $$p \leq 0.1098$$; S1 Text). Lastly, we changed the time period of interest for the disaster severity variable from 2015–2017 to 2017 only. Again, the FI*disaster severity interaction was not statistically significant (interaction $$p \leq 0.1923$$; S1 Text).
**Table 3**
| Unnamed: 0 | Multivariate Model | Multivariate Model.1 |
| --- | --- | --- |
| Factor | RR (95% CI) | p |
| Food insecurity | | |
| None or mild | 1 (ref) | |
| Moderate | 1.36 (1.32–1.41) | <0.0001 |
| Severe | 1.58 (1.52–1.65) | <0.0001 |
| Gender | | |
| Male | 1 (ref) | |
| Female | 1.03 (1.00–1.05) | 0.0714 |
| Urbanicity | | |
| Urban | 1 (ref) | |
| Suburban | 0.99 (0.95–1.03) | 0.6260 |
| Small village | 0.97 (0.95–1.00) | 0.0486 |
| Rural | 1.01 (0.97–1.05) | 0.7181 |
| Marital status | | |
| Single | 1 (ref) | |
| Married | 1.06 (1.01–1.102) | 0.0089 |
| Separated | 1.15 (1.03–1.28) | 0.0141 |
| Divorced | 1.13 (0.99–1.28) | 0.0677 |
| Widowed | 1.34 (1.13–1.58) | 0.0005 |
| Domestic partner | 1.03 (0.98–1.08) | 0.2488 |
| Education | | |
| Tertiary | 1 (ref) | |
| Secondary | 1.03 (0.98–1.09) | 0.1762 |
| Elementary or less | 1.06 (1.00–1.13) | 0.0436 |
| Employment | | |
| Full-time (employer) | 1 (ref) | |
| Full-time (self-employed) | 0.97 (0.93–1.02) | 0.2841 |
| Part-time (seeking full-time) | 0.94 (0.90–0.98) | 0.0062 |
| Part-time (not seeking full-time) | 0.95 (0.91–0.99) | 0.0206 |
| Unemployed | 1.01 (0.97–1.06) | 0.5286 |
| Out of workforce | 0.87 (0.83–0.91) | <0.0001 |
| Number of children in household <15 years | 1.01 (1.00–1.01) | 0.0034 |
| Age | 1.02 (1.02–1.03) | <0.0001 |
| Log annual household income | 0.96 (0.95–0.97) | <0.0001 |
| Disaster severitya | 1.00 (1.00–1.00) | 0.3042 |
| Food insecurity*disaster severityb | | |
| None or mild*disaster severity | 1 (ref) | |
| Moderate*disaster severity | 1.00 (1.00–1.00) | 0.1817 |
| Severe *disaster severity | 1.00 (1.00–1.00) | 0.0063 |
## Discussion
Youth are vulnerable with respect to both FI and mental health. Within this context, they are susceptible to the additional burdens caused by climate-related disasters such as floods, storms, and heatwaves. Through our multilevel analysis, we explored whether the context of climate-related disaster severity significantly worsened the FI-mental health relationship among this age group. Few studies have explored the relationship among these three variables to date, particularly at a global level.
Our analysis revealed two main findings. For one, we observed a significant association between FI and poor mental health among youth globally. This result was in line with several previous analyses of GWP data, conducted among both youth [21, 22] and adults [8, 49–52]. Our analysis controlled for a number of indicators of poverty and material deprivation, such as annual household income, education, and employment status. This suggests that the relationship between FI and mental health among youth is not solely driven by poverty and deprivation, indicating that other psychosocial or biological mechanisms are likely also at play [22]. For example, individuals may develop what is known as toxic stress, a condition that occurs under chronic pressures, such as prolonged experiences of FI, and lack of adequate supports [54–56]. Toxic stress is has been implicated in poor mental health in the context of children and youth [56–58]. Additionally, those experiencing FI tend to consume inexpensive foods of poor nutritional quality and thereby are at a greater risk of malnutrition [7]. Malnutrition itself can contribute to poor mental health through reduced function of the brain and gut microbiome [59, 60].
Figs 1 and 2 show that the association between FI and poor mental health is potentially stronger among youth in Western countries such as Canada, the United States, Australia, and much of Eastern Europe. This finding aligns with previous work showing that the relationship between FI and poor mental health is stronger in higher-income countries and weaker in lower-income countries [50]. Experiences of FI tend to be more normalised in lower-income countries, and therefore its effect on one’s mental health may not be as pronounced in comparison to the higher-income countries where FI is less common [50]. In addition, many of these countries rely on collectivistic cultures, which have been linked to greater levels of connection and support [61, 62], therefore potentially mitigating the negative mental health impacts of FI. Individuals in lower-income countries may also regularly face not only FI but also many other large-scale stressors, including poverty, conflict, and insecurity, meaning that FI itself likely plays a small role in overall mental health.
The nature of the observed association between FI and mental health suggests the potential for a causal relationship. Firstly, our models showed that FI had the strongest relationship with mental health of all included covariates, indicating its relative importance in predicting mental health. Similarly, Frongillo et al. [ 50] found that FI produced the largest standardised regression coefficient after adjusting for various economic and social development indicators such as gross domestic product and income inequality. Second, in line with previous analyses, we observed a dose-response relationship between FI and mental health at the global scale. Jones et al. [ 52] suggested two mechanisms through which this dose-response relationship may manifest; greater levels of FI may amplify existing psychological stressors and/or invoke multiple new pathways that compound with existing stressors to harm mental health. There is a large body of evidence linking experiences of hunger, a defining feature of severe FI as measured by the FIES [41], to serious mental health problems among children and youth, including depression and suicide-related outcomes [63–65]. Physiologically, hunger may heighten the body’s response to stress in the hypothalamic-pituitary-adrenal axis, leading to subsequent negative mental health outcomes [63]. Depression and suicide-related outcomes potentially represent the pathways invoked through more severe cases of FI. Third, we found that the FI-mental health association was highly consistent. The association remained significant when the sample was stratified by country income level and UN sub-region. In their analysis of FI, state fragility, and mental health among youth, Elgar et al. [ 22] also reported significant associations between FI and poor mental health across all UN sub-regions. These findings suggest that the FI-mental health relationship persists among youth across cultural contexts.
Our second major finding was that while climate-related disaster severity did not modify the relationship between FI and mental health when treated as a dichotomous variable, we did see significant effect modification when it was treated as a continuous variable. Specifically, we observed a very weak yet significant interaction between FI and disaster severity in our sensitivity analysis where disaster severity was represented as a continuous measure of deaths per 1,000,000 population. This finding suggests that at the country level, the relationship between youth FI and poor mental health may grow stronger with higher levels of disaster severity. It is well known that FI negatively affects mental health through worries and shame related to food as well as through the negative mental health effects of malnutrition [8, 9]. Climate-related disasters are also directly associated with poor mental health [26–28]. Therefore, youth experiencing both climate-related disasters and FI may have compounding burdens on their mental health. Although our findings suggest that an interaction may exist, it is important to acknowledge that this effect was very weak and may not be practically meaningful. Ultimately, future research is needed to better understand and quantify this effect and the pathways through which it may operate.
## Strengths and limitations
Our study had several strengths. For one, to our knowledge it was the first of its kind to explore the relationship among youth FI, mental health, and climate-related disasters. Another strength was the use of the GWP dataset, which provided nationally representative data from 142 countries. While lack of a representative sample is a common limitation in the study of climate-related disasters, the GWP dataset allowed us to conduct a representative analysis on the global scale. Lastly, we used the FIES, a valid and reliable scale designed to provide a globally relevant measure of FI.
This analysis also had several limitations. For one, the cross-sectional nature of the data prevented us from identifying true causal relationships. It is possible that poor mental health itself may lead to FI, for example through the inability to hold steady employment or to cope with challenging situations [12–14]. As this study used pre-existing datasets, there were likely several unmeasured confounders. For example, the GWP did not report on pre-existing mental health conditions, which may predict both FI and current mental health status. Further, measures of both FI and poor mental health were self-reported, introducing the potential for social desirability bias. We also made assumptions regarding the use of EM-DAT mortality data. Foremost, the EM-DAT provided country-level data only, therefore we assigned the same value of disaster severity to all individuals within a given country. Climate-related disasters are often localised, meaning that our measure of disaster severity may not have reflected the true experiences of youth included in our study sample. To avoid ecological fallacy [66], we interpreted the effects of the disaster severity variable exclusively at the country level; however, future research should aim to collect this information at the individual level.
## Conclusions
Using data from the Gallup World Poll, we conducted a novel global analysis investigating the role of climate-related disaster severity in the relationship between food insecurity and poor mental health among 28,292 youth in 142 countries. In line with previous research, we observed a significant dose-response association between FI and poor mental health that transcended geographic and cultural contexts. Policies aimed at reducing levels of FI, such as promoting diversification in food production, improving infrastructure for those transporting and accessing food, promoting domestic production and food fortification, and reducing poverty and income inequality [2], may also be broadly effective in in improving youth mental health globally. In addition, while we observed a significant interaction effect where contexts of higher disaster severity strengthened the relationship between FI and poor mental health, its magnitude of effect was very weak. Our study provides an important starting point for understanding the pathways through which climate-related disasters and FI may synergistically harm mental health, yet further research is needed to improve our understanding of this complex system and ultimately support policy decisions.
## References
1. **Declaration and Plan of Action**. *Rome* (1996.0)
2. 2Food and Agriculture Organization (IT), International Fund for Agricultural Development (IT). The State of Food Security and Nutrition in the World 2021: Transforming food systems for food security, improved nutrition and affordable healthy diets for all [Internet]. Rome, Italy: FAO; 2021 [cited 2022 Mar 18]. Available from: https://www.fao.org/documents/card/en/c/cb4474en/. *The State of Food Security and Nutrition in the World 2021: Transforming food systems for food security, improved nutrition and affordable healthy diets for all* (2021.0)
3. Jomaa L, Naja F, Cheaib R, Hwalla N. **Household food insecurity is associated with a higher burden of obesity and risk of dietary inadequacies among mothers in Beirut, Lebanon**. *BMC Public Health* (2017.0) **17** 1-14. PMID: 28049454
4. Men F, Gundersen C, Urquia ML, Tarasuk V. **Association between household food insecurity and mortality in Canada: A population-based retrospective cohort study**. *Can Med Assoc J* (2020.0) **192** E53-60. DOI: 10.1503/cmaj.190385
5. Seligman HK, Laraia BA, Kushel MB. **Food insecurity is associated with chronic disease among low-income NHANES participants**. *J Nutr* (2010.0) **140** 304-10. DOI: 10.3945/jn.109.112573
6. Stuff JE, Casey PH, Connell CL, Champagne CM, Gossett JM, Harsha D. **Household food insecurity and obesity, chronic disease, and chronic disease risk factors**. *J Hunger Environ Nutr* (2007.0) **1** 43-62
7. Kirkpatrick SI, Tarasuk V. **Food insecurity is associated with nutrient inadequacies among Canadian adults and adolescents**. *J Nutr* (2008.0) **138** 604-12. DOI: 10.1093/jn/138.3.604
8. Elgar FJ, Pickett W, Pförtner TK, Gariépy G, Gordon D, Georgiades K. **Relative food insecurity, mental health and wellbeing in 160 countries**. *Soc Sci Med* (2021.0) **268** 113556. DOI: 10.1016/j.socscimed.2020.113556
9. Weaver LJ, Hadley C. **Moving beyond hunger and nutrition: A systematic review of the evidence linking food insecurity and mental health in developing countries**. *Ecol Food Nutr* (2009.0) **48** 263-84. DOI: 10.1080/03670240903001167
10. Radimer KL, Olson CM, Greene JC, Campbell CC, Habicht JP. **Understanding hunger and developing indicators to assess it in women and children**. *J Nutr Educ* (1992.0) **24** 36S-44S
11. Coates J, Frongillo EA, Rogers BL, Webb P, Wilde PE, Houser R. **Commonalities in the Experience of Household Food Insecurity across Cultures: What Are Measures Missing?**. *J Nutr* (2006.0) **136** 1438S-1448S. DOI: 10.1093/jn/136.5.1438S
12. Melchior M, Caspi A, Howard LM, Ambler AP, Bolton H, Mountain N. **Mental health context of food insecurity: A representative cohort of families with young children**. *Pediatrics* (2009.0) **124** e564-72. DOI: 10.1542/peds.2009-0583
13. Tarasuk V, Mitchell A, McLaren L, McIntyre L. **Chronic physical and mental health conditions among adults may increase vulnerability to household food insecurity**. *J Nutr* (2013.0) **143** 1785-93. DOI: 10.3945/jn.113.178483
14. Wehler C, Weinreb LF, Huntington N, Scott R, Hosmer D, Fletcher K. **Risk and protective factors for adult and child hunger among low-income housed and homeless female-headed families**. *Am J Public Health* (2004.0) **94** 109-15. DOI: 10.2105/ajph.94.1.109
15. Bocquier A, Vieux F, Lioret S, Dubuisson C, Caillavet F, Darmon N. **Socio-economic characteristics, living conditions and diet quality are associated with food insecurity in France**. *Public Health Nutr* (2015.0) **18** 2952-61. DOI: 10.1017/S1368980014002912
16. Pryor L, Lioret S, van der Waerden J, Fombonne É, Falissard B, Melchior M. **Food insecurity and mental health problems among a community sample of young adults**. *Soc Psychiatry Psychiatr Epidemiol* (2016.0) **51** 1073-81. DOI: 10.1007/s00127-016-1249-9
17. Masa R, Khan Z, Chowa G. **Youth food insecurity in Ghana and South Africa: Prevalence, socioeconomic correlates, and moderation effect of gender**. *Child Youth Serv Rev* (2020.0) **116** 105180
18. Baer TE, Scherer EA, Fleegler EW, Hassan A. **Food Insecurity and the Burden of Health-Related Social Problems in an Urban Youth Population**. *J Adolesc Health* (2015.0) **57** 601-7. DOI: 10.1016/j.jadohealth.2015.08.013
19. 19On my mind: promoting, protecting and caring for children’s mental health [Internet]. New York, NY: UNICEF; 2021 p. 262. (The State of the World’s Children 2021). Available from: https://www.unicef.org/media/114636/file/SOWC-2021-full-report-English.pdf. *On my mind: promoting, protecting and caring for children’s mental health* (2021.0) 262
20. 20World Health Organization (CH). Suicide worldwide in 2019 [Internet]. Geneva, CH: World Health Organization; 2021 [cited 2022 Mar 18] p. 35. (Global Health Estimates). Available from: https://www.who.int/publications-detail-redirect/9789240026643. *Suicide worldwide in 2019* (2021.0) 35
21. Asfahani F, Kadiyala S, Ghattas H. **Food insecurity and subjective wellbeing among Arab youth living in varying contexts of political instability**. *J Adolesc Health* (2019.0) **64** 70-8. DOI: 10.1016/j.jadohealth.2018.08.010
22. Elgar FJ, Sen A, Gariépy G, Pickett W, Davison C, Georgiades K. **Food insecurity, state fragility and youth mental health: A global perspective**. *SSM—Popul Health* (2021.0) **14** 100764. DOI: 10.1016/j.ssmph.2021.100764
23. Watts N, Amann M, Arnell N, Ayeb-Karlsson S, Beagley J, Belesova K. **The 2020 report of The Lancet Countdown on health and climate change: Responding to converging crises**. *The Lancet* (2021.0) **397** P129-170. DOI: 10.1016/S0140-6736(20)32290-X
24. 24Intergovernmental Panel on Climate Change (CH). Climate change 2014: Synthesis report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Internet]. Geneva, CH: Intergovernmental Panel on Climate Change; 2014 p. 151. Available from: https://www.ipcc.ch/site/assets/uploads/2018/02/SYR_AR5_FINAL_full.pdf. *Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change* (2014.0) 151
25. **2020 Global report on food crises: Joint analysis for better decisions**. *Rome IT and Washington DC: Food and Agriculture Organization (FAO), World Food Programme (WFP), and International Food Policy Research Institute (IFPRI)* (2020.0)
26. Beaglehole B, Mulder RT, Frampton CM, Boden JM, Newton-Howes G, Bell CJ. **Psychological distress and psychiatric disorder after natural disasters: Systematic review and meta-analysis**. *Br J Psychiatry* (2018.0) **213** 716-22. DOI: 10.1192/bjp.2018.210
27. Charlson F, Ali S, Benmarhnia T, Pearl M, Massazza A, Augustinavicius J. **Climate change and mental health: A scoping review**. *Int J Environ Res Public Health* (2021.0) **18** 4486. DOI: 10.3390/ijerph18094486
28. Sharpe I, Davison CM. **Climate change, climate-related disasters and mental disorder in low- and middle-income countries: a scoping review**. *BMJ Open* (2021.0) **11** e051908. DOI: 10.1136/bmjopen-2021-051908
29. Smith KRA, Woodward A, Campbell-Lendrum D, Chadee DD, Honda Y, Liu Q, Field CB, Barros VR, Dokken DJ, Mach KJ, Mastrandrea MD, Bilir TE. *Impacts, adaptation, and vulnerability Part A: Global and sectoral aspects, contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change* (2014.0) 709-54
30. Rother HA, Hayward RA, Paulson JA, Etzel RA, Shelton M, Theron LC. **Impact of extreme weather events on Sub-Saharan African child and adolescent mental health: The implications of a systematic review of sparse research findings✰**. *J Clim Change Health* (2022.0) **5** 100087
31. Friel S, Berry H, Dinh H, O’Brien L, Walls HL. **The impact of drought on the association between food security and mental health in a nationally representative Australian sample**. *BMC Public Health* (2014.0) **14** 1102. PMID: 25341450
32. Kousky C.. **Impacts of natural disasters on children**. *Future Child* (2016.0) **26** 73-92
33. 33Gallup worldwide research methodology country dataset details [Internet]. Washington, DC: Gallup, Inc.; 2019 p. 131. Available from: https://www.gallup.com/services/177797/country-data-set-details.aspx. (2019.0) 131
34. 34Gallup worldwide research methodology and codebook [Internet]. Washington, DC: Gallup, Inc.; 2019
Aug p. 75. Available from: https://news.gallup.com/poll/165404/world-poll-methodology.aspx. *Gallup worldwide research methodology and codebook* (2019.0) 75
35. 35United Nations Department of Economic and Social Affairs. Definition of youth [Internet]. n.d. [cited 2020 Oct 31]. Available from: https://www.un.org/esa/socdev/documents/youth/fact-sheets/youth-definition.pdf
36. 36Centre for Research on the Epidemiology of Disasters (CRED), Guha-Sapir D. EM-DAT: the Emergency Events Database—Université catholique de Louvain (UCL)—Brussels, Belgium [Internet]. [cited 2020 May 13]. Available from: https://www.emdat.be/
37. Bettin G, Zazzaro A. **The impact of natural disasters on remittances to low- and middle-income countries**. *J Dev Stud* (2018.0) **54** 481-500
38. Datar A, Liu J, Linnemayr S, Stecher C. **The impact of natural disasters on child health and investments in rural India**. *Soc Sci Med* (2013.0) **76** 83-91. DOI: 10.1016/j.socscimed.2012.10.008
39. Heo J hoon Kim MH, Koh SB, Noh S, Park JH, Ahn JS. **A prospective study on changes in health status following flood disaster**. *Psychiatry Investig* (2008.0) **5** 186-92. DOI: 10.4306/pi.2008.5.3.186
40. Liu A, Tan H, Zhou J, Li S, Yang T, Wang J. **An epidemiologic study of posttraumatic stress disorder in flood victims in Hunan China**. *Can J Psychiatry* (2006.0) **51** 350-4. DOI: 10.1177/070674370605100603
41. Ballard TJ, Kepple AW, Cafiero C. *The Food Insecurity Experience Scale* (2013.0) 61
42. 42Food and Agriculture Organization (IT). Methods for estimating comparable rates of food insecurity experienced by adults throughout the world [Internet]. Rome, IT: Food and Agriculture Organization of the United Nations; 2016 p. 60. Available from: http://www.fao.org/3/a-i4830e.pdf. *Methods for estimating comparable rates of food insecurity experienced by adults throughout the world* (2016.0) 60
43. Cafiero C, Viviani S, Nord M. **Food security measurement in a global context: The food insecurity experience scale**. *Measurement* (2018.0) **116** 146-52
44. Frongillo EA. **Validity and cross-context equivalence of experience-based measures of food insecurity**. *Glob Food Secur* (2022.0) **32** 100599
45. Bland JM, Altman DG. **Statistics Notes: Cronbach’s alpha**. *BMJ* (1997.0) **314** 572. PMID: 9055718
46. Brooks N, Adger WN. *Country level risk measures of climate-related natural disasters and implications for adaptation to climate change* (2003.0)
47. 47World Bank Group. Data | Population, total [Internet]. n.d. [cited 2021 Mar 8]. Available from: https://data.worldbank.org/indicator/SP.POP.TOTL
48. Ward PS, Shively GE. **Disaster risk, social vulnerability, and economic development**. *Disasters* (2017.0) **41** 324-51. DOI: 10.1111/disa.12199
49. Na M, Miller M, Ballard T, Mitchell DC, Hung YW, Melgar-Quiñonez H. **Does social support modify the relationship between food insecurity and poor mental health? Evidence from thirty-nine sub-Saharan African countries**. *Public Health Nutr* (2019.0) **22** 874-81. DOI: 10.1017/S136898001800277X
50. Frongillo EA, Nguyen HT, Smith MD, Coleman-Jensen A. **Food insecurity is more strongly associated with poor subjective well-being in more-developed countries than in less-developed countries**. *J Nutr* (2019.0) **149** 330-5. DOI: 10.1093/jn/nxy261
51. Frongillo EA, Nguyen HT, Smith MD, Coleman-Jensen A. **Food insecurity is associated with subjective well-being among individuals from 138 countries in the 2014 Gallup World Poll**. *J Nutr* (2017.0) **147** 680-7. DOI: 10.3945/jn.116.243642
52. Jones AD. **Food insecurity and mental health status: A global analysis of 149 countries**. *Am J Prev Med* (2017.0) **53** 264-73. DOI: 10.1016/j.amepre.2017.04.008
53. 53Esri. Light Gray Canvas (Local Language) [Internet]. 2022 [cited 2022 Apr 24]. Available from: https://www.arcgis.com/home/item.html?id=ee8678f599f64ec0a8ffbfd5c429c896. *Light Gray Canvas (Local Language)* (2022.0)
54. Knowles M, Rabinowich J, Ettinger de Cuba S, Cutts DB, Chilton M. **“Do you wanna breathe or eat?”: Parent perspectives on child health consequences of food insecurity, trade-offs, and toxic stress**. *Matern Child Health J* (2016.0) **20** 25-32. DOI: 10.1007/s10995-015-1797-8
55. Maynard M, Andrade L, Packull-McCormick S, Perlman CM, Leos-Toro C, Kirkpatrick SI. **Food insecurity and mental health among females in high-income countries**. *Int J Environ Res Public Health* (2018.0). DOI: 10.3390/ijerph15071424
56. Shonkoff JP, Garner AS. **Committee on Psychosocial Aspects of Child and Family Health, Committee on Early Childhood, Adoption, and Dependent Care, Section on Developmental and Behavioral Pediatrics. The lifelong effects of early childhood adversity and toxic stress**. *Pediatrics* (2012.0) **129** e232-46. PMID: 22201156
57. Bucci M, Marques SS, Oh D, Harris NB. **Toxic Stress in Children and Adolescents. Adv Pediatr**. (2016.0) **63** 403-28
58. Joos CM, McDonald A, Wadsworth ME. **Extending the toxic stress model into adolescence: Profiles of cortisol reactivity**. *Psychoneuroendocrinology* (2019.0) **107** 46-58. DOI: 10.1016/j.psyneuen.2019.05.002
59. Firth J, Gangwisch JE, Borsini A, Wootton RE, Mayer EA. **Food and mood: How do diet and nutrition affect mental wellbeing?**. *BMJ* (2020.0) **369** m2382. DOI: 10.1136/bmj.m2382
60. Kaplan BJ, Crawford SG, Field CJ, Simpson JSA. **Vitamins, minerals, and mood**. *Psychol Bull* (2007.0) **133** 747-60. DOI: 10.1037/0033-2909.133.5.747
61. Kwak K.. **Adolescents and Their Parents: A Review of Intergenerational Family Relations for Immigrant and Non-Immigrant Families**. *Hum Dev* (2003.0) **46** 115-36
62. Wellman B, Wortley S. **Different Strokes From Different Folks: Community Ties and Social Support**. *Am J Sociol—AMER J SOCIOL* (1990.0) **96**
63. Koyanagi A, Stubbs B, Oh H, Veronese N, Smith L, Haro JM. **Food insecurity (hunger) and suicide attempts among 179,771 adolescents attending school from 9 high-income, 31 middle-income, and 4 low-income countries: A cross-sectional study**. *J Affect Disord* (2019.0) **248** 91-8. DOI: 10.1016/j.jad.2019.01.033
64. McIntyre L, Wu X, Kwok C, Patten SB. **The pervasive effect of youth self-report of hunger on depression over 6 years of follow up**. *Soc Psychiatry Psychiatr Epidemiol* (2017.0) **52** 537-47. DOI: 10.1007/s00127-017-1361-5
65. Romo ML, Abril-Ulloa V, Kelvin EA. **The relationship between hunger and mental health outcomes among school-going Ecuadorian adolescents**. *Soc Psychiatry Psychiatr Epidemiol* (2016.0) **51** 827-37. DOI: 10.1007/s00127-016-1204-9
66. Blakely TA, Woodward AJ. **Ecological effects in multi-level studies**. *J Epidemiol Community Health* (2000.0) **54** 367-74. DOI: 10.1136/jech.54.5.367
|
---
title: 'Cardiovascular risk factors and outcomes in COVID-19: A hospital-based study
in India'
authors:
- Arvind K. Sharma
- Vaseem Naheed Baig
- Sonali Sharma
- Gaurav Dalela
- Raja Babu Panwar
- Vishwa Mohan Katoch
- Rajeev Gupta
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021757
doi: 10.1371/journal.pgph.0000234
license: CC BY 4.0
---
# Cardiovascular risk factors and outcomes in COVID-19: A hospital-based study in India
## Abstract
### Background & objectives
Presence of cardiovascular (CV) risk factors enhance adverse outcomes in COVID-19. To determine association of risk factors with clinical outcomes in India we performed a study.
### Methods
Successive virologically confirmed adult patients of COVID-19 at a government hospital were recruited at admission and data on clinical presentation and in-hospital outcomes were obtained. The cohort was classified according to age, sex, hypertension, diabetes and tobacco use. In-hospital death was the primary outcome. Logistic regression was performed to compared outcomes in different groups.
### Results
From April to September 2020 we recruited 4645 (men 3386, women 1259) out of 5103 virologically confirmed COVID-19 patients ($91.0\%$). Mean age was 46±18y, hypertension was in $17.8\%$, diabetes in $16.6\%$ and any tobacco-use in $29.5\%$. Duration of hospital stay was 6.8±3.7 days, supplemental oxygen was in $18.4\%$, non-invasive ventilation in $7.1\%$, mechanical ventilation in $3.6\%$ and $7.3\%$ died. Unadjusted and age-sex adjusted odds ratio(OR) and $95\%$ confidence intervals(CI) for in-hospital mortality, respectively, were: age ≥60y vs <40y, OR 8.47($95\%$ CI 5.87–12.21) and 8.49(5.88–12.25), age 40-59y vs <40y 3.69(2.53–5.38) and 3.66(2.50–5.33), men vs women 1.88(1.41–2.51) and 1.26(0.91–1.48); hypertension 2.22(1.74–2.83) and 1.32(1.02–1.70), diabetes 1.88(1.46–2.43) and 1.16(0.89–1.52); and tobacco 1.29(1.02–1.63) and 1.28(1.00–1.63). Need for invasive and non-invasive ventilation was greater among patients in age-groups 40–49 and ≥60y and hypertension. Multivariate adjustment for social factors, clinical features and biochemical tests attenuated significance of all risk factors.
### Conclusion
Cardiovascular risk factors, age, male sex, hypertension, diabetes and tobacco-use, are associated with greater risk of in-hospital death among COVID-19 patients.
## Introduction
Presence of cardiovascular risk factors (smoking, diabetes, hypertension, sedentary lifestyle and obesity, etc.) and clinical cardiovascular disease are associated with adverse outcomes in COVID-19 [1]. It has been reported that these factors lead to rapid progression of clinical manifestations, more severe pulmonary disease, greater requirement for oxygen and ventilatory support and greater mortality [1, 2]. It has also been reported that presence of hypertension, diabetes and cardiovascular disease is associated with a two-fold increase in risk of death in COVID-19 [2]. In a meta-analysis of 109 studies and 20,296 patients, the risk of mortality was higher in patients with increasing age, male sex (relative risk, RR 1.45, $95\%$ confidence intervals, CI, 1.23–1.71), diabetes (1.59, 1.41–1.78), hypertension, tobacco use and congestive heart failure (4.76, 1.34–16.97) [3]. In another meta-analysis of 45 studies with 18,300 patients a significant association of in-hospital death was observed with age (coefficient 1.06, $95\%$ CI 1.04–1.09), diabetes (1.04, 1.02–1.07) and hypertension (1.01, 1.01–1.03), but remained significant only for diabetes after statistical adjustment [4]. A third meta-analysis of 51 studies and 48,317 patients, mostly from high and upper middle income countries, reported relative risk of developing severe disease or deaths as significantly higher in patients with hypertension (RR 2.50, $95\%$ CI 2.15–2.90), diabetes (2.25, 1.89–2.69) and cardiovascular disease (3.11, 2.55–3.79); the risk being significantly greater in older than younger individuals [5]. Population-based studies have identified importance of cardiovascular risk factors and disease in COVID-19 incidence and outcomes [6, 7].
Burden of COVID-19 related mortality is high in lower-middle and low-income countries of Asia and Africa [8]. Prevalence of cardiovascular risk factors is high in these countries.9 India has high burden of cardiovascular risk factors [9], and COVID-19 cases and deaths [10]. There are limited data on association of cardiovascular risk factors with disease incidence and outcomes in India [11–13]. A macrolevel study in India reported that states with higher prevalence of cardiovascular risk factors- aging, hypertension, diabetes and obesity- had significantly higher COVID-19 incidence and deaths [14]. Although higher age is well-known COVID-19 risk factor, controversy exists regarding relative importance of risk factors- hypertension, diabetes or tobacco use [2]. Therefore, to evaluate association of cardiovascular risk factors (age, male sex, hypertension, diabetes and any tobacco use) in virologically confirmed COVID-19 cases successively admitted to a government hospital in India we performed a registry-based study.
## Methods
We conducted a hospital based observational study on patients with laboratory confirmed COVID-19 admitted to a 1200-bed dedicated COVID-19 government hospital (Rajasthan University of Health Sciences Hospital, Jaipur, India) from April to mid-September 2020. Initial data on some of these patients have been reported earlier [15–17]. The registry has been approved by the college administration and institutional ethics committee (CDSCO Registration Number: CR/762/Inst/RJ/2015). It is registered with Clinical Trials Registry of India at www.ctri.nic.in with number REF/$\frac{2020}{06}$/034036. Individual patient consent was waivered by the ethics committee and anonymized data have been used with no patient identifiers.
## Patient data
Successive adult patients (more than 18 years) presenting to the hospital for admission with suspicion of COVID-19 infection were enrolled in the study and those who tested positive for COVID-19 on nasopharyngeal and oropharyngeal reverse transcriptase-polymerase chain reaction (RT-PCR) test were included. Details of methodology have been reported [16]. All RT-PCR positive patients admitted from 1 April to 15 September have been included. Patients recruited into the study in mid-September were followed up to discharge or death and outcome events were recorded.
A questionnaire was developed and details of sociodemographic, clinical, laboratory, treatments and outcomes variables were recorded using patient-history and medical files. Demographic details were obtained at the time of admission. These included name, age, sex, residence address, educational status. History of tobacco use (smoking, smokeless tobacco) and past hypertension, diabetes, cardiovascular diseases and other chronic diseases was also recorded at admission. Current smokers and users of smokeless tobacco were categorized as tobacco use. Hypertension and diabetes were diagnosed from history of known disease or on treatment. Details of physical examination at the time of admission were obtained from patient case files. These included history of duration of symptoms at admission, pulse, blood pressure (BP), respiratory rate, and surface oxygen concentration (SpO2) using digital devices. We could not obtain details of body mass index as height and weight was not routinely recorded on admission. Details of investigations at admission were obtained from the case files and biochemistry, microbiology and pathology departments as reported earlier [15, 16]. We do not have data on serial investigations. We obtained data on duration of hospital stay from medical record department. For patients discharged alive from the hospital, we obtained data on number of patients who needed oxygen support (nasal prongs, facial mask or high-flow nasal cannula), non-invasive ventilation (CPaP or BiPaP support) or invasive ventilation after endotracheal intubation. Binary outcomes were obtained for all patients and included recovery, referral to non-government hospitals on request of family, or death. In-hospital death was the primary outcome while requirement for invasive and non-invasive ventilation were secondary outcomes. All these data are being routinely sent to the Department of Health, Government of Rajasthan, India, but are not currently accessible.
## Statistical analyses
The data were computerized and processing was performed using commercially available statistical software, SPSS v.20.0. Numerical data are expressed as numbers ±1 SD and categorical data as percent. Significance of intergroup differences were calculated using unpaired t-test or ANOVA for continuous variables and χ2 test for categorical variables. To evaluate association of COVID-19 related in-hospital deaths and other adverse outcomes (invasive ventilation, non-invasive ventilation) with age, male sex, hypertension, diabetes and tobacco use, we performed stepwise logistic regression. Univariate and multivariate odds ratios (OR) and $95\%$ confidence intervals ($95\%$ CI) were calculated. In the first step we calculated univariate odds ratio. Age- and sex-adjusted odds ratios were calculated in the second step. For multivariate adjusted odds ratios we added household size, educational status, comorbidities, risk factors (other than the risk factor in question) and clinical severity (oxygen level at admission, need for oxygenation) and calculated OR and $95\%$ CI. P value <0.05 is considered significant.
## Results
Data were obtained from April to mid-September 2020. During this period, a total of 7349 patients were hospitalized with confirmed or suspected COVID-19, 5103 patients ($69.0\%$) tested positive for the disease on RT-PCR test and for the present study 4645 adult men and women ≥18 years ($91.0\%$ of confirmed cases), men 3386 ($72.9\%$) and women 1259 ($27.1\%$), in whom detailed clinical data were available have been included. The mean age of the cohort was 46±18 years, $54\%$ were less than 50 years and about half lived in large family households (>5 persons). Prevalence of low educational status was higher in women while tobacco use was more in men. Comorbidities were present in $28.6\%$ with hypertension ($17.8\%$) and diabetes ($16.6\%$) being the most common. Other comorbidities were chronic pulmonary disease, tuberculosis, coronary heart disease and neurological disease (Fig 1). Data on hematological investigations were available in 4456 ($95.9\%$) and for biochemical tests in 867 ($18.7\%$). All patients received standard treatment according to guidelines available from Indian Council of Medical Research and the State government [18]. The average length of stay in hospital was 6.8±3.7 days. Oxygen requirement was in 861 ($18.4\%$), non-invasive ventilation or high flow oxygen in 334 ($7.1\%$) and mechanical ventilation in 169 ($3.6\%$). In-hospital mortality was in 340 patients ($7.3\%$).
**Fig 1:** *Distribution of cardiovascular risk factors among men and women in the study cohort.*
Clinical characteristics, important clinical findings, selected investigations and outcomes in patients aged >40 years, 40–59 years and ≥60 years and are shown in Table 1. Older participants were less educated, lived in larger households and had greater prevalence of tobacco use, hypertension, diabetes, and cardiovascular disease with significantly higher systolic blood pressure (BP) and more hypoxia (Table 1). There were no differences in hematological and biochemical parameters. Oxygen requirement, non-invasive as well as invasive ventilation were more in older age-groups (40–59 and ≥60 years) with graded escalation. Deaths were significantly higher in age-group 40–59 ($7.1\%$) and ≥60 ($15.0\%$) when compared to <40 years ($2.1\%$) (Fig 2). Table 2 shows that women were less literate and had lower prevalence of hypertension, diabetes and tobacco use. Oxygen requirement was significantly more in women while requirement of non-invasive or invasive ventilation were not different. Hematological and biochemical parameters were not significantly different and are not shown. Number of in-hospital deaths were significantly more in men ($$n = 282$$, $8.3\%$) as compared to women ($$n = 58$$, $4.6\%$) (Fig 2) with univariate OR 1.88, ($95\%$ CI 1.41–3.51).
**Fig 2:** *In-hospital deaths in COVID-19 patients with or without the risk factor.* TABLE_PLACEHOLDER:Table 1 TABLE_PLACEHOLDER:Table 2 Risk factors, clinical findings and outcomes in patients with hypertension, diabetes and smoking/tobacco use are shown in Tables 3–5. Hematological and biochemical parameters did not show significant inter-group differences (data not shown). Patients with known hypertension were older and had higher prevalence of diabetes, cardiovascular disease, hypothyroidism and smoking/tobacco (Table 3). Need for non-invasive ventilation was also more in hypertension (OR 1.82, CI 1.41–2.35). Deaths were significantly greater in patients with hypertension ($12.6\%$) versus in those without hypertension ($6.1\%$) (Fig 2) (univariate OR 2.22, CI 1.74–2.83). Patients with diabetes were older with greater prevalence of smoking/tobacco, hypertension, pulmonary disease and cardiovascular disease with higher admission BP (Table 4). As compared to non-diabetics, number of deaths were significantly greater in patients with diabetes ($$n = 89$$, $11.5\%$) versus without diabetes ($$n = 251$$, $6.4\%$) (Fig 2) (univariate OR 1.88, CI 1.46–2.43). Smoking/tobacco-user group (smokers, smokeless tobacco) had more men, with greater prevalence of hypertension, diabetes and cardiovascular disease (Table 5). Compared with non-tobacco users the need for oxygen (OR 1.23, CI 1.05–1.44) and non-invasive ventilation (OR 1.31, CI 1.04–1.66). Deaths were significantly among tobacco users ($$n = 117$$ ($8.5\%$) as compared to non-tobacco users ($$n = 223$$, $6.7\%$) (Fig 2) with OR 1.29 ($95\%$ CI 1.02–1.63).
We also performed age-sex adjusted and multivariate analyses to determine association of various cardiovascular risk factors with COVID-19 related in-hospital mortality (Table 6) and other outcomes (Table 7). Age ≥60 years vs <40 years emerged as the most important risk factor with significantly greater deaths on univariate (OR 8.47, $95\%$ CI 5.87–12.21), sex-adjusted (OR 8.49, CI 5.88–12.25) and multivariate analyses (OR 7.25, CI 4.92–10.66). In age-group 40–59 years also deaths were significantly higher than <40 y on univariate and multivariate analyses (Table 6). On univariate analyses (Table 6), male sex (OR 1.88, 1.41–2.51), hypertension (2.22, 1.74–2.83), diabetes (1.88, 1.46–2.43) and tobacco (1.29, 1.02–1.63) were associated with significantly more deaths ($p \leq 0.001$) (Fig 3). There was moderate attenuation of significance with age and sex adjusted analyses, but hypertension (1.32, 1.02–1.70) and tobacco use (1.28, 1.00–1.63) continued to be significant. Following multivariate analyses significance of all the risk factors completely attenuated (Fig 3). Analyses of secondary outcomes show that in patients age ≥60 years as well in age-group 40–59 years, compared with <40 years, the need for invasive ventilation as well as non-invasive ventilation were higher (Table 7). Hypertension was significantly associated with greater risk of invasive ventilation on univariate and adjusted analyses and greater risk of non-invasive ventilation on univariate analyses. Diabetes patients had greater risk of non-invasive and invasive ventilation on univariate analyses which attenuated on age-sex adjusted and multivariate analyses (Table 7).
**Fig 3:** *Odds ratio and $95\%$ confidence intervals for COVID-19 related deaths in patients in old vs young, men vs women, hypertension, diabetes and tobacco groups on univariate (black markers), age and sex adjusted (grey markers) and multivariate (open markers) logistic regression.* TABLE_PLACEHOLDER:Table 6 TABLE_PLACEHOLDER:Table 7
## Discussion
This study shows that multiple cardiovascular risk factors- hypertension, diabetes, tobacco use, increasing age and male sex are associated with greater risk of death and adverse outcomes among hospitalized COVID-19 patients on univariate analyses. Age is the most important predictor and there is graded increment of deaths and other adverse outcomes with increasing age. Significance of hypertension and tobacco-use is retained even after age and sex adjustment highlighting greater importance of these factors.
Our results are similar to most of the previous meta-analyses that have identified age as the most important risk factor for adverse outcomes in COVID-19 [2–5]. Cardiovascular risk factors- hypertension and diabetes- have been identified as important in many previous studies [3–5]. In the present study, although both of these factors are associated with greater risk of death and some adverse outcomes (Table 6), there is substantial attenuation after age-adjustment for hypertension and complete attenuation in diabetes. These findings indicate that age is important intermediate pathway of increased risk-factor associated mortality in COVID-19. A study limitation is that we used self-reported presence of hypertension and diabetes at the time of admission as risk factor. Given the fact that in India only about half of the patients with hypertension and two-thirds with diabetes are aware of their condition [19], the prevalence of these conditions might have been higher in our cohort. However, many individuals with hypertension present with low BP in acute COVID-19 and therefore estimation of prevalence of hypertension based on measured BP would have been erroneous. Moreover, our unadjusted OR of 2.22 (CI 1.74–2.51) and age-sex adjusted OR of 1.32 (CI 1.02–1.70) is similar to many previous studies and meta-analyses have calculated hypertension related OR in COVID-19 between 1.90 (CI 1.69–2.35) [3] and 2.50 (CI 2.15–2.90) [5], similar to the present study. We did not inquire the type of anti-hypertensive patients in our study cohort. Certain BP medications such as renin-angiotensin system blockers are known to be useful in COVID-19 [20, 21].
Previous meta-analyses including studies from India have identified diabetes as equally important as hypertension for adverse COVID-19 related outcomes [11, 12]. In the present study the unadjusted OR for diabetes and deaths were 1.88 (CI 1.46–2.43), however, the risk significantly attenuated after age and sex adjustment to OR 1.16 (CI 0.89–1.52) which is different from the previous studies. In the present study, we included patients with known diabetes only and this is a study limitation [22]. It is likely that using biomarkers for diabetes diagnosis (HbA1c, glucose tolerance test, etc.) we would have diagnosed more individuals with diabetes, but these criteria are fraught with inconsistency during any acute illness. We also found significant association of smoking/tobacco use with death and other adverse outcomes in our cohort. This association holds even after multivariate adjustments and shows that tobacco is an important risk factor. This is different from some previous studies that have reported disparate results [23]. On the other hand, a large meta-analysis that included 109 studies with 517,020 patients reported that smoking was associated with increased risk of admission to ICU and increased mortality (OR 1.58, CI 1.38–1.81) [24]. Meta-regression analyses identified that the increased risk of smoking was mediated via increased age, hypertension and diabetes. Our finding is similar to data from Chinese cohorts (high rates of smoking) in the aforementioned meta-analysis [24].
The study has several other limitations. This is a single-centre study and the results may not be externally valid within India or other countries as *Rajasthan is* one of the less-developed states in the nation with lower prevalence of hypertension and diabetes [19, 25]. On the other hand, this is the largest study from India and much larger than many other studies from developed countries, the data were obtained from a government hospital thus assuring wider population representation and better data granularity. Secondly, this is not a population-based study as many studies from Europe and North America [7], and we may have missed data on milder forms of disease. Thirdly, we do not have data on obesity or body-mass index which is an important COVID-19 risk factor in hospital- and population-based studies [7, 24]. Fourthly, we also do not have data on biochemical investigations for all the patients, although data on white cell count are available for more than $90\%$. Also, we do not have data of radiological evaluation of all the patients as it is well known that computerized tomographic images provide important prognostic information [26]. Fifthly, the rate of progression of illness as well as greater details of causes of deaths are not available and this is a study limitation as discussed earlier [17]. We did not analyze data on patients with known cardiovascular disease, chronic respiratory disease, cancers and chronic kidney disease because of small numbers of these patients (Fig 1). And finally, we did not obtain data regarding post-COVID syndrome which is emerging as important health problem especially in persons with comorbidities [27]. On the other hand, this is the largest study from India and with robust data has important clinical implications, especially in view of the ongoing third wave in the country [8].
In conclusion, this study shows that older patients, males, and those with hypertension, diabetes and any tobacco use have greater risk of death and adverse outcomes from COVID-19. Attenuation of risk with age-adjustment shows that increasing age is the most important factor of risk. It is recommended that individuals with cardiovascular risk factors, especially older men and women, should be focus of public health measures and must be informed regarding increased risk of death in COVID-19. Moreover, these high risk individuals must aggressively follow all non-pharmacological physical measures for prevention [28]. These groups should also be prioritized for primary vaccinations and vaccine-boosters [29]. Clinicians are advised to seek early evidence of deterioration of pulmonary function and signs of cardiovascular and extrapulmonary manifestation of acute COVID-19 in these patients and provide optimum management [30, 31]. It is likely that with proper preventive and therapeutic interventions the higher risk of adverse outcomes in COVIDF-19 patients with cardiovascular risk factors can be mitigated.
## References
1. Nishiga M, Wang DW, Lewis DB, Wu JC. **COVID-19 and cardiovascular disease: from basic mechanisms to clinical perspectives**. (2020.0) 543-558. DOI: 10.1038/s41569-020-0413-9
2. Matsushita K, Ding N, Kou M, Hu X, Chen M, Gao Y. **The relationship of COVID-19 severity with cardiovascular disease and its traditional risk factors: a systematic review and meta-analysis**. (2020.0) **15** 64-81. DOI: 10.5334/gh.814
3. Chidambaram V, Tun NL, Haque WZ, Majella MG, Sivakumar RK, Kumar A. **Factors associated with disease severity and mortality among patients with COVID-19: a systematic review and meta-analysis.**. (2020.0) **15** 0241541. DOI: 10.1371/journal.pone.0241541
4. Silverio A, Maio MD, Citro R, Esposito L, Iuliano G, Bellino M. **Cardiovascular risk factors and mortality in hospitalized patients with COVID-19: a systematic review and meta-analysis of 45 studies and 18300 patients**. (2021.0) **21** 23. DOI: 10.1186/s12872-020-01816-3
5. Bae SA, Kim SR, Kim MN, Shim WJ, Park SM. **Impact of cardiovascular disease and risk factors on fatal outcomes in patients with COVID-19 according to age: a systematic review and meta-analysis**. (2021.0) **107** 373-380. DOI: 10.1136/heartjnl-2020-317901
6. Collard D, Nurmohamed NS, Kaiser Y, Reeskamp LF, Dormans T, Moeniralam H. **Cardiovascular risk factors and COVID-19 outcomes in hospitalized patients: a prospective cohort study**. (2021.0) **11** 045482. DOI: 10.1136/bmjopen-2020-045482
7. Gao M, Piernas C, Astbury NM, Hippisley-Cox J, O’Rahilly S, Aveyard P. **Associations between body mass index and COVID-19 severity in 6.9 million people in England: a prospective, community-based, cohort study**. (2021.0) **9** 350-359. DOI: 10.1016/S2213-8587(21)00089-9
8. Ritchie H, Mathieu E, Rodes-Guirao L, Appel C, Giattino C, Ortiz-Ospina E. *Coronavirus pandemic (COVID-19).* (2022.0)
9. **Global burden of cardiovascular diseases and risk factors, 1990–2019: Update from the Global Burden of Disease 2019 Study.**. (2020.0) **76** 2982-3021. DOI: 10.1016/j.jacc.2020.11.010
10. Ritchie H, Mathieu E, Rodes-Guirao L, Appel C, Giattino C, Ortiz-Ospina E. **India: Coronavirus Pandemic Country Profile.**. (2022.0)
11. Singh AK, Gilles CL, Singh R, Singh A, Chudasma Y, Coles B. **Prevalence of co-morbidities and their association with mortality in patients with COVID-19: a systematic review and meta-analysis**. (2020.0) **22** 1915-1924. DOI: 10.1111/dom.14124
12. Nandy K, Salunke A, Pathak SK, Pandey A, Doctor C, Punj K. **Coronavirus disease (COVID-19): a systematic review and meta-analysis to evaluate the impact of various comorbidities and serious events.**. (2020.0) **14** 1017-1025. DOI: 10.1016/j.dsx.2020.06.064
13. Chakafana G, Mutithu D, Hoevelmann J, Ntusi N, Sliwa K. **Interplay of COVID-10 and cardiovascular diseases in Africa: an observational snapshot**. (2020.0) **109** 1460-1468. DOI: 10.1007/s00392-020-01720-y
14. Gaur K, Khedar RS, Mangal K, Sharma AK, Dhamija RK, Gupta R. **Macrolevel association of COVID-19 with non-communicable disease risk factors in India.**. (2021.0) **15** 343-350. DOI: 10.1016/j.dsx.2021.01.005
15. Sharma AK, Ahmed A, Baig VN, Dhakar P, Dalela G, Kacker S. **Characteristics and outcomes of hospitalized young adults with mild to moderate COVID-19 at a university hospital in India.**. (2020.0) **68** 62-65. PMID: 32738843
16. Sharma S, Sharma AK, Dalela G, Dhakar P, Singh TV, Baig VN. **Association of SARS CoV-2 cycle threshold (Ct) with clinical outcomes: a hospital-based study.**. (2021.0) **69** 86-90
17. Sharma AK, Gupta R, Baig VN, Singh TV, Chakraborty S, Sunda JP. **Educational status and COVID-19 related outcomes in India: Hospital-based cross-sectional study**. (2022.0). DOI: 10.1136/bmjopen-2021-055403
18. 18Government of India, Ministry of Health and Family Welfare. Clinical management protocol: COVID-19. Available at: http://www.rajswasthya.nic.in/PDF/COVID%20-19/FOR%20HOSPITALS/27.06.2020.pdf. Accessed 30 April 2021.
19. Gupta R, Gaur K, Ram CVS. **Emerging trends in hypertension epidemiology in India.**. (2019.0) **33** 575-587. DOI: 10.1038/s41371-018-0117-3
20. Savoia C, Volpe M, Kreutz R. **Hypertension, a moving target in COVID-19: current views and perspectives**. (2021.0) **128** 1062-1079. DOI: 10.1161/CIRCRESAHA.121.318054
21. Tavares CAM, Bailey MA, Girardi ACC. **Biological context linking hypertension and higher risk for COVID-19 severity.**. (2020.0) **11** 599729. DOI: 10.3389/fphys.2020.599729
22. Gupta A, Gupta R, Sharma KK, Lodha S, Achari V, Asirvatham AJ. **Prevalence of diabetes and cardiovascular risk factors in middle-class urban populations in India.**. (2014.0) **2** e000048. DOI: 10.1136/bmjdrc-2014-000048
23. Dorjee K, Kim H, Bonomo E, Dolma R. **Prevalence and predictors of death and severe disease in patients hospitalized due to COVID-19: a comprehensive systematic review and meta-analysis of 77 studies and 38,000 patients.**. (2020.0) **15** e0243191. DOI: 10.1371/journal.pone.0243191
24. Zhang H, Ma S, Han T, Qu G, Cheng C, Uy JP. **Association of smoking history with severe and critical outcomes in COVID-19 patients: a systematic review and meta-analysis.**. (2021.0) **43** 101313
25. Gupta R, Gaur K. **Epidemiology of ischemic heart disease and diabetes in India: An overview of the twin epidemic**. (2020.0) **17** e100620186664
26. Machnicki S, Patel D, Singh A, Talwar A, Mina B, Oks M. **The usefulness of chest CT imaging in patients with suspected or diagnosed COVID-19: a review of literature.**. (2021.0) **160** 652-670. DOI: 10.1016/j.chest.2021.04.004
27. Nalbandian A, Sehgal K, Gupta A, Madhavan MV, McGroder C, Stevens JS. **Post-acute COVID-19 syndrome**. (2021.0) **27** 601-615. DOI: 10.1038/s41591-021-01283-z
28. Kucharski A, Klepac P, Conlan AJ, Kissler SM, Tang ML, Fry H. **Effectiveness of isolation, testing, contact tracing, and physical distancing on reducing transmission of SARS-CoV-2 in different settings: a mathematical modelling study**. (2020.0) **20** 1151-1160. DOI: 10.1016/S1473-3099(20)30457-6
29. Schmidt H, Weintraub R, Williams MA, Miller K, Buttenheim A, Sadecki E. **Equitable allocation of COVID-19 vaccines in the United States**. (2021.0) **27** 1298-1307. DOI: 10.1038/s41591-021-01379-6
30. Azevodo RB, Botelho BG, de Hollanda JVG, Ferreira LVL, de Andrade LZJ, Oei SSML. **COVID-19 and the cardiovascular system: a comprehensive review.**. (2021.0) **35** 4-11. PMID: 32719447
31. Gupta A, Madhavan MV, Sehgal K, Nair N, Mahajan S, Sehrawat TS. **Extrapulmonary manifestations of COVID-19**. (2020.0) **26** 1017-1032. DOI: 10.1038/s41591-020-0968-3
|
---
title: 'Vaccine rollout strategies: The case for vaccinating essential workers early'
authors:
- Nicola Mulberry
- Paul Tupper
- Erin Kirwin
- Christopher McCabe
- Caroline Colijn
journal: PLOS Global Public Health
year: 2021
pmcid: PMC10021761
doi: 10.1371/journal.pgph.0000020
license: CC BY 4.0
---
# Vaccine rollout strategies: The case for vaccinating essential workers early
## Abstract
In vaccination campaigns against COVID-19, many jurisdictions are using age-based rollout strategies, reflecting the much higher risk of severe outcomes of infection in older groups. In the wake of growing evidence that approved vaccines are effective at preventing not only adverse outcomes, but also infection, we show that such strategies are less effective than strategies that prioritize essential workers. This conclusion holds across numerous outcomes, including cases, hospitalizations, Long COVID (cases with symptoms lasting longer than 28 days), deaths and net monetary benefit. Our analysis holds in regions where the vaccine supply is limited, and rollout is prolonged for several months. In such a setting with a population of 5M, we estimate that vaccinating essential workers sooner prevents over 200,000 infections, over 600 deaths, and produces a net monetary benefit of over $500M.
## Introduction
As of March 2021, several vaccines have received widespread approval for use against COVID19 [1]. Vaccination rollouts are underway in much of the world, but the quantity of vaccine doses available differs greatly between jurisdictions. The question of how to deploy vaccines, taking into account their efficacy in preventing symptomatic disease, their efficacy in blocking transmission, and the demographics and underlying contact structure of the population, poses substantial and ongoing challenges. Many jurisdictions are using primarily age-based rollout strategies, where the oldest are vaccinated first and the youngest last, with the rationale that the risk of severe outcomes from COVID-19 steeply increases with age. Such strategies may appear optimal if we are only considering the ability of the vaccines to prevent illness, and/or if everyone is likely to be exposed.
Data on the reduction of infection and transmission from vaccination are accumulating as more countries deploy vaccines. Recent phase 3 trials have shown that the Moderna vaccine [2], the PfizerBioNTech vaccine [3], and the AstraZeneca [4] vaccine are effective at preventing symptomatic infection and severe illness. There are broadly two ways that a vaccine can attain this kind of result: either by preventing infection from occurring in the first place (known as “sterilizing immunity”) or allowing infection but preventing disease [5]. In the first case the vaccine necessarily prevents onward transmission, but in the latter case the vaccine may or may not prevent subsequent transmission, depending in part on whether it decreases viral load. The emerging data show both a high rate of sterilizing immunity and a reduction of viral load in the minority of those vaccinated who do become infected. A $\frac{2}{3}$ reduction in infection was observed in asymptomatic infection [6], which together with the already documented reduction in symptomatic cases [2] gives a high overall reduction in infection. Similar preliminary results are emerging for the Pfizer vaccine [7,8] and for the AstraZeneca vaccine [9]. As well, substantial reduction in viral loads have been found among those who become infected after receiving either the AstraZeneca vaccine [10] or the BioNTech/Pfizer BNT162b2 vaccine [11,12], suggesting reduced transmission even when infection does occur.
A natural goal for minimizing the impact of the pandemic is to prevent as many deaths due to COVID-19 as possible. This is a primary motivation for vaccination plans that start with the oldest individuals and then go down through the age cohorts, since risk of mortality increases sharply with age [13]. But another important consideration is that, in a small but significant number of cases at all ages, COVID causes long-lasting symptoms that can be debilitating [14]. There is a syndrome that has come to be known as Long COVID [15]: extreme fatigue and other COVID symptoms that may last for weeks or months [16], and may turn out to be chronic to the best of our knowledge now [14]. The symptoms are similar to those described by survivors of SARS [17], and fit the clinical definition of myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) [18]. In addition, there are COVID complications [19]: severe secondary conditions caused by COVID-19 infection, including diabetes [20], organ damage [21,22], and neurological and psychiatric disorders [23]. These can also be caused by other severe viral infections such as SARS [17,24] and *Zika virus* [25,26], and early evidence suggests that they are not rare for hospitalized COVID cases. Both of these outcomes, Long COVID and COVID complications, which we will together refer to as chronic outcomes, are likely to be lasting and serious consequences of the pandemic for many individuals [27]. Accordingly, vaccination strategies need to factor them into consideration, despite uncertainty in the duration, severity, and frequency of their occurrence.
There is an emerging literature on optimizing vaccine rollout for COVID-19 given limited vaccine resources. An important early study by Bubar et al [28] modeled the pandemic using an age-stratified SEIR model and compared five vaccine prioritization strategies. They found that targeting 20–49-year-olds reduced the overall number of infections, but led to higher mortality among the elderly. Their results stem from the conditions under which their model is run, which allow for a higher prevalence than has been reached in many jurisdictions where non-pharmaceutical interventions (NPIs) are deployed before hospital capacity is exceeded. Matrajt et al [29] use modelling to explore age-based vaccination during vaccine rollout with no NPIs in place, similarly therefore not accounting for reductions in transmission due to either ongoing or repeatedly-introduced measures reducing case numbers. Both these studies [28,29] do not include essential worker categories. Another approach is that of Chen et al [30] who used an agent-based model with a detailed social contact network to study vaccine prioritization, with Virginia as an example. They found that targeting individuals with many contacts rather than a purely age-based strategy lead to substantially better outcomes. Jentsch et al [31] also do not consider essential workers, but supports vaccinating younger age groups earlier under some circumstances. Other studies are also split, and favour vaccinating essential workers early under some circumstances [32,33] or a primarily age-based prioritization in others [34].
We reassess vaccine rollout strategies in the light of the emerging data on vaccine efficacy and in light of the importance of chronic outcomes of COVID-19 infection. We compare age-based vaccine rollouts with strategies that prioritize those we call essential workers. Throughout this work, we use the phrase “essential workers” to mean those who have high contact at work. This is distinct from “essential services” and could include teachers, taxi drivers, retail workers, food production workers, law enforcement and public safety, first responders, social workers, agriculture, transportation and many more [35]. Our analysis is differentiated by our contextualisation to settings where COVID prevalence is maintained by NPIs, our consideration of chronic outcomes of COVID, and our use of the Net Benefit framework.
## Methods
We use an age- and contact-structured compartmental model to investigate the impact of different vaccination strategies in British Columbia, incorporating chronic outcomes of COVID-19 infection, along with infections, hospitalizations and deaths. The model has susceptible, exposed, infectious and recovered individuals and was originally developed to explore vaccination by age, considering “leaky” or “all or nothing” vaccination and taking existing seroprevalence into account [28]. See Section S1 Text in the Supplementary Material for details and parameters of the model. Our model is an extension of the framework introduced by Bubar et al [28], but now considers both age and essential worker status; in addition, we explore Long COVID and chronic outcomes. Furthermore, we attach health economic outputs by applying a Net Benefit framework, including the expected cost due to hospitalization and chronic conditions resulting from infection, as well as decrements in health utility, which are measured using quality-adjusted life years (QALYs).
In total, we model 15 population groups by age and work status {0–9, 10–19, 20–29, …, 70–79, 80+, 20-29e, 30-39e, …, 70-79e}, where the e superscript denotes an “essential worker” group. Our model of vaccine efficacy works in two stages: a vaccine may prevent infection altogether, but even when it does not, it may still prevent severe outcomes (symptomatic disease, hospitalization, death and chronic outcomes). We have made the code necessary to reproduce our results publicly available at https://github.com/nmulberry/essential-workers-vaccine.
Our simulation approach is motivated by the vaccination programs in British Columbia, across Canada, and in similar jurisdictions. Such jurisdictions have had a relatively small portion of the population naturally infected at the time of writing, and have begun vaccination with the very elderly during a time when social distancing measures have kept the reproduction number low while those over 80 years of age are vaccinated. We initialize our simulations using reported case counts from British Columbia. In line with observed events, we hold R at 1.05 from January 1, 2021 for 60 days while those aged 80+ are vaccinated. After 60 days we raise R, typically to 1.15, 1.3, 1.5 in the main text and with some higher examples in the supplement. This rise in transmission models either relaxation of distancing measures, reduced compliance to widespread distancing measures, or rising frequencies of higher-transmission COVID-19 variants of concern [36]. During the next 210 simulated days, we proceed with the specified vaccination scenario. Both among 80+ and the other age groups we model age-specific hesitancy, with some portion of each age group declining the vaccine.
We consider five vaccination scenarios for the period of time after the 80+ age group is vaccinated. In scenario A, available vaccines are distributed to age groups in order of decreasing age. In scenario B, after the 80+ group is vaccinated, the vaccine is distributed to everyone else with no preference for age. In scenarios C, D, and E, after the 80+ group, essential workers are then vaccinated without regard for age. In scenario C, the rest of the population is vaccinated in decreasing order of age. In scenario D, the rest of the population is vaccinated without regard for age. In scenario E, the 70–79 cohort is vaccinated next and then the rest of the population is vaccinated without regard to age. For each scenario, the rate of vaccination is initially about $0.075\%$ population per day until the entire 80+ cohort has been vaccinated. The rate of vaccination then increases to about $0.45\%$ population per day until the end of the simulations (this value is fixed across all scenarios A–D, and sensitivity to this value is explored in S1 Text Supplemental Material (Fig L). These rates were chosen to match the projected timelines released by the British Columbia Centre for Disease Control as of February 2021 [37].
Finally, we considered some economic measures of the cost of the pandemic from a health system payer perspective: health utility losses measured in QALYs lost and net monetary benefit (NMB). Estimating health utility in terms of QALYs allows us to quantify and compare loss of quality and duration of life due to illness, disability, and death [38]. QALYs are estimated such that the maximum value of 1 indicates a year in perfect health, whereas a value of 0 indicates no health (death). Utility decrements due to acute COVID infection, hospitalizations, chronic outcomes, and death are estimated as the difference between expected health, and health utility due to COVID. A key parameter in our estimates is the standardized mortality ratio (SMR): the increased hazard of dying COVID patients have of non-COVID causes (on average) relative to other patient of a similar age.
Following Briggs et al [39], since COVID fatalities often occur to patients with preexisting conditions that may shorten their lifespan, we took an SMR of 2. Details of the calculations are in the S1 Text Supplementary Material. These estimates are synthesized into NMB by converting QALYs to monetary value, to allow a unified way of evaluating vaccination strategies [40].
Following the data from studies on the Pfizer and Moderna vaccines on the impact of vaccination on infection and on viral load, and data from the clinical trials on the impact on severe disease [2,3,6–8], we vary the effectiveness of the vaccine in preventing infection from $60\%$ to $90\%$, and set the efficacy of it preventing illness and death in the case of infection at $90\%$. This yields an overall effectiveness at preventing symptomatic infection (including severe outcomes) between $96\%$ and $99\%$.
## Results
We find that vaccinating essential workers earlier gives large reductions in infections, hospitalizations, deaths, and instances of Long COVID (cases with symptoms lasting longer than 28 days), across a range of scenarios for transmission and vaccine efficacy. Except for deaths, results were similarly good with a strategy that vaccinates younger people sooner without targeting essential workers. However, with scenarios vaccinating the elderly later, they have very slightly higher rates of adverse outcomes, as expected, depending on how they are prioritized.
Fig 1 illustrates the high impact of including younger age groups sooner in the program than in a primarily oldest-first strategy (A), either through expanding to all age groups after those over 80 have been vaccinated (strategy B) or through vaccinating essential workers of any age group after those over 80 and then continuing from oldest to youngest (C), or expanding to all ages after essential workers (D). Vaccinating essential (high-contact) workers early has a strong effect, and can be done within the context of a broader oldest-first vaccination scheme. This does not require additional doses of vaccine. The shading in Fig 1 indicates which age groups are affected in four example simulations in the model, with four distinct vaccination strategies.
**Fig 1:** *Vaccinating younger adults sooner, and vaccinating essential workers sooner, reduces cases, deaths, hospitalizations and Long COVID (COVID symptoms lasting longer than 28 days).In our model the rollout consists of an initial phase during which those over 80 are vaccinated and R = 1.05, followed by a second stage where the transmission rate is higher (R = 1.3 here). In these simulations vaccine efficiency is 90% for preventing disease and 75% at preventing infection and therefore subsequent transmission. Top row: Illustration of the vaccination programs by age group over time: Fraction of the age group vaccinated versus time. Next two rows: Simulated cases, hospitalization, deaths and Long COVID prevalence over time.*
Fig 2 shows the simulated total cases, deaths, hospitalizations and Long COVID cases for a range of transmission-blocking efficacy and for two values of the underlying R. At the time of writing, many jurisdictions are vaccinating on a primarily oldest-to-youngest vaccination program [41–43], although some health care workers and staff in long term care homes and some other vulnerable communities are included in the early stages. In our model, a strictly age-based approach leads to considerably more cases and more Long COVID across a range of values of the vaccine’s efficacy against transmission and across a range of R values. When R is kept very low (1.15), for example through continued strong social distancing, all of the simulated strategies do well at reducing deaths. When R rises to 1.3, strategies placing essential workers after 80+, either continuing with an age-based rollout or opening to all adults aged 20–69 after those 70+, have an advantage in reducing deaths in addition to strong advantages for infections, hospitalizations and Long COVID. Other differences between the age ordering are very small compared to the difference between “oldest first” strategies and any alternative that prioritizes essential workers or even all younger adults earlier in the program.
**Fig 2:** *Vaccinating essential workers earlier in the program has benefits for cases, hospitalizations, death and Long COVID (COVID symptoms for longer than 28 days).This does not depend sensitively on the efficacy of the vaccine against transmission, nor on the underlying transmission rate. Similar results are attained by vaccinating over all age groups (strategy B), with the exception that this gives less reduction in deaths.*
In S1 Text Supplemental Material we explored the sensitivity of our vaccine strategy comparisons to the portion of workplace contacts taking place among essential workers (Fig G), the efficacy against transmission (Fig H), the portion of workers considered “essential” (Fig I) and to the contact matrix (Fig J). We explore a wider set of strategies (Fig B). Consistently, vaccinating essential workers earlier has considerable benefits compared to oldest-first strategies while the relative merits of different age-based rollouts depend on assumptions. For example, when the portion of workplace contacts among nonessential workers is higher, strategy C (80+, EW, 70–79,..) which is predominantly oldest first loses some of its benefit because essential workers have less of the overall contact (Fig G). While the efficacy against transmission has a high impact on the outcomes, it does not much impact the relative performance of the strategies unless R is high (Fig H). Finally, we determine the optimal strategy simply from the point of view of minimizing deaths; despite the fact that the mortality risk motivates oldest-first vaccination, this strategy is only the best for deaths as an outcome when efficacy against transmission is extremely low (0.1–0.2) and when R is high (Figs K and L). See the S1 Text Supplemental Material for further details.
Fig 3 shows health utility losses in QALYs and their source (cases, chronic impacts, deaths and hospitalizations). Here chronic impacts includes both chronic Long COVID symptoms (similar to ME/CFS) and other long-term complications. Deaths are naturally a large contribution to health utility loss; the next biggest contribution by far are the chronic impacts where we find that vaccinating essential workers sooner has profound benefits For example, if R rises to 1.3 and the vaccine is $75\%$ effective in preventing transmission, the combined reduction in deaths, chronic impacts, cases and hospitalizations when essential workers are vaccinated after those aged 80+ means that over 11000 QALYs are gained (Fig 3, bottom middle panel). Of these, 3000 are due to chronic impacts. When efficacy against transmission is lower, because there are more infections, it remains very beneficial from a health utility point of view to vaccinate essential workers sooner. As with hospitalization, deaths and Long COVID, the differences between age rollouts beyond whether essential workers are included early is relatively small.
**Fig 3:** *Health losses (QALYs).R and the efficacy against transmission vary and are labelled on the side (R) and the top (efficacy). Community infections refers to QALYs lost from acute COVID infections that do not result in hospitalization.*
We find similar effects when we estimate the incremental costs of the pandemic including both direct costs of hospitalization and chronic outcomes and health utility decrements converted to monetary values (Fig 4). Vaccinating essential workers sooner reduces the overall NMB loss due to the pandemic by 50 to $65\%$. This results in a potential savings of (for example) over $400M (if $R = 1.3$); see Fig B in the S1 Text Supplemental Material. In all scenarios, NMB lost was largest for the age-based immunization strategy (A). The largest improvement in NMB was achieved with strateges C, D and E, irrespective of vaccine efficacy. Fig 4 highlights the potential for a substantial impact of chronic consequences of infection, both related to health utility losses, and future health system cost.
**Fig 4:** *Net Monetary Benefit (loss), due to the cost of treating chronic outcomes, hospitalizations during the pandemic, and health utilty converted from QALYs lost shown in Fig 3.Transmission-blocking efficacy ve and R vary as labelled over the panels.*
Finally, we explore immunity at the point where transmission begins to decline (while keeping R constant through, for example, NPIs). Fig 5 shows the fractions of the population who were infected (top panels), or who were protected by either natural infection or vaccination (bottom panels), at the time when simulated infections began to decline, for three transmission-preventing efficacies and for a range of R. Consistently, the oldest to youngest strategy requires more immunity than strategies vaccinating essential workers or younger people sooner.
**Fig 5:** *Immunity and vaccination at the time when infections begin to decline (turn around time), as a function of R.(A) Proportion ever infected by turn around time. (B) Proportion ever infected or vaccinated by turn around time. The efficacy of vaccination against infection (ve) is varied from 0.6 (left) to 0.9 (right).*
## Discussion
Strategies that vaccinate essential workers early lead to substantial reductions in the number of infections, hospitalizations, deaths, and cases of Long COVID relative to a strategy of oldest first. Essential workers cannot effectively isolate under any social distancing regime. Our finding that there is a strong benefit across many outcomes if these groups are vaccinated early hold for any group of similarly high-contact individuals, and is consistent with Centre for Disease Control recommendations for vaccination of essential workers [44]. Vaccinating essential workers is important from an equity standpoint, as the COVID-19 pandemic has disproportionately impacted essential workers. Across neighbourhoods in Toronto, for example, per-capita COVID cases and deaths were 2.5 to 3 times higher in neighbourhoods with high (vs low) concentrations of essential workers [45]. Essential workers often have lower incomes, and may be hired as contractors; they may not have paid sick leave, and may have limited ability to negotiate safe working conditions [45]. While many others are able to work safely at home, these individuals cannot. Our findings suggest that prioritizing them for vaccination would help to reduce this substantial disparity, and would not come at a cost of increased adverse outcomes.
We have considered contexts in which interventions are brought in to reduce case numbers when the health care system begins to be stretched for capacity, as has occurred in most jurisdictions in Canada, the US, and Europe (i.e. R is not left much higher than 1 for long periods). Larger transmission rates lead to the rapid overwhelming of the healthcare system, and governments then introduce lockdown or stricter social distancing measures in response. For example, in the United Kingdom, when cases reached a rolling average of 55,000 new cases per day (approximately 80 per 100K at the end of 2020), strict lockdowns were implemented with $78\%$ of the population under severe restrictions [46]. Across Canada, lockdown measures were introduced when daily incidence reached approximately 20–50 cases per 100K. Sustaining $R = 2$ before vaccination rollout is at a level to substantially protect the population results in incidence 40–60 times that, or at least 10 times the UK value. Even at a reduced effective hospitalization rate (with older groups vaccinated), health care systems would be overwhelmed if these incidence levels were reached. Accordingly, R is not likely to remain high enough, in advance of vaccination, for oldest-first strategies to be best; distancing and other measures have consistently been put in place to prevent this.
The mechanisms behind our results are illustrated in Fig 5, which shows that the oldest to youngest strategy requires the most immunity (through either infection or vaccination) out of all strategies considered. This is because individuals who have a comparatively low likelihood of exposure and transmission are vaccinated, but their protection contributes less to the population’s collective protection than vaccination of those who are at risk of exposure and transmission. Accordingly, vaccinating 80+, essential workers and then all ages requires the fewest infections and the least immunity in order for cases to decline; in this strategy those likely to transmit are protected efficiently. The immunity required for infections to decline is more sensitive to the vaccination strategy when R is relatively low; above $R = 2$, the required protection for all strategies is just under $\frac{1}{2.}$ However, vaccinating essential workers early allows far more of that fraction to be protected by vaccination, whereas in the oldest-first strategy more infection is required to achieve declining infections.
Using doses of vaccines effectively is particularly relevant where vaccine supply is low and rollout is slow—something likely to be relevant in low- and middle-income settings for many months to come. This urgency is amplified by the increasingly-recognized importance of long COVID and the severity and extent of chronic impacts that COVID-19 infection can have in adults of any age. This provides a counterpoint to what has become the conventional wisdom that age-based rollouts are the most effective at saving lives.
We modelled a rise in transmission 60 days into the simulation, in part because of increasing evidence that at the time of writing variants of concern (VOC) were rising in frequency [47,48] during vaccine rollout in Canada and similar jurisdictions and these VOC have increased transmissibility [49]. As their frequency rises, VOC are therefore likely to drive higher transmission rates, particularly in the current context where many areas are considering relaxing restrictions following declining cases and encouraged by the availability of vaccines. If VOC transmission is contained, the relaxation of measures itself is likely to result in higher transmission.
Even without considering Long COVID and COVID complications, vaccinating essential workers sooner has strong benefits in terms of reducing infections, hospitalizations, deaths, and in terms of net monetary benefit (NMB). However, taking chronic outcomes into account makes the advantages of prioritizing high-contact individuals even more stark, showing that they potentially save hundreds of millions of dollars of additional NMB (in a population of approximately 5M). These long-term consequences of COVID infection could impact the future health of $0.5\%$ of the population under an oldest-first vaccination strategy, and far fewer ($0.25\%$) if essential workers and/or younger adults are vaccinated earlier. Despite uncertainty in the likelihood and duration of longterm consequences of COVID infection, Long COVID and COVID complications need to be included in considerations of vaccine priority.
Our results about the relative benefits of different rollouts hold for jurisdictions where social distancing and other non-pharmaceutical interventions keep the basic reproductive number relatively low, and where vaccine supply is limited. There are several additional limitations. We have assumed that the vaccines in question are effective at reducing transmission as well as severe outcomes. Our results should not be applied to jurisdictions where these assumptions—about epidemic trajectory, vaccine supply and vaccine efficacy—do not hold. We have performed a sensitivity analysis to ensure that our results are robust to key unknown parameters (see S1 Text Supplemental Material). In some parameter regions, such as when R is very low ($R = 1.15$), the benefit conferred by vaccinating essential workers early is small, especially in terms of overall deaths. However, we note that there was no scenario or outcome considered in which such strategies performed worse than the oldest-first strategies. Furthermore, our calculations of the QALYs lost and NMB of different strategies were based on estimates of SMR, life expectancy at different ages, and other parameters specific to particular chronic conditions. All these parameters are subject to a great deal of uncertainty, and updating them as more information is available about mortality and morbidity due to COVID may change the total estimated cost of chronic conditions, but not the fact that such costs are large and need to be factored into consideration.
Strategies that explicitly target high-contact workers were adopted in several Canadian provinces [50,51]. These provinces additionally reserved first doses for younger adults ahead of second doses for older adults. Although it is difficult to assess the impact of vaccination strategies alone, since there are many additional factors affecting infection and mortality rates, our findings suggest that these vaccination strategies were key to the sustained and rapid decline in COVID-19 cases across Canada in late Spring–Summer, 2021. As of the time of writing, COVID-19 cases and deaths across the country remain low, even following significant re-opening and emergence of highly transmissible variants.
## References
1. Zimmer C, Corum J, Wee SL. **Coronavirus Vaccine Tracker**. *The New York Times* (2021.0)
2. Baden LR, El Sahly HM, Essink B, Kotloff K, Frey S, Novak R. **Efficacy and safety of the mRNA-1273 SARS-CoV-2 vaccine**. *New England Journal of Medicine* (2020.0). DOI: 10.1056/NEJMoa2035389
3. Oliver SE, Gargano JW, Marin M, Wallace M, Curran KG, Chamberland M. **The Advisory Committee on Immunization Practices’ Interim Recommendation for Use of Pfizer-BioNTech COVID-19 Vaccine—United States, December 2020**. *MMWR Morb Mortal Wkly Rep* (2020.0) **69** 1922-1924. DOI: 10.15585/mmwr.mm6950e2
4. Voysey M, Clemens SAC, Madhi SA, Weckx LY, Folegatti PM, Aley PK. **Safety and efficacy of the ChAdOx1 nCoV-19 vaccine (AZD1222) against SARS-CoV-2: an interim analysis of four randomised controlled trials in Brazil, South Africa, and the UK**. *Lancet* (2021.0) **397** 99-111. DOI: 10.1016/S0140-6736(20)32661-1
5. Caddy SL. **Coronavirus: few vaccines prevent infection–here’s why that’s not a problem**. *The Conversation* (2021.0)
6. 6ModernaTX I. mRNA-1273: Sponsor Briefing Document Addendum;. Vaccines and Related Biological Products Advisory Committee. https://www.cdc.gov/me-cfs/. Available from: https://www.fda.gov/media/144453/download.
7. Mitnick J, Regalado A. *A leaked report shows Pfizer’s vaccine is conquering covid-19 in its largest real-world test* (2021.0)
8. Hall VJ, Foulkes S, Saei A, Andrews N, Oguti B, Charlett A. *Effectiveness of BNT162b2 mRNA Vaccine Against Infection and COVID-19 Vaccine Coverage in Healthcare Workers in England, Multicentre Prospective Cohort Study (the SIREN Study)* (2021.0)
9. 9Single Dose Administration, And The Influence Of The Timing Of The Booster Dose On Immunogenicity and Efficacy Of ChAdOx1 nCoV-19 (AZD1222) Vaccine.
10. Emary KRW, Golubchik T, Aley PK, Ariani CV, Angus BJ, Bibi S. *Efficacy of ChAdOx1 nCoV-19 (AZD1222) Vaccine Against SARS-CoV-2 VOC 202012/01 (B.1.1.7)* (2021.0)
11. Petter E, Mor O, Zuckerman N, Oz-Levi D, Younger A, Aran D. *Initial real world evidence for lower viral load of individuals who have been vaccinated by BNT162b2* (2021.0)
12. Levine-Tiefenbrun M, Yelin I, Katz R, Herzel E, Golan Z, Schreiber L. *Decreased SARSCoV-2 viral load following vaccination* (2021.0). DOI: 10.1038/s41591-021-01316-7
13. Mallapaty S. **The coronavirus is most deadly if you are older and male—new data reveal the risks**. *Nature* (2020.0) **585** 16-17. DOI: 10.1038/d41586-020-02483-2
14. Yong E. **Long-Haulers Are Redefining COVID-19**. *The Atlantic.* (2020.0)
15. **Long COVID: let patients help define long-lasting COVID symptoms**. *Nature* (2020.0) **586**. DOI: 10.1038/d41586-020-02796-2
16. Logue JK, Franko NM, McCulloch DJ, McDonald D, Magedson A, Wolf CR. **Sequelae in Adults at 6 Months After COVID-19 Infection**. *JAMA Netw Open* (2021.0) **4** e210830. DOI: 10.1001/jamanetworkopen.2021.0830
17. Lam MHB, Wing YK, Yu MWM, Leung CM, Ma RC, Kong AP. **Mental morbidities and chronic fatigue in severe acute respiratory syndrome survivors: long-term follow-up**. *Archives of internal medicine* (2009.0) **169** 2142-2147. DOI: 10.1001/archinternmed.2009.384
18. 18Centers for Disease Control and Prevention. Myalgic encephalomyelitis/chronic fatigue syndrome;. Accessed: 2021-02-14. https://www.cdc.gov/me-cfs/.
19. 19Office for National Statistics (UK). The prevalence of long COVID symptoms and COVID-19 complications; Accessed: 2021-2-17. https://www.ons.gov.uk/news/statementsandletters/theprevalenceoflongcovidsymptomsandcovid19complications.
20. Rubino F, Amiel SA, Zimmet P, Alberti G, Bornstein S, Eckel RH. **New-Onset Diabetes in Covid-19**. *N Engl J Med* (2020.0) **383** 789-790. DOI: 10.1056/NEJMc2018688
21. 21Pinto DS. Coronavirus disease 2019 (COVID-19): Myocardial infarction and other coronary artery disease issues;. Accessed: 2021-2-17. https://www.uptodate.com/contents/coronavirus-disease-2019-covid-19-myocardial-infarction-and-other-coronar.
22. Rismanbaf A, Zarei S. **Liver and Kidney Injuries in COVID-19 and Their Effects on Drug Therapy; a Letter to Editor**. *Arch Acad Emerg Med* (2020.0) **8** e17. PMID: 32185369
23. Taquet M, Geddes JR, Husain M, Luciano S, Harrison PJ. **Six-month Neurological and Psychiatric Outcomes in 236,379 Survivors of COVID-19.**. *medRxiv* (2021.0)
24. Gu J, Gong E, Zhang B, Zheng J, Gao Z, Zhong Y. **Multiple organ infection and the pathogenesis of SARS**. *J Exp Med* (2005.0) **202** 415-424. DOI: 10.1084/jem.20050828
25. Souza INO, Barros-Aragao FGQ, Frost PS, Figueiredo CP, Clarke JR. **Late Neurological Consequences of Zika Virus Infection: Risk Factors and Pharmaceutical Approaches**. *Pharmaceuticals* (2019.0) **12**. DOI: 10.3390/ph12020060
26. Cao-Lormeau VM, Blake A, Mons S, Lastere S, Roche C, Vanhomwegen J. **Guillain-Barré Syndrome outbreak associated with Zika virus infection in French Polynesia: a case-control study**. *Lancet* (2016.0) **387** 1531-1539. DOI: 10.1016/S0140-6736(16)00562-6
27. Spencer CA. **‘Long-haul’ covid-19 complications are real. I faced similar problems after surviving Ebola**. *The Washington Post* (2020.0)
28. Bubar KM, Reinholt K, Kissler SM, Lipsitch M, Cobey S, Grad YH. **Model-informed COVID-19 vaccine prioritization strategies by age and serostatus**. *Science* (2021.0)
29. Matrajt L, Eaton J, Leung T, Brown ER. **Vaccine optimization for COVID-19: Who to vaccinate first?**. *Sci Adv* (2020.0) **7**. DOI: 10.1126/sciadv.abf1374
30. Chen J, Hoops S, Marathe A, Mortveit H, Lewis B, Venkatramanan S. **Prioritizing allocation of COVID-19 vaccines based on social contacts increases vaccination effectiveness**. *medRxiv* (2021.0). DOI: 10.1101/2021.02.04.21251012
31. Jentsch PC, Anand M, Bauch CT. **Prioritising COVID-19 vaccination in changing social and epidemiological landscapes: a mathematical modelling study**. *Lancet Infect Dis* (2021.0). DOI: 10.1016/S1473-3099(21)00057-8
32. Rodríguez J, Patón M, Acuña JM. **COVID-19 vaccination rate and protection attitudes can determine the best prioritisation strategy to reduce fatalities**. *medRxiv* (2021.0)
33. Buckner JH, Chowell G, Springborn MR. **Dynamic prioritization of COVID-19 vaccines when social distancing is limited for essential workers**. *Proc Natl Acad Sci U S A* (2021.0) **118**. DOI: 10.1073/pnas.2025786118
34. Babus A, Das S, Lee S. *The Optimal Allocation of Covid-19 Vaccines* (2020.0). DOI: 10.1213/XAA.0000000000001133
35. 35Emergency Management BC, Government Communications and Public Engagement. COVID-19 essential services;. Accessed: 2021-2-21. https://www2.gov.bc.ca/gov/content/safety/emergency-preparedness-response-recovery/covid-19-provincial-support/essential-services-covid-19.
36. 36Mayo Clinic. COVID-19 variants: What’s the concern?; 2021. Accessed: 2021-2-23. https://www.mayoclinic.org/diseases-conditions/coronavirus/expert-answers/covid-variant/faq-20505779.
37. 37COVID-19 Immunization Plan;. Accessed: 2021-1-22. https://www2.gov.bc.ca/gov/content/covid-19/vaccine/plan#phases.
38. Whitehead SJ, Ali S. **Health outcomes in economic evaluation: the QALY and utilities**. *Br Med Bull* (2010.0) **96** 5-21. DOI: 10.1093/bmb/ldq033
39. Briggs AH, Goldstein DA, Kirwin E, Meacock R, Pandya A, Vanness DJ. **Estimating (quality-adjusted) life-year losses associated with deaths: With application to COVID-19**. *Health Economics* (2020.0). DOI: 10.1002/hec.4208
40. Stinnett AA, Mullahy J. **Net health benefits: a new framework for the analysis of uncertainty in cost-effectiveness analysis**. *Med Decis Making* (1998.0) **18** S68-80. DOI: 10.1177/0272989X98018002S09
41. 41Government Communications, Engagement P. COVID-19 Immunization Plan;. Accessed: 2021-2-22. https://www2.gov.bc.ca/gov/content/safety/emergency-preparedness-response-recovery/covid-19-provincial-support/vaccines.
42. 42Massachusetts’ COVID-19 vaccination phases;. Accessed: 2021-2-22. https://www.mass.gov/info-details/massachusetts-covid-19-vaccination-phases.
43. 43Australian Government Department of Health. When will I get a COVID-19 vaccine?; 2020. Accessed: 2021-2-22. https://www.health.gov.au/initiatives-and-programs/covid-19-vaccines/getting-vaccinated-for-covid-19/when-will-i-get-a-covid-19-vaccine.
44. 44Centre for Disease Control. CDC’s COVID-19 Vaccine Rollout Recommendations; 2021. Accessed: 2021-3-1. https://www.cdc.gov/coronavirus/2019-ncov/vaccines/recommendations.html.
45. Rao A, Ma H, Moloney G, Kwong JC, Juni P, Sander B. **A disproportionate epidemic: COVID-19 cases and deaths among essential workers in Toronto, Canada**. *medRxiv* (2021.0). DOI: 10.1016/j.annepidem.2021.07.010
46. 46A timeline of UK lockdown measures since the pandemic began;. Accessed: 2021-3-19. https://www.expressandstar.com/news/uk-news/2021/01/04/a-timeline-of-uk-lockdown-measures-since-the-pandemic-began/.
47. Washington NL, Gangavarapu K, Zeller M, Bolze A, Cirulli ET, Schiabor Barrett KM. **Genomic epidemiology identifies emergence and rapid transmission of SARS-CoV-2 B.1.1.7 in the United States**. *medRxiv* (2021.0). DOI: 10.1101/2021.02.06.21251159
48. Brown KA, Gubbay J, Hopkins J, Patel S, Buchan SA, Daneman N. *Rapid Rise of S-Gene Target Failure and the UK variant B.1.1.7 among COVID-19 isolates in the Greater Toronto Area, Canada* (2021.0)
49. 49WHO—SARS-CoV-2 Variants. World Health Organization; 2021. urlhttps://www.who.int/csr/don/31-december-2020-sars-cov2-variants/en/.
50. 50Ontario’s COVID-19 vaccination plan;. Accessed: 2021-07-11. https://covid-19.ontario.ca/ontarios-covid-19-vaccination-plan.
51. 51Front-line workers prioritized as COVID-19 vaccine rollout accelerates;. Accessed: 2021-07-11. https://news.gov.bc.ca/releases/2021PREM0021-000504.
|
---
title: The effects of COVID-19 lockdown measures on health and healthcare services
in Uganda
authors:
- David Musoke
- Sarah Nalinya
- Grace Biyinzika Lubega
- Kevin Deane
- Elizabeth Ekirapa-Kiracho
- David McCoy
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021763
doi: 10.1371/journal.pgph.0001494
license: CC BY 4.0
---
# The effects of COVID-19 lockdown measures on health and healthcare services in Uganda
## Abstract
Many countries across the world instituted lockdowns as a measure to prevent the spread of COVID-19. However, these lockdowns had consequences on health systems. This study explored effects of the COVID-19 lockdown measures on health and healthcare services in Uganda. The qualitative study employed focus group discussions (FGDs), household interviews, and key informant interviews (KIIs) in both an urban (Kampala district) and rural (Wakiso district) setting in central Uganda. Fourteen FGDs were conducted among community members, local leaders, community health workers, and health practitioners. Interviews were conducted among 40 households, while 31 KIIs were held among various stakeholders including policy makers, non-governmental organisations, and the private sector. Data was analysed by thematic analysis with the support of NVivo 2020 (QSR International). Findings from the study are presented under four themes: maternal and reproductive health; child health; chronic disease services; and mental health. Maternal and reproductive health services were negatively affected by the lockdown measures which resulted in reduced utilisation of antenatal, postnatal and family planning services. These effects were mainly due to travel restrictions including curfew, and fear of contracting COVID-19. The effects on child health included reduced utilisation of services which was a result of difficulties faced in accessing health facilities because of the travel restrictions. Patients with chronic conditions could not access health facilities for their routine visits particularly due to suspension of public transport. Depression, stress and anxiety were common due to social isolation from relatives and friends, loss of jobs, and fear of law enforcement personnel. There was also increased anxiety among health workers due to fear of contracting COVID-19. The COVID-19 lockdown measures negatively affected health, and reduced access to maternal, reproductive and child health services. Future interventions in pandemic response should ensure that their effects on health and access to health services are minimised.
## Introduction
As COVID-19 progressively spread around the world at the beginning of 2020, many countries scrambled to control the spread of the virus by implementing a range of interventions [1]. These included standard infection prevention and control initiatives such as hand washing with soap and use of alcohol-based hand sanitizers. Later, the wearing of face masks was added to the measures [2]. However, as the pandemic worsened, other more stringent measures to enable social distancing were implemented. These included lockdown measures consisting of curfews, travel restrictions, closure of various institutions, prohibition of public gatherings, and closure of borders [2].
Uganda commenced its response to COVID-19 in early March 2020 after the WHO declared COVID-19 a public health emergency of international concern even before the first case of the disease was confirmed in the country. Several measures were therefore instituted to prevent the spread of COVID-19 in the country. These measures included a national lockdown consisting of night curfew from 19:00hrs to 05:30hrs, suspension of various forms of public transport, restricted movement of people, closure of schools and other institutions of learning, suspension of all forms of public gatherings, and closure of national borders [2]. Several other countries within sub Saharan Africa and beyond also instituted lockdown measures to reduce the spread of COVID-19 [3]. By May 2020 when the first phase of the lockdown was gradually eased, Uganda had a daily average of less than 45 reported new COVID-19 cases. Although the country thereafter briefly experienced a surge in cases, the strict second lockdown that commenced in June 2020 contributed to reduction in reported cases [4]. Uganda arguably minimised and contained community transmission of COVID-19 through the strategy of early implementation of the COVID-19 control measures including the lockdown in 2020 and 2021 [5].
Lockdown measures have short and long-term effects on health and health service delivery in many countries, with children and women most affected [6]. For example, in one modelling study, it was estimated that the impact of COVID-19 and its control measures could increase maternal mortality by up to $38.6\%$, and a $44.7\%$ increase in child mortality across 118 low- and middle-income countries (LMICs) primarily through the indirect effects of the disease [5]. Recent studies that investigated the effects of COVID-19 on maternal health in Uganda found that several indicators worsened, with an increase in pregnancy complications, as well as maternal mortality and low birth weight, likely due to limited access to care [7, 8]. In addition, child health and well-being were anticipated to be negatively affected by the various lockdown measures in Uganda [9]. Indeed, available evidence indicates an increase in physical, sexual and emotional violence against children [5, 9, 10], child labour [10, 11], as well as food insecurity and malnutrition [10, 11]. There is generally limited empirical data documenting the effects of COVID-19 lockdown particularly in LMICs, hence more evidence is needed to concretize the effects of lockdown measures on the health system. Such evidence will not only be useful in the control of future waves of COVID-19 but also other pandemics. This study therefore explored the effects of COVID-19 lockdown measures on health and healthcare services in Uganda. Our findings provide important insights for various stakeholders on the scope of the health challenges experienced during the COVID-19 lockdowns in both rural and urban settings, in addition to informing the development of interventions to mitigate the negative effects of lockdown measures in the future.
## Study area and setting
The study was carried out in both an urban and rural setting. The urban setting was Rubaga Division, Kampala District, while the rural one was Kasanje Town Council, Wakiso District. Kampala and Wakiso districts were selected as the 2 study sites because they had registered the highest cases of COVID-19 in Uganda during the first (November to December 2020) and second (April to June 2021) waves. Rubaga Division in Kampala has 13 parishes which are predominantly urban informal settlements. The division has a projected population of approximately 427,300 people ($53.8\%$ female and $46.2\%$ male) [12]. The major economic activities in the area include small scale business such as general retail shops, furniture shops, restaurants and auto-mobile repairs. Kasanje Town Council in Wakiso has seven parishes which are predominantly rural and a few peri-urban. Kasanje Town Council has a projected population of 46,042 people ($49.9\%$ female and $50.1\%$ male) [13]. The main economic activities in the area include subsistence agriculture, tourism, small-scale trade and brick laying. The town council has one public health facility (Kasanje Health Centre III), two government-aided (private not-for-profit) health facilities, nine registered (private) clinics, and 22 registered (private) drug shops [13]. Rubaga Division has two government health facilities (Kitebi Health Centre III and Kawaala Health Centre III), 11 private-not for profit facilities, and several private facilities [14]. The health facilities in both study sites provide both preventive and curative health services such as maternal and child health, as well as services for communicable and non-communicable diseases. Prior to the pandemic, inhabitants of the study area had routine access to health services in both public and private sector health facilities. In addition, services for management of childhood illnesses (malaria, pneumonia and diarrhoea) could be obtained from community health workers.
## Study design and participants
The study was qualitative and employed three data collection methods: focus group discussions (FGDs), household interviews, and key informant interviews (KIIs). The different methods were used to get a diverse range of views from participants, so as to allow for triangulation. FGD, household survey and KII guides were used to explore and obtain insights from different stakeholders about the impacts of the COVID-19 prevention measures including lockdown on health and healthcare services in Uganda. The tools included questions on: COVID-19 in Uganda; the lockdown measures implemented by the government; perceptions on the consequences on these measures; effects of the measures on the health system including delivery of services; effects of the measures on women and children; and effects of the measures on household members as well as the broader community. A total of 14 FGDs were conducted, 7 in the urban setting in Kampala district and 7 in rural Wakiso district. The FGDs comprised of 8 to 16 participants, and were categorized into female youth, male youth, adult females, adult males, local community leaders, community health workers (CHWs), and health practitioners. The youth involved in the FGDs were between 18 and 30 years. Categorisation of the FGD participants by age and sex ensured a gendered space for exploration to enhance free expression during the discussions.
The study also used household interviews to obtain their experiences and perspectives on the impact of the COVID-19 prevention measures. A total of 40 household interviews, 20 from each site (urban and rural), were conducted using an interview guide. One interview was conducted per household, and households that participated in the study were identified by community mobilisers to represent high, middle and low-income households. Socio-economic status was assessed based on various household factors including type of employment, location of residence, and house structure. The mobilisers, who had lived in the area for over 20 years, were knowledgeable about the community hence able to identity households that belonged to each of the 3 categories. Before commencement of data collection, the research assistants did an assessment (including observation and asking some questions) to ensure that the households fit within the selected category. In the urban area, 6 high-income, 7 middle-income and 7 low-income households were involved, while 5 high-income, 7 middle-income and 8 low-income households participated from the rural setting. Household heads or their spouses in the respective households took part in the study. These study participants were predominantly female due to their availability at home at the time of data collection. KIIs were conducted to obtain the perspectives of other stakeholders such as policy makers, implementers, international agencies, non-governmental organisations, researchers and the private sector on the impacts of COVID-19 prevention measures. A total of 31 KIIs were conducted, and key informants were purposively selected due to their involvement in the response to COVID-19 control at national and sub-national levels. A summary of the study participants is provided in Table 1.
**Table 1**
| Kampala focus group discussions | Category | Number |
| --- | --- | --- |
| | Health practitioners | 16.0 |
| | Community health workers | 13.0 |
| | Community leaders | 11.0 |
| | Female youth | 12.0 |
| | Female adults | 15.0 |
| | Male youth | 12.0 |
| | Male adults | 15.0 |
| | Total | 94.0 |
| Wakiso focus group discussions | | |
| | Health practitioners | 8.0 |
| | Community health workers | 12.0 |
| | Community leaders | 12.0 |
| | Female youth | 13.0 |
| | Female adults | 13.0 |
| | Male youth | 12.0 |
| | Male adults | 11.0 |
| | Total | 81.0 |
| Kampala household survey | | |
| | Low-income | 7.0 |
| | Middle income | 7.0 |
| | High income | 6.0 |
| | Total (17 female, 3 male) | 20.0 |
| Wakiso household survey | | |
| | Low-income | 8.0 |
| | Middle income | 7.0 |
| | High income | 5.0 |
| | Total (15 female, 5 male) | 20.0 |
| Key informant interviews | | |
| | Government ministries | 4.0 |
| | International agencies | 7.0 |
| | Non-governmental organisations | 5.0 |
| | Local government officials | 6.0 |
| | Hospital staff | 4.0 |
| | Professional bodies | 1.0 |
| | Research institutions | 4.0 |
| | Total (13 female, 18 male) | 31.0 |
## Data collection
The seven zones from Rubaga Division in Kampala which participated in the study were: Nakulabye, Lubya, Mapeera, Lugala, Lusaze, Namungoona and Kasubi IV. These zones were purposively selected as they were largely comprised of urban informal settlements, with a high population and overcrowded households which increased their vulnerability to COVID-19. In addition, these zones consisted of people with varying socio-economic status which offered varied experiences and perspectives about the impacts of the COVID-19 preventive measures. Participants of the health practitioners’ FGD in Kampala were from Kawaala Health Centre IV which is the largest public health facility in the division and a designated COVID-19 sample collection and case-management facility.
Participants of the community FGDs in Kasanje Town Council, Wakiso District were predominantly from rural zones. The ten zones which participated in the study were: Kasanje, Sokolo, Bweyogerere, Buyege, Jjungo, Koba, Bukalaza, Kikalala, Taba and Sakabusolo. Community FGD participants were selected by the community mobilisers. All the community FGDs were conducted at public places within the respective communities particularly schools and offices of local council chairpersons. Participants of the health practitioners’ FGD in Wakiso were from Kasanje Health Centre III which was also purposively selected as it was a COVID-19 testing site in the area. Participants of the health practitioner FGDs were purposively selected by management of the respective health facility from various departments including paediatrics, laboratory, maternity. The FGDs for health practitioners were conducted at the respective health facilities. The FGDs involved participants creating a timeline of various COVID-19 prevention measures from March 2020 to July 2021. The participants of each FGD then made presentations of how these measures impacted health. The FGDs were audio recorded, with all community FGDs conducted in Luganda, the most widely used local language in both study sites. The health practitioner FGDs were conducted in English, and all FGDs were facilitated by 2 female research assistants with prior experience in qualitative research. The research assistants were not known to study participants which facilitated open and objective interaction during data collection.
The participants of the household interviews in both study sites were from the same zones as those who participated in the FGDs. Key informants were mainly selected purposively based on their involvement, expertise and influence regarding response to COVID-19 and its impact on health. Consultation was done by the research team among key stakeholders in the COVID-19 response including from the Ministry of Health and local authorities to ensure appropriate key informants were selected. After they were identified, the research assistants made an appointment with each key informant at a time of their convenience. All tools were developed in English and later translated to Luganda (except for the KII guide which was used in English) before they were piloted in a zone in Wakiso district that was not involved in the study. All key informant interviews were conducted virtually via Zoom due to COVID-19 travel restrictions at the time.
## Data management and analysis
The research team that participated in data management and analysis (SN, GBL, DMu) all have expertise and experience in qualitative research. The audio recordings of the FGDs and household interviews were transcribed verbatim and proof-read by a research assistant to ensure that they were accurate. Since the FGDs were mostly conducted in Luganda, the transcripts were translated into English, and the translation verified by another researcher. Both the research assistant who transcribed and translated the transcripts, and the researcher who reviewed and verified the translated transcripts (SN) are proficient in both English and the local language, and experienced in qualitative data analysis. The audio recordings of the KIIs were auto transcribed by a software since they were conducted in English. *All* generated transcripts were then proof-read by another research assistant and edits made to ensure accuracy. The transcripts were then imported into NVivo 2020 (QSR International) software where data analysis was done. Thematic analysis was used to guide the analysis process using the inductive approach [15]. Two researchers read through the transcripts several times to familiarise themselves with the data. Thereafter, the researchers developed codes from the transcripts which were then discussed by the entire team and subsequently refined and revised. The codes were then defined, and several quotes representing different codes were highlighted to develop a code book. The code book was reviewed by the entire research team and modifications made as agreed. Codes that were linked or those that covered a similar subject were grouped to form sub-themes. Related sub-themes were then grouped to develop themes. The final themes obtained from the analysis are the major findings presented in this paper.
## Ethical considerations
The study obtained two stage ethical approval, first from Makerere University School of Public Health Research and Ethics Committee (# 923), and then the Uganda National Council for Science and Technology (# SS881ES). Written informed consent was obtained from all participants, and participation was voluntary. Anonymity was ensured as the participants’ identifying information was not audio recorded during data collection, neither were their names taken. Data was only accessed by the research team to maintain confidentiality and not used for any other purpose. The FGDs were held in accordance with the COVID-19 prevention guidelines in the country. Indeed, both the research assistants and participants were provided with face masks, hand sanitization was carried out frequently, and social distancing was observed. In addition, the household interviews were carried out in open spaces which ensured social distancing as a preventive measure for COVID-19. The FGDs and household interviews were also conducted during a low community COVID-19 transmission period between the first and second waves of the pandemic.
## Results
A total of 175 FGD participants were involved in the 14 FGDs, while there were 40 household survey participants and 31 key informants that also took part in the study. Findings from the study on the impact of COVID-19 lockdown measures of health and health care services are presented under four themes: maternal and reproductive health; child health; chronic disease services; and mental health.
## Maternal and reproductive health
Study participants reported that maternal and reproductive health services were negatively affected by the COVID-19 lockdown measures. These participants revealed that pregnant women and new mothers faced several challenges regarding their scheduled antenatal and postnatal visits. The hindrances included: bureaucracies involved in obtaining travel permits from the authorities; difficulty in accessing transport initially because of the suspension of public transport and curfews, and secondly due to the increased transport fares when public transport was allowed to operate at half capacity. These challenges related to travel affected both urban and rural settings, and were worse especially during the first phase of the lockdown. The need to obtain travel permits and other challenges in seeking healthcare particularly during the first phase of the lockdown resulted in many pregnant mothers delivering either at home, with local traditional birth attendants, or on the way to health facilities.
Health facility related factors such as fear of contracting COVID-19 or being forced to test for COVID-19 while attending antenatal or postnatal visits were reported by study participants. Respondents in urban areas were particularly affected by the temporary suspension of routine maternal health services (including antenatal and postnatal care) in some COVID-19 designated health facilities. Indeed, the government decision to transform some health facilities into strictly COVID-19 treatment sites meant that patients seeking other healthcare services had to look for alternative facilities.
Participants of the health practitioner FGDs and some key informants agreed that provision and access to antenatal and postnatal services was greatly affected by the COVID-19 lockdown. This was attributed to the reduced health workforce as a result of the high cost of transport during lockdown, and the fear of contracting COVID-19 from clients. This created a very high workload for the few health workers available at the facilities in both the urban and rural areas. In addition, more priority was given to COVID-19 rather than maternal health services at the time especially during the early waves of the pandemic. Rural health facilities were more affected by the shortage of health workers due to their remote locations, coupled with the absence of transportation means to take them to work.
Community FGDs and household interviews revealed that access and utilisation of family planning services also declined during the lockdown. This was due to change in family dynamics including the presence of men at home most of the time, and reduced priority for family planning services. Indeed, many families considered getting through the pandemic as their main concern, with less attention given to child spacing. Men being present in some homes meant that some women could not go for family planning services without their partner knowing as was the case before lockdown.
Participants of the health practitioner FGDs said that reduced access to family planning services (even when the services existed at health facilities) led to a rise in the number of unwanted pregnancies during the lockdown. In addition, participants of the FGDs and household survey, particularly those in rural areas, added that pregnancies and unsafe abortions among children under 18 years increased during the lockdown. This finding was attributed to children being out of school for a long time during the pandemic.
Furthermore, participants of the health practitioner FGDs in both the urban and rural settings reported that the adolescent sexual health education clinics at health facilities were suspended for a while during the lockdown. This suspension, which was primarily due to priority being given to COVID-19, limited the routine special support given to this group regarding sexual and reproductive health.
## Child health
Findings from the FGDs and household survey showed that child health was generally negatively affected by the lockdown measures as a result of the challenges that parents and guardians faced. Participants reported that many children, especially those from low-income households, lacked sufficient and nutritious food especially during the early months of the lockdown. The key informants agreed with this finding and added that it was because many parents survived on income earned on a day-to-day basis hence had little or no savings during lockdown. Indeed, it was reported that lockdown gave such parents no opportunity to prepare for how they would take care of their families. Children from urban households where most of the food was normally bought from markets were more affected than those in rural areas where food was predominantly grown at subsistence level.
The key informants as well as participants of the health practitioners and community health worker (CHW) FGDs reported that fewer routine paediatric services were provided due to the COVID-19 lockdown. For example, CHWs reported that they could not carry out their usual treatment of malaria, pneumonia and diarrhoea among children under 5 years due to various reasons such as lack of medicines, and fear of moving around in the community. Health practitioners in both urban and rural settings also reported that services such as immunisation and other paediatric activities were negatively affected by the lockdown. This was largely attributed to health systems and non-health related challenges such as difficulty in obtaining transport to health facilities, increased health practitioner workload, and lack of health facility supplies.
Participants of the community FGDs and household survey revealed that many community members resorted to self-medication including using local herbs and other local remedies to treat their children when they became ill. These practices reportedly reduced access to health services during the lockdown period.
## Chronic disease services
The FGDs and household survey revealed that patients with chronic conditions such as HIV/AIDS, tuberculosis and non-communicable diseases, who needed regular care and medicines, faced numerous challenges during the lockdown. Overall, the condition of patients with chronic conditions worsened, while others died primarily because they could not access the healthcare services that they needed. For instance, due to the abrupt onset of the lockdown, many HIV/AIDS patients were stuck in areas far from where they were registered to receive their medication hence they could not access medicine refills. The suspension of public transport also made it difficult for patients to access health facilities in both the urban and rural settings hence many missed their appointments.
Both community and health practitioner FGDs reported that there were medicine stock-outs for chronic conditions such as HIV/AIDS, diabetes and hypertension in the public health facilities. The major reason for these stock-outs included the reduced priority accorded to these conditions, as well as disruption in the routine supply chain system due to the pandemic. Rural health facilities which were harder to reach were more affected by stock-outs than urban ones.
In addition, participants of the FGDs and household survey, especially those in urban areas, reported that chronic conditions worsened due to insufficient health education about them during the lockdown. This was because most of the health education messages especially through radio, television and CHWs mainly focused on COVID-19. Other factors such as lack of exercise were cited as risk factors which increased the incidence of NCDs during the lockdown.
## Mental health
The study participants revealed that the lockdown led to social isolation from relatives and friends. This situation meant that many mental health support networks were disrupted, thus increasing depression, stress and anxiety within the community. These community networks were predominantly informal including family, friends and religious groups. For example, some participants said that suspension of religious gatherings took away their opportunity to connect with friends and reduce stress through routine communal prayer. Children’s mental health in both the urban and rural areas was also greatly affected during the lockdown due to social isolation from peers leading to depression.
In addition, loss of jobs and family income due to the lockdown increased stress among community members, especially those in urban areas who had informal jobs. This result increased depression among the community during the pandemic.
Participants of the community FGDs reported that fear of law enforcement, especially related to curfews, increased anxiety in the community in both the urban and rural settings. Community members were always worried that they would be beaten by law enforcers if they did not observe the curfew protocol or were found participating in communal gatherings. This concern greatly affected their mental health.
Participants of the health practitioner FGDs reported that the fear of contracting COVID-19 from the patients that they treated increased stress and anxiety among health workers. The health practitioners also said that the health facilities would sometimes run out of personal protective equipment, making them even more vulnerable to contracting the virus. Participants of the health practitioner FGDs in both the urban and rural settings added that they were stigmatised by the community because people perceived them to have COVID-19 hence they were avoided.
## Discussion
This study sought to describe the effects and impact of COVID-19 lockdown measures on health and health services in rural and urban Ugandan communities with the aim of providing evidence to guide future decision making during pandemics. Participants highlighted that the health of the population including access to various services were negatively affected by the COVID-19 lockdown measures. Participants reported facing difficulties in accessing maternal and reproductive health services such as antenatal care, postnatal care and family planning mainly due to transport challenges. Child health was also reported to have been affected by increased child malnutrition, reduced access to services and mental challenges. In addition, routine paediatric services such as immunisation and treatment of childhood illnesses declined during the lockdown. Participants also revealed that the prevention and management of chronic conditions such as HIV/AIDS and NCDs, as well as mental health was negatively affected by the implementation of the lockdown across the country. These effects, observed in both the urban and rural settings, were mainly due to measures such as suspension of both public and private transport, and night-time curfew that greatly reduced access to health services. The study findings emphasize that whereas lockdown measures may be needed to control disease spread during a pandemic, negative consequences affecting health may emerge [3]. These findings highlight the importance of careful consideration of lockdown measures to maximize benefit and minimise harm to the population during the management of pandemics.
Maternal health was negatively affected due to implementation of COVID-19 lockdown measures where access to antenatal and postnatal services declined. These findings concur with those from a study conducted in three East African countries where midwives reported that suspension of transport during the lockdown posed challenges for pregnant women accessing essential maternal health services [16] and may have increased pregnancy-related complications and maternal mortality. The suspension of both private and public transport in Uganda greatly hindered access to essential maternal health services including health facility deliveries, yet no alternative measures were instituted to ensure continuity of care. A study conducted in four sub-Saharan African countries showed that Kenya instituted measures such as multi-month dispensing of antenatal care related supplements and medications as well as telephone antenatal care sessions [17], strategies that were not considered in Uganda. Although the Ugandan government promised that pregnant women would get access to ambulances to transport them to health facilities if they contacted their local leaders, anecdotal evidence suggests that such ambulances were not readily available, forcing many women to resort to traditional birth attendants [16]. The decline of postnatal services during the COVID-19 lockdown was also affected by reduction in home visiting carried out by CHWs. These findings emphasize the importance of planning for the continuity of various maternal health related services during future responses to pandemics.
Our study revealed that access to family planning services declined due to lockdown measures which in turn led to an influx of unplanned pregnancies during the pandemic in both urban and rural areas. Indeed, there was reduced prioritization of family planning services during the lockdown periods in the country. In addition to transport challenges in accessing health facilities for family planning services, outreach activities including those supported by CHWs were reduced [16]. Such reduction in services during lockdown greatly affected access to health services including for family planning. As an example, adequately trained and equipped CHWs are proven to be a viable option especially in areas which do not have an established tele-medicine structure [18] yet their work was greatly affected by the lockdown. The extended closure of schools during lockdown also contributed to the increase in teenage pregnancies especially in rural areas. The extended stay at home exposed youth to risky sexual behaviours including cross-generational relationships [10]. Yet during the lockdown, there was absence of suitable support systems that could have raised their awareness about the dangers of engaging in sex and how to deal with such sexual encounters. Strategies to keep children and youth engaged during lockdowns to minimize their engagement in sexual activities should be planned for in the future management of pandemics with extended school closure.
The findings of our study showed a decline in child health such as an increase in malnutrition among children in both the urban and rural settings. Malnutrition among children in LMICs was projected in studies conducted during the early stages of the pandemic [19–23]. Indeed, our findings are similar to those from earlier studies conducted both in Uganda [6, 23–25] and other LMICs [26–28] which found increased levels of malnutrition among children. Malnutrition in children may have lowered their immunity and made them more susceptible to other diseases. Our study also revealed that routine paediatric services such as immunisation and treatment of childhood illnesses were negatively affected. Immunisation services offered at both health facilities and in the community were significantly slowed during the lockdown [17]. It is believed that low rates of immunisation across the country during the pandemic may have resulted in the recent polio outbreak in Uganda [29, 30]. Due to transport challenges in accessing health services during the lockdown, home treatment of childhood infections using herbs were common. The use of herbs in management of diseases has been reported in other studies conducted in Uganda [31]. Although lockdown measures were aimed at curbing the spread of COVID-19, future interventions should be implemented in ways that minimize severe and long-term impact on children’s health and wellbeing.
The lockdown measures were reported to have devastating effects on the health of those with chronic conditions such as HIV/AIDS, tuberculosis and NCDs. The increased challenges in accessing drugs among patients were highlighted in our study. Regarding access to HIV/AIDS drugs, our findings differ from those obtained from a study conducted in South-Western Uganda which found sustained access to health services among patients [32]. This difference could have been due to the presence of a community HIV/AIDS programme which ensured that HIV/AIDS drugs were delivered to the patients. However, the study [32] found increased stigma among HIV/AIDS patients which could have been due to the pandemic including the fear of dying from COVID-19 hence potentially leading to poor mental health outcomes. This finding therefore calls for enhancing attention to mental health services among HIV/AIDS patients and other vulnerable groups during pandemics. For other chronic conditions such as NCDs, other studies conducted in Uganda also established that lockdown measures disrupted the supply-chain of medicines and reduced health-seeking behaviours [33–35]. People with chronic conditions consistently need their medicines to remain in optimal health particularly during such pandemics as they are a vulnerable group. Equitable access to drugs and other health services among such patients should therefore be streamlined before, during and after implementation of lockdown measures.
Poor mental health outcomes due to social isolation, loss of income, and fear of contracting COVID-19 in both urban and rural settings emerged as a major finding in our study. This could be attributed to the abrupt and strict implementation of the lockdown measures which left people with little time to mentally and financially prepare for the extended periods of staying at home. The novelty of COVID-19, reports of high morbidity and mortality rates globally, and information overload from media may have contributed to the anxiety and fear of contracting the disease. Similar findings were documented in studies from other LMICs which showed increased prevalence of depression and anxiety during the pandemic [36–38]. The lockdown measures instituted in Uganda also worsened the already weak mental health infrastructure in the country. Therefore, strategies to support mental health during future lockdown periods such as improving family and social support systems are vital for general wellbeing.
Our study had some strengths and limitations. The triangulation of data sources and methods was a strength of our study. Indeed, our data was obtained using three different methods (KIIs, FGDs and household survey) and from a diverse range of participants at national, sub-national and community levels. This provided a comprehensive range of varied insights and perspectives on the effects of the COVID-19 lockdown. Furthermore, our study collected data from both urban and rural settings hence providing demographically varied experiences of the lockdown. A limitation of our study is that only 2 districts in the central region were involved hence some findings may not be generalisable to other areas in the country. Purposively sampling districts with high levels of COVID-19 could have also over-estimated the health impacts due to the lockdown. In addition, KIIs were conducted virtually due to the COVID-19 restrictions which may have limited the advantages of face-to-face interviews such as observing and probing on non-verbal cues.
## Conclusion
The COVID-19 lockdown measures negatively affected health and reduced access to maternal, reproductive and child health services. In addition, continuity of health services for patients with chronic conditions reduced. Future lockdowns aimed at minimising disease spread during pandemic response should ensure that their effects on health and access to health services are minimised. In addition, interventions should be accompanied by strategies to monitor unexpected consequences and propose mitigation measures. Policy makers and other stakeholders should also build the adaptive capacity of the health system to ensure it remains resilient during pandemics.
## References
1. 1World Health Organization. Coronavirus. Geneva, Switzerland: World Health Organization, 2020. [Cited 2021 10 October]. https://www.who.int/health-topics/coronavirus.. *Coronavirus* (2020.0)
2. 2Government of Uganda. COVID-19 Response Information Hub. Kampala: Government of Uganda. 2020. [Cited 2022 15 February]. https://covid19.gou.go.ug.
3. Haider N, Osman AY, Gadzekpo A, Akipede GO, Asogun D, Ansumana R. **Lockdown measures in response to COVID-19 in nine sub-Saharan African countries**. *BMJ Glob Health* (2020.0) **5** e003319. DOI: 10.1136/bmjgh-2020-003319
4. 4Ministry of Health—Uganda. COVID-19 Statistics. 2022. [Cited 2022 15 March]. https://covid19.gou.go.ug/statistics.html.
5. Roberton T, Carter ED, Chou VB, Stegmuller AR, Jackson BD, Tam Y. **Early estimates of the indirect effects of the COVID-19 pandemic on maternal and child mortality in low-income and middle-income countries: a modelling study**. *The Lancet Glob Health* (2020.0) **8** e901-8. DOI: 10.1016/S2214-109X(20)30229-1
6. Russo G, Jesus TS, Deane K, Osman AY, McCoy D. **Epidemics, Lockdown Measures and Vulnerable Populations: A Mixed-Methods Systematic Review of the Evidence of Impacts on Mother and Child Health in Low-and Lower-Middle-Income Countries**. *Int J Health Policy Manag* (2021.0). DOI: 10.34172/ijhpm.2021.155
7. Burt JF, Ouma J, Lubyayi L, Amone A, Aol L, Sekikubo M. **Indirect effects of COVID-19 on maternal, neonatal, child, sexual and reproductive health services in Kampala, Uganda**. *BMJ Glob health* (2021.0) **6** e006102. DOI: 10.1136/bmjgh-2021-006102
8. Atim MG, Kajogoo VD, Amare D, Said B, Geleta M, Muchie Y. **COVID-19 and Health Sector Development Plans in Africa: The Impact on Maternal and Child Health Outcomes in Uganda**. *Risk Manag Healthc Policy* (2021.0) **14** 4353-4360. DOI: 10.2147/RMHP.S328004
9. De Cao E, Sanders M. **The potential impact of COVID-19 on child abuse and neglect: The role of childcare and unemployment**. *VoxEU* (2020.0) 8
10. Sserwanja Q, Kawuki J, Kim JH. **Increased child abuse in Uganda amidst Covid‐19 pandemic**. *J Paediatr Child Health* (2021.0) **57** 188-191. DOI: 10.1111/jpc.15289
11. Nuwematsiko R, Nabiryo M, Bomboka JB, Nalinya S, Musoke D, Okello D. **Unintended socio-economic and health consequences of COVID-19 among slum dwellers in Kampala, Uganda**. *BMC Public Health* (2022.0) **22** 88. DOI: 10.1186/s12889-021-12453-6
12. 12Uganda Bureau of Statistics UBOS 2020. [Cited 2021 12 December]. https://africaopendata.org/en/dataset/uganda-bureau-of-statistics-statistical-abstract-2020.
13. 13Kasanje Town Council, 2022. [Cited 2022 03 February]. https://kasanje.go.ug/kasanje-town-council/.
14. 14Ministry of Health—Uganda 2018. [Cited 2022 03 February]. https://www.health.go.ug/cause/nkwanzi-rakai-lwengo-kalangala-mukono-buikwe-mpigi-butambala-butam-butamba-wakiso-mubende-lyantonde-n-n-n-sembabule-buvuma-kampala-m-m-a-complete-list-of-all-health-facilities-in-uganda/.
15. Braun V, Clarke V. **Using thematic analysis in psychology**. *Qual Res. Psych* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa
16. Pallangyo E, Nakate MG, Maina R, Fleming V. **The impact of covid-19 on midwives’ practice in Kenya, Uganda and Tanzania: A reflective account**. *Midwifery* (2020.0) **89** 102775. DOI: 10.1016/j.midw.2020.102775
17. Plotkin MK, Williams KM, Mbinda A, Oficiano VN, Nyauchi B, Walugembe P. **Keeping essential reproductive, maternal and child health services available during COVID-19 in Kenya, Mozambique, Uganda and Zimbabwe: analysis of early-pandemic policy guidelines**. *BMC Public Health* (2022.0) **22** 577. DOI: 10.1186/s12889-022-12851-4
18. Perry HB, Zulliger R, Rogers MM. **Community health workers in low-, middle-, and high-income countries: an overview of their history, recent evolution, and current effectiveness**. *Annu Rev Public Health* (2014.0) **35** 399-421. DOI: 10.1146/annurev-publhealth-032013-182354
19. Headey D, Heidkamp R, Osendarp S, Ruel M, Scott N, Black R. **Impacts of COVID-19 on childhood malnutrition and nutrition-related mortality**. *Lancet* (2020.0) **396** 519-521. DOI: 10.1016/S0140-6736(20)31647-0
20. Ntambara J, Chu M. **The risk to child nutrition during and after COVID-19 pandemic: what to expect and how to respond**. *Public Health Nutr* (2021.0) **24** 3530-3536. DOI: 10.1017/S1368980021001610
21. Osendarp S, Akuoku JK, Black RE, Headey D, Ruel M, Scott N. **The COVID-19 crisis will exacerbate maternal and child undernutrition and child mortality in low-and middle-income countries**. *Nature Food* (2021.0) **2** 476-84
22. Akseer N, Kandru G, Keats EC, Bhutta ZA. **COVID-19 pandemic and mitigation strategies: implications for maternal and child health and nutrition**. *Am J Clin Nutr* (2020.0) **112** 251-256. DOI: 10.1093/ajcn/nqaa171
23. Baral S, Rao A, Rwema JOT, Lyons C, Cevik M, Kågesten AE. **Competing health risks associated with the COVID-19 pandemic and early response: A scoping review**. *PLoS One* (2022.0) **17** e0273389. DOI: 10.1371/journal.pone.0273389
24. Mbazzi FB, Nalugya R, Kawesa E, Nimusiima C, King R, Van Hove G. **The impact of COVID-19 measures on children with disabilities and their families in Uganda**. *Disability & Society* (2020.0) 1-24. DOI: 10.1080/09687599.2020.1867075
25. Kansiime MK, Tambo JA, Mugambi I, Bundi M, Kara A, Owuor C. **COVID-19 implications on household income and food security in Kenya and Uganda: Findings from a rapid assessment**. *World Dev* (2021.0) **137** 105199. DOI: 10.1016/j.worlddev.2020.105199
26. Wamoyi J, Ranganathan M, Stöckl H. **COVID-19 social distancing measures and informal urban settlements**. *Bull World Health Organ* (2021.0) **99** 475-476. DOI: 10.2471/BLT.20.265942
27. Jayatissa R, Herath HP, Perera AG, Dayaratne TT, De Alwis ND, Nanayakkara HP. **Impact of COVID-19 on child malnutrition, obesity in women and household food insecurity in underserved urban settlements in Sri Lanka: a prospective follow-up study**. *Public Health Nutr* (2021.0) **24** 3233-3241. DOI: 10.1017/S1368980021001841
28. Aborode AT, Ogunsola SO, Adeyemo AO. **A crisis within a crisis: covid-19 and hunger in African children**. *Am J Trop Med Hyg* (2021.0) **104** 30-31. DOI: 10.4269/ajtmh.20-1213
29. Alleman MM, Jorba J, Henderson E, Diop OM, Shaukat S, Traoré MA. **Update on vaccine-derived poliovirus outbreaks—worldwide, January 2020–June 2021**. *MMWR Morb Mortal Wkly Rep* (2021.0) **70** 1691-1699. DOI: 10.15585/mmwr.mm7049a1
30. 30Ministry of Health–Uganda. 2022. MOH launches house to house polio vaccination campaign. [Cited 2022 26 March]. https://www.health.go.ug/2022/01/18/moh-launches-house-to-house-polio-vaccination-campaign/.
31. Taremwa IM, Ashaba S, Kyarisiima R, Ayebazibwe C, Ninsiima R, Mattison C. **Treatment-seeking and uptake of malaria prevention strategies among pregnant women and caregivers of children under-five years during COVID-19 pandemic in rural communities in South West Uganda: a qualitative study**. *BMC Public Health* (2022.0) **22** 373. DOI: 10.1186/s12889-022-12771-3
32. West NS, Ddaaki W, Nakyanjo N, Isabirye D, Nakubulwa R, Nalugoda F. **“A Double Stress”: The Mental Health Impacts of the COVID-19 Pandemic Among People Living with HIV in Rakai, Uganda**. *AIDS Behav* (2022.0) **26** 261-265. DOI: 10.1007/s10461-021-03379-6
33. Musinguzi G, Ndejjo R, Aerts N, Wanyenze RK, Sodi T, Bastiaens H. **The Early Impact of COVID-19 on a Cardiovascular Disease Prevention Program in Mukono and Buikwe Districts in Uganda: A Qualitative Study**. *Global Heart* (2021.0) 16. DOI: 10.5334/gh.917
34. Tumwesigye NM, Denis O, Kaakyo M, Biribawa C. *Effects of the COVID-19 pandemic on health services and mitigation measures in Uganda* (2021.0)
35. Schwartz JI, Muddu M, Kimera I, Mbuliro M, Ssennyonjo R, Ssinabulya I, Semitala FC. **Impact of a COVID-19 national lockdown on integrated care for hypertension and HIV**. *Glob Heart* (2021.0) **16** 9. DOI: 10.5334/gh.928
36. Mamun MA, Sakib N, Gozal D, Bhuiyan AI, Hossain S, Bodrud-Doza M. **The COVID-19 pandemic and serious psychological consequences in Bangladesh: a population-based nationwide study**. *J Affect Disord* (2021.0) **279** 462-472. DOI: 10.1016/j.jad.2020.10.036
37. Kola L, Kohrt BA, Hanlon C, Naslund JA, Sikander S, Balaji M. **COVID-19 mental health impact and responses in low-income and middle-income countries: reimagining global mental health**. *Lancet Psychiatry* (2021.0) **8** 535-550. DOI: 10.1016/S2215-0366(21)00025-0
38. Kumar A, Nayar KR. **COVID 19 and its mental health consequences**. *J Ment Health* (2021.0) **30** 1-2. DOI: 10.1080/09638237.2020.1757052
|
---
title: 'Prevalence of overweight and obesity in Nigeria: Systematic review and meta-analysis
of population-based studies'
authors:
- Innocent Ijezie Chukwuonye
- Kenneth Arinze Ohagwu
- Okechukwu Samuel Ogah
- Collins John
- Efosa Oviasu
- Ernest Ndukaife Anyabolu
- Ignatius Ugochukwu Ezeani
- Gabriel Uche Paschal Iloh
- Miracle Erinma Chukwuonye
- Caleb Ogechi Raphael
- Uwa Onwuchekwa
- Umezurike Hughes Okafor
- Clement Oladele
- Emmanuel Chukwuebuka Obi
- Chimezie Godswill Okwuonu
- Okechukwu Iheji
- Ogbonna Collins Nwabuko
- Martin Anazodo Nnoli
- Ikechi G. Okpechi
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021772
doi: 10.1371/journal.pgph.0000515
license: CC BY 4.0
---
# Prevalence of overweight and obesity in Nigeria: Systematic review and meta-analysis of population-based studies
## Abstract
In Nigeria, several studies have assessed the prevalence of overweight/obesity with different reports. The purpose of this study was to use a systematic review and meta-analysis to analyze these overweight and obesity reports from different locations in Nigeria over the last ten years. In addition, there was a dearth of systematic reviews and meta-analyses on the prevalence, trends, and demographic characteristics of overweight and obesity in the country. This was a systematic review and meta-analysis of cross-sectional population-based studies among adult Nigerians on the prevalence of overweight/ obesity (defined by body mass index) published from January 2010 to December 2020. Relevant abstracts were scrutinized and articles that included adults of all age groups and were not restricted to a particular group of people (e.g. university community) were selected. Each article was scrutinized by more than 2 authors before selection. The prevalence of overweight/obesity among all participants, among men and among women in Nigeria and its 6 geopolitical zones was determined. All analyses were performed using STATA version 14 (Stata Corp. College Station, Texas, USA). Thirty-three studies were selected and the number of participants was 37,205. The estimated prevalence of overweight and obesity was $27.6\%$, and $14.5\%$ respectively. The prevalence of overweight among men and among women was $26.3\%$ and $28.3\%$ respectively and, the prevalence of obesity among men and women was $10.9\%$ and $23.0\%$ respectively. The prevalence of overweight in the 6 geopolitical zones was Southeast $29.3\%$, Southwest $29.3\%$, South-south $27.9\%$, Northwest $27.2\%$, North-central $25.3\%$, Northeast $20.0\%$ and obesity South-south $24.7\%$, Southeast $15.7\%$, Southwest $13.9\%$, Northwest $10.4\%$, North-central $10.2\%$, Northeast $6.4\%$. Egger’s tests showed no statistically significant publication bias among the studies that reported the overweight and obesity prevalence respectively ($$p \leq 0.225$$, P 0.350). The prevalence of overweight/obesity in *Nigeria is* high. The southern geopolitical zones had a higher prevalence of overweight/obesity.
## Introduction
The prevalence of overweight and obesity is on the increase worldwide, with serious public health implications. In the last three and half decades, the prevalence of obesity has increased steadily, with regard to the standard established by the World Health Organization (WHO) body mass index (BMI) categorization of obesity. The steady increase in the prevalence of overweight and obesity is global and the rate of increase in African countries like *Nigeria is* not lower than that observed in developed countries of the world [1,2]. In 2016, the WHO reported that about 1.9 billion adults were overweight (using BMI classification) and about a third of these (650 million) were obese globally. The prevalence of overweight was $38\%$ ($9\%$ among men and $40\%$ among women), while the prevalence of obesity was $13\%$ ($11\%$ among men and$15\%$ among women) in adults aged 18 years and above in the WHO report [3,4].
The Global Burden of Disease Study in 2017 evaluated 84 risk factors and obesity was reported as one of five leading environmental, behavioral, and metabolic risks that drive injury and disease worldwide. Obesity was also observed to have the greatest relative increase in exposure since 1990 [5]. Obesity and being overweight are associated with a greater risk of non-communicable diseases such as cardiovascular diseases, diabetes mellitus, metabolic syndrome, chronic kidney disease, cancer, and musculoskeletal disorders. Cardiovascular disease was responsible for $41\%$ of obesity-related deaths and $34\%$ of obesity-related disability-adjusted life-years in obese people worldwide. In 2015, diabetes was the second largest cause of death from obesity-related causes. [ 6]. In Nigeria, some of the co-morbidities reported included type 2 diabetes mellitus, hypertension, and dyslipidemia [7].
The theoretical framework for available multilevel factors driving adult obesity classifies the determinants of obesity into three levels: individual levels (genetic, ethnicity, socioeconomic, etc.), environmental factors, and lifestyle/behavioral/social factors. Changes in the risk factors at these different levels in the system affect the development of obesity in individuals [8].
In Nigeria, some risk factors for obesity have been reported and these; include gender, age, locality (urban community), decreased physical activity, educational status, high income, and diet [9–12]. Increased dietary consumption of energy-dense foods, high levels of refined sugar and saturated fats (fast food) and sedentary lifestyles are recognized as some of the major causes of the increased prevalence of obesity in Nigeria [10]. There has also been a rapid increase in the number of eateries that sell fast food in most urban communities in the country within the last three decades with associated increased patronage by the upper and middle class that can afford it. A study in Nigeria reported that the prevalence of obesity in low, middle, and upper-income classes were $12.2\%$, $16\%$, and $20\%$, respectively [13], indicating that the prevalence was higher in the upper and middle class in the country.
Nigeria has strategic direction documents on promoting physical activities, nutritional counseling, adhering to dietary guidelines, and implementing mandatory nutritional labeling. All these are captured in the country’s health and nutritional policies. The problem however is that more attention is currently being paid to undernutrition [14]. In order to convince policy-makers to pay more attention to overweight and obesity reliable statistics highlighting obesity as a serious public health problem in Nigeria are needed. The goal of this study was to assess the prevalence of overweight and obesity in Nigeria and its six geopolitical zones using data from multiple population-based studies conducted across the country. In addition, we also intended to test the hypothesis that the prevalence of obesity had increased in the last decade when compared to preceding decades. A recent reliable estimate of the prevalence of overweight and obesity among the adult population in the country will contribute to the statistics needed to sway policymakers in the country to take urgent and substantial action on the increasing prevalence of obesity.
## Methodology
This was a systematic review and meta-analysis study and the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [15] were adopted and the PRISMA checklist adhered to http://www.prisma-statement.org/. The study focused on overweight and obesity defined by BMI and BMI was classified as follows: overweight, BMI of 25–29.9 kg/m2; and obesity, BMI of 30 kg/m 2 and above [16].
## Literature review
A literature search for population-based studies on overweight and obesity, published from January 1, 2010, to December 31st, 2020 on adult Nigerians using the search terms, “obesity” “body mass index” and “overweight”. These search terms were used in Google Scholar, PubMed, and Embase search engines to retrieve all potentially relevant English articles. The key search words were combined with Nigeria. Bibliographies of some of the authors were also searched and retrieved abstracts were also scrutinized. In order to eliminate difficulties in analyzing the data, we only paid attention to surveys that made use of BMI in the definition of overweight and obesity, or where both BMI and another method were used only the BMI results were extracted. Waist circumference is another common method of assessing obesity in Nigeria. However, in most studies that were available on abdominal obesity; the protocols for measuring waist circumference were not the same, making it difficult to compare most of the studies. In addition, there is no universally acceptable cut-off criterion for defining abdominal obesity for men and women, due to the existence of different criteria (e.g. The Adult Treatment Panel III and the International Diabetes Foundation) [17]. This is why we focused only on studies that used BMI in defining overweight and obesity.
## Validation of search results
Search results were validated in three stages, as follows:
## Data extraction
Data extracted from the studies that met the selection criteria included the study community, the geopolitical zone of Nigeria where the study was carried out, year of publication of the article, study design, sample size, mean age (years) of participants in the study, and the prevalence of overweight and obesity among adults in the study based on BMI only.
## Statistical analysis
The standard error (SE) and effect size (ES) of the prevalence estimates were calculated using metaprop_one which is the update command of metaprop for performing a meta-analysis of proportions. Heterogeneity chi-square χ2 test and τ2 Tau2 statistic (τ2) was used to assess heterogeneity and the estimate of between-study variance. P-values of less than 0.05 were considered as heterogeneity. The I2 statistic was also done for each of the pooled estimates to test for variation in ES attributable to heterogeneity. As the differences between the studies were very large (92–$98\%$ inconsistency), a random-effects model was used to pool the prevalence of overweight and obesity in Nigeria. Metareg, which performs random-effects meta-regression, was done to assess heterogeneity and combinability. Freeman-Tukey transformations were done to stabilize the geopolitical variances to arrive at the overall prevalence of overweight and obesity in Nigeria. Geopolitical zone-wise pooled estimates weighted by population size in each study place within a given zone (Southeast; South-south, South-west; North-central, Northeast and North-west) for the prevalence of overweight and obesity were also calculated. All analyses were done using STATA version 14 (Stata Corp. College Station, Texas, USA).
## Publication bias
Inspection of the funnel plot and Egger’s bias test were used to assess potential bias in the study [18].
The publication bias among studies included for overweight and obesity was determined using the funnel plot and Egger’s tests. The results of Egger’s tests for the funnel plot showed that there was no statistically significant publication bias in the studies that reported the overweight and obesity prevalence, respectively ($$p \leq 0.225$$ for overweight and $$p \leq 0.350$$ for obesity). ( See Figs 6 and 7 for overweight people and Figs 8 and 9 for obese people).
**Fig 6:** *Distribution of studies included in the prevalence of overweight among adults in Nigeria in the meta funnel plot.* **Fig 7:** *Egger’s test for detection of publication bias for studies included in the prevalence of overweight among adults in Nigeria.* **Fig 8:** *Distribution of studies included in the prevalence of obesity among adults in Nigeria in the meta funnel plot.* **Fig 9:** *Egger’s test for detection of publication bias for studies included in the prevalence of obesity among adults in Nigeria.*
## Study selection
The number of abstracts on population-based studies on overweight and obesity identified in the study from the databases was 1,148 and 153 articles were identified from bibliographies. A total of 922 abstracts were excluded from 1,031 non-duplicate abstracts and a total of 109 original articles were retrieved. Thirty of the 109 retrieved articles were from the Southeast (SE), 17 from the South-south (SS), 33 from the Southwest (SW), 13 from the North-central (NC), 6 from the Northeast (NE), and 10 from the Northwest (NW) geopolitical zone. Thirty-three articles that fulfilled all the inclusion criteria were finally selected (PRIMA diagram Fig 1).
**Fig 1:** *PRISMA flow diagram.Graphical representation of the flow of citations reviewed in the course of the systematic review and meta-analysis. Abbreviation: Preferred items for systematic review and meta-analysis (PRISMA).*
## Study characteristics
The total number of the participants that took part in the 33 included articles was 37,205 (SE = 17,422, SS = 5,313, SW = 8,488, NC = 1,414, NE = 2,822, NW 1,566) The 33 studies were population-based observational cross-sectional designs. They were all community-based studies; however, 2 of the studies were state-wide studies [10,11] and there was no geopolitical zone, regional or nationwide study. Ten of the articles [9–11,19–25] ($30.30\%$) were from the SE, 5 ($15.15\%$) were from the SS [12,26–29], 9 ($27.27\%$) were from the SW [30–38], 3 ($9.09\%$) were from the NC [39–41], 2 ($6.06\%$) were from the NE [42,43]and 4 ($12.12\%$) from the NW geopolitical zone [44–47]. Five of the 33 studies [19,27,29,32,34] did not report the prevalence of overweight among the participants. The prevalence of overweight and obesity among the men and women genders with the sample sizes of the men and women that took part in the study was reported by 10 and 15 studies, respectively (Table 1).
**Table 1**
| Zones | Ref. | Author (year) | Sample size | Study design | Mean age(years) | Obesity prevalence (%) | Overweight prevalence (%) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| South East | 9 | Ijoma et al. (2019) | 605 | Cross sectional | 44.5 | 19.5(M = 7.9,F = 24.9) | 29.4(M = 28.3,F = 30.0) |
| South East | 10 | Chukwuonye et al. (2015) | 2928 | Cross sectional | 41.7 ± 18.5 | 12 b.3(M = 7.8,F = 16.4) | 28.2(M = 28.8,F = 27.7) |
| South East | 11 | Chigbu et al. (2018) | 6628 | Cross sectional | | 6.8 | 19.0 |
| South East | 19 | Gladys et al. (2011) | 218 | Cross sectional | | 13.3(M = 10.1,F = 14.8) | |
| South East | 20 | Ezeala-Adikaibe et al. (2016) | 774 | Cross sectional | 43.9 | 17.8(M = 7.2,F = 23.7) | 27.9(M = 25.5,F = 29.2) |
| South East | 21 | Fatai and Udoji (2015) | 1521 | Cross sectional | 43.98 | 26.9(M = 19.6,F = 36.0 | 31.2(M = 32.3,F = 29.8) |
| South East | 22 | Ijoma et al. (2020) | 210 | Cross sectional | 51.24 ± 16.24 | 10.9(M = 10.9,F = 10.9) | 28.0(M = 27,F = 28) |
| South East | 23 | Okafor et al. (2011) | 898 | Cross sectional | 48.7 ± 12.9 | 21.2 | 37.8 |
| South East | 24 | Ulasi et al. (2010) | 1458 | Cross sectional | 43.8 ± 13.7 | 17.3 | 31.6 |
| South East | 25 | Ulasi et al. (2013) | 2182 | Cross sectional | 43.7 ± 13.2 | 14.9 | 31.9 |
| South-south | 12 | Adienbo et al. (2012) | 304 | Cross sectional | 37.66 ± 14.94 | 49.34(M = 35.51,F = 64.49) | 22.4 |
| South-south | 26 | Nwafor et al. (2015) | 250 | Cross sectional | | 16.4(M = 5.6,F = 10.8) | 41.2(M = 15.2,F = 26.0) |
| South-south | 27 | Egbe et al. (2014) | 1134 | Cross sectional | | 27.4(M = 22.3,F = 34.2) | |
| South-south | 28 | Isara et al. (2015) | 845 | Cross sectional | 56.4 ± 16.3 | 10.6 | 21.8 |
| South-south | 29 | Ekpenyong et al. (2012) | 2780 | Cross sectional | | 25.00 | |
| South-west | 30 | Chinedu et al. (2013) | 489 | Cross sectional | | 18.0 | 31.0 |
| South-west | 31 | Raimi and Dada (2018) | 552 | Cross sectional | 39.9 ±15.5 | 18.3(M = 8.8,F = 27.7) | 34.8(M = 30.9,F = 37.6) |
| South-west | 32 | Oluyombo et al. (2015) | 750 | Cross sectional | 61.7 ± 18.2 | 8.5 | |
| South-west | 33 | Abiodun et al. (2014) | 776 | Cross sectional | 42.6 ± 14.3 | 17.5 | 29.9 |
| South-west | 34 | Oluyombo et al. (2016) | 1083 | Cross sectional | 55.1 ± 19.9 | 5.7 | |
| South-west | 35 | Amira et al. (2012) | 1368 | Cross sectional | 41.9 ± 12.9 | 22.2(M = 15.7,F = 29.5) | 32.7(M = 33.3,F = 31.9) |
| South-west | 36 | Akinwale et al. (2013) | 2434 | Cross sectional | | 19.1 | 36.2 |
| South-west | 37 | Adebayo et al. (2014) | 777 | Cross sectional | 36.3 ± 14.3 | 8.4(M = 10.3,F = 6.9) | 20.8(M = 22.3,F = 19.0) |
| South-west | 38 | Asekun-Olarinmoye et al. (2013) | 259 | Cross sectional | 49.7 ± 1.6 | 11.5 | 19.6 |
| North central | 39 | Etukumana et al | 750 | Cross sectional study | 39.42±16.17 | 8 (M = 3.2, F 13.1) | 23.3 |
| North central | 40 | Sola et al. (2011) | 435 | Cross sectional | 24.2 ± 0.2 | 4.0 | 22.0 |
| North central | 41 | Adediran et al. (2012) | 229 | Cross sectional | | 22.3(M = 8,F = 36.2) | 32.3 |
| North-east | 42 | Oyeyemi et al. (2012) | 1818 | Cross sectional | 32.3 ± 10.0 | 8.1 | 22.8 |
| North-east | 43 | Adedoyin et al. (2012) | 1004 | Cross sectional | 41.5 ± 13.5 | 3.8 | 15.4 |
| North-west | 44 | Wahab et al. (2011) | 300 | Cross sectional | 37.6 ± 10.6 | 21.0(M = 9.3,F = 29.8) | 53.3(M = 41.9,F = 62.0) |
| North-west | 45 | Dahiru and Ejembi, (2013) | 199 | Cross sectional | 39.9 ± 15.6 | 7.0 | 26.9 |
| North-west | 46 | Makusidi et al. (2013) | 535 | Cross sectional | 37.0 ± 17.0 | 6.7 | 12.3 |
| North-west | 47 | Ramalan et al. (2019) | 532 | Cross sectional | 38.9 ± 15.9 | 9. 2(M = 4.3,F = 13.6) | 20.9(M = 15.4,F = 24.4) |
Two articles were excluded during the data extraction. One of the articles [48] was from the same community-based study from which an article had been selected [10]. The second study included pregnant women in the study and also excluded people living with diabetes mellitus and other chronic diseases [49].
## The prevalence of overweight in Nigeria
The prevalence of overweight from the studies ranged from $12.3\%$ in NW 46 to $41.2\%$ in SS26 geopolitical zones. Heterogeneity was significantly present among the geo-political zones. After stabilizing the regional data using Freeman-Tukey transformations, the overall prevalence of overweight found in Nigeria from this study was $27.6\%$ ($95\%$ CI: 24.8–30.5; I2 = 96.75, $P \leq 0.001$). The prevalence of overweight people was highest in SE and SW, with both having $29.3\%$ [$95\%$ CI: (24.7–34.2) and (24.8–34.0) respectively]. While NE had the lowest prevalence $20.0\%$ ($95\%$ CI: 18.6–21.5) (See Table 2, Figs 2 and 3 for more details).
**Fig 2:** *The prevalence of overweight among adults in Nigerians.* **Fig 3:** *The prevalence of overweight in the six geopolitical zones in Nigeria.* TABLE_PLACEHOLDER:Table 2
## The prevalence of obesity in Nigeria
The prevalence of obesity from the studies ranged between $4.0\%$ in NC 38 to $49.3\%$ in SS9 (The overall pooled crude prevalence of obesity in Nigeria was $14.5\%$ ($95\%$ CI: 11.8–17.4; I2 = $98.2\%$, $P \leq 0.001$). There was a significant difference in the pooled prevalence across the geopolitical zones. SS zone had the highest prevalence of $24.7\%$ ($95\%$ CI: 15.9–34.6; I2 = $98.1\%$, $P \leq 0.001$). NE had the lowest prevalence of obese people at $6.4\%$ ($95\%$ CI: 5.5–7.3; I2 = $0\%$). ( See Table 3, Figs 4 and 5 for more details).
**Fig 4:** *The prevalence of obesity among adults in Nigeria.* **Fig 5:** *The prevalence of obesity among adults in the six geopolitical zones of Nigeria.* TABLE_PLACEHOLDER:Table 3
## The prevalence of overweight among men and among women in Nigeria
The pooled prevalence of overweight among men and women was determined by 10 studies that met the inclusion criteria. The prevalence of overweight among men was $26.3\%$ ($95\%$ CI: 22.9–29.9; I2 = $82.83\%$, $P \leq 0.001$) in Nigeria. Among women the prevalence was $28.3\%$ ($95\%$ CI: 25.6–31.2; I2 = $77.61\%$, $P \leq 0.001$). Geopolitical zone-wise, SS had the lowest prevalence of overweight among men at $15.2\%$ ($95\%$ CI: 9.1–24.3), while the South-east region had the highest at $29.2\%$ ($95\%$ CI: 26.9–31.6; I2 = $31.33\%$, $P \leq 0.05$). Among women, the prevalence of overweight was $24.4\%$ (20.0–29.4) in the NW geopolitical zone as the lowest and $29.2\%$ ($95\%$ CI: 19.4–40.0) in the SW as the highest. ( See Tables 4 and S1 and S1 and S2 Figs).
**Table 4**
| Unnamed: 0 | Male | Male.1 | Female | Female.1 |
| --- | --- | --- | --- | --- |
| | Prevalence % (95% CI) | I2%, p-value | Prevalence % (95% CI) | I2%, p-value |
| Nationwide overweight | 26.3 (22.9–29.9) | 82.83, 0.00 | 28.3 (25.6–31.2) | 77.61, 0.00 |
| Geopolitical zone | Geopolitical zone | Geopolitical zone | | |
| North-central | - | - | - | - |
| North-east | - | - | - | - |
| North-west | 15.40 (11.1–20.9) | - | 24.4 (20.0–29.4) | - |
| South-east | 29.2 (26.9–31.6) | 31.33, 0.21 | 28.6 (27.1–30.2) | 0.00, 0.81 |
| South-west | 28.8 (21.9–36.1) | - | 29.2 (19.4–40.0) | - |
| South-south | 15.2 (9.1–24.3) | - | 26.0 (20.0–33.2) | - |
## The prevalence of obesity among men and among women in Nigeria
The pooled prevalence of obesity among men and among women was determined by 15 studies. A higher pooled prevalence of obesity was observed among women $23.0\%$ ($95\%$ CI: 17.2–29.4; I2 = $97.0\%$, $P \leq 0.001$), compared to the men $10.9\%$ ($95\%$ CI: 17.2–29.4; I2 = $94.2\%$, $P \leq 0.001$) in Nigeria. Regarding the different geo-political zones the rates of obesity prevalence were $13.6\%$ ($95\%$ CI: 10.3–17.8) in the NW as the lowest and $34.8\%$ ($95\%$ CI: 11.3–63.3) in the SS as the highest. Among the men the lowest prevalence was $4.0\%$ ($95\%$ CI: 2.4–6.0) in the NC while the highest was $19.8\%$ ($95\%$ CI: 8.0–35.1) in the SS. ( See Tables 5 and S2 and S3 and S4 Figs).
**Table 5**
| Unnamed: 0 | Male | Male.1 | Female | Female.1 |
| --- | --- | --- | --- | --- |
| | Prevalence % (95% CI) | I2%, p-value | Prevalence % (95% CI) | I2%, p-value |
| Nationwide obesity | 10.9 (17.2–29.4) | 94.2, 0.00 | 23.0 (17.2–29.4) | 97.0, 0.00 |
| Geopolitical zone | Geopolitical zone | | | |
| North-central | 4.0(2.4–6.0) | - | 17.8 (14.5–21.4) | - |
| North-east | - | - | - | - |
| North-west | 4.3 (7.6–14.8) | - | 13.6 (10.3–17.8) | - |
| South-east | 10.3 (5.7–15.9) | 93.1, 0.00 | 20.8 (14.1–28.4) | 95.7, 0.00 |
| South-west | 11.7 (7.8–16.3) | - | 20.1 (7.1–37.5) | - |
| South-south | 19.8 (8.0–35.1) | - | 34.8 (11.3–63.3) | - |
## Discussion
This systematic review and meta-analysis highlighted the prevalence of overweight and obesity among adults in Nigeria and its 6 geopolitical zones based on published studies from January 1, 2010, to December 31, 2020. It is the first systematic review and meta-analysis in the country that delved into the prevalence of overweight and obesity among adults in each of the 6 geopolitical zones in Nigeria from the literature search. The total number of participants that took part in this study was 37,205 and the estimated pooled prevalence of overweight among adults in Nigeria ranged from $12.3\%$ to $41.2\%$ and the prevalence of obesity among adults in Nigeria ranged from $4.0\%$ to $49.3\%$. In an earlier systematic review [17], the prevalence of overweight among adults in Nigeria ranged from $20.3\%$–$35.1\%$, while the prevalence of obesity ranged from $8.1\%$–$22.2\%$. The observed differences between both studies might be due to the reported rising level of overweight and obesity [3] and also because this study captured more recent studies and covered a wider period.
The estimated pooled prevalence of overweight and obesity among adults in Nigeria in this study was $27.6\%$, and $14.5\%$ respectively. There is a dearth of systematic reviews and meta-analyses on the prevalence of overweight and obesity in Nigeria. In an earlier meta-analysis study by Abubakari et al, [50] the prevalence of obesity among Nigerian adults was $8.8\%$ (CI 7.0–10.6) in 2000, and obesity in Ghanaian adults (> or = 25 years) was $14.1\%$ (CI 13.1–$15.1\%$) in 1998, A comparison of the reports showed that the prevalence of obesity had risen in Nigeria. A meta-analysis in Ghana by Ofori-arenso et al [51] reported a prevalence of overweight and obesity among adults in Ghana of $25.4\%$ and $17.1\%$, respectively. The observed results were close to those observed in this study. The prevalence of overweight and obesity in Nigeria and Ghana were similar which suggested this might be the pattern in the West African region. In a meta-analysis by Kassie et al [4] in Ethiopia, involving published studies on the prevalence of overweight and obesity among adults in Ethiopia covering almost the same period as our study (from January 2010 –March 2020) the estimated pooled prevalence rate of overweight and obesity was $19\%$ and $5.4\%$ respectively. The observed results were much lower than those observed in Nigeria and Ghana (West African countries) and tend to suggest that the prevalence of overweight and obesity among adults in West African countries might be markedly higher than that obtainable in countries in the Horn of Africa. Ofori-arenso et al [51] and Kassie et al [4] also reported an increased prevalence of overweight and obesity in Ghana and Ethiopia respectively. These findings strongly suggest that there is a rapid rise in the prevalence of overweight and obesity in most or all African countries primarily due to lifestyle modifications and other factors. The prevalence rate of overweight and obesity observed among adults in Nigeria in this study was not lower than that reported by the WHO [3] in 2016. This was a pointer that the prevalence of overweight and obesity in Nigeria and some other African countries like Ghana was on par with the WHO [3] reports.
The prevalence of overweight among men and among women was $26.3\%$ and $28.3\%$ respectively. In addition, the prevalence of obesity among men and women was $10.9\%$ and $23.0\%$ respectively. These results showed that the prevalence of overweight and obesity was higher among women. These findings were in keeping with those observed by the WHO [3] and other studies [4,10] from Africa. Magemba et al [52] reported that the use of hormonal contraceptives and marriage were among the risk factors for overweight and obesity among women in Zimbabwe. However, more research is needed to determine the reasons for the difference in the prevalence of obesity between men and women in African countries. This is the first study from the literature search that is reporting the prevalence of overweight/obesity among men and women in Nigeria and there was none to compare our results with from the literature search.
Heterogeneity was observed in the prevalence of overweight/obesity among adults in the 6 geopolitical zones in Nigeria. The prevalence of overweight was (SE $29.3\%$, SW $29.3\%$, SS $27.9\%$, NW $27.2\%$, NC $25.3\%$, NE $20.0\%$) and obesity (SS $24.7\%$, SE $15.7\%$, SW $13.9\%$, NW $10.4\%$, NC $10.2\%$, NE $6.4\%$). In both overweight and obesity, the differences between the regions were statistically significant ($p \leq 0.05$). The southern geopolitical zones of the country had higher prevalence rates of overweight/obesity when compared to the northern geopolitical zones. However, there was no previous study to compare the results from the literature search. The reasons for the higher prevalence of overweight and obesity in the southern geopolitical zones were multifactorial, and these included higher patronage of fast food in the southern geopolitical zone, increased sedentary lifestyle due to more affluence and industrialization. In addition, differences in dietary habits and a higher level of education in the southern region may also be part of the risk factors.
## Conclusion
The prevalence of overweight and obesity in Nigeria was high and had increased over the decades. There is a need to stem the trend because the cost implications are huge. The cost implications of overweight and obesity can be classified as direct or indirect costs. The costs of preventive, diagnostic, and treatment services constitute the direct cost and the cost of morbidity and mortality constitute the indirect cost. Morbidity costs are defined as the income lost from decreased productivity, restricted activity, absenteeism, and hospital admission days. The value of future income lost by the premature death of obese patients is known as mortality costs [17]. In the United States, obesity-related medical care costs in 2008, were estimated to be $147 billion and the annual nationwide productivity costs of obesity-related absenteeism ranged between $3.38 billion ($79 per obese individual) and $6.38 billion ($132 per obese individual) [53]. The direct and indirect cost of obesity in *Nigeria is* not known but is expected to be huge considering the high prevalence of obesity in Nigeria and also the fact that *Nigeria is* the most populous black nation on Earth. Obesity’s increased prevalence in Nigeria, however, is matched by rising levels of obesity’s co-morbidities, such as hypertension and diabetes mellitus [54,55]. There is a need for the various levels of governments and other key stakeholders in Nigeria and other African countries to invest more in preventive, diagnostic, and treatment of obesity and its comorbidities.
## Limitations
Five out of the 33 selected studies did not report the prevalence of overweight among the study participants. In addition, only 10 and 15 studies adequately reported the prevalence of overweight and obesity, respectively, among men and women in their various study populations.
## Recommendation
Based on the high and rising levels of overweight and obesity observed in this study, we urge that policymakers in Nigeria and other sub-Saharan African countries pay more attention to overweight and obesity due to the fact that they pose serious public health problems.
## References
1. Templin T, Cravo Oliveira Hashiguchi T, Thomson B, Dieleman J, Bendavid E. **The overweight and obesity transition from the wealthy to the poor in low- and middle-income countries: A survey of household data from 103 countries**. *PLoS Med* (2019.0) **16** e1002968. DOI: 10.1371/journal.pmed.1002968
2. **Trends in obesity and diabetes across Africa from 1980 to 2014: an analysis of pooled population-based studies**. *Int J Epidemiol.* (2017.0) **46** 1421-1432. DOI: 10.1093/ije/dyx078
3. 3WHO. Media Centre [Accessed December 3, 2020]: Obesity and overweight 2016. Available at: http://www.who.int/mediacentre/factsheets/fs311/en/.
4. Kassie AM, Abate BB, Kassaw MW. **Prevalence of overweight/obesity among the adult population in Ethiopia: a systematic review and meta-analysis**. *BMJ Open* (2020.0) **10** e039200. DOI: 10.1136/bmjopen-2020-039200
5. **Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016.**. *Lancet* (2017.0) **390** 1345-1422. DOI: 10.1016/S0140-6736(17)32366-8
6. **Health effects of dietary risks in 195 countries, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. GBD 2017 Diet Collaborators.**. *Lancet* (2019.0) **393** 1958-1972. DOI: 10.1016/S0140-6736(19)30041-8
7. Iloh G.U.P, Ikwudinma A.O, Obiegbu N.P.. **Obesity and Its Cardio-metabolic Co-morbidities Among Adult Nigerians in a Primary Care Clinic of a Tertiary Hospital in South-Eastern, Nigeria.**. *Family Med Prim Care.* (2013.0) **2** 20-6. DOI: 10.4103/2249-4863.109936
8. Sartorius B, Veerman LJ, Manyema M, Chola L, Hofman K. **Determinants of Obesity and Associated Population Attributability, South Africa: Empirical Evidence from a National Panel Survey, 2008–2012.**. *PLoS One.* (2015.0) **10** e0130218. DOI: 10.1371/journal.pone.0130218
9. Ijoma UN, Chime P, Onyekonwu C. **Factors associated with overweight and obesity in an urban area of south east Nigeria.**. *Food Nutr Sci* (2019.0) **10** 735-49
10. Chukwuonye II, Chuku A, Onyeonoro U. **Body Mass Index, Prevalence and Predictors of Obesity in Urban and Rural Communities in Abia State South Eastern Nigeria.**. *J Diabetes Metab* (2015.0) **6** 570. DOI: 10.4172/2155-6156.1000570
11. Chigbu CO, Parhofer KG, Aniebue UU, Berger U. **Prevalence and sociodemographic determinants of adult obesity: a large representative household survey in a resource-constrained African setting with double burden of undernutrition and overnutrition.**. *J Epidemiol Community Health.* (2018.0) **72** 702-707. DOI: 10.1136/jech-2018-210573
12. Adienbo OM, Hart VO, Oyeyemi WA. **High prevalence of obesity among indigenous residents of a Nigerian ethnic group: The Kalabaris in the Niger Delta Region of South-South Nigeria.**. *Greener J Med Sci* (2012.0) **2** 152-6
13. Chukwuonye II, Chuku A, Okpechi IG, Onyeonoro UU, Madukwe OO, Okafor GO. **Socioeconomic status and obesity in Abia State**. *South East Nigeria Diabetes Metab Syndr Obes* (2013.0) **6** 371-8. DOI: 10.2147/DMSO.S44426
14. 14Food and Agriculture Organization of United Nations Global Forum on Food Security and Nutrition • FSN Forum Discussion 14.06.2016–07.07.2016. Are there any successful policies and programmes to fight overweight and obesity? www.fao.org/fsnforum/activities/discussions/overweight_obesity.
15. Moher D, Liberati A, Tetzlaff J, Altman DG. **Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement.**. *J Clin Epidemiol.* (2009.0) **62** 1006-1012. DOI: 10.1016/j.jclinepi.2009.06.005
16. 16World Health Organization. Physical Status: The Use and Interpretation of Anthropometry. Geneva: WHO; 1995.. *Physical Status: The Use and Interpretation of Anthropometry* (1995.0)
17. Chukwuonye II, Chuku A, John C, Ohagwu KA, Imoh ME, Isa SE. **Prevalence of overweight and obesity in adult Nigerians–a systematic review**. *Diabetes Metab Syndr Obes* (2013.0) **6** 43-47. DOI: 10.2147/DMSO.S38626
18. Egger M, Davey-Smith G, Altman D. *Systematic reviews in health care: meta-analysis in context* (2008.0)
19. Ahaneku GI, Osuji CU, Anisiuba BC, keh VO, Oguejiofor OC, Ahaneku JE. **Evaluation of blood pressure and indices of obesity in a typical rural community in eastern Nigeria**. *Ann Afr Med Apr-Jun* (2011.0) **10** 120-6. DOI: 10.4103/1596-3519.82076
20. Ezeala-Adikaibe BA, Orjioke C, Ekenze OS, Ijoma U, Onodugo O, Okudo G. **Population-based prevalence of high blood pressure among adults in an urban slum in Enugu, South East Nigeria.**. *J Hum Hypertens* (2016.0) **30** 285-91. DOI: 10.1038/jhh.2015.49
21. Maruf FA, Udoji NV. **Prevalence and Socio-Demographic Determinants of Overweight and Obesity in a Nigerian Population.**. *J Epidemiol.* (2015.0) **25** 475-481. DOI: 10.2188/jea.JE20140099
22. Ijoma U., Njoku P., Arodiwe E, Ijoma C. **Increasing Trend of Overweight and Obesity in a Rural Community in South East Nigeria**. *Open Journal of Epidemiology* (2020.0) **10** 323-333. DOI: 10.4236/ojepi.2020.103026
23. Okafor C.I, Fasanmade O, Ofoegbu E, Ohwovoriole A.E. **Comparison of the performance of two measures of central adiposity among apparently healthy Nigerians using the receiver operating characteristic analysis.**. *Indian J Endocrinol Metab* (2011.0) **15** 320-6. DOI: 10.4103/2230-8210.85588
24. Ulasi II, Ijoma CK, Onodugo OD. **A community-based study of hypertension and cardio-metabolic syndrome in semi-urban and rural communities in Nigeria.**. *BMC Health Serv Res* (2010.0) **10** 71. DOI: 10.1186/1472-6963-10-71
25. Ulasi II, Ijoma CK, Onodugo OD, Arodiwe EB, Ifebunandu NA, Okoye JU. **Towards prevention of chronic kidney disease in Nigeria: A community-based survey in southeast Nigeria**. *Kidney Int suppl* (2013.0) **3** 195-201
26. Nwafor A, Mmom F.C, Obia O, Obiandu C, Hart V.O, Chinko B.C. **Relationship between Blood Pressure, Blood Glucose and Body Mass Index and Coexisting Prehypertension and Prediabetes among Rural Adults in Niger Delta Region, Nigeria, British**. *Journal of Medicine & Medical Research* (2015.0) **9** 1-12
27. Egbe EO, Asuquo OA, Ekwere EO, Olufemi F, Ohwovoriole AE. **Assessment of anthropometric indices among residents of Calabar, South-East Nigeria.**. *Indian J Endocrinol Metab* (2014.0) **18** 386-93. DOI: 10.4103/2230-8210.131196
28. Isara AR, Okundia PO. **The burden of hypertension and diabetes mellitus in rural communities in southern Nigeria**. *Pan Afr Med J* (2015.0) **20** 103. DOI: 10.11604/pamj.2015.20.103.5619
29. Ekpenyong CE, Udokang NE, Akpan EE, Samson TK. **Double burden, Noncommunicable diseases and risk factors evaluation in Sub-Saharan Africa: The Nigerian experience.**. *European J. Sus. Dev* (2012.0) **1** 249-270
30. Chinedu SN, Ogunlana OO, Azuh DE, Iweala EE, Afolabi IS, Uhuegbu CC. **Correlation between body mass index and waist circumference in Nigerian adults: implication as indicators of health status.**. *J Public Health Res.* (2013.0) **2** e16. DOI: 10.4081/jphr.2013.e16
31. Rami T.H, Dada S.A.. **Lower BMI cut-off than the World Health Organization based classification is appropriate for Nigerians**. *Journal of Diabetes and Endocrinology* (2018.0) **9** 1-10. DOI: 10.5897/JDE2017.0117
32. Oluyombo R, Olamoyegun M.A, Olaifa O, Iwuala S.O, Babatunde O.A. **Cardiovascular risk factors in semi-urban communities in southwest Nigeria: Patterns and prevalence.**. *J Epidemiol Glob Health.* (2015.0) **5** 167-74. DOI: 10.1016/j.jegh.2014.07.002
33. Abiodun OA, Jagun OA, Olu-Abiodun OO, Sotunsa JO. **Correlation between Body mass index, Waist Hip ratio, blood sugar levels and blood pressure in apparently healthy adult Nigerians.**. *IOSR Journal of Dental and Medical Sciences* (2014.0) **13** 56
34. Oluyombo R, Akinwusi PO, Olamoyegun MO, Ayodele OE, Fawale MB, Okunola OO. **Clustering of cardiovascular risk factors in semi-urban communities in south-western Nigeria**. *Cardiovasc J Afr* (2016.0) **27** 322-327. DOI: 10.5830/CVJA-2016-024
35. Amira CO, Sokunbi DOB, Dolapo D, Sokunbi A. **Prevalence of obesity, overweight and proteinuria in an urban community in South West Nigeria.**. *Nigerian Medical Journal* (2011.0) **52** 110-113
36. Akinwale OP, Oyefara LJ, Adejoh P, Adeneye AA, Adeneye AK, Musa ZA. **Survey of Hypertension, Diabetes and Obesity in Three Nigerian Urban Slums.**. *Iran J Public Health.* (2013.0) **42** 972-9. PMID: 26060658
37. Adebayo RA, Balogun MO, Adedoyin RA, Obashoro-John OA, Bisiriyu LA, Abiodun OO. **Prevalence and pattern of overweight and obesity in three rural communities in southwest Nigeria**. *Diabetes Metab Syndr Obes* (2014.0) **7** 153-8. DOI: 10.2147/DMSO.S55221
38. Asekun-Olarinmoye E, Akinwusi P, Adebimpe W, Isawumi M, Hassan M, Olowe O. **Prevalence of hypertension in the rural adult population of Osun State, southwestern Nigeria.**. *Int J Gen Med.* (2013.0) **6** 317-322. DOI: 10.2147/IJGM.S42905
39. Etukumana EA, Puepet FH, Obadofin M. **Prevalence of overweight and obesity among rural adults in North Central Nigeria.**. *Nigerian Journal of Family Practice* (2013.0) **3** 41-46
40. Sola AO, Steven AO, Kayode JA, Olayinka AO. **Underweight, overweight and obesity in adults Nigerians living in rural and urban communities of Benue State.**. *Ann Afr Med.* (2011.0) **10** 139-43. DOI: 10.4103/1596-3519.82081
41. Adediran OS, Okpara IC, Adeniyi OS, Jimoh AK. **Obesity prevalence and its associated factors in an urban and rural area of Abuja Nigeria.**. *Glob Adv Res J Med Med Sci* (2012.0) **1** 237-241
42. Oyeyemi AL, Adegoke BO, Oyeyemi AY, Deforche B, De Bourdeaudhuij I, Sallis JF. **Environmental factors associated with overweight among adults in Nigeria.**. *Int J Behav Nutr Phys Act* (2012.0) **9** 32. DOI: 10.1186/1479-5868-9-32
43. Adedoyin RA, Mbada CE, Ismaila SA, Awotidebe OT, Oyeyemi AL, Ativie RN. **Prevalence of cardiovascular risk factors in a low income semi-urban community in the North-East Nigeria.**. *TAF Prev Med Bull* (2012.0) **11** 463-70. DOI: 10.5455/pmb.1-1320075671
44. Wahab KW, Sani MU, Yusuf BO, Gbadamosi M, Gbadamosi A, Yandutse MI. **Prevalence and determinants of obesity—a cross-sectional study of an adult Northern Nigerian population.**. *Int Arch Med.* (2011.0) **4** 10. DOI: 10.1186/1755-7682-4-10
45. Dahiru T, Ejembi CL. **Clustering of cardiovascular disease risk-factors in semi-urban population in Northern Nigeria.**. *Niger Clin Pract* (2013.0) **16** 511-6. DOI: 10.4103/1119-3077.116903
46. Makusidi MA, Liman HM, Yakubu A, Isah MD, Jega RM, Adamu H. **Prevalence of Non-communicable Diseases and it’s awareness among inhabitants of Sokoto Metropolis; Outcome of a Screening Program for Hypertension, Obesity, Diabetes Mellitus and Overt Proteinuria.**. *Arab Journal of Nephrology and Transplantation.* (2013.0) **6** 189-91. PMID: 24053748
47. Ramalan MA, Gezawa ID, Uloko AE, Musa BM. **Prevalence and risk factors for overweight and obesity among suburban semi-nomadic fulani’s of Northwestern Nigeria**. *Nigerian Journal of Medicine* (2019.0) **28** 360-7
48. Chukwuonye II, Chuku A, Onyeonoro UU, Okpechi IG, Madukwe OO, Umeizudike TI. **Prevalence of abdominal obesity in Abia State, Nigeria: results of a population-based house-to-house survey**. *Diabetes Metab Syndr Obes* (2013.0) **6** 285-91. DOI: 10.2147/DMSO.S43545
49. Ugwuja EI, Ogbonnaya LU, Obuna AJ, Awelegbe F, Uro-Chukwu H. **Anaemia in Relation to Body Mass Index (BMI) and Socio-Demographic Characteristics in Adult Nigerians in Ebonyi State.**. *J Clin Diagn Res.* (2015.0) **9** LC04-LC07. DOI: 10.7860/JCDR/2015/9811.5485
50. Abubakari AR, Bhopal RS. **Systematic review on the prevalence of diabetes, overweight/obesity and physical inactivity in Ghanaians and Nigerians.**. *Public Health.* (2008.0) **122** 173-182. DOI: 10.1016/j.puhe.2007.06.012
51. Ofori-Asenso R, Agyeman AA, Amos Laar A, Boateng D. **Overweight and obesity epidemic in Ghana—a systematic review and meta-analysis.**. *BMC Public Health* (2016.0) **16** 1239. DOI: 10.1186/s12889-016-3901-4
52. Mangemba NT, Sebastian MS. **Societal risk factors for overweight and obesity in women in Zimbabwe: a cross-sectional study.**. *BMC Public Health.* (2020.0) **20** 103. DOI: 10.1186/s12889-020-8215-x
53. Trogdon JG, Finkelstein EA, Hylands T, Dellea PS. **Indirect costs of obesity: a review of the current literature.**. *Kamal-Bahl.* (2008.0) **9** 489-500. DOI: 10.1111/j.1467-789X.2008.00472.x
54. Adeloye D, Owolabi EO, Ojji DB, Auta A, Dewan MT, Olanrewaju TO. **Prevalence, awareness, treatment, and control of hypertension in Nigeria in 1995 and 2020: A systematic analysis of current evidence.**. *J Clin Hypertens (Greenwich).* (2021.0) **23** 963-977. DOI: 10.1111/jch.14220
55. Uloko AE, Musa BM, Ramalan MA, Gezawa ID, Puepet FH, Uloko AT. **Prevalence and Risk Factors for Diabetes Mellitus in Nigeria: A Systematic Review and Meta-Analysis.**. *Diabetes Ther* (2018.0) **9** 1307-1316. DOI: 10.1007/s13300-018-0441-1
|
---
title: 'Ethnic-specific associations between dietary consumption and gestational diabetes
mellitus incidence: A meta-analysis'
authors:
- Harriett Fuller
- J. Bernadette Moore
- Mark M. Iles
- Michael A. Zulyniak
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021780
doi: 10.1371/journal.pgph.0000250
license: CC BY 4.0
---
# Ethnic-specific associations between dietary consumption and gestational diabetes mellitus incidence: A meta-analysis
## Abstract
Globally, one in seven pregnant women are diagnosed with gestational diabetes mellitus (GDM), conferring short- and long-term health risks to both mother and child. While dietary prevention strategies are common in clinical practice, their effectiveness in different ethnicities is uncertain. To better inform prevention strategies, here the effects of unhealthy and healthy diets on GDM risk within distinct ethnic or cultural populations and geographic regions were evaluated and summarised. Pubmed, Scopus, Cochrane and OVID were systematically searched to identify randomised controlled trials (RCTs) and observational studies that investigated diet and GDM. A grouped analysis of common ‘healthy’ and ‘unhealthy’ diets was performed first, before analysing individual dietary patterns (e.g., prudent, Mediterranean). Random effect models and dose response analyses were performed where possible. PROSPERO (CRD42019140873). Thirty-eight publications provided information on 5 population groups: white European (WE), Asian, Iranian, Mediterranean and Australian. No associations were identified between healthy diets and GDM incidence in RCTs in any population. However, when synthesizing observational studies, healthy diets reduced odds of GDM by $23\%$ ($95\%$ CI: 0.70–0.89, $p \leq 0.001$, I2 = $75\%$), while unhealthy diets increased odds of GDM by $61\%$ ($95\%$ CI: 1.41–1.81, $p \leq 0.0001$, I2 = $0\%$) in WE women. No evidence of consistent effects in other populations were observed, even when adequately powered. Diet consistently associated with GDM risk in WEs but not in other populations. Heterogenous use and reporting of ethnically and culturally appropriate diets and dietary assessment tools, particularly in RCTs, raises uncertainty regarding the lack of association found in non-WE populations. Future studies require the use of culturally appropriate tools to confidently evaluate dietary and metabolic mediators of GDM and inform culturally-specific dietary prevention strategies.
## Introduction
Gestational diabetes mellitus (GDM), hyperglycaemia that develops during pregnancy, is diagnosed in $14\%$ of pregnancies globally and is associated with numerous health risks [1]. Pregnancy and long-term complications for the mother include antepartum and postpartum haemorrhage, post-natal depression, and risk of type 2 diabetes (T2D; 7-fold) and cardiovascular disease (CVD; 2.3 fold) [2–6]. For offspring of GDM-mothers, risks include adulthood obesity and T2D [3] and future risk of GDM (~ 9 fold) [7]. These morbidities underline the need for effective GDM prevention [8]. Women from ethnic minority groups (e.g., African, Asian) disproportionally suffer from GDM compared to white European (WE) women ($15\%$ vs $5\%$ prevalence) [9], largely independent of country of residence or general health. This suggests that underlying features and characteristics unique to ethnic minority groups and their cultures [10] are driving GDM risk disparity, and positions ethnicity as a major disease determinant. However, despite evidence of ethnic-specific associations between diet and maternal health during pregnancy [11, 12], and the benefit of ethnic-specific diets versus standard care [13], the majority of evidence on dietary prevention of GDM is from WE populations [14].
Indeed, in a systematic review of 18 randomised controlled trials (RCTs), fewer than $50\%$ reported population ethnicity (of which all were primarily of WE decent); and only 1 study investigated an ethnic-specific intervention [13]. While one meta-analysis of 11 RCTs across 7 countries ($$n = 2$$,838) noted ethnicity as a moderator of diet-GDM associations (RR: 0.75, $95\%$ CI: 0.60,0.95), it did not report ethnic-specific effects [15]. Evidence of ethnic-specific associations are more often reported in observational studies but evidence between observational studies are rarely compared. For example, in a WE cohort, meat-based and plant-based diets increased and decreased odds of GDM (odds ratio(OR)meat: 1.38; $95\%$ CI: 1.14, 1.68; and ORplant = 0.84; $95\%$ CI: 0.68, 1.04) [16], whereas in a Chinese cohort comparable diet patterns showed the opposide (ORmeat = 0.89; $95\%$ CI: 0.58, 1.36; ORplant = 1.04: $95\%$ CI; 0.90, 1.20) [17]. In spite of the urgent need to inform global health strategies, no meta-analysis has yet evaluated the effect of diet on GDM development in an ethnic-specific manner. Therefore, national GDM prevention strategies are often biased towards evidence from WE studies [18–22].
To address this and better inform prevention stratergies we have evaluated and summarised the effects of healthy and unhealthy dietary patterns on GDM incidence within distinct populations.
## Search strategy and selection criteria
Searches were structured using PICO and MESH indexing with key terms (and synonyms thereof) for pregnancy (P), diet (I), ethnicity (C) and gestational diabetes (O); and study designs (PROSPERO:CRD42019140873) (Appendix A in S1 Text). Ovid, Cochrane (including trial registries), Scopus, and PubMed databases were searched from inception until January 2021. Citation lists of included studies were searched for additional relevant studies until no further articles were identified.
Eligible studies at title-abstract screening were human randomised control trials (RCTs) and observational studies (with the exception of non-nested case-control) that explored the association of diet and GDM (ORs, RRs, or raw data) published in English. Studies were excluded if they: (i) follow-up was < 1 trimester or started in the third trimester [23]; (ii) included unhealthy participants; (iii) combined diet with other lifestyle interventions; (iv) did not report participant ethnicity or nationality. Abstracts were screened in duplicate, with disagreements mediated by a third reviewer. Where effect estimates were adjusted for ethnicity without stating the ethnicity, it was assumed to be the ethnic majority (≥$60\%$ of population). Where this was not possible to confirm, corresponding authors were contacted. If no additional data were obtained the study was excluded.
## Data analysis
Data extraction was performed in duplicate for all variables (Tables A and B in S1 Text). RCTs and observational studies were analysed separately. Healthy and unhealthy diet categories were defined based on: (i) study authors’ definition, or (ii) common definitions according to major health bodies (e.g., WHO, WCRF, NHS UK, ADA) [19, 21, 22, 24]. *In* general, ‘healthy’ diets were characterised by fruit and vegetables, wholegrains, fish, lean meats, and unsaturated fats; while ‘unhealthy’ diets were characterised by red/ processed meats, fried foods, confectionary, sugar sweetened beverages (SSBs), saturated fats, and added sugars. Where it was difficult to group with confidence, diets were unclassified. Unclassified diets required ≥2 studies to be considered.
Where exposure data were presented categorically, highest consumers were compared to lowest consumers. Where multiple models were detailed, effect estimates from fully-adjusted were extracted. To allow for a qualitative comparison between RCTs and observational evidence, risk ratios (RRs) were converted to odds ratios (Ors) with $95\%$ confidence intervals (CI). All meta-analyses used generic inverse variance weighted random-effects (DerSimonian-Laird;DL) in Review Manager 5.3 (Cochrane) [25]. In meta-analyses with < 5 studies, the Hartung-Knapp-Sidik-Jonkman (HKSJ) random effects model [meta package (V4.9–6); R(v.1.2.5019)] was also used [26–28]. Bonferroni correction was applied as necessary. Uncertainty intervals for I2 values were also calculated [29]. Sensitivity analyses were performed where I2≥$40\%$ [30] and for GDM confounders: (i) timing of diet recall/intervention, (ii) maternal age, and (iii) overweight/obese. To account for differences in lean and overweight/obese classification between ethnic groups, a cut-off of BMI≥25 kg/m2 was used to classify overweight/obese in non-Asian populations, while in Asian populations a BMI cut-off of ≥23 kg/m2 was used [31].Where possible, dose response relationships were examined using the dosresmeta package in R (v.1.2.5019). See Supplementary Methods in S1 Text. Figures were produced within R studio through the use of the rworldmap and ggplot2 packages [32, 33].
## Risk of Bias (ROB) assessment
ROB was assessed using a modified 2016 Academy of Nutrition and Dietetics tool [34]. Comprised of 10 validity questions, it is designed for nutrition research and is translatable to RCTs and observational studies. Full methodological details in supplementay material (Appendix B in S1 Text).
## Power analysis
Post-hoc power analyses were undertaken using fixed-effects (τ2 = 0) or random-effects (τ2>0) methods [29]. Power ≥$80\%$ was considered adequate and based on recent meta-analyses investigating the association between common healthy and unhealthy diets on GDM in Europe, US and Asian populations [12, 35, 36], we considered a meaningful change in effect size as a difference in OR ≥ |0.20| between the high and low exposure groups. Exposed and unexposed were calculated as an average of the exposed/ unexposed sizes of all studies for the exposure.
## Results
After the removal of duplicates, 3393 studies were identified, from which 53 studies progressed to full text screening (Fig 1). Of these, 25 studies reported heathy diets [‘healthy recommendations’ ($$n = 13$$), ‘Mediterranean diet’ ($$n = 6$$), ‘prudent diet’ ($$n = 4$$), ‘plant-based’ ($$n = 6$$) or ‘healthy snack’ ($$n = 1$$)], and 13 reported unhealthy diets [‘western diet‘ ($$n = 6$$), ‘fried/fast food’ ($$n = 4$$), ‘sweet and seafood’ ($$n = 2$$), and ‘unhealthy diet score’ ($$n = 1$$)]. Additional diets considered neither healthy or unhealthy were grouped as ‘unclassified’ [meat pattern, high protein diet, ‘traditional Asian’, ‘high-fat’, ‘high-carbohydrate’, ‘high-animal protein’, ‘high-vegetable protein’ and ‘high-fish’ diets] when ≥2 studies were identified (Appendix C in S1 Text). This provided a total of 38 studies (6 RCTs and 32 observational;$$n = 251$$,778 participants) for review and meta-analyses (Fig A in S1 Text), 28 of which included more than one exposure, across 4 distinct ethnic and geographic groups: white European (WE; $83\%$), Asian ($9\%$; East and South Asian), Mediterranean (i.e. southern European populations) [37] ($5\%$), and Australian Nationals ($3\%$;Australian residents with an almost equal proportion of Asians and WEs). ( Fig B in S1 Text) Maternal age of study participants ranged 24–36 years and $\frac{14}{38}$ studies reported an average BMI of overweight or obese. All included RCTs used a 75g oral glucose tolerance test (OGTT) for GDM diagnoses, while in observational studies, the majority of studies ($\frac{20}{32}$; $62.5\%$) used OGTT with clearly defined criteria to diagnose GDM. A chi-square test reported no significant difference ($P \leq 0.05$) in the use of OGTT between ethnic groups in observational studies.
**Fig 1:** *Prisma diagram.Outline of identification of studies in Ovid (AMID, CAB abstracts, EBM, EMBASE, Global Health, Health Care Management Information Consortium, MIDRIS, OVID Medline R), Cochrane (including trial registries), Scopus, and PubMed and record of screening process of articles.*
## RCTs
Six RCTs ($$n = 3$$,041), including 4 population groups (Asian, Australian, Mediterranean, WE) evaluated the impact of a healthy diet compared to control (Table A in S1 Text). No effect of healthy diets on GDM was found across across ethnic groups (Fig 2). For individual diets, stratified by ethnicity, no associations were identified for healthy recommendations ($$n = 4$$) or healthy snacks ($$n = 1$$). However, one study reported a protective effect of the Mediterranean diet against GDM (OR = 0.67; $95\%$CI: 0.48, 0.94) in a Mediterranean population (Fig C in S1 Text). No differences were observed with the HKSJ model. No RCTs reported on the effect of an unhealthy diet (Table C in S1 Text). Only 3 of the 5 studies in non-white Europeans used a culturally validated questionnaire (Table I in S1 Text). The analysis was adequately powered to detect a $4\%$ change in odds (Table D in S1 Text).
**Fig 2:** *Forest plot of reported effect sizes of healthy dietary interventions on GDM in ethnic–specific RCTs.Forest plot to investigate the association between heathy diets and GDM, investigated within distinct ethnic groups. TE: treatment effect, SE: standard error, IV: inverse variance method, CI: Confidence interval.*
## Observational studies
In total 32 observational studies ($$n = 248$$,737), reported on 5 ethnic groups (Asian, Australian Nationals, Iranian, Mediterranean and WE) across 20 countries (Table B in S1 Text). Within these studies, 17 dietary patterns were reported and grouped as healthy, unhealthy, or unclassified. The majority of observational studies in non-white Europeans ($\frac{14}{16}$) reported the use of a validated dietary assessment tool (Table I in S1 Text).
## Healthy diets
20 studies reported the consumption of healthy diets (healthy recommendations, Mediterranean, prudent, and plant-based diets) in 4 ethnic groups. When assessed as a single multi-ethnic group, high adherence to a healthy diet associated with a reduction in odds of GDM (OR = 0.78; $95\%$CI: 0.70,0.88; I2 = $74\%$), compared to lowest adherence (Fig 3). An inspection by ethnicity showed that healthy diets are protective in WEs (OR = 0.75; $95\%$CI: 0.65 0.88; I2 = $79\%$) but not other ethnic groups. A subsequent analysis of distinct healthy diets, found healthy recommendation and Mediterranean diet patterns to be protective against GDM (Fig D in S1 Text), with both associations driven by WEs (OR = 0.70; $95\%$CI: 0.56,0.86; I2 = $70\%$ and OR = 0.66; $95\%$ CI: 0.50, 0.85; I2 = $65\%$). When the more stringent HKSJ model was applied, only healthy recommendations retained significance (Table C in S1 Text). No associations were observed in Asian populations even after stratification by East Asian (Chinese and Japanese) and South/South-East Asian regions (Indian subcontinent and Malaysian) (Table E in S1 Text). All analyses were adequately powered to detect a change in odds ≥ $6\%$ (Table C in S1 Text).
**Fig 3:** *Forest plot of reported associations between healthy dietary interventions and odds of GDM in ethnically–defined observational studies.Forest plot illustrating the association between habitual healthy dietary intake and GDM in distinct ethnicities. TE: treatment effect, SE: standard error, IV: inverse variance method, CI: Confidence interval.*
## Unhealthy diets
Thirteen studies reported the association between an unhealthy diet (Western, fried/ fast food, unhealthy diet score, sweet and seafood pattern) and GDM across 4 ethnic groups. In a large multi-ethnic cohort, high adherence to an unhealthy diet associated with increased odds of GDM (OR = 1.44; $95\%$CI: 1.25,1.67, I2 = $12\%$), compared to low adherence (Fig 4). This association was also observed in distint ethnic group [WEs (OR = 1.59; $95\%$CI: 1.41, 1.81, I2 = 0), Mediterranean (OR = 1.69, $95\%$ CI: 1.21, 2.35, I2 = 0), and one Iranian (OR = 2.12; $95\%$CI: 1.12, 4.01, I2 = NA)] aside from Asians. When considering individual unhealthy diets, the Western diet increased odds of GDM (OR = 1.51; $95\%$CI: 1.23,1.86; I2 = $7\%$) in the overall population, primarily driven by WEs (OR = 1.60; $95\%$CI: 1.26,2.02; I2 = $3\%$) and one Mediterranean study (OR = 1.56; $95\%$ CI: 1.00,2.43; I2 = NA) ‒ using HKSJ, WEs achieved $$P \leq 0.07$$ (Fig E and Table C in S1 Text). The fried/fast food diet pattern associated with increased odds of GDM (OR = 1.66; $95\%$CI: 1.42,1.92; I2 = $0\%$), with comparable effect sizes across ethnic groups. All analyses were adequately powered to detect a change in odds ≥ $9\%$ (Table D in S1 Text).
**Fig 4:** *Forest plot of reported associations between unhealthy dietary interventions an odds of GDM in ethnically–defined observational studies.Forest plot illustrating the association between habitual unhealthy dietary intake and GDM in distinct ethnicities. TE: treatment effect, SE: standard error, IV: inverse variance method, CI: Confidence interval.*
## Unclassified diets
Four dietary patterns (meat-based, high-protein, traditional Asian, and high-fish) were unclassified.(Fig F in S1 Text) The meat-based pattern ($$n = 7$$) associated with increased odds of GDM when evaluated in a multi-ethnci cohort (OR = 1.41, $95\%$ CI:1.22,1.63, I2 = 0), with only small deviations of effect sizes between ethnicities. A high-fish diet ($$n = 5$$) was protective against GDM in WEs (OR = 0.85, $95\%$ CI: 0.75, 0.98, I2 = 0) but not in Asians. No associations were observed between (i) high-protein diet and GDM or (ii) a traditional Asian ($$n = 2$$) diet and GDM in Asians. All analyses were adequately powered ($80\%$) with varied thresholds for each analysis (0.5–$18\%$ change in odds) (Table D in S1 Text).
## Macronutrients
Four diets classified diet exposure based on a dominant macronutrient (% total energy): animal protein, vegetable protein, fat, and carbohydrate. ( Fig G in S1 Text) Animal protein (OR = 1.49, $95\%$CI:1.25,1.77,I2 = 0), carbohydrate (OR = 0.49, $95\%$ CI:0.38,0.63; I2 = 0), and fat diet patterns (OR = 1.50, $95\%$CI: 1.22,1.83, I2 = 0) all associated with odds of GDM. Within ethnic groups, comparable effect sizes were observed for Asians and WEs. All associations remained significant after Bonferroni correction following the HKSJ approach, with the exception of the animal protein diet in Asians. No associations were observed for the vegetable protein exposure. All analyses were adequately powered ($80\%$) with varied thresholds for each analysis (0.5–$13\%$ change in odds) (Table D in S1 Text).
## Dose response analyses
When evaluated as a multi-ethnic population, no associations were observed. When ethnic groups were investigated independently, a positive non-linear dose-response relationship was found for WEs consuming a fried food diet but it did not pass Bonferroni correction.
## Post-hoc analysis—Combined RCT & observational analysis
Effect estimates from RCTs and observational studies were of similar magnitude with overlapping confidence intervals, comparable I2 values, and ROB scores (Fig H in S1 Text). This suggests that the primary source of heterogeneity may not be study design. Therefore, a post-hoc analysis with combined RCTs and observational studies was undertaken. Power improved in all ethnic subgroups. This analysis uncovered a novel association between healthy diets and GDM in Australian nationals (OR = 0.92, $95\%$ CI: 0.88,0.97; I2 = 0) with only negligible changes in effect sizes and no other new associations identified (Table C in S1 Text).
## Sensitivity analyses–Confounders
Sensitivity analyses were undertaken on all exposures with I2≥ $40\%$. Sensitivity analyses were performed based upon: (i)diet during pregnancy; (ii)adjustment for obstetric risk factors (parity, gravidity or multiple pregnancy); and (iii) pre-pregnancy BMI; and (iv) maternal age. When only considering dietary intake during pregnancy, no association was found between healthy diets and GDM in any ethnic group. In addition, when considering the overall unhealthy or the western diets, no association was found with GDM in overweight/obese WEs. Interestingly, a high protein diet increased odds of GDM in older women that was driven by WEs (OR = 1.28; $95\%$CI: 1.09–1.52; I2 = 0). ( Tables D–H in S1 Text). All sensitivity analyses were well powered ($0.80\%$) to detect an effect size ≥$10\%$ with the majority suitably powered to detect an effect size ≥ $5\%$. Two exceptions were the assessments of the plant-based diet in overweight women and healthy diets in overweight/obese Asian women (Table D in S1 Text). A single sensitivity analysis of the effect of the western diet during pregnancy had inadequate power to detect a change in odds of $20\%$.
## Risk of Bias (ROB)
Evidence of moderate ROB was found but no study exceeded the Academy of Nutriton and Dietetics’ exclusion threshold. Within RCTs $45\%$ and $35\%$ scored low or neutral ROB, with $21\%$ at risk of bias. Within observational studies, $55\%$ and $22\%$ scored a low or neutral risk of bias, with $24\%$ at risk of bias. Areas of concern were participant selection, management of withdrawals, and study group comparability. Carbohydrate, fat, and Mediterranean diets were at highest risk of bias while the prudent and fast food diets were at low risk. No evidence of publication bias was identified. ( Figs I–L in S1 Text).
## Discussion
The aim of this work was to offer clarity regarding the ability of diet to mitigate GDM in different ethnic groups. Evidence from RCTs and observational studies reporting healthy, unhealthy, and unclassified dietary patterns were systematically reviewed and meta-analysed. The results confirm a protective association between healthy diets and an adverse association between unhealthy diets and GDM in WEs and Ausrlian nationals, with evidence in ethnic minority groups hindered by fewer studies (<$20\%$ of studies), and limited use of ethnically informed methods. In minority populations where ≥2 studies were available, only carbohydrate rich diets were associated (protectively) with GDM in Asians.
The majority of RCTs we collected ($\frac{5}{6}$ studies) commenced during pregnancy. A meta-analysis of 5 RCTs ($$n = 1$$,155) agree with our findings and report no significant effect of dietary interventions on GDM risk in WEs. Interestingly, a recent meta-analysis with 37 RCTs reported that diets designed to manage gestational weight gain (GWG), reduced GDM incidence in Asian countries but not WE-majority countries [38]. This is contrary to our results but may explained by the aims of the interventions, with GDM diets focussed on reducing glycaemic loads rather than total caloric intake. Therefore, it may be that (as a mediator of dysglycemia) controlling GWG is key and that future dietary interventions to reduce GDM risk in Asian countries require greater emphasis on weight management.
Observational studies often focus on pre-pregnancy diet, making them crucial to understand how pre-conception dietary habits influence GDM. A systematic review of 34 observational cohort and case-control studies, reported that high consumption of cholesterol, heme iron, and processed meat increased risk of GDM, while patterns rich in fruit, wholegrains and vegetables reduced risk of GDM [39]. However, a high heterogeneity between ethnic groups was observed by the authors—likely due to confounding from ethnic-specific food preferences, cooking methods, and meal times. To address this, we performed our meta-analyses in ethnic-specific subgroups; thereby, minimsing confounding within each ethnic analysis while permitting a comparison of effect sizes between them. Fifteen dietary exposures were identified in multiple studies, and classified as either ‘healthy’, ‘unhealthy’ or ‘unclassified’ (i.e., neither healthy nor unhealthy) diets. Following stratification by ethnicity, the protective effect of healthy diets against GDM was confirmed in WEs and the hazardous effect of unhealthy and meat-based diets in WEs; however, consistent evidence of an association within non-WE groups was not found. Interestingly, all associations were unaffected by mother’s age and BMI, suggesting that modified guidelines for WE women at high-risk of GDM due to age or BMI may not be required. The presence of an association in WEs within observational studies but not RCTs could be a result of increased power or it could highlight the importance of a healthy diet prior to conception. However, future RCTs investigating dietary interventions during ‘family planning’ are required to test this hypothesis.
Interestingly, examining macronutrient-specific diets, those characterised by animal protein or fat increased odds of GDM by up to $50\%$ in WE and Asians, whereas carbohydrate-rich diets reduced odds of GDM by ≈$50\%$. Unfortunately, because all exposures were quantified as % energy intake, it was not possible to tease apart whether the protective effect on GDM was driven by reductions in protein and fat or increased consumptionof carbohydrate, or a combination thereof. Interestingly, while animal protein associated with GDM risk in WEs and Asians, no association was observed with the meat-based dietary pattern in Asians. While the animal-protein diet may have been carbohydrate-rich and negated the effects of high-animal protein, an alternative explanation may be that ethnic-specific foods and cooking methods are difficult to capture with some dietary recall tools. With limited associations identified in non-WE studies and recognising the similarities in effect sizes in this analysis of RCTs and observational studies, study types were combined [40]. This uncovered an association between healthy diets and GDM in Australian nationals, a heterogenous group comprised of WE and Asian mothers but no other additional associations were reported.
Despite numerous significant associations between dietary patterns and GDM in WEs, no consistent evidence was found in non-WE populations. The reason for this is unclear but inconsistent reporting (and use) of ethnically and culturally-informed assessment tools may have contributed to this [41]. Ethnically sensitive interventions consider dietary habits, food preparation, and cultural beliefs that are relevant to the study population to improve accuracy of dietary assessment [42, 43], rather than a single FFQ used across multiple ethnic groups that can introduce bias and ethnic-specific differences under/over-reporting [44, 45]. While many studies reported ethnically-modified approaches, some of the details regarding their modification and validity were unclear, particularly in randomised controlled trials where only 3 of 5 ($60\%$) of studies in non-white European populations reported using a culturally validated questionnaire. This agrees with a systematic review ($$n = 42$$ studies) that reported a lack of validation of dietary assessment tools (only $17\%$) in studies undertaken in minority ethnic groups [46]. Interestingly, the majority of non-white observational studies included in this review ($\frac{14}{16}$; $87.5\%$) did use culturally appropriate dietary assessment tools, validated within a relevant population.
The use of metabolomics-based strategies may offer a method to characterise diet, metabolism, and the role of bioavailability across ethnic groups and expose underlying ethnic-specific requirements. Previous work in the Born in Bradford cohort has demonstrated that pregnant WEs and SAs have significantly different metabolic profiles (i.e., namely, lipoproteins, lipids, glycolysis metabolites, and amino acids) [47]. Work by the multi-ethnic NUTRIGEN consortium also suggests that a healthy plant-based diet consumed during pregnancy effects infant birth weight differently in SAs (increased birthweight) and WEs (reduced birth weight) [48]. This evidence suggests that ethnically-distinct dietary and underlying metabolic qualities exist between ethnicities and further highlights the need for ethnically-tailored interventions.
This is the first study to meta-analyse the effect of numerous dietary patterns and interventions consumed before or during pregnancy on GDM within distinct ethnic groups and geographic regions, using both RCTs and observational study data. Strengths included the fact that all data were examined using both standard methods (DL) as well as supplementary and more conservative analyses (HKSJ) when ≤5 studies were available [26–28]; along with robust sensitivity analyses for confounding factors. Moreover, a single ROB assessment that is translatable for both RCTs and observational studies permitted comparison of bias between study design. Finally, power analyses confirmed adequate power in WE and Asian analyses with limited power in other minority ethnic groups. However, this study did have limitations. First, only studies written in English were included, which many have limited the scope. Second, all observational studies are limited by confounding; however, the similar effect sizes between RCTs and observational studies that we observed may suggest that the adjustment for confounding in the included observational studies limited confounding reasonably well. Finally, there is a risk of type 2 errors due to the widespread use of the Nurses Cohort Study (NCS); however, sensitivity analysis demonstrated no impact on results when NCS studies were removed.
Through the use of both RCTs and observational studies, this meta-analysis confirms the presence of a protective effect of healthy diets against GDM in WE women and an increase in risk in mothers consuming unhealthy diets. Current evidence in ethnic minority groups is less certain because of fewer studies and, with limited evidence of an assocoaion in non-white ethnic groups. However, inconsistent reporting of ethnically appropriate diets or assessment tools, challenges the certainty of evidence within studies of minority ethnic groups. In summary, our work highlights that future studies in ethnic minority groups, using ethnically informed diets and tools, particularly RCTs, are urgently needed to accurately evaluate the effect of diets on GDM so that appropriate strategies for these high-risk populations can be confidently assessed and defined.
## References
1. 1World Health Organisation. WHO recommendation on the diagnosis of gestational diabetes in pregnancy. 2018.
2. Bellamy L, Casas J-P, Hingorani AD, Williams D. **Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis**. *The Lancet* (2009.0) **373** 1773-9. DOI: 10.1016/S0140-6736(09)60731-5
3. Kanguru L, Bezawada N, Hussein J, Bell J, Kanguru L, Bezawada N. **The burden of diabetes mellitus during pregnancy in low- and middle- income countries: a systematic review**. *Global Health Action* (2014.0) **7**. DOI: 10.3402/gha.v7.23987
4. Barakat S, Martinez D, Thomas M, Handley MA. **What do we know about Gestational Diabetes Mellitus and Risk for Postpartum Depression among Ethnically Diverse Low-Income Women in the United States?**. *NIH Public Access* (2014.0) **17** 587-92
5. Kuller LH, Catov J. **Invited Commentary: Gestational Hypertension and Diabetes—A Major Public Health Concern**. *American Journal of Epidemiology* (2017.0) **186** 1125-8. DOI: 10.1093/aje/kwx265
6. Kramer CK, Campbell S, Retnakaran R. **Gestational diabetes and the risk of cardiovascular disease in women: a systematic review and meta-analysis**. *Diabetologia* (2019.0) **62** 905-14. DOI: 10.1007/s00125-019-4840-2
7. Retnakaran R, Shah BR. **Sex of the baby and future maternal risk of Type 2 diabetes in women who had gestational diabetes**. *DIABETICmedicine* (2016.0) **33** 956-60. DOI: 10.1111/dme.12989
8. Egeland GM, Skjaerven R, Irgens LM. **Birth characteristics of women who develop gestational diabetes: population based study**. *Bmj* (2000.0) **321** 546-7. DOI: 10.1136/bmj.321.7260.546
9. McIntyre HD, Catalano P, Zhang C, Desoye G. **Gestational diabetes mellitus**. *Nature Reviews Disease Primers* (2019.0) 0123456789
10. Egede LE. **Race, ethnicity, culture, and disparities in health care**. *J Gen Intern Med* (2006.0) **21** 667-9. DOI: 10.1111/j.1525-1497.2006.0512.x
11. Stuebe AM, Oken E, Gillman MW. **Associations of diet and physical activity during pregnancy with risk for excessive gestational weight gain**. *Am J Obstet Gynecol* (2009.0) **201** 58. DOI: 10.1016/j.ajog.2009.02.025
12. Hassani Zadeh S, Boffetta P, Hosseinzadeh M. **Dietary patterns and risk of gestational diabetes mellitus: A systematic review and meta-analysis of cohort studies**. *Clin Nutr ESPEN* (2020.0) **36** 1-9. DOI: 10.1016/j.clnesp.2020.02.009
13. Valentini R, Dalfrà MG, Masin M, Barison A, Marialisa M, Pegoraro E. **A Pilot Study on Dietary Approaches in Multiethnicity: Two Methods Compared**. *International Journal of Endocrinology* (2012.0) **2012** 985136. DOI: 10.1155/2012/985136
14. Yamamoto JM, Kellett JE, Balsells M, Garcia-Patterson A, Hadar E, Sola I. **Gestational Diabetes Mellitus and Diet: A Systematic Review and Meta-analysis of Randomized Controlled Trials Examining the Impact of Modified Dietary Interventions on Maternal Glucose Control and Neonatal Birth Weight**. *Diabetes Care* (2018.0) **41** 1346-61. DOI: 10.2337/dc18-0102
15. Guo XY, Shu J, Fu XH, Chen XP, Zhang L, Ji MX. **Improving the effectiveness of lifestyle interventions for gestational diabetes prevention: a meta-analysis and meta-regression**. *BJOG: an international journal of obstetrics and gynaecology* (2018.0). DOI: 10.1111/1471-0528.15467
16. Bao W, Bowers K, Tobias DK, Olsen SF, Chavarro J, Vaag A. **Prepregnancy low-carbohydrate dietary pattern and risk of gestational diabetes mellitus: a prospective cohort study**. *American Journal of Clinical Nutrition* (2014.0) **99** 1378-84. DOI: 10.3945/ajcn.113.082966
17. Mak JKL, Pham NM, Lee AH, Tang L. **Dietary patterns during pregnancy and risk of gestational diabetes: a prospective cohort study in Western China**. *Nutrition Journal* (2018.0) **17** 1-11. DOI: 10.1186/s12937-018-0413-3
18. Balasubramanian G, Morampudi S, Godwa A, Zomorodi B, Patil AS. *The Challenges and Recommendations for Gestational Diabetes Mellitus Care in India: A Review* (2017.0) **8**
19. 19National Health Service (NHS). Treatment—Gestational Diabetes. 2019.
20. 20National Institute of Diabetes and Digestive and Kidney Diseases. Gestational diabetes 2017 [updated May 1, 2017. Available from: http://tiny.cc/xbqtcz.
21. 21World Health Organisation (WHO). Diabetes. 2020.
22. 22American Diabetes Association (ADA). Gestational diabetes—Gestational diabetes and a healthy baby? 2020.
23. Shepherd E, Jc G, Tieu J, Han S, Ca C, Middleton P. **Combined diet and exercise interventions for preventing gestational diabetes mellitus**. *Cochrane Database of Systematic Reviews* (2017.0). DOI: 10.1002/14651858.CD010443.pub3
24. 24Centre for Disease Control (CDC). Gestational Diabetes. 2019.
25. 25The Cochrane Publication. Review Manager Web (Revman Web). 2019.
26. Tobias DK, Zhang C, Chavarro J, Bowers K, Rich-edwards J, Rosner B. **Prepregnancy adherence to dietary patterns and lower risk of gestational diabetes mellitus**. *American Journal of Clinical Nutrition* (2012.0) **96** 289-95. DOI: 10.3945/ajcn.111.028266
27. Inthout J, Ioannidis JPA, Borm GF. **The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method**. *BMC medical research methodology* (2014.0) 1-12. DOI: 10.1186/1471-2288-14-1
28. Jackson D, Law M, Rücker G, Schwarzer G. **The Hartung-Knapp modification for random-effects meta-analysis: A useful refinement but are there any residual concerns?**. *Statistics in medicine* (2017.0) 3923-34. DOI: 10.1002/sim.7411
29. Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. *Introduction to Meta-Analysis* (2009.0) 262-73
30. Higgins JPT, Green S. **Identifying and measuring heterogenity**. *Cochrane Handbook for Systematic Reviews of Interventions (Version 510)* (2011.0) 9.5.2-9.5.2
31. **Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies**. *Lancet (London, England)* (2004.0) **363** 157-63. DOI: 10.1016/S0140-6736(03)15268-3
32. South A.. **rworldmap: a new R package for mapping global data**. *R Journal* (2011.0) **3**
33. Wickham H.. *ggplot2: elegant graphics for data analysis* (2016.0)
34. **Academy of N, Dietetics**. *Evidence Analysis Manual: Step in the Academy Evidence Analysis Process* (2016.0)
35. Chen Z, Qian F, Liu G, Li M, Voortman T, Tobias DK. **Prepregnancy plant-based diets and the risk of gestational diabetes mellitus: a prospective cohort study of 14,926 women**. *The American Journal of Clinical Nutrition* (2021.0) **114** 1997-2005. PMID: 34510175
36. Mijatovic-Vukas J, Capling L, Cheng S, Stamatakis E, Louie J, Cheung NW. **Associations of Diet and Physical Activity with Risk for Gestational Diabetes Mellitus: A Systematic Review and Meta-Analysis**. *Nutrients* (2018.0) **10** 698. PMID: 29849003
37. Seldin MF, Shigeta R, Villoslada P, Selmi C, Tuomilehto J, Silva G. **European Population Substructure: Clustering of Northern and Southern Populations**. *PloS Genetics* (2006.0) **2**. DOI: 10.1371/journal.pgen.0020143
38. Bennett CJ, Walker RE, Blumfield ML, Gwini S, Ma J, Wang F. **Interventions designed to reduce excessive gestational weight gain can reduce the incidence of gestational diabetes mellitus: A systematic review and meta-analysis of randomised controlled trials**. *Diabetes Research and Clinical Practice* (2018.0) **141** 69-79. DOI: 10.1016/j.diabres.2018.04.010
39. Schoenaker MCS-M. **The Role of Energy, Nutrients, Foods, and Dietary Patterns in the Development of Gestational Diabetes Mellitus: A Systematic Review of Observational Studies**. *Diabetes Care* (2016.0) **39** 16-23. DOI: 10.2337/dc15-0540
40. Joonseok K, Jaehyong C, Kwon SY, McEnvoy JW, Blaha M, Blumenthal RS. **Association of Multivitamin and Mineral Supplementation and Risk of Cardiovascular Disease**. *Circ Cardiovasc Qual Outcomes* (2018.0) 1-14
41. Yuen L, Wong VW, Simmons D, Simmons D. **Ethnic Disparities in Gestational Diabetes**. *Current Diabetes Report* (2018.0) **18**. DOI: 10.1007/s11892-018-1040-2
42. Resnicow K, Baranowski T. **Cultural Sensitivity in Public Health: Defined and Demystified**. *Ethn Dis* (1999.0). PMID: 10355471
43. Kelemen LE, Anand SS, Vuksan V, Yi Q, Teo KK, Devanesen S. **Development and evaluation of cultural food frequency questionnaires for South Asians, Chinese, and Europeans in North America**. *Journal of the American Dietetic Association* (2003.0) **103** 1178-84. DOI: 10.1016/s0002-8223(03)00985-4
44. Mchiza ZJ, Goedecke JH, Lambert EV. **Accuracy of reporting food energy intake: influence of ethnicity and body weight status in South African women**. *South African Journal of Clinical Nutrition* (2010.0) **23**
45. Hébert JR, Peterson KE, Hurley TG, Stoddard AM, Cohen N, Field AE. **The Effect of Social Desirability Trait on Self-reported Dietary Measures among Multi-Ethnic Female Health Center Employees**. *Annals of Epidemiology* (2001.0) **11** 417-27. DOI: 10.1016/s1047-2797(01)00212-5
46. Almiron-Roig E, Aitken A, Galloway C, Ellahi B. **Dietary assessment in minority ethnic groups: a systematic review of instruments for portion-size estimation in the United Kingdom**. *Nutrition Reviews* (2017.0) **75** 188-213. DOI: 10.1093/nutrit/nuw058
47. Taylor K L., Santos Ferreira D, West J, Yang T, Caputo M, A. Lawlor D. **Differences in Pregnancy Metabolic Profiles and Their Determinants between White European and South Asian Women: Findings from the Born in Bradford Cohort**. *Metabolites* (2019.0) **9** 190. DOI: 10.3390/metabo9090190
48. Zulyniak MA, de Souza RJ, Shaikh M, Desai D, Lefebvre DL, Gupta M. **Does the impact of a plant-based diet during pregnancy on birth weight differ by ethnicity? A dietary pattern analysis from a prospective Canadian birth cohort alliance**. *BMJ open* (2017.0) **7** e017753. DOI: 10.1136/bmjopen-2017-017753
|
---
title: “Must you make an app?” A qualitative exploration of socio-technical challenges
and opportunities for designing digital maternal and child health solutions in Soweto,
South Africa
authors:
- Sonja Klingberg
- Molebogeng Motlhatlhedi
- Gugulethu Mabena
- Tebogo Mooki
- Nervo Verdezoto
- Melissa Densmore
- Shane A. Norris
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021787
doi: 10.1371/journal.pgph.0001280
license: CC BY 4.0
---
# “Must you make an app?” A qualitative exploration of socio-technical challenges and opportunities for designing digital maternal and child health solutions in Soweto, South Africa
## Abstract
Participatory and digital health approaches have the potential to create solutions to health issues and related inequalities. A project called Co-Designing Community-based ICTs Interventions for Maternal and Child Health in South Africa (CoMaCH) is exploring such solutions in four different sites across South Africa. The present study captures initial qualitative research that was carried out in one of the urban research sites in Soweto. The aim was twofold: 1) to develop a situation analysis of existing services and the practices and preferences of intended end-users, and 2) to explore barriers and facilitators to utilising digital health for community-based solutions to maternal and child health from multiple perspectives. Semi-structured interviews were conducted with 28 participants, including mothers, other caregivers and community health workers. Four themes were developed using a framework method approach to thematic analysis: coping as a parent is a priority; existing services and initiatives lack consistency, coverage and effective communication; the promise of technology is limited by cost, accessibility and crime; and, information is key but difficult to navigate. Solutions proposed by participants included various digital-based and non-digital channels for accessing reliable health information or education; community engagement events and social support; and, community organisations and initiatives such as saving schemes or community gardens. This initial qualitative study informs later co-design phases, and raises ethical and practical questions about participatory intervention development, including the flexibility of researcher-driven endeavours to accommodate community views, and the limits of digital health solutions vis-à-vis material needs and structural barriers to health and wellbeing.
## Introduction
The increased use of information and communication technologies (ICTs) for health, also referred to as digital health, is creating opportunities to support maternal and child health. Digital health has the potential to address health inequalities and community health challenges if undertaken in a contextually relevant manner so as not to exacerbate inequities [1–3]. Existing uses in South Africa and other African countries include reminders about antenatal visits, pregnancy registration, child health information, supporting health behaviours through mobile messages, and technologies specifically adopted in response to the coronavirus disease 2019 (COVID-19) pandemic and its impact on health care services [4–9]. Most notably, South Africa introduced a text messaging service called MomConnect in 2014 for the purposes of registering pregnancies and providing health information to pregnant and postpartum women [10, 11].
Participatory approaches have gained popularity in health research, providing opportunities to incorporate contextual insights and preferences of intended beneficiaries throughout the process of designing studies or health interventions [12]. The rationale for doing so includes ethical arguments for democratic involvement of intended beneficiaries, minimising research waste by ensuring the practical relevance of research outputs to stakeholders expected to implement findings, and improving the quality of said outputs through grounding the research process in lived experiences [12, 13]. In South Africa, participatory design and research approaches have been used to, for example, develop interventions for optimising immunisation coverage and service delivery [14], and to support mothers of pre-term infants [15].
A multi-site study, Co-Designing Community-based ICTs Interventions for Maternal and Child Health in South Africa (CoMaCH), was initiated partly due to the unequal nature of digital health development, taking a participatory and community-centred approach to this domain. The project team includes a multidisciplinary, cross-cultural, and cross-geographical consortium of researchers, technology designers, healthcare professionals and community stakeholders who are exploring the potential of ICTs to enhance maternal and child health and wellbeing in South Africa [16, 17].
This article describes the formative qualitative research phase that was undertaken to contextualise and inform the subsequent co-design process [18] in one of the research sites, the urban township setting of Soweto in Johannesburg Metropolitan Municipality. Previous qualitative research on digital health has been limited in settings like Soweto [19, 20]. A better understanding of the socio-technical challenges and opportunities that digital health technologies afford is needed in the context of maternal and child health in Soweto, especially with regards to the COVID-19 pandemic and resulting shifts to digital technologies [9, 21]. The aim was thus to develop a situation analysis of existing services and the practices and preferences of intended end-users, and explore barriers and facilitators to utilising ICTs for community-based solutions to maternal and child health from multiple perspectives in Soweto. An additional aim of the analysis was to generate transferable insights about the ethics and power dynamics of incorporating participant views into planning and design processes.
## Materials and methods
The qualitative formative phase involved consulting intended users regarding the context, preferences and previous experiences of digital health. Three groups in Soweto were identified as relevant and feasible to reach for gaining multiple perspectives of the context: recent mothers taking part in a separate health intervention trial run by the research team’s research unit [22]; adjacent caregivers, meaning family members or other individuals with caregiving roles in the mothers’ households; and community health workers employed in the aforementioned trial [23]. Sampling involved purposive considerations of balancing participant groups, and practical considerations of participants’ availability and ability to provide relevant insights. The sample size was pragmatically estimated as a number likely to yield sufficient data for fulfilling the study’s aims. Recruitment was carried out by TM, who invited participants and scheduled interviews. Most participants already had experience of research activities run by the research unit, either through previous studies or having worked as affiliated community health workers. Ethical approval was obtained from the University of the Witwatersrand (M200872, MED20-08-043) and Cardiff University, and written informed consent for participation and audio-recording was given by all participants. Each participant received ZAR 156 (approximately US$ 10) to ensure that no costs (e.g. transport) were incurred from participation.
Semi-structured interviews were conducted at the research unit using interview guides (S1 and S2 Files) shared with CoMaCH research sites in other South African provinces, but with openness to emergent, locally relevant topics. Semi-structured interviews were selected to elicit rich accounts of specific topics of interest, while maintaining flexibility beyond pre-determined questions. As the research unit is situated on a hospital campus, and has the capacity to carry out research following COVID-19 protocols (e.g. symptom screening), these interviews were conducted in-person in line with guidance and approvals from the University. All research staff and participants wore masks at all times, in accordance with South African COVID-19 regulations at the time. The interviews were carried out in private rooms with open windows and sufficient space for physical distancing. MM, GM and SK conducted interviews over two weeks in May 2021 with discussions and debriefs between interviews. Due to MM’s and GM’s working relationships with the community health worker participants, those interviews were carried out in English by SK where possible (7 out of 8 interviews), while MM and GM conducted interviews with ten recent mothers and ten adjacent caregivers, respectively, in English, isiZulu and Setswana as relevant. All interviews were audio-recorded, transcribed verbatim and translated into English where relevant by professional transcribers, and checked and corrected by SK, MM and GM. Audio recordings were stored on a password-protected University server. Copies were securely sent for transcription without any participant identifiers. Some transcription and translation corrections were required, and interpretations arising from the analysis were continuously discussed among the team and checked against recordings to ensure these were not based on incorrectly transcribed or translated data.
To accommodate the multidisciplinary nature of the study, and the three distinct participant groups, qualitative analysis was done using the procedures outlined in the Framework Method [24] approach to thematic analysis [25], which accommodates teamwork and comparisons of different participant sub-groups. The analysis comprised seven stages undertaken collaboratively by the research team: transcription, familiarisation with the interviews, coding, developing a working analytical framework, applying the analytical framework, charting data into the framework matrix, and interpreting the data [24]. SK led the analysis using analysis software MAXQDA, with contributions from MM, GM and TM to coding and the analytical framework, and discussions among the author team regarding the framework matrix and interpreting the data. Coding was initiated through a priori codes based on the interview guide and specific topics that the team had identified as relevant during the interviews. As coding progressed, the initial framework was expanded and refined inductively with new and rephrased codes. Based on the final code framework, themes were developed to capture salient insights and patterns of meaning corresponding with the study aim, and illustrative quotes were charted into the framework matrix to convey nuances and contrasts between different participant groups. The analysis predominantly drew on manifest data content. Higher level abstractions relying on more latent subtleties were treated with caution due to the multilingual and cross-cultural dimensions of the study and discussed in depth among the authors.
This qualitative study employs a subjectivist inductive approach, meaning that the general direction is from context-specific qualitative data to theory and practice [26]. Nevertheless, existing theory and pre-determined topics of interest are also relevant to utilise deductively, and the analysis is thus best described as a pragmatic and theory-informing data analysis, as opposed to being either fully inductive theory development or fully theory-informed [26]. The qualitative approaches utilised here involve assumptions about subjective, local and specific reality and knowledge, while also considering material dimensions (e.g. infrastructure). The study therefore draws on critical realist ontology and epistemology without suggesting a hierarchy of the ontological layers of real, actual and empirical [27, 28].
Procedures to ensure the quality and rigor of the research were undertaken in line with Tracy’s criteria [29] and reported according to the Standards for Reporting Qualitative Research (SRQR) [30], as detailed in the accompanying checklist. Meaningful coherence with the qualitative research paradigm was prioritised over rigid adherence with standardised criteria. For example, we did not pursue inter-rater reliability or member checking, which involve more positivist assumptions about reality and knowledge [31]. Instead, we focused on ensuring the credibility of the analysis through ‘critical friend’ approaches within the research team [31].
Due to the study’s focus on eliciting the views of intended beneficiaries, a relevant interpretive tool is the typology of participation developed by Arnstein through a ladder analogy [32]. The ladder of citizen participation describes different levels from non-participation through degrees of tokenism to the highest degrees of citizen power and shared control over a process. Arnstein’s power-centred theorising has been extended and critiqued in health-related applications [33], but the essence of the ladder as an illustration of power imbalances was taken as a starting point for the analysis of community-based priorities, challenges and facilitators for maternal and child health in the context of societal and health inequalities.
## Results
Twenty-eight adult participants were interviewed. There were no refusals to participate, but some interviews were rescheduled due to participants’ other commitments. Ten participants were recent mothers from different neighbourhoods in Soweto, ten were adjacent caregivers (e.g. grandmothers, aunts and sisters from the same household), and eight were community health workers. All participants were Black African women, which reflects the demographics of Soweto and the legacy of colonial and apartheid urban planning. The absence of men as participants is due to both the design of the trial recent mothers and community health workers were recruited from, and the available and actual choices of recent mothers in identifying adjacent caregivers. Through the Framework Method, four themes were developed with different nuances in each participant group, as demonstrated with quotes in Table 1, and pooled descriptions of each theme below.
**Table 1**
| Unnamed: 0 | Theme | Theme.1 | Theme.2 | Theme.3 |
| --- | --- | --- | --- | --- |
| Group | 1. Coping as a parent is a priority | 2. Existing services and initiatives lack consistency, coverage and effective communication | 3. The promise of technology is limited by cost, accessibility and crime | 4. Information is key but difficult to navigate |
| Recent mothers (RM) | Young mothers can be overwhelmed by the new responsibilities and health and welfare needs arising from parenthood, as well as the emotional aspects of parenthood. Suggested solutions often involved community assistance organisations or sharing accurate health information through e.g. community meetings and pamphlets.“I think that we must have an organisation that will help kids or mothers…so that they can read how they must treat kids, what they must do for a child, so that there must be people that can help you buy other things that you cannot buy for your child, that we cannot do for our kids.” (RM2)“I never received my parents’ love so I didn’t know how to love [my child].” (RM7) | Primary health care is relied on but not reliable, and staff do not make sure that health information is understood.“They are lazy, they don’t want to explain, I don’t understand that clinic, when you take your child there to get and injection you stand in a line, take a card, get an injection and then then you leave, they don’t explain to you.” (RM1)“Seriously they’re very poor because last time when I took him for the 2-year-old shot he didn’t get the injection, they said the government hasn’t yet delivered the material.” (RM3)“Our nurses in our clinics mostly don’t know how to communicate with us as parents. Not all of us went to school for knowing how to treat a condition.” (RM6) | Mobile data is expensive, which means that many only have data for part of the month. Internet-based solutions are not going to reach everyone, and face-to-face interactions are preferred by many. While public Wi-Fi exists in many parts of Soweto, it is not easily accessible, and hanging around in public to access Wi-Fi is not safe or feasible.“Participant: Yes, we buy data on our phonesInterviewer: And how long does data last you?Participant: A week, I buy month end” (RM10)“I cannot go sit in the streets for me to get access [to Wi-Fi].” (RM9)“Okay, to help moms, I think workshops are quite helpful, for face-to-face interactions, and sometimes for, like they show you things, like how they are done… Must you make an app?” (RM5) | Written information can be difficult to make use of without clear and friendly guidance. Some recent mothers are actively finding information from different sources but there are also concerns about misinformation online or from other people.“So I didn’t understand that [maternity case] book to be honest…They did not explain anything… It’s because we didn’t ask, actually when you go to the clinic when you are pregnant they give you that paper and then you do what you are supposed to and then you go home, we didn’t ask.” (RM1)“I use these groups on Facebook, mommy groups, groups for moms that have children, then we ask, and then we help each other out.” (RM5)“Some of us are lazy to go [to clinics], trusting that elders will be there to assist with home remedies, only to find that would worsen the issue.” (RM9) |
| Adjacent caregivers (AC) | Adjacent caregivers consider young women and recent mothers to have a limited understanding of the responsibilities and practicalities of parenthood but many try to support them where possible even if there is also some judgment involved.“Eish ya I would say that to help them we need to tell them that… yes a child is a blessing but…it’s tough these days, so I would teach them that you need to look after the ones you have.” (AC1)“We would be happy if they could…teach us how to raise our grandchildren, what we need to do so that we can understand better.” (AC4) | Queuing, rude or too busy staff, and shortages of medicines compromise health care access and waste people’s time.“Eish it’s poor…and you wake up early to go to the clinic, by 5:30 you’re supposed to have someone join the queue for you, otherwise you’ll finish at 4pm, and they’re always short-staffed; we need to ask someone to wake up early and go and join the queue for the mother and baby.” (AC4)“Sometimes you’re supposed to go for de-worming, and they don’t have de-worming, but if they would have told us via SMS it would be much better than to go and queue. You join the queue when you arrive, and when it’s your time to go in there’s no de-worming, there’s no immunisation.” (AC8) | It cannot be assumed that everyone has access to or is able to use a smartphone or the internet even if there is great potential in technology-based solutions and many would prefer them.“Honestly, when it comes to phones, I’m not very savvy there.” (AC7)“I think technology would be better… [The clinics] haven’t reached that stage of sending WhatsApp (Laughter)… I’m just saying that they haven’t yet reached that stage of sending SMSes, oh no.” (AC8) | Information, awareness and education are seen to be the solutions to many health-related issues, but the way in which information is provided matters.“Education is the best, it surpasses everything… Limited information is dangerous and it doesn’t move us forward.” (AC4)“There are houses where they don’t even have TVs, there’s no phone because not one person in that house works. So, like if… okay in this hall there’s such and such a thing on this date, if you could announce that to people and they go and they listen to the presentation, that would be better.” (AC5) |
| Community health workers (CHW) | Community health workers see the challenges recent mothers and young women go through and try to support as much as possible with limited formal welfare, social services or community-based initiatives in place.“I think young women especially are facing a lot… How do you provide for those kids, how do you make sure they have a good health, they have good nutrition because that obviously affects their health. So, how do you provide for those things? I think that is the challenge that they face.” (CHW2)“They need this information, they really need this information because, first of all, when these girls get pregnant they are not so happy they are pregnant, it is most likely it is an accident they are pregnant. So, when they are pregnant they are depressed, they are nervous, they are anxious. They are thinking ‘I’m barely making it at home how am I going to take care of a child?´.” (CHW5) | Different solutions are needed for different people and situations because no single system works consistently or for everyone. Community health workers frequently engage in problem-solving to ensure they can help and support trial participants as much as possible despite gaps or challenges.“I also think the clinics need to play the part and they are actually not… I have a participant that is 6 months [pregnant], she has not started her clinic visits. She has gone to three different clinics, they send her to this one, to this one, ‘no not booking, we are renovating.’ And this woman is at risk because she once had a miscarriage so she is high risk already.” (CHW2)“She was still young and she was worried about her exams, how is she going to go about it… I had to call the school, which is not part of it, but I went the extra mile to call the school and find out the information in how we can assist her after giving birth.” (CHW3) | Community health workers are sometimes not able to reach people or capture research data efficiently due to phone network issues and lack of data or electricity. Bringing electronic devices into communities has safety implications even if it facilitates their work.“We don’t take tablets to the field most of the time because we are trying to be careful, because you must remember that one of our staff members was mugged, her tablet was taken away.” (CHW7)“It is different each day because today you have network, then by tomorrow you don’t. There is no reliability in our work.” (CHW2) | Communication needs to be tailored to specific needs and situations, and face-to-face meetings and community events are often preferred for reaching particular groups, supporting mothers and ensuring the information is understood. Using pictures and translating information are also important considerations.“I can say home visit is better because others, they will say ‘I don’t have money for WhatsApp or to buy data’. I think home visit is better because you know how to check her baby, if it is eating, or if she is pregnant… Support is better than just over the phone, yes sometimes you can show the support over the phone but others, they prefer you to come. So, home visit is the best.” (CHW7)“There should be one person that they can go to and talk to… about their problems and then maybe there will be pictures of communicating maybe, just to say if you are experiencing this, or doing this to the child, just by pictures I think that would be well. And the person should speak the very same language they are speaking so it will be clear.” (CHW4) |
The themes are formulated as specific conclusions from the analysis, comprising: 1) coping as a parent is a priority; 2) existing services and initiatives lack consistency, coverage and effective communication; 3) the promise of technology is limited by cost, accessibility and crime; and 4) information is key but difficult to navigate. In addition to these four themes, the dataset captured participants’ ideas and solutions to maternal and child health, and these are described under the sub-heading of ‘Suggested solutions for maternal and child health in Soweto’.
## Coping as a parent is a priority
The interviews highlighted how the topic of maternal and child health forms only one part of a complex reality experienced by young women in Soweto. Recent mothers are navigating the emotional dimensions of parenthood along with the more practical aspects. In addition, welfare, income and employment featured strongly in the interviews because both emotional and financial coping are at the core of motherhood for the participants of this study. Notably, this includes concerns around mental health, as evidenced by the quote about first-time mothers and post-natal depression.
While there were differences between how different participant groups described parenting and coping, there was a shared emphasis on young women needing considerable and tailored support. However, adjacent caregivers and community health workers tended to express concerns, and even judgement, towards young mothers in Soweto, describing pregnancies as unwanted.
## Existing services and initiatives lack consistency, coverage and effective communication
There was a recognition among participants that there are many resources and services available, but these do not necessarily cover all the needs of participants or work in a reliable way. Primary health care is relied on for most health-related needs but there are several issues with how services work, and delays in accessing care and medication deter people from spending time trying to get help through clinics. Existing technological solutions, such as MomConnect, are seen as helpful but tend to work inconsistently in practice. Out of the participant groups, the community health workers were the most knowledgeable and optimistic about existing services but recognised the issues mothers may have in accessing these. Recent mothers, more than adjacent caregivers, tended to be familiar with many existing services and solutions but had needs beyond what these currently cover, as described in more detail under the theme about coping as a parent.
## The promise of technology is limited by cost, accessibility and crime
There was a general recognition of technology offering many solutions in terms of efficient communication and information, and its use was seen as somewhat inevitable, as explained by one of the community health workers: “People tend to respond better if they have technology, and they are using technology so much because people can’t live without their phones.” ( CHW3) However, key barriers were highlighted across participant groups in terms of the costs of data, access to and ability to use technology, and the risks of devices being stolen or the need to be in potentially dangerous public spaces in order to access public Wi-Fi. Community health workers knew of more existing digital health solutions and were, for example, more experienced than other groups in using specific websites and online forums for finding health information. Some recent mothers were similarly active in finding information online, but others, especially the adjacent caregivers, were not familiar or comfortable with making use of digital health resources.
## Information is key but difficult to navigate
The ideas participants shared for improving maternal and child health centred on access to information and health education. However, there were also many issues with navigating reliable sources of information or health information being provided in an accessible format, including in the participants’ own language. Many therefore suggested alternatives to ICTs in terms of mobilising and reaching communities in more traditional ways such as through meetings, events, pamphlets, flyers, radio or TV.
Concerning the reliability of health information, participants flagged both traditional practices and information on the internet as potentially harmful or inaccurate. Young mothers were not always sure they could trust the advice they received from elders or neighbours, whereas older adjacent caregivers suggested that with more support and information, they could be better equipped to help the younger generations when it comes to parenting and health.
While education and awareness were seen as essential, it was also pointed out that the availability of information does not mean it is well understood, let alone acted upon. People’s lack of awareness, or even more judgemental notions of ignorance or laziness, were cited across participant groups as reasons why existing services were not made use of, or why there was a need for more health information. This also reflected a degree of cynicism about information leading to change, as expressed by one recent mother: “What help is it to know something… I am like ‘why must I have this information when I won’t benefit anything, where I won’t help anyone with that’, it’s just better not to know it.” ( RM1) However, it is important to note here that while participants tended to describe individuals’ role in navigating information or services, their challenges also reflect broader and structural barriers to accessing or utilising services, as described under the second theme.
## Suggested solutions for maternal and child health in Soweto
Apart from either explicit or implicit references to mental health concerns, recent mothers and adjacent caregivers in Soweto found it challenging to identify specific health challenges in their community when asked directly. Community health workers tended to focus on the topics covered in the intervention they are delivering: nutrition, physical activity, mental health, pregnancy and prevention or management of non-communicable diseases. *In* general, access to health information, welfare, and any other support required to act on health information were discussed across interviews and topics, but these were not necessarily anchored in any specific aspect of health or illness. The suggested solutions to maternal and child health provided by study participants are described below according to broad categories.
## Digital health solutions
The range of digital health solutions suggested at this initial phase of the research included: electronic booking systems for clinics; reminders of appointments and children’s vaccinations; online support for locating the nearest clinic and checking whether specific services or medications are available; more use of technology and visual aids at clinics (e.g. showing information about anthropometry on screens); improving existing services such as MomConnect; online services for government agencies; digital food assistance vouchers; expanding public Wi-Fi; apps, online groups and forums providing health information and answers to specific question; and, free data for health-related purposes. These suggestions have fed into the co-design phases of the CoMaCH project.
## Workshops and word of mouth
Due to the challenges related to internet access, and the perceived availability of many people due to high levels of unemployment in Soweto, a typical suggestion was health-themed workshops or other events for providing health education and information, including parenting education and support for recent mothers. In response to the long waiting times at clinics, it was also suggested that such health education sessions could take place at clinics while pregnant women or mothers with their children are waiting to be seen by health care providers.
## Home visits, face-to-face support and community engagement
One solution that recent mothers and adjacent caregivers frequently proposed was home visits by community health workers, and community health workers themselves also emphasised the value of meeting face-to-face or doing home visits, as opposed to digital health solutions. It is important to note that both public sector and research-affiliated community health workers are deployed across Soweto already, but many participants were unfamiliar with this cadre of health workers.
In addition, participants across the different groups emphasised approaches like going from door to door to share information about upcoming health information events, and the potential to reach many people through community engagement or sensitisation events. Community health workers provided advice on getting buy-in from communities through engaging leaders and designing health promotion efforts in a participatory way: “When you want to do events, you bring [the community] in to brainstorm… You give the job to the community, and that is how things will work.” ( CHW7)
## Flyers, pamphlets, posters and media
Many participants mentioned the use of printed information, such as flyers, newspapers, pamphlets and posters, and more interactive solutions such as radio, TV and social media. However, few participants had specific ideas for what approaches would be interesting or engaging enough to get people’s attention, and some acknowledged that people may easily ignore information shared via these channels as the theme of information being key but difficult to navigate illustrates.
## Community food assistance and gardening
Community-based organisations or churches providing basic food assistance were commonly mentioned as the only existing non-governmental health resources in Soweto, especially in the context of the COVID-19 pandemic and increasing food insecurity due to lockdowns and unemployment. Some of the suggested solutions for promoting maternal and child health therefore also revolved around expanding food assistance, and setting up community gardens on vacant plots of land.
## Group saving schemes and community organisations
The interviews included some examples of existing group saving schemes, typically referred to as stokvel in South Africa. These were mentioned as a potential solution for maternal and child health challenges in Soweto through providing financial and social support to families. Similarly, some participants suggested setting up their own community-based organisation to help provide families with childcare, food, children’s clothes, and other basic needs, or help people set up small businesses and support each other. There were also calls to improve existing community organisations, for example by ensuring that early childhood development centres have qualified staff.
## Discussion
This article provides a situation analysis and explores barriers and facilitators to utilising ICTs for community-based solutions to maternal and child health from the perspectives of recent mothers, adjacent caregivers and community health workers in Soweto, South Africa. The qualitative analysis resulted in four themes: 1) coping as a parent is a priority; 2) existing services and initiatives lack consistency, coverage and effective communication; 3) the promise of technology is limited by cost, accessibility and crime; and 4) information is key but difficult to navigate. The message is clear that internet-based technology has great potential for health promotion, but it is not consistently accessed by everyone, and thus may fall short of the need to address rather than exacerbate inequalities [1]. This is in line with amplification theory characterisations of technology as enforcing and amplifying, rather than challenging or transforming, existing institutional or structural forces [34].
The findings also echo previous qualitative research from Soweto in that health concerns, especially any future or relatively abstract health issues, must be seen against the background of more urgent personal needs, such as welfare and income, and underlying structural factors that health interventions rarely address [22, 35, 36]. Coping as parents was a common concern with multiple dimensions such as financial, emotional and practical coping. Participants therefore called for material and social support as well as better health information and improved primary health care services.
The experiences captured in Soweto reflect some broader shortcomings of existing digital health resources, and the challenges of navigating what some authors refer to as information ecology [37, 38]. These findings complement evaluations of MomConnect [39, 40] suggesting that there is a general interest in utilising the service if the issues users are facing can be addressed. Identified technical issues include problems with registering for or consistently using MomConnect, and challenges related to language or literacy when engaging with written information [41, 42]. Our participants expressed more critical views of MomConnect than specific evaluations of the programme seem to have received, suggesting that the participants did not perceive the research team to be representing MomConnect, and felt more comfortable being critical about its shortcomings.
These findings echo recognised impediments to trust in, and utilisation of, digital health technologies, including cost, unequal access, defective technology, poor information quality and inadequate publicity hindering optimal use of already existing digital health solutions [9, 43]. Apart from new solutions, enhancing the quality and reliability of existing services would therefore be a promising and economical avenue for improving digital health promotion in South Africa. It may also be necessary to tailor features of existing services from a user experience perspective, as despite the many requests for a service that can provide health information, participants did not seem to be familiar with such features of MomConnect. Indeed, a national evaluation of the interactive features of South Africa’s MomConnect mobile messaging programme found that only about $8\%$ of registered users engaged with the programme’s features that enabled them to request specific health information [44]. The present study cannot fully explain why utilisation of seemingly promising and acceptable features of existing services remains low, but we recommend engaging users in the development and improvement processes.
While this formative phase involved an openness to emerging ideas and solutions, the scope of solutions was already delineated to digital health, and the topic was broadly described as ‘maternal and child health’, in line with an international funding call. A fully open-ended participatory process for setting the research agenda or intervention focus [45] was not pursued, in part because of pandemic-related delays and challenges. This raises questions about the extent to which intended beneficiaries’ views are listened to and incorporated into the design of research or health interventions [46, 47]. If end-users do not have agenda-setting powers, the scope of solutions is going to be limited to what researchers were able to imagine a priori [32, 48]. This is particularly important to consider in a context where the relatively low priority of health issues vis-à-vis socioeconomic concerns is already well known, and where challenges of technology, including power supply and internet access, easily justify questions like the one posed by a recent mother: “Must you make an app?” ( RM5).
Another clash of practical and ethical considerations is the mismatch between participant recommendations, established evidence about health interventions, and problematic histories of paternalistic interventions. For example, participants called for ‘educating’ the community, and addressing health issues through providing information and raising awareness about specific topics. However, it is a typical assumption that giving people information will lead to health-related behaviour change [49, 50]. Based on decades of health psychology, intervention research and implementation science, some recommendations from participants would likely not lead to effective health interventions or solutions. Nevertheless, public awareness is an important component of wider health promotion efforts [49], and participants’ concerns about misinformation or harmful advice should be taken seriously. Finding the balance between listening to participant views, and appropriately incorporating those with existing theoretical and empirical research knowledge is an inevitable challenge of participatory research and health interventions [36, 51, 52]. This, again, points to the responsibility of researchers to also welcome participant views in decision-making, and provide accessible feedback about the process (including why something may not work) in order to be accountable to the groups contributing their time, experiences and ideas to the process.
To return to Arnstein’s ladder of citizen participation [32], unless participatory research or intervention design processes are situated on the highest rungs, with participating communities enjoying full authority in terms of setting the agenda according to their needs and preferences, the decision-making will inevitably be dominated by the more powerful. Intentional efforts by research teams, who tend to hold more power than participants, are needed to start bridging such power imbalances. Researchers engaging in participatory or co-design processes thus have a responsibility to communicate clearly and transparently about the topic, the scope of feasible solutions or the process itself, and the extent to which people’s real concerns can be accommodated within the parameters of the study or intervention design process [48, 53].
In an effort to address problematic power imbalances and avoid paternalistic health interventions, researchers have drawn on Freire’s philosophy [54, 55], aiming to foster a level of critical consciousness about oppressive social structures as a precondition for successful health interventions [56, 57]. Indeed, interventions drawing on notions of critical consciousness have had some success in starting to address structural factors, such as norms around health behaviours, through participatory approaches [58]. Critical consciousness thus enables a recognition of not merely the challenges individuals or communities face, but a deeper understanding of relational factors and power dynamics at play in producing and maintaining (health) inequalities [59]. For developing digital health interventions, critical consciousness could entail awareness and activism around the wider societal inequalities that inevitably influence health promotion efforts, as well as exposure to the more technical and design-specific aspects of digital health [46] (e.g. scale, sustainability and impact [60]). As Fig 1 illustrates, the ladder of participation can incorporate these elements as well, indicating the desired direction towards participation on equal terms. This involves intentional power sharing on the part of the researchers, and critical consciousness on the part of communities, resulting in the ladder analogy becoming more like a staircase.
**Fig 1:** *Staircase of participation [32, 54].*
The participants provided many potential solutions, both digital and beyond, demonstrating the utility of relatively open-ended qualitative research in informing intervention design. Another strength of the study was drawing on diverse perspectives, both through the different participant groups and the multidisciplinary research team composition, as it enabled a holistic consideration of what may be relevant, effective or feasible. To honour the participants’ contributions, the insights about non-digital solutions are feeding into another Soweto-based research initiative to develop community engagement with health research and interventions [61]. For example, participants’ interest in events where reliable health information is provided can be responded to through the research unit’s wider community engagement and research dissemination efforts, and community gardening initiatives are being piloted and supported by the unit.
Limitations of the study include data quality concerns due to transcription and translation errors, which are challenging to fully mitigate when several languages are used simultaneously. These challenges have been addressed through continuous discussions about reflexivity and interpretations within the author team [62]. Furthermore, the analysis could have benefitted from more contextual and sociodemographic data. Such data could have provided further nuance but were not collected as the study was not designed for sociodemographic representativeness or sub-group analyses. Another shortcoming is the absence of men among participants, limiting the scope of the study in terms of insights about digital health preferences and utilisation of existing services, as these may involve gendered differences [63]. While men’s perspectives were not intentionally excluded, our approach did not involve an active effort to include men, which has been found to be necessary for other qualitative child health research in Soweto [64] and is currently being pursued in other studies at the research unit. The participation and inclusion of men, as well as comparisons between urban and rural research sites, has also been discussed more thoroughly in later phases of the CoMaCH research and co-design process [18].
## Conclusions
This study captured insights about the potential of digital and non-digital solutions for maternal and child health in Soweto. Participatory approaches can help realise such potential in more equitable ways, and this study also sheds light on many context-specific barriers to equitable implementation of digital health solutions. These include factors like the cost of data and limited access to devices, such as smartphones, and underlying socioeconomic circumstances that necessitate more holistic consideration and interventions, such as access to welfare and social support. The findings also highlight the value of conducting formative qualitative research to inform co-design processes for developing health interventions. Further research on men’s engagement with digital health in this setting is needed.
## References
1. 1World Health Organization. WHO guideline: recommendations on digital interventions for health system strengthening Web Supplement 2: Summary of findings and GRADE tables. Geneva; 2019. https://www.who.int/publications/i/item/9789241550505
2. Mohan D, Scott K, Shah N, Bashingwa JJH, Chakraborty A, Ummer O. **Can health information through mobile phones close the divide in health behaviours among the marginalised? An equity analysis of Kilkari in Madhya Pradesh, India**. *BMJ Glob Heal* (2021.0) **6** e005512. DOI: 10.1136/bmjgh-2021-005512
3. Goto R, Watanabe Y, Yamazaki A, Sugita M, Takeda S, Nakabayashi M. **Can digital health technologies exacerbate the health gap? A clustering analysis of mothers’ opinions toward digitizing the maternal and child health handbook**. *SSM—Popul Heal* (2021.0) **16** 100935. DOI: 10.1016/J.SSMPH.2021.100935
4. Coleman J, Bohlin KC, Thorson A, Black V, Mechael P, Mangxaba J. **Effectiveness of an SMS-based maternal mHealth intervention to improve clinical outcomes of HIV-positive pregnant women**. *AIDS Care* (2017.0) **29** 890-897. DOI: 10.1080/09540121.2017.1280126
5. 5Wardle CJ, Green M, Mburu CW, Densmore M. Exploring co-design with breastfeeding mothers. Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ‘18). New York: Association for Computing Machinery; 2018. pp. 1–12.
6. Swartz A, LeFevre AE, Perera S, Kinney MV, George AS. **Multiple pathways to scaling up and sustainability: an exploration of digital health solutions in South Africa**. *Global Health* (2021.0) **17**. DOI: 10.1186/s12992-021-00716-1
7. Atnafu A, Otto K, Herbst CH. **The role of mHealth intervention on maternal and child health service delivery: findings from a randomized controlled field trial in rural Ethiopia**. *mHealth* (2017.0) **3** 39-39. DOI: 10.21037/mhealth.2017.08.04
8. Willcox M, Moorthy A, Mohan D, Romano K, Hutchful D, Mehl G. **Mobile Technology for Community Health in Ghana: Is Maternal Messaging and Provider Use of Technology Cost-Effective in Improving Maternal and Child Health Outcomes at Scale?**. *J Med Internet Res* (2019.0) **21**. DOI: 10.2196/11268
9. Mbunge E, Batani J, Gaobotse G, Muchemwa B. **Virtual healthcare services and digital health technologies deployed during coronavirus disease 2019 (COVID-19) pandemic in South Africa: a systematic review**. *Glob Heal J* (2022.0) **6** 102-113. DOI: 10.1016/j.glohj.2022.03.001
10. Pillay Y, Motsoaledi PA. **Digital health in South Africa: innovating to improve health**. *BMJ Glob Heal* (2018.0) **3** e000722. DOI: 10.1136/bmjgh-2018-000722
11. Peter J, Barron P, Pillay Y. **Using mobile technology to improve maternal, child and youth health and treatment of HIV patients**. *South African Med J* (2015.0) **106** 3-4. DOI: 10.7196/SAMJ.2016.v106i1.10209
12. Slattery P, Saeri AK, Bragge P. **Research co-design in health: A rapid overview of reviews**. *Heal Res Policy Syst* (2020.0) **18** 1-13. DOI: 10.1186/s12961-020-0528-9
13. Kaisler RE, Missbach B. **Co-creating a patient and public involvement and engagement “how to” guide for researchers**. *Res Involv Engagem* (2020.0) **6** 1-10. DOI: 10.1186/s40900-020-00208-3
14. Timothy A, Coetzee D, Morgan C, Kelaher M, Bailie RS, Danchin M. **Using an adaptive, codesign approach to strengthen clinic-level immunisation services in Khayelitsha, Western Cape Province, South Africa**. *BMJ Glob Heal* (2021.0) **6** e004004. DOI: 10.1136/bmjgh-2020-004004
15. 15Mburu CW, Wardle CJ, Joolay Y, Densmore M. Co-designing with mothers and neonatal unit staff: Use of technology to support mothers of preterm infants. Proceedings of the Second African Conference for Human Computer Interaction: Thriving Communities (AfriCHI ‘18). New York: Association for Computing Machinery; 2018. pp. 1–10.
16. 16CoMaCH Network. CoMaCH Network: Co-Designing Community-based ICTs Interventions for Maternal and Child Health in South Africa. 2021 [cited 10 Nov 2021]. https://comach.melissadensmore.com/
17. 17Till S, Densmore M, Coleman TL, Dias NV. Poster: Lessons from Doing Fieldwork with Bandwidth-constrained communities Online. Proc 2021 4th ACM SIGCAS Conf Comput Sustain Soc COMPASS 2021. 2021; 445–448.
18. 18Till S, Farao J, Coleman TL, Shandu LD, Khuzwayo N, Muthelo L, et al. Community-based co-design across geographic locations and cultures: methodological lessons from co-design workshops in South Africa. Proceedings of the Participatory Design Conference 2022—Volume 1 (PDC ‘22). New York: Association for Computing Machinery; 2022.
19. Dietrich JJ, Lazarus E, Andrasik M, Hornschuh S, Otwombe K, Morgan C. **Mobile Phone Questionnaires for Sexual Risk Data Collection Among Young Women in Soweto, South Africa**. *AIDS Behav* (2018.0) **22** 2312-2321. DOI: 10.1007/s10461-018-2080-y
20. Dietrich JJ, Benadé GL, Mulaudzi M, Kagee A, Hornschuh S, Makhale LM. **“You Are on the Right Track With the App:” Qualitative Analysis of Mobile Phone Use and User Feedback Regarding Mobile Phone Sexual Risk Assessments for HIV Prevention Research**. *Front Digit Heal* (2021.0) **3**. DOI: 10.3389/fdgth.2021.576514
21. Lubinga E, Sitto K, Molebatsi K. **Health disparities and the digital divide within South African disadvantaged communities during the COVID-19 pandemic**. *Catalan J Commun Cult Stud* (2021.0) **13** 285-302. DOI: 10.1386/CJCS_00054_1
22. Draper CE, Prioreschi A, Ware LJ, Lye S, Norris SA. **Pilot implementation of Bukhali: A preconception health trial in South Africa**. *SAGE open Med* (2020.0) **8**. DOI: 10.1177/2050312120940542
23. Draper CE, Bosire E, Prioreschi A, Ware LJ, Cohen E, Lye SJ. **Urban young women’s preferences for intervention strategies to promote physical and mental health preconception: A Healthy Life Trajectories Initiative (HeLTI)**. *Prev Med Reports* (2019.0) **14**. DOI: 10.1016/j.pmedr.2019.100846
24. Gale NK, Heath G, Cameron E, Rashid S, Redwood S. **Using the framework method for the analysis of qualitative data in multi-disciplinary health research**. *BMC Med Res Methodol* (2013.0) **13** 1-8. DOI: 10.1186/1471-2288-13-117
25. Kiger ME, Varpio L. **Thematic analysis of qualitative data: AMEE Guide No. 131**. *Med Teach* (2020.0) **42** 846-854. DOI: 10.1080/0142159X.2020.1755030
26. Varpio L, Paradis E, Uijtdehaage S, Young M. **The Distinctions between Theory, Theoretical Framework, and Conceptual Framework**. *Acad Med* (2020.0) 989-994. DOI: 10.1097/ACM.0000000000003075
27. Wiltshire G.. **A case for critical realism in the pursuit of interdisciplinarity and impact**. *Qual Res Sport Exerc Heal* (2018.0) **10** 525-542. DOI: 10.1080/2159676X.2018.1467482
28. Ronkainen NJ, Wiltshire G. **Rethinking validity in qualitative sport and exercise psychology research: a realist perspective**. *Int J Sport Exerc Psychol* (2021.0) **19** 13-28. DOI: 10.1080/1612197X.2019.1637363
29. Tracy SJ. **Qualitative Quality: Eight “Big-Tent” Criteria for Excellent Qualitative Research**. *Qual Inq* (2010.0) **16** 837-851. DOI: 10.1177/1077800410383121
30. O’Brien BC, Harris IB, Beckman TJ, Reed DA, Cook DA. **Standards for Reporting Qualitative Research**. *Acad Med* (2014.0) **89** 1245-1251. DOI: 10.1097/ACM.0000000000000388
31. Smith B, McGannon KR. **Developing rigor in qualitative research: problems and opportunities within sport and exercise psychology**. *Int Rev Sport Exerc Psychol* (2018.0) **11** 101-121. DOI: 10.1080/1750984X.2017.1317357
32. Arnstein SR. **A Ladder Of Citizen Participation**. *J Am Inst Plann* (1969.0) **35** 216-224. DOI: 10.1080/01944366908977225
33. Tritter JQ, McCallum A. **The snakes and ladders of user involvement: Moving beyond Arnstein**. *Health Policy (New York)* (2006.0) **76** 156-168. DOI: 10.1016/j.healthpol.2005.05.008
34. 34Toyama K. Technology as amplifier in international development. Proceedings of the 2011 iConference (iConference ‘11). New York; 2011. pp. 75–82.
35. Klingberg S, Van Sluijs EMF, Draper CE. **Parent perspectives on preschoolers’ movement and dietary behaviours: a qualitative study in Soweto, South Africa**. *Public Health Nutr* (2021.0) **24** 3637-3647. DOI: 10.1017/S1368980020003730
36. Blue S, Shove E, Carmona C, Kelly MP. **Theories of practice and public health: understanding (un)healthy practices**. *Crit Public Health* (2014.0) **26** 36-50. DOI: 10.1080/09581596.2014.980396
37. Pekkarinen S, Hasu M, Melkas H, Saari E. **Information ecology in digitalising welfare services: a multi-level analysis**. *Inf Technol People* (2021.0) **34** 1697-1720. DOI: 10.1108/ITP-12-2019-0635
38. 38Bagalkot N, Verdezoto N, Ghode A, Purohit S, Murthy L, MacKintosh N, et al. Beyond Health Literacy: Navigating Boundaries and Relationships during High-risk Pregnancies: Challenges and Opportunities for Digital Health in North-West India. Proceedings of the 11th Nordic Conference on Human-Computer Interaction: Shaping Experiences, Shaping Society (NordiCHI ‘20). New York: Association for Computing Machinery; 2020.
39. Skinner D, Delobelle P, Pappin M, Pieterse D, Esterhuizen TM, Barron P. **User assessments and the use of information from MomConnect, a mobile phone text-based information service, by pregnant women and new mothers in South Africa**. *BMJ Glob Heal* (2018.0) **3** e000561. DOI: 10.1136/bmjgh-2017-000561
40. LeFevre AE, Dane P, Copley CJ, Pienaar C, Parsons AN, Engelhard M. **Unpacking the performance of a mobile health information messaging program for mothers (MomConnect) in South Africa: evidence on program reach and messaging exposure**. *BMJ Glob Heal* (2018.0) **3** e000583. DOI: 10.1136/bmjgh-2017-000583
41. 41Schneidermann N. Texting Like A State: mHealth and the first thousand days in South Africa. In: Somatosphere [Internet]. 2018 [cited 6 Feb 2022]. http://somatosphere.net/2018/texting-like-a-state.html/
42. 42Densmore M. Experiences with bulk SMS for health financing in Uganda. CHI ‘12 Extended Abstracts on Human Factors in Computing Systems (CHI EA ‘12). New York; 2012. pp. 383–398.
43. Adjekum A, Blasimme A, Vayena E. **Elements of Trust in Digital Health Systems: Scoping Review**. *J Med Internet Res* (2018.0) **20**. DOI: 10.2196/11254
44. Xiong K, Kamunyori J, Sebidi J. **The MomConnect helpdesk: how an interactive mobile messaging programme is used by mothers in South Africa**. *BMJ Glob Heal* (2018.0) **3** e000578. DOI: 10.1136/bmjgh-2017-000578
45. Dowhaniuk N, Ojok S, McKune SL. **Setting a research agenda to improve community health: An inclusive mixed-methods approach in Northern Uganda**. *PLoS One* (2021.0) **16**. DOI: 10.1371/journal.pone.0244249
46. Rogers Y, Marsden G. **Does He Take Sugar? Moving Beyond the Rhetoric of Compassion**. *Interactions* (2013.0) **20** 48-57. DOI: 10.1145/2486227.2486238
47. Gonsalves PP, Hodgson ES, Kumar A, Aurora T, Chandak Y, Sharma R. **Design and Development of the “POD Adventures” Smartphone Game: A Blended Problem-Solving Intervention for Adolescent Mental Health in India**. *Front Public Heal* (2019.0) **7**. DOI: 10.3389/fpubh.2019.00238
48. Steen M.. **Co-Design as a Process of Joint Inquiry and Imagination**. *Des Issues* (2013.0) **29** 16-28. DOI: 10.1162/DESI_A_00207
49. Kelly MP, Barker M. **Why is changing health-related behaviour so difficult?**. *Public Health* (2016.0) **136** 109-116. DOI: 10.1016/j.puhe.2016.03.030
50. Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W. **The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: Building an international consensus for the reporting of behavior change interventions**. *Ann Behav Med* (2013.0) **46** 81-95. DOI: 10.1007/s12160-013-9486-6
51. Jennings HM, Morrison J, Akter K, Kuddus A, Ahmed N, Kumer Shaha S. **Developing a theory-driven contextually relevant mHealth intervention**. *Glob Health Action* (2019.0) **12**. DOI: 10.1080/16549716.2018.1550736
52. 52Hekler EB, Klasnja P, Froehlich JE, Buman MP. Mind the theoretical gap: Interpreting, using, and developing behavioral theory in HCI research. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ‘13). New York: Association for Computing Machinery; 2013. pp. 3307–3316.
53. 53Stowell E, Lyson MC, Saksono H, Wurth RC, Jimison H, Pavel M, et al. Designing and evaluating mHealth interventions for vulnerable populations: A systematic review. Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ‘18). New York: Association for Computing Machinery; 2018. pp. 1–17.
54. Freire P.. *Pedagogy of the oppressed* (1972.0)
55. Jemal A.. **Critical Consciousness: A Critique and Critical Analysis of the Literature**. *Urban Rev* (2017.0) **49** 602-626. DOI: 10.1007/s11256-017-0411-3
56. Morrison J, Akter K, Jennings HM, Kuddus A, Nahar T, King C. **Implementation and fidelity of a participatory learning and action cycle intervention to prevent and control type 2 diabetes in rural Bangladesh**. *Glob Heal Res Policy* (2019.0) **4**. DOI: 10.1186/s41256-019-0110-6
57. Campbell C, MacPhail C. **Peer education, gender and the development of critical consciousness: participatory HIV prevention by South African youth**. *Soc Sci Med* (2002.0) **55** 331-345. DOI: 10.1016/s0277-9536(01)00289-1
58. Morrison J, Akter K, Jennings HM, Nahar T, Kuddus A, Shaha SK. **Participatory learning and action to address type 2 diabetes in rural Bangladesh: A qualitative process evaluation**. *BMC Endocr Disord* (2019.0) **19** 118. DOI: 10.1186/s12902-019-0447-3
59. Mosse D.. **A Relational Approach to Durable Poverty, Inequality and Power**. *J Dev Stud* (2010.0) **46** 1156-1178. DOI: 10.1080/00220388.2010.487095
60. Peter J.. **Achieving scale, sustainability and impact: a donor perspective on a mobile health messaging service and help desk (MomConnect) for South African mothers**. *BMJ Glob Heal* (2018.0) **3** e000562. DOI: 10.1136/bmjgh-2017-000562
61. Klingberg S, Adhikari B, Draper CE, Bosire EN, Tiigah P, Nyirenda D. **Engaging communities in non-communicable disease research and interventions in low- and middle-income countries: a realist review protocol**. *BMJ Open* (2021.0) **11** e050632. DOI: 10.1136/bmjopen-2021-050632
62. Subramani S.. **Practising reflexivity: Ethics, methodology and theory construction**. *Methodol Innov* (2019.0) **12**. DOI: 10.1177/2059799119863276
63. Chakraborty A, Mohan D, Scott K, Sahore A, Shah N, Kumar N. **Does exposure to health information through mobile phones increase immunisation knowledge, completeness and timeliness in rural India?**. *BMJ Glob Heal* (2021.0) **6** e005489. DOI: 10.1136/BMJGH-2021-005489
64. Klingberg S, van Sluijs EMF, Draper CE. **“The thing is, kids don’t grow the same”: Parent perspectives on preschoolers’ weight and size in Soweto, South Africa**. *PLoS One* (2020.0) **15** e0231094. DOI: 10.1371/journal.pone.0231094
|
---
title: 'Population benefits of addressing programmatic and social determinants of
gender disparities in tuberculosis in Viet Nam: A modelling study'
authors:
- Katherine C. Horton
- Richard G. White
- Nguyen Binh Hoa
- Hai Viet Nguyen
- Roel Bakker
- Tom Sumner
- Elizabeth L. Corbett
- Rein M. G. J. Houben
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021793
doi: 10.1371/journal.pgph.0000784
license: CC BY 4.0
---
# Population benefits of addressing programmatic and social determinants of gender disparities in tuberculosis in Viet Nam: A modelling study
## Abstract
High prevalence of infectious tuberculosis among men suggests potential population-wide benefits from addressing programmatic and social determinants of gender disparities. Utilising a sex-stratified compartmental transmission model calibrated to tuberculosis burden estimates for Viet Nam, we modelled interventions to increase active case finding, to reduce tobacco smoking, and to reduce alcohol consumption by 2025 in line with national and global targets. For each intervention, we examined scenarios differentially targeting men and women and evaluated impact on tuberculosis morbidity and mortality in men, women, and children in 2035. Active case finding interventions targeting men projected greater reductions in tuberculosis incidence in men, women, and children ($16.2\%$, uncertainty interval, UI, 11.4–$23.0\%$, $11.8\%$, UI 8.0–$18.6\%$, and $21.5\%$, UI 16.9–$28.5\%$, respectively) than those targeting women ($5.2\%$, UI 3.8–$7.1\%$, $5.4\%$, UI 3.9–$7.3\%$, and $8.6\%$, UI 6.9–$10.7\%$, respectively). Projected reductions in tuberculosis incidence for interventions to reduce male tobacco smoking and alcohol consumption were greatest for men ($17.4\%$, UI 11.8–$24.7\%$, and $11.0\%$, UI 5.4–$19.4\%$, respectively), but still substantial for women ($6.9\%$, UI 3.8–$12.5\%$, and $4.4\%$, UI 1.9–$10.6\%$, respectively) and children ($12.7\%$, UI 8.4–$19.0\%$, and $8.0\%$, UI 3.9–$15.0\%$, respectively). Comparable interventions targeting women projected limited impact, with declines of $0.3\%$ (UI $0.2\%$-$0.3\%$) and $0.1\%$ (UI $0.0\%$-$0.1\%$), respectively. Addressing programmatic and social determinants of men’s tuberculosis burden has population-wide benefits. Future interventions to increase active case finding, to reduce tobacco smoking, and to reduce harmful alcohol consumption, whilst not ignoring women, should focus on men to most effectively reduce tuberculosis morbidity and mortality in men, women, and children.
## Introduction
In 2019, an estimated 10.0 million people developed tuberculosis (TB), and 1.4 million people with TB died, making TB the leading infectious cause of death worldwide [1]. Substantial gender disparities exist in the epidemiological burden of TB. Nearly twice as many cases of TB are reported among men as among women each year [1]. Prevalence surveys indicate greater disparity in the underlying burden of disease, with men accounting for $70\%$ of adults with undiagnosed TB in low- and middle-income settings [2]. Men’s disadvantage in TB extends to diagnostic and treatment pathways, with men having less access to timely diagnosis [2,3] and worse treatment outcomes than women [4]. As a result of these disparities, over $60\%$ of adult TB deaths occur among men each year [1], and most new infections among men, women, and children are likely attributable to contact with men [5,6].
A gender-responsive approach is needed to address factors that increase men’s risk of infection with *Mycobacterium tuberculosis* (Mtb) and progression to TB [6–12] whilst limiting men’s access to timely and appropriate care [13–15]. Applying a gender lens to the World Health Organization (WHO) End TB Strategy [16] highlights opportunities to address programmatic and social determinants of gender disparities in TB burden and care. Prioritisation of early diagnosis and treatment for all TB patients through integrated, patient-centred care [16] offers opportunity to confront gender disparities throughout the TB care cascade. Calls for bold policies and systems to address social determinants of TB, particularly among vulnerable groups [16], encompass strategies to tackle those social determinants that contribute to gender disparities in TB.
Viet *Nam is* an exemplar of the potential impact of a gender-responsive approach to TB. The country is a high TB burden country, with an estimated incidence of 176 per 100,000 in 2019 [1]. The first national TB prevalence survey in 2007 identified five male cases of undiagnosed TB for every female case, one of the highest sex ratios in the world [17]. Gaps in disease detection and reporting were significantly higher among men than women [17], and modelling indicates that men with infectious TB remained undiagnosed for one year longer than women [3]. By the second national TB prevalence survey in 2017, overall TB prevalence had fallen and gender gaps in detection and reporting had been eliminated, perhaps as a result of active case finding focused on high-risk, predominantly male populations [18]. However, prevalence remained four times higher among men than women [18], possibly due to social determinants such as tobacco smoking and harmful alcohol consumption, both well-documented proximal determinants of TB [19–22] with particularly high rates among men [23,24].
To evaluate population benefits of addressing programmatic and social determinants of gender disparities in tuberculosis, we developed a sex-stratified compartmental transmission model incorporating gendered programmatic and social determinants of *Mycobacterium tuberculosis* infection and TB disease. We assessed the potential impact of future interventions to increase active case finding, to reduce tobacco smoking, and to reduce harmful alcohol consumption, in line with national and global targets [25,26]. For each intervention, we examined scenarios differentially targeting men and women and evaluated impact on tuberculosis morbidity and mortality in men, women, and children. While our analyses focused on Viet Nam, we expect lessons on the impact of programmatic and social determinants of gender disparities in TB burden will have wider relevance across high TB burden settings.
## Data
We developed a sex-stratified dynamic compartmental model of Mtb transmission and TB disease incorporating sex-disaggregated demographics and both programmatic and social determinants of gender disparities in TB in Viet Nam. Data supporting the findings of this study are available within the article and its supplemental materials (S1 Text).
The sex ratio at birth in Viet *Nam is* skewed towards an excess of male births [27], and life expectancy has been 8–10 years longer for women than men for the past 50 years [28]. Two-thirds of HIV infections occur among men, and fewer men living with HIV are on treatment relative to women living with HIV [29]. Social contact patterns show a high proportion of sex-assortative mixing amongst adults [30]. Gaps in TB disease detection and reporting were $72\%$ higher among men than women in 2007 [17], and modelling indicates that men with smear-positive TB remained undiagnosed for one year longer than women [3]. In 2020, $40\%$ of men and less than $2\%$ of women were current tobacco smokers [23]. In the same year, the proportion of current alcohol drinkers was $48\%$ in men and $10\%$ in women, and men who were current alcohol drinkers consumed over five times as many standard drinks daily as women (51 grams vs. 8 grams) [24]. The adult prevalence of diabetes is similar in men and women ($1\%$ in 2000 projected to rise to $2\%$ by 2030) [31], as is the proportion of the population that is underweight, as indicated by body mass index less than 18.5 ($25\%$ in 2002) [32], so neither of these risks was explicitly considered due to the focus on gender disparities in our model.
## Model structure
Our model is an extension of the TIME model [33] (S1 Text). The core model has four states related to Mtb infection and TB disease: susceptible (uninfected), latent Mtb infection, active smear-positive TB disease, and active smear-negative TB disease. States of infection and disease are stratified by treatment history and drug resistance, and the entire model is stratified by sex, age, and HIV and antiretroviral treatment (ART) status.
Sex-specific risks of Mtb infection and progression to active disease are incorporated in the model (S1 Text). Demographic data and data on HIV incidence, HIV progression, and ART uptake are sex-specific. Heterogeneous mixing reflects age- and sex-assortative mixing patterns between men (males age ≥ 15 years), women (females age ≥ 15 years), and children (both sexes age < 15 years). Increased risks of Mtb infection and progression to active disease as a result of tobacco smoking are applied to a proportion of men and women based on annual estimates of the prevalence of current smokers. Increased risk of progression to active disease is raised further for a proportion of men and women based on annual estimates of alcohol consumption. A constant relative risk term is applied to the risk of infection to reflect residual sex-specific effects not otherwise specified. The annual rate at which individuals enter the TB care cascade is also sex-specific.
## Model calibration
We calibrated the model to population size estimates and projections [34,35] and epidemiological calibration targets, including TB morbidity and mortality [18,36–38] and case notification rates [39] (S1 Text). Targets included indicators specific to age [36,39], sex [18,37–39], HIV status [36,39], and MDR status [40], where possible. We utilised an adaptive approximate Bayesian computation (ABC) Markov chain Monte Carlo (MCMC) method [41,42], using a modified version of the easyABC package that accepts seed parameter values [43] in R [44]. MCMC chains were thinned to select 1,000 parameter sets with posterior estimates consistent with all calibration targets. Results for the baseline calibration present median values, with uncertainty intervals (UIs) based on minimum and maximum values.
## Estimating the potential impact of future interventions
We assessed the potential impact of future interventions to increase active case finding for TB, to reduce tobacco smoking, and to reduce harmful alcohol consumption over the period 2021–2025. We modelled increased active case finding for TB by increasing the rate at which adults enter the TB care cascade such that, when implemented across the adult population, interventions result in 11,000 additional case notifications in 2021 increasing to 28,500 additional case notifications in 2025, per targets in the Viet Nam National Strategic Plan 2021–2025 [25] (S1 Text). We independently modelled interventions to reduce tobacco smoking and harmful alcohol consumption over the period 2021–2025 to reach a $30\%$ reduction in the prevalence of current tobacco use and a $10\%$ reduction in harmful alcohol consumption, both relative to 2015, by 2025, in line with global targets [26] (S1 Text). We assumed sigmoidal scale-up of each intervention over the period 2021–2025 with no further scale-up after 2025.
For each intervention, we examined scenarios differentially targeting men and women to evaluate the impact of meeting intervention targets only in men, meeting intervention targets only in women, and meeting intervention targets in both men and women. We calculated the projected impact of each scenario on overall TB morbidity and mortality, as well as TB burden in men, women, and children.
For each intervention scenario, we calculated the percent decline in projected TB incidence and mortality in 2035 relative to the baseline calibration. We also projected the expected number of incident TB cases and deaths averted over the period 2021–2035 for each intervention scenario relative to the baseline calibration. Results are reported as median values, with UIs based on minimum and maximum values.
## Inclusivity in global research
Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S1 Checklist).
## Baseline calibration
The calibrated model fits population size estimates and projections as well as epidemiological calibration targets (S1 Text). Model estimates reflect substantial declines in TB burden in recent decades (Fig 1). Gender disparities in TB burden changed little between 2000 and 2020 in our model (S1 Text). TB incidence, prevalence, and mortality were approximately three times higher in men than in women, while case notification rates were approximately twice as high in men as in women.
**Fig 1:** *Epidemiological estimates and projections for men (light blue) and women (red) for incidence (top left), mortality (top right), case notification rate (bottom left), and prevalence (bottom right) for the calibrated model.Figures show median model estimates (line), model uncertainty (shaded area), and sex-specific calibration targets (error bars).*
## Potential impact of future interventions
For interventions to increase active case finding for TB, projected declines in TB incidence and mortality were greatest for scenarios targeting both men and women and higher for scenarios targeting only men than those targeting only women (Fig 2). For interventions to reduce tobacco smoking and to reduce harmful alcohol consumption, projected declines in TB incidence and mortality were similar for scenarios targeting both men and women and scenarios targeting only men; projected reductions were limited for scenarios targeting only women (Fig 2).
**Fig 2:** *Projections for total population incidence (top) and mortality (bottom) for interventions to increase active case finding for TB (left), to reduce tobacco smoking (middle), and to reduce harmful alcohol consumption (right).Each plot shows median model estimates (lines) and model uncertainty (shaded area) for the baseline calibration (dark blue) and intervention scenario targeting men (light blue), targeting women (red), and targeting both men and women (purple).*
Active case finding intervention scenarios targeting only men projected greater reductions in TB incidence in men, women, and children ($16.2\%$, uncertainty interval, UI, 11.4–$23.0\%$, $11.8\%$, UI 8.0–$18.6\%$, and $21.5\%$, UI 16.9–$28.5\%$, respectively) than those targeting only women ($5.2\%$, UI 3.8–$7.1\%$, $5.4\%$, UI 3.9–$7.3\%$, and $8.6\%$, UI 6.9–$10.7\%$, respectively) (Fig 3). Projected reductions in TB mortality were greater for men and children in scenarios targeting only men ($26.0\%$, UI 21.7–$31.9\%$, and $19.7\%$, UI 15.6–$25.5\%$, respectively) compared to scenarios targeting only women ($4.9\%$, UI 3.7–$6.6\%$, and $7.9\%$, UI 6.4–$9.8\%$, respectively), and greater for women in scenarios targeting only women ($17.0\%$, UI 15.1–$19.3\%$) compared to scenarios targeting only men ($11.1\%$, UI 7.6–$17.4\%$).
**Fig 3:** *Percent decline in TB incidence (top) and mortality (bottom) in 2035, by target population (men, women, both men and women), for interventions to increase active case finding for TB (left), to reduce tobacco smoking (middle), and to reduce harmful alcohol consumption (right).Each plot shows median model estimates (bar) and model uncertainty (error bars) for the total population (dark blue), men (light blue), women (red), and children (green).*
For interventions to reduce tobacco smoking and to reduce harmful alcohol consumption, projected reductions in TB incidence for scenarios targeting only men were greatest for men ($17.4\%$, UI 11.8–$24.7\%$, and $11.0\%$, UI 5.4–$19.4\%$, respectively), but still substantial for women ($6.9\%$, UI 3.8–$12.5\%$, and $4.4\%$, UI 1.9–$10.6\%$, respectively) and children ($12.7\%$, UI 8.4–$19.0\%$, and $8.0\%$, UI 3.9–$15.0\%$, respectively) (Fig 3). Projected declines in mortality for scenarios targeting only men were similarly distributed across men, women, and children. Comparable interventions targeting only women projected limited impact on TB incidence and mortality.
Between 2021 and 2035, the baseline calibration projected a total estimated 1.68 million (UI 1.47–1.93 million) incident TB cases and 300,000 (UI 250,000–330,000) TB deaths. The greatest proportion of incident cases and deaths were projected to be averted by interventions targeting both men and women. Across interventions, scenarios targeting only men are projected to avert a higher percentage of incident cases and deaths than scenarios targeting only women (Table 1). Distributions of incident cases and deaths averted in men, women, and children are similar to distributions of reductions in incidence and mortality in 2035 across those populations.
**Table 1**
| Targeted population | Percent of projected incident cases averted | Percent of projected deaths averted |
| --- | --- | --- |
| Interventions to increase active case finding for TB | Interventions to increase active case finding for TB | Interventions to increase active case finding for TB |
| Men | 9.4% (6.6–13.1%) | 13.4% (11.0%-16.8%) |
| Women | 3.2% (2.4–4.2%) | 4.5% (3.7%-5.4%) |
| Men and women | 12.7% (8.4–19.2%) | 18.1% (12.5%-25.6%) |
| Interventions to reduce tobacco smoking | Interventions to reduce tobacco smoking | Interventions to reduce tobacco smoking |
| Men | 10.2% (6.9–14.9%) | 8.1% (5.4%-12.1%) |
| Women | 0.2% (0.1–0.2%) | 0.1% (0.1%-0.2%) |
| Men and women | 10.3% (7.0–15.0%) | 8.2% (5.5%-12.2%) |
| Interventions to reduce harmful alcohol consumption | Interventions to reduce harmful alcohol consumption | Interventions to reduce harmful alcohol consumption |
| Men | 6.6% (3.1–11.5%) | 5.2% (2.4%-9.1%) |
| Women | 0.1% (0.0–0.1%) | 0.0% (0.0%-0.1%) |
| Men and women | 6.6% (3.1–11.5%) | 5.2% (2.4%-9.1%) |
## Discussion
Our analyses using mathematical modelling shows that addressing programmatic and social determinants of gender disparities in TB burden and care in Viet Nam will have population-wide benefits. Interventions to increase active case finding for TB, to reduce tobacco smoking, and to reduce harmful alcohol consumption are projected to reduce TB morbidity and mortality in men, women, and children, especially when interventions are targeted towards men.
Over the past decade, the scale up of active case finding activities in Viet Nam has closed the gender gap in access to diagnosis and treatment by focusing on high risk populations including predominantly male populations such as prison inmates and coal miners [18]. Our analyses show clear population-wide benefits of efforts to further improve access to TB care, particularly for interventions targeting men. Future active case finding interventions, whilst not ignoring women, should focus efforts on men to most effectively reduce TB morbidity and mortality in men, women, and children. Appropriately, the National Strategic Plan for the National TB Programme in Viet Nam highlights vulnerable, predominantly male populations including tobacco smokers, alcohol consumers, those with silicosis, inmates and staff in correctional facilities, and migrants. Active case finding strategies to reach these and other hard-to-reach male populations must consider gendered barriers that disadvantage men in access TB care, notably financial and work-related concerns, stigma, and negative masculinities. Evidence from other settings indicates a range of strategies that offer convenient access to diagnosis and treatment may be needed across community [45–47], health care [48], occupational [49,50], transport [51], and leisure settings, though these approaches must be tailored to reach men in different geographic settings, age groups, and socioeconomic strata within Viet Nam.
Efforts are also needed to address social determinants of gender disparities in TB that contribute to men’s disproportionate incidence of disease. Our analyses show that the benefit of interventions to reduce tobacco smoking and to reduce harmful alcohol consumption comes almost entirely from their implementation in men. When focusing on TB morbidity and mortality, reducing tobacco smoking and harmful alcohol consumption in men has a greater impact on women’s health than reducing tobacco smoking and harmful alcohol consumption in women. The prevalence of tobacco smoking and alcohol consumption are, respectively, 20 and 10 times higher among men than women, and while the prevalence of tobacco smoking has declined slightly in recent decades [52], alcohol consumption has increased substantially [24]. Whilst efforts to change social determinants of TB are challenging for a national TB programme to instigate, such efforts are in line with the WHO End TB Strategy, as well as the multidimensional approach outlined in the Sustainable Development Goals [53]. Although the potential impact on TB epidemiology is less pronounced for interventions to reduce tobacco smoking and to reduce harmful alcohol consumption, compared to interventions to increase active case finding, the wider benefits of these interventions should make them appealing to TB programmes as well as non-communicable disease programmes.
We have not specified intervention strategies given the limited evidence on effective strategies to reach men with interventions examined here. For the same reason, we have considered neither the economic costs nor operational feasibility of interventions. Our aim was to examine the potential population-wide benefits of these interventions, focusing on the relative impact of strategies differentially targeting the sexes, in order to highlight the potential and guide the development of future interventions. Further work will be needed to identify acceptable and feasible, effective and cost-effective strategies.
Our work has several limitations. Few sex-stratified data points were available for model calibration, emphasising the need for further disaggregation of routine surveillance data and TB burden estimates by sex. Similarly, we were not able to examine more nuanced age-sex interactions in natural history and gendered risks, including tobacco smoking and harmful alcohol consumption, due to the lack of empirical data. The sex-specific risks we have included in the model are not comprehensive, and we have not explored their interactions. However, to our knowledge, this study presents the first comprehensive sex-stratified dynamic transmission model of TB to explore potential solutions for the substantial and consistent sex disparities in TB burden across low- and middle-income countries [2,54]. By incorporating programmatic and social determinants of gender disparities in TB, we have been able to identify context-specific drivers of those disparities in TB burden and generate TB burden estimates and trends not currently available with sex disaggregation.
Gender disparities in the epidemiological burden of TB are pronounced in Viet Nam and globally. As countries strive to reduce TB morbidity and mortality in line with the End TB Strategy and Sustainable Development Goals [16,53], a gender-responsive approach that considers programmatic and social determinants of TB is essential to accelerate progress toward these targets. Our work adds evidence that future interventions to increase active case finding, to reduce tobacco smoking, and to reduce harmful alcohol consumption, whilst not ignoring women, should focus on men to most effectively reduce TB morbidity and mortality in men, women, and children.
## References
1. 1World Health Organization. Global tuberculosis report 2020 Geneva, Switzerland: World Health Organization; 2020. https://apps.who.int/iris/bitstream/handle/10665/329368/9789241565714-eng.pdf?ua=1.
2. Horton KC, Macpherson P, Houben R, White R, Corbett E. **Sex differences in tuberculosis burden and notifications in low and middle-income countries: a systematic review and meta-analysis**. *PLoS Med* (2016.0) **13** e1002119. DOI: 10.1371/journal.pmed.1002119
3. Horton KC, Sumner T, Houben RM, Corbett EL, White RG. **A Bayesian approach to understanding sex differences in tuberculosis disease burden**. *Am J Epidemiol* (2018.0) **187** 2431-8. DOI: 10.1093/aje/kwy131
4. Chidambaram V, Tun NL, Majella MG, Ruelas Castillo J, Ayeh SK, Kumar A. **Male sex is associated with worse microbiological and clinical outcomes following tuberculosis treatment: A retrospective cohort study, a systematic review of the literature, and meta-analysis**. *Clin Infect Dis* (2021.0) **73** 1580-8. DOI: 10.1093/cid/ciab527
5. Dodd PJ, Looker C, Plumb I, Bond V, Schaap A, Shanaube K. **Age- and sex-specific social contact patterns and incidence of Mycobacterium tuberculosis infection**. *Am J Epidemiol* (2016.0) **183** 156-66. DOI: 10.1093/aje/kwv160
6. Horton KC, Hoey AL, Béraud G, Corbett EL, White RG. **Systematic Review and Meta-Analysis of Sex Differences in Social Contact Patterns and Implications for Tuberculosis Transmission and Control**. *Emerg Infect Dis* (2020.0) **26** 910. DOI: 10.3201/eid2605.190574
7. Neyrolles O, Quintana-Murci L. **Sexual inequality in tuberculosis**. *PLoS Med* (2009.0) **6** e1000199. DOI: 10.1371/journal.pmed.1000199
8. Nhamoyebonde S, Leslie A. **Biological differences between the sexes and susceptibility to tuberculosis**. *J Infect Dis* (2014.0) **209** S100-S6. DOI: 10.1093/infdis/jiu147
9. Lönnroth K, Thuong L, Linh P, Diwan V. **Delay and discontinuity—a survey of TB patients search of a diagnosis in a diversified health care system**. *Int J Tuberc Lung Dis* (1999.0) **3** 992-1000. PMID: 10587321
10. Corbett EL, Churchyard GJ, Clayton TC, Williams BG, Mulder D, Hayes RJ. **HIV infection and silicosis: the impact of two potent risk factors on the incidence of mycobacterial disease in South African miners**. *AIDS* (2000.0) **14** 2759-68. DOI: 10.1097/00002030-200012010-00016
11. Baussano I, Williams BG, Nunn P, Beggiato M, Fedeli U, Scano F. **Tuberculosis incidence in prisons: a systematic review**. *PLoS Med* (2010.0) **7** e1000381. DOI: 10.1371/journal.pmed.1000381
12. Hudelson P.. **Gender differentials in tuberculosis: The role of socio-economic and cultural factors**. *Tubercle and Lung Disease* (1996.0) **77** 391-400. DOI: 10.1016/s0962-8479(96)90110-0
13. Mavhu W, Dauya E, Bandason T, Munyati S, Cowan F, Hart G. **Chronic cough and its association with TB–HIV co-infection: factors affecting help-seeking behaviour in Harare, Zimbabwe**. *Trop Med Int Health* (2010.0) **15**. DOI: 10.1111/j.1365-3156.2010.02493.x
14. Chikovore J, Hart G, Kumwenda M, Chipungu GA, Desmond N, Corbett L. **Control, struggle, and emergent masculinities: a qualitative study of men’s care-seeking determinants for chronic cough and tuberculosis symptoms in Blantyre, Malawi**. *BMC Public Health* (2014.0) **14** 1053. DOI: 10.1186/1471-2458-14-1053
15. Courtenay WH. **Constructions of masculinity and their influence on men’s well-being: a theory of gender and health**. *Soc Sci Med* (2000.0) **50** 1385-401. DOI: 10.1016/s0277-9536(99)00390-1
16. 16World Health Organization. The End TB Strategy
Geneva, Switzerland: World Health Organization; 2015. https://apps.who.int/iris/rest/bitstreams/1271371/retrieve.. *The End TB Strategy* (2015.0)
17. Hoa NB, Sy DN, Nhung NV, Tiemersma EW, Borgdorff MW, Cobelens FG. **National survey of tuberculosis prevalence in Viet Nam**. *Bull World Health Organ* (2010.0) **88** 273-80. DOI: 10.2471/BLT.09.067801
18. Nguyen HV, Tiemersma EW, Nguyen HB, Cobelens FG, Finlay A, Glaziou P. **The second national tuberculosis prevalence survey in Vietnam**. *PLoS One* (2020.0) **15** e0232142. DOI: 10.1371/journal.pone.0232142
19. Pedrazzoli D, Boccia D, Dodd PJ, Lönnroth K, Dowdy DW, Siroka A. **Modelling the social and structural determinants of tuberculosis: opportunities and challenges**. *Int J Tuberc Lung Dis* (2017.0) **21** 957-64. DOI: 10.5588/ijtld.16.0906
20. Bates MN, Khalakdina A, Pai M, Chang L, Lessa F, Smith KR. **Risk of tuberculosis from exposure to tobacco smoke: a systematic review and meta-analysis**. *Arch Intern Med* (2007.0) **167** 335-42. DOI: 10.1001/archinte.167.4.335
21. Lin H-H, Ezzati M, Murray M. **Tobacco smoke, indoor air pollution and tuberculosis: a systematic review and meta-analysis**. *PLoS Med* (2007.0) **4** e20. DOI: 10.1371/journal.pmed.0040020
22. Lönnroth K, Williams BG, Stadlin S, Jaramillo E, Dye C. **Alcohol use as a risk factor for tuberculosis–a systematic review**. *BMC Public Health* (2008.0) **8** 289. DOI: 10.1186/1471-2458-8-289
23. 23Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2016 (GBD 2016) Burden by Risk 1990–2016 Seattle, WA: Institute for Heatlh Metrics and Evaluation; 2017. https://ghdx.healthdata.org/sites/default/files/record-attached-files/IHME_GBD_2016_RISK_FACTORS_1990_2016_1.zip.
24. 24Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2016 (GBD 2016) Alcohol Use Estimates 1990–2016 Seattle, WA, USA: Institute for Health Metrics and Evaluation; 2018. https://ghdx.healthdata.org/sites/default/files/record-attached-files/IHME_GBD_2016_ALCOHOL_USE_DATA.zip.
25. 25Viet Nam National Tuberculosis Program. National Strategic Plan, 2021–2025. Hanoi, Viet Nam: Ministry of Health, 2020.
26. 26World Health Organization. Global action plan for the prevention and control of noncommunicable diseases 2013–2020 Geneva, Switzerland: World Health Organization; 2013. https://apps.who.int/iris/rest/bitstreams/442296/retrieve.
27. Becquet V, Guilmoto CZ, Dutreuilh C. **Sex imbalance at birth in Vietnam: Rapid increase followed by stabilization**. *Population* (2018.0) **73** 519-44
28. 28United Nations Department of Economic and Social Affairs. World population prospects: The 2019 revision New York, NY, USA: Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat; 2020. https://population.un.org/wpp/Download/Standard/Population/.
29. 29UNAIDS. Country profile: Viet Nam Geneva, Switzerland: UNAIDS; 2020 [cited 2020 14 November]. https://www.unaids.org/en/regionscountries/countries/vietnam.
30. Horby P, Thai PQ, Hens N, Yen NTT, Thoang DD, Linh NM. **Social contact patterns in Vietnam and implications for the control of infectious diseases**. *PLoS One* (2011.0) **6** e16965. DOI: 10.1371/journal.pone.0016965
31. Wild S, Roglic G, Green A, Sicree R, King H. **Global prevalence of diabetes: estimates for the year 2000 and projections for 2030**. *Diabetes Care* (2004.0) **27** 1047-53. DOI: 10.2337/diacare.27.5.1047
32. Tuan NT, Tuong PD, Popkin B. **Body mass index (BMI) dynamics in Vietnam**. *Eur J Clin Nutr* (2008.0) **62** 78-86. DOI: 10.1038/sj.ejcn.1602675
33. Houben R, Lalli M, Sumner T, Hamilton M, Pedrazzoli D, Bonsu F. **TIME Impact–a new user-friendly tuberculosis (TB) model to inform TB policy decisions**. *BMC Medicine* (2016.0) **14** 56. DOI: 10.1186/s12916-016-0608-4
34. 34United Nations Department of Economic and Social Affairs Population Division. File POP/8-2: Male population by broad age group, major area, region and country, 1950–2100 (thousands) 2017 [cited 2018 19 September]. https://esa.un.org/unpd/wpp/DVD/Files/1_Indicators%20(Standard)/EXCEL_FILES/1_Population/WPP2017_POP_F08_2_TOTAL_POPULATION_BY_BROAD_AGE_GROUP_MALE.xlsx.
35. 35United Nations Department of Economic and Social Affairs Population Division. File POP/8-3: Female population by broad age group, major area, region and country, 1950–2100 (thousands) 2017 [cited 2018 19 September]. https://esa.un.org/unpd/wpp/DVD/Files/1_Indicators%20(Standard)/EXCEL_FILES/1_Population/WPP2017_POP_F08_3_TOTAL_POPULATION_BY_BROAD_AGE_GROUP_FEMALE.xlsx.
36. 36World Health Organization. WHO TB burden estimates Geneva, Switzerland: World Health Organization; 2020. http://www.who.int/tb/country/data/download/en/.
37. 37Glaziou P. Revised estimates for first national prevalence survey. 2019.
38. Marks G, Nhung N, Nguyen T, Hoa N, Khoa T, Son N. **Prevalence of latent tuberculous infection among adults in the general population of Ca Mau, Viet Nam**. *Int J Tuberc Lung Dis* (2018.0) **22** 246-51. DOI: 10.5588/ijtld.17.0550
39. 39World Health Organization. Case notifications
Geneva, Switzerland: World Health Organization; 2020. http://www.who.int/tb/country/data/download/en/.. *Case notifications* (2020.0)
40. Nhung N, Hoa N, Sy D, Hennig C, Dean A. **The fourth national anti-tuberculosis drug resistance survey in Viet Nam**. *Int J Tuberc Lung Dis* (2015.0) **19** 670-5. DOI: 10.5588/ijtld.14.0785
41. Marjoram P, Molitor J, Plagnol V, Tavaré S. **Markov chain Monte Carlo without likelihoods**. *PNAS* (2003.0) **100** 15324-8. DOI: 10.1073/pnas.0306899100
42. Harris RC, Sumner T, Knight GM, Evans T, Cardenas V, Chen C. **Age-targeted tuberculosis vaccination in China and implications for vaccine development: a modelling study**. *Lancet Glob Health* (2019.0) **7** e209-e18. DOI: 10.1016/S2214-109X(18)30452-2
43. Jabot F, Faure T, Dumoulin N. **Easy ABC: performing efficient approximate B ayesian computation sampling schemes using R**. *Methods Ecol Evol* (2013.0) **4** 684-7. DOI: 10.1111/2041-210X.12050
44. 44R Core Team. A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2015.. *A language and environment for statistical computing* (2015.0)
45. 45Alkazi A. OA-21-627-22-Investigating the feasibility of universal screenings and instituional treatment support of the homeless population through a mobile digital X-ray and CBNAAT algorithm in India. Union World Conference on Lung Health; Virtual conference; 2020.
46. Sharma M, Barnabas R, Celum C. **Community-based strategies to strengthen men’s engagement in the HIV care cascade in sub-Saharan Africa**. *PLoS Med* (2017.0) **14** e1002262. DOI: 10.1371/journal.pmed.1002262
47. Hensen B, Taoka S, Lewis JJ, Weiss HA, Hargreaves J. **Systematic review of strategies to increase men’s HIV-testing in sub-Saharan Africa**. *AIDS* (2014.0) **28** 2133-45. DOI: 10.1097/QAD.0000000000000395
48. 48Ramapepe M, Rozario A, Stender S, Matsinyane P. Assessing sensitivity of symptoms for efficiency in TB case detection in a male clinic in Lesotho. Union World Conference on Lung Health; Guadalajara, Mexico; 2017.
49. 49Chukwueme N, Gidado M, Mitchell E, Abdur-Razzaq H, Adegbola A, Ogbudebe C, et al. Strategies for reaching men through occupational screening in Lagos, Nigeria. Union World Conference on Lung Health; Liverpool, United Kingdom; 2016.
50. 50USAID Applying Science to Strengthen and Improve Systems (ASSIST) Project. A guide to increasing TB screening amongst the fishing communities: USAID ASSIST Experience from Northern Uganda Chevy Chase, MD: USAID Applying Science to Strengthen and Improve Systems (ASSIST) Project; 2017. https://www.usaidassist.org/sites/default/files/guide_to_tb_screening_among_fishing_community_march_2017.pdf.
51. 51Jenkins H. EP04-134-21-Geospatial determinants of TB active case finding among men and working age adults in Lima, Peru. Union World Conference on Lung Health; Virtual conference; 2020.
52. 52Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2015 (GBD 2015) Smoking Prevalence 1980–2015 Seattle, United States: Institute for Health Metrics and Evaluation; 2017. https://ghdx.healthdata.org/sites/default/files/record-attached-files/IHME_GBD_2015_SMOKING_PREVALENCE_1980_2015_1.zip.
53. 53United Nations General Assembly. Transforming our world: the 2030 Agenda for Sustainable Development 2015 [cited 2018]. http://www.un.org/ga/search/view_doc.asp?symbol=A/RES/70/1&Lang=E.
54. 54World Health Organization. Global tuberculosis report 2019 Geneva, Switzerland: World Health Organization; 2019 [cited 2019]. https://apps.who.int/iris/bitstream/handle/10665/329368/9789241565714-eng.pdf?ua=1.
|
---
title: 'Stolen childhood taking a toll at young adulthood: The higher risk of high
blood pressure and high blood glucose comorbidity among child brides'
authors:
- Biplab Datta
- Ashwini Tiwari
- Lynn Glenn
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021810
doi: 10.1371/journal.pgph.0000638
license: CC BY 4.0
---
# Stolen childhood taking a toll at young adulthood: The higher risk of high blood pressure and high blood glucose comorbidity among child brides
## Abstract
Despite notable progress being made in preventing child marriage, a significant proportion of women worldwide are still married before reaching adulthood. Though many aspects of child marriage have been widely studied, little is known on the later life health outcomes of child brides, let alone the critical need for healthcare during adulthood. This paper examines whether child brides at a young adult age bear a greater risk of high blood pressure (HBP) and high blood glucose (HBG) comorbidity than those who were married as adults. Using nationally representative data from India, we categorized married young adult (aged 20-34 years) women in four categories: neither HBP nor HBG, HBP only, HBG only, and both HBP and HBG. We estimated multinomial logistic regressions to obtain unadjusted and adjusted relative risk ratios in favor of these mutually exclusive outcomes for the child marriage indicator. Around $0.5\%$ of the women in our sample had high blood pressure and high blood glucose comorbidity. While the prevalence of comorbidity was $0.4\%$ among women who were married as adults, comorbidity was $40\%$ higher ($p \leq 0.000$) among women who were married as children. The relative risk of the comorbidity among child brides was 1.4 ($95\%$CI: 1.2–1.7) times that of their peers who were not married as children. The findings, thus, suggest that child brides at young adult age are at greater risk of having high blood pressure and high blood glucose comorbidity. Concerted public health efforts, therefore, are necessary to improve their long-term health and wellbeing.
## Introduction
Girl child marriage is associated with a range of adverse consequences including higher risk of reproductive morbidity and mortality, lack of access to healthcare services, lower educational attainment and limited economic opportunities, higher risk of intimate partner violence, and lack of voice and agency [1, 2]. Though many aspects of child marriage have been widely studied in existing literature, there is a dearth of evidence on the later life health impacts of child marriage. Currently an estimated 650 million females around the world were married as children and the number could further increase by 170 million by 2030 if adequate measures are not taken [3]. Child brides thus constitute a significant share of currently living adult women, whose long-term health and wellbeing are at stake.
Child marriage is a key driver of adolescent childbearing that can impact long-term health outcomes of women [4]. Several studies document the association between adolescent childbearing and later life risks of hypertension and cardiovascular diseases [5, 6]. A couple of recent studies also report a higher risk of hypertension and other self-reported chronic conditions being associated with child marriage [7, 8]. Hypertension is often associated with diabetes mellitus comorbidity [9, 10]. Specifically, diagnosis of hypertension is one of the most common co-existing conditions among individuals with elevated glycated hemoglobin (HbA1c) and type 2 diabetes [11, 12]. The coexistence of these two conditions poses a greater risk of life-threatening macrovascular complications [13]. With millions of currently living young adult women married as children and millions of girls at risk of child marriage, examining the risk of hypertension and diabetes or raised blood glucose comorbidity among child brides is of critical public health interest. Early diagnosis and intervention could play an important role in reducing future disease burden and improving population health. We explore the relationship between child marriage and comorbidity of raised blood pressure and raised blood glucose using data from India, where incidence of child marriage, despite recent progress, is markedly high and the burden of noncommunicable, chronic diseases is gradually rising.
One third of the world’s child brides (approximately 223 million) live in India [14]. Around $40\%$ of the young adult women (age 20 to 34 years) in India were married as children, and nearly $80\%$ of the child brides in this age group entered motherhood at an adolescent age [15]. The prevalence of raised blood pressure and raised blood glucose among adult (age 18+) women in India was estimated $27.0\%$ and $10.2\%$ respectively [16]. In a sub-national survey, the prevalence of comorbid diabetes and hypertension was estimated at $1.5\%$ among adult (age 18+) Indian women [17]. While a socioeconomic gradient in incidence of child marriage is observed in India [18], socioeconomic status conditions (SES) such as educational attainment have not demonstrated as significant predictors of hypertension and diabetes comorbidity [17]. A positive socioeconomic gradient, however, is observed for standalone diabetes and hypertension morbidities [19]; although there exist notable variations in the relationship between diabetes prevalence and low SES conditions across administrative regions in India [20].
As child brides experience various sociodemographic disadvantages in India, examining their risk of having hypertension and diabetes comorbidity entails a novel and relevant angle for prevention and control strategies. Though the prevalence and disparities in comorbid hypertension and diabetes among reproductive age women has been studied in some developed [21] and developing [22] countries, to our knowledge no studies to date explore the association between child marriage and hypertension and diabetes comorbidity. Understanding this association is important since identifying the at-risk population is critical for strategizing efficient public health interventions, particularly in low resource settings. This study, therefore, aims to examine whether the prevalence of high blood pressure (HBP) and high blood glucose (HBG) comorbidity in young adult women is higher among those who were married as children in India compared to their peers who were married as adults.
## Ethics statement
We used publicly available anonymized secondary data for our analysis. Participation in the survey was voluntary and informed consent was obtained prior interview. The survey protocol was approved by the International Institute for Population Sciences (IIPS) Institutional Review Board and ICF Institutional Review Board [15].
## Data
We obtained data on 262,205 married women of young adult age (20 to 34 years) from the 2015–16 wave of the India National Family Health Survey (NFHS-4). The NFHS-4 was conducted using a two-stage sampling framework and covers urban and rural areas of all 36 states (including union territories) of India. The NFHS-4 collects a variety of sociodemographic and anthropometric information form a nationally representative sample of women with a response rate of 97 percent.
## Measures
In the NFHS-4, respondent’s blood pressure was measured three times during a single visit with at least five minutes interval between each measure. Respondents were also asked if they were taking any medication to control HBP. An individual was identified as having HBP if the average systolic blood pressure (SBP) was greater than or equal to 140mmHg or if the average diastolic blood pressure (DBP) was greater than or equal to 90mmHg or if the individual was taking anti-hypertensive medication at the time of the survey [15].
Respondent’s random blood glucose level was also measured in the NFHS-4. An individual was identified as having HBG if the random blood glucose level was 141mg/dl or more [15]. Based on the blood pressure and blood glucose measures, individuals were categorized in four mutually exclusive categories: no HBP or HBG, HBP only, HBG only, and both HBP and HBG.
The NFHS-4 further reports respondent’s age at first marriage for women who were married at the time of the survey, from which we identified whether an individual was married before the legal age of 18 years. Self-reported age at first marriage could be subject to misreporting. To mitigate any bias from probable misreporting, we used an alternative measure of child marriage, captured by adolescent motherhood. Child marriage is regarded as the major contributing factor of adolescent pregnancy and childbearing [23]. Since childbirth in India mostly occurs within marriage [24], adolescent motherhood, defined as giving birth by age 19 years, is most likely to follow marriage before age 18 years. The NFHS-4 collects detailed birth history for each respondent, from which we were able to calculate age of women at first birth. Using this information, we created a binary variable indicating if a respondent gave birth during her adolescence (i.e., ≤19 years) and used that as an alternative measure of child marriage.
## Statistical analysis
We first estimated the share of women across the four mutually exclusive HBP and HBG comorbidity categories by child marriage. We conducted Wald tests to examine the equality of the share across women married as adults and married as children. We then estimated a multinomial logistic regression to obtain relative risk ratios (RRR) of the following mutually exclusive outcomes—i) HBP only, ii) HBG only, and iii) both HBP and HBG, relative to the base outcome of “neither HBP nor HBG” as follows: P(Yi=d|CM)=exp(β0d+γdCMi)1+∑$k = 24$exp(β0k+γkCMi) [1] The mutually exclusive outcomes are denoted by d in Eq 1. Our explanatory variable, CMi, is a binary variable indicating whether respondent i was married as a child (i.e., before age 18 years) or not. We then estimated a multivariable specification to obtain adjusted relative risk ratios (ARRR) as follows: P(Yi=d|CM,X)=exp(β0d+γdCMi+xi′βjd+State)1+∑$k = 24$exp(β0k+γkCMi+xi′βjk+Statek) [2] *In this* specification, X is a vector of sociodemographic correlates that include: age group—20 to 22 (reference group), 23 to 25, 26 to 28, 29 to 31, and 32 to 34 years; educational attainment—no education (reference group), primary, secondary, and higher; household size—3 or less (reference group), 4 to 5, 6 to 8, and 9+; household wealth index quintiles—1st quintile (poorest) (reference group), 2nd quintile, 3rd quintile, 4th quintile, and 5th quintile (least poor); religion—Hindu (reference group), Muslim, Christian, Sikh, Buddhist, and other; caste—not backward class (reference group), scheduled caste, scheduled tribe, other backward class; and place of residence—rural (reference group), and urban. These covariates are common in literature exploring the sociodemographic factors associated with hypertension and diabetes as well as with child marriage [17, 22, 25]. In addition to these sociodemographic correlates, to account for state level variations in policies and healthcare provisions, we further controlled for state fixed effects. Of note, the correlates were added in the model for the purpose of examining whether the relationship between child marriage and the HBP and HBG comorbidity persisted after controlling for relevant socioeconomic attributes.
Next, in addition to the sociodemographic covariates in vector X, we controlled for factors that may impact HBP and/or HBG conditions in the following specification: P(Yi=d|CM,X,R)=exp(β0d+γdCMi+xi′βjd+ri′δjd+State)1+∑$k = 24$exp(β0k+γkCMi+xi′βjk+ri′δjk+Statek) [3] R in Eq 3 is a vector of factors that include: nutritional status—normal (BMI 18.5 to 24.9 kg/m2—reference group), thin (BMI less than 18.5 kg/m2), overweight (BMI 25.0 to 29.9 kg/m2), and obese (BMI greater than 30.0 kg/m2); parity or the number of children born—none (reference group), 1, 2, and 3+; tobacco or alcohol consumption, oral contraception use, and pregnancy status. We examined whether the key results (i.e., ARRR for the child marriage indicator) persisted after controlling for these factors.
Lastly, we estimated the regression models with the alternative measure of child marriage, that is adolescent motherhood. We replaced CMi in Eqs 1, 2 and 3 with AMi, a binary variable indicating whether respondent i gave birth during adolescent age. All estimates accounted for complex survey weights and analyses were conducted using Stata 17.0 software.
## Results
Around $40\%$ of the young adult women in our sample were married before their 18th birthday. Table 1 presents the background characteristics of study participants. The prevalence of child marriage was higher in rural areas than in urban areas. Across household wealth categories, the prevalence was the highest among the poorest quintile and the lowest among the richest quintile. Child marriage was more prevalent among scheduled castes and scheduled tribes. Child brides were less likely to attain higher education than their peers. Child brides, on the other hand, were more likely to consume tobacco or alcohol and give birth of two or more children.
**Table 1**
| Unnamed: 0 | All (%) | Married as adults (%) | Married as children (%) |
| --- | --- | --- | --- |
| Age group | | | |
| 20–22 | 16.45 | 16.29 | 16.69 |
| 20–22 | [0.11], (16.24, 16.67) | [0.14], (16.02, 16.57) | [0.18], (16.33, 17.05) |
| 23–25 | 22.31 | 23.89 | 20.02 |
| 23–25 | [0.12], (22.07, 22.55) | [0.16], (23.57, 24.22) | [0.18], (19.66, 20.37) |
| 26–28 | 22.29 | 22.9 | 21.39 |
| 26–28 | [0.12], (22.05, 22.53) | [0.16], (22.59, 23.22) | [0.19], (21.03, 21.76) |
| 29–31 | 21.01 | 20.02 | 22.45 |
| 29–31 | [0.12], (20.78, 21.25) | [0.15], (19.72, 20.33) | [0.19], (22.08, 22.82) |
| 32–34 | 17.93 | 16.88 | 19.45 |
| 32–34 | [0.11], (17.71, 18.15) | [0.15], (16.59, 17.18) | [0.18], (19.10, 19.80) |
| Education | | | |
| No education | 24.8 | 17.19 | 35.84 |
| No education | [0.15], (24.49, 25.10) | [0.15], (16.90, 17.49) | [0.24], (35.38, 36.31) |
| Primary | 13.99 | 10.64 | 18.86 |
| Primary | [0.11], (13.78, 14.20) | [0.11], (10.42, 10.87) | [0.19], (18.49, 19.23) |
| Secondary | 48.6 | 52.61 | 42.78 |
| Secondary | [0.18], (48.24, 48.96) | [0.23], (52.16, 53.06) | [0.25], (42.30, 43.27) |
| Higher | 12.61 | 19.56 | 2.52 |
| Higher | [0.14], (12.33, 12.89) | [0.21], (19.14, 19.97) | [0.10], (2.32, 2.72) |
| Household size | | | |
| 3 or less | 12.69 | 14.63 | 9.87 |
| 3 or less | [0.12], (12.46, 12.92) | [0.16], (14.32, 14.94) | [0.15], (9.58, 10.16) |
| 4–5 | 41.21 | 39.34 | 43.93 |
| 4–5 | [0.16], (40.90, 41.52) | [0.20], (38.95, 39.74) | [0.24], (43.46, 44.39) |
| 6–8 | 30.95 | 30.46 | 31.67 |
| 6–8 | [0.16], (30.64, 31.26) | [0.21], (30.06, 30.87) | [0.21], (31.26, 32.08) |
| 9+ | 15.15 | 15.56 | 14.54 |
| 9+ | [0.14], (14.87, 15.42) | [0.17], (15.24, 15.89) | [0.19], (14.17, 14.91) |
| Wealth index quintiles | | | |
| 1st quintile (poorest) | 19 | 14.27 | 25.86 |
| 1st quintile (poorest) | [0.15], (18.70, 19.29) | [0.14], (13.99, 14.55) | [0.23], (25.42, 26.31) |
| 2nd quintile | 20.28 | 16.96 | 25.12 |
| 2nd quintile | [0.14], (20.01, 20.56) | [0.15], (16.66, 17.26) | [0.22], (24.70, 25.55) |
| 3rd quintile | 21.11 | 20.08 | 22.6 |
| 3rd quintile | [0.15], (20.82, 21.40) | [0.17], (19.74, 20.43) | [0.22], (22.18, 23.03) |
| 4th quintile | 21.02 | 23.6 | 17.28 |
| 4th quintile | [0.17], (20.69, 21.36) | [0.21], (23.20, 24.01) | [0.22], (16.85, 17.71) |
| 5th (least poor) | 18.58 | 25.09 | 9.14 |
| 5th (least poor) | [0.19], (18.21, 18.96) | [0.25], (24.60, 25.57) | [0.19], (8.77, 9.50) |
| Religion | | | |
| Hindu | 81.2 | 80.69 | 81.94 |
| Hindu | [0.24], (80.72, 81.67) | [0.27], (80.16, 81.22) | [0.31], (81.34, 82.54) |
| Muslim | 13.66 | 13.17 | 14.37 |
| Muslim | [0.23], (13.22, 14.11) | [0.25], (12.68, 13.66) | [0.29], (13.81, 14.93) |
| Christian | 1.99 | 2.34 | 1.47 |
| Christian | [0.06], (1.87, 2.11) | [0.08], (2.18, 2.50) | [0.07], (1.33, 1.62) |
| Sikh | 1.54 | 2.14 | 0.67 |
| Sikh | [0.04], (1.46, 1.62) | [0.05], (2.04, 2.25) | [0.04], (0.59, 0.74) |
| Buddhist | 0.88 | 1 | 0.7 |
| Buddhist | [0.06], (0.76, 1.00) | [0.08], (0.85, 1.16) | [0.06], (0.59, 0.81) |
| Other | 0.74 | 0.66 | 0.85 |
| Other | [0.07], (0.59, 0.88) | [0.07], (0.52, 0.79) | [0.11], (0.64, 1.06) |
| Caste | | | |
| Not backward class | 25.67 | 27.66 | 22.78 |
| Not backward class | [0.22], (25.23, 26.10) | [0.27], (27.14, 28.18) | [0.29], (22.21, 23.34) |
| Scheduled caste | 20.88 | 19.56 | 22.8 |
| Scheduled caste | [0.21], (20.46, 21.30) | [0.24], (19.08, 20.03) | [0.28], (22.25, 23.36) |
| Scheduled tribe | 9.57 | 8.49 | 11.14 |
| Scheduled tribe | [0.14], (9.30, 9.84) | [0.14], (8.23, 8.76) | [0.19], (10.76, 11.52) |
| Other backward class | 43.88 | 44.29 | 43.28 |
| Other backward class | [0.24], (43.41, 44.35) | [0.28], (43.74, 44.84) | [0.30], (42.70, 43.87) |
| Residence | | | |
| Rural | 68.32 | 63.38 | 75.5 |
| Rural | [0.25], (67.84, 68.81) | [0.30], (62.80, 63.96) | [0.29], (74.94, 76.06) |
| Urban | 31.68 | 36.62 | 24.5 |
| Urban | [0.25], (31.19, 32.16) | [0.30], (36.04, 37.20) | [0.29], (23.94, 25.06) |
| Obs. | 262205 | 160505 | 101700 |
The mean SBP and DBP in the sample was 112 mmHg and 75 mmHg respectively, and the mean blood glucose level was 101 mg/dl. Around $6\%$ of the women in the sample had HBP and nearly $4\%$ had HBG. $91.2\%$ of the women, who were not married as children, neither had HBP nor HBG. The prevalence of “not having HBP or HBG” was 1.6 percentage points lower ($p \leq 0.000$) among those who were married before age 18 years. The prevalence of “HBP only” was higher ($p \leq 0.000$) among child brides and there was not much difference in the prevalence of “HBG only” among child brides and those who were married as adults. While $0.4\%$ of the women who were married as adults (at or after age 18 years) had both HBP and HBG, the share of HBP and HBG comorbidity was $40\%$ higher ($p \leq 0.000$) among women who were married as children (Table 2).
**Table 2**
| Unnamed: 0 | All (%) | Married as adults (%) | Married as children (%) |
| --- | --- | --- | --- |
| Neither high blood pressure nor blood glucose | 90.55 | 91.21 | 89.58† |
| Neither high blood pressure nor blood glucose | [0.09] | [0.11] | [0.14] |
| Neither high blood pressure nor blood glucose | (90.37, 90.72) | (90.99, 91.43) | (89.30, 89.86) |
| High blood pressure only | 5.61 | 5.06 | 6.41† |
| High blood pressure only | [0.07] | [0.09] | [0.11] |
| High blood pressure only | (5.47, 5.75) | (4.89, 5.23) | (6.19, 6.63) |
| High blood glucose only | 3.38 | 3.33 | 3.46 |
| High blood glucose only | [0.06] | [0.08] | [0.08] |
| High blood glucose only | (3.27, 3.49) | (3.18, 3.48) | (3.30, 3.62) |
| Both high blood pressure & blood glucose | 0.46 | 0.4 | 0.56† |
| Both high blood pressure & blood glucose | [0.02] | [0.02] | [0.03] |
| Both high blood pressure & blood glucose | (0.43, 0.50) | (0.35, 0.45) | (0.49, 0.62) |
| Obs. | 262205 | 160505 | 101700 |
Panel A in Table 3 presents the RRRs in favor of the mutually exclusive HBP and HBG outcomes associated with child marriage. Relative to the “neither HBP nor HBG” outcome, the relative risk of the “HBP and HBG comorbidity” for women who were married as children were 1.4 times that of those who were married as adults. The relative risks of “HBP only” and “HBG only” outcomes were also greater for child brides compared to their peers.
**Table 3**
| Unnamed: 0 | Outcome 1: HBP only | Outcome 2: HBG only | Outcome 3: Both HBP & HBG |
| --- | --- | --- | --- |
| A. Unadjusted | | | |
| Child marriage | 1.291*** | 1.057 | 1.407*** |
| Child marriage | (1.228, 1.357) | (0.990, 1.130) | (1.189, 1.666) |
| B. Adjusted for sociodemographic correlates | | | |
| Child marriage | 1.250*** | 1.074** | 1.464*** |
| Child marriage | (1.185, 1.318) | (1.003, 1.149) | (1.229, 1.743) |
| C. Adjusted sociodemographic correlates correlates & risk factors | | | |
| Child marriage | 1.229*** | 1.079** | 1.464*** |
| Child marriage | (1.163, 1.298) | (1.006, 1.158) | (1.221, 1.754) |
The ARRRs for child marriage, adjusted for sociodemographic correlates, are presented in Panel B in Table 3. The adjusted relative risks associated with child marriage for all three outcomes were very similar to the unadjusted relative risks reported in Panel A. Relative to the base outcome, a child bride’s adjusted relative risk of having HBP and HBG comorbidity were 1.5 times that of those who were married as adults. The ARRRs for child marriage, across all three outcomes, adjusted of sociodemographic correlates and risk factors, persisted when we account for respondent’s nutritional status, current pregnancy status, oral contraception use, tobacco or alcohol consumption and number of children born (Panel C of Table 3).
Results for the adolescent motherhood, the alternative measure of child marriage, are presented in Table 4. The correlation between age at first birth and age at first marriage in our sample was 0.83. The tetrachoric correlation between the indicator variables of child marriage and adolescent motherhood was 0.88—suggesting a strong relationship between the two measures. Adolescent mothers’ adjusted relative risk of HBP and HBG comorbidity was 1.3 to 1.4 times that of those who did not give birth during adolescence. These results further reinforced the relationship between child marriage and HBP and HBG comorbidity among young adult women in India.
**Table 4**
| Unnamed: 0 | Outcome 1: HBP only | Outcome 2: HBG only | Outcome 3: Both HBP & HBG |
| --- | --- | --- | --- |
| A. Unadjusted | | | |
| Adolescent motherhood | 1.230*** | 1.064 | 1.295*** |
| Adolescent motherhood | (1.171, 1.293) | (0.992, 1.141) | (1.091, 1.536) |
| B. Adjusted for sociodemographic correlates | | | |
| Adolescent motherhood | 1.203*** | 1.067 | 1.337*** |
| Adolescent motherhood | (1.141, 1.267) | (0.992, 1.149) | (1.123, 1.590) |
| C. Adjusted sociodemographic correlates & risk factors | | | |
| Adolescent motherhood | 1.212*** | 1.103** | 1.401*** |
| Adolescent motherhood | (1.146, 1.283) | (1.018, 1.196) | (1.165, 1.685) |
## Discussion
Child brides represent a marginalized and vulnerable segment of the global population susceptible to various forms of socioeconomic deprivations in many developing countries. The physical and mental wellbeing of child brides at adult age are often subject to societal negligence, partly because of lack of evidence on the long-term health consequences of child marriage. In this paper, we document a disproportionately higher risk of hypertension and HBG comorbidity among child brides at young adulthood, compared to their peers married as adults. After accounting for household wealth, education, and other sociodemographic correlates as well as observed risk factors for hypertension and diabetes, our results demonstrate a robust relationship between child marriage and comorbidity of HBP and HBG.
There is a growing body of literature that highlights how exposure to physical, psychosocial, and environmental factors over the life course plays critical role in shaping current health outcomes [26]. Studies document relationships between chronic conditions, such as cardiovascular diseases and type 2 diabetes, and early life process during childhood and adolescence [27]. Child marriage forces a girl to play adult roles before reaching adulthood, which could have detrimental physical and psychological consequences on health across the life course. Our findings that child brides bearing a higher risk of HBP and HBG comorbidity compared to their peers add to the growing body of literature exploring the life course approaches to chronic diseases.
Of importance, associations between pre-existing diabetes and hypertension and poor maternal health outcomes are well established, such as a heightened risk of congenital anomalies, fetal demise, preterm births, perinatal mortality, and pre-eclampsia [28]. Studies also document child marriage’s association with higher fertility, unwanted pregnancy, and pregnancy termination [25, 29]. The association between child marriage and HBP and HBG comorbidity, therefore, is intertwined with maternal health outcomes. Despite experiencing a dramatic decline in maternal mortality over the past decade, an estimated 45,000 maternal deaths occurred in India in 2015, which was around $19\%$ of the global maternal deaths [30]. As such, the presentation of chronic disease among child brides may have significant ramifications for maternal health and well-being if not addressed via prevention or behavioral intervention efforts.
Undiagnosed diabetes and hypertension are critical public health concerns worldwide. High blood sugar or hyperglycemia awareness is considerably low in the Indian population contributing to suboptimal control rate [31]. Research also documents low awareness and control of hypertension or HBP in India [32]. Uncontrolled and untreated diabetes and hypertension can lead to an increased risk of cardiovascular diseases [33]. In this study, we identified a group of population (i.e., child brides) who are at heightened risk of HBP and HBG comorbidity. Therefore, targeted interventions for health promotion in this group could be an effective strategy for prevention and control of certain chronic conditions. In addition to the challenges associated with screening and diagnosis, providing adequate care for managing HBP and HBG is a challenge in low-and-middle income countries (LMICs). Treatment of these conditions in the LMICs is associated with catastrophic health expenditure and could disrupt household consumption of essential commodities [34, 35]. Since child marriage is linked to poverty and adverse economic consequences, the disease burden of HBP and HBG could further exacerbate the economic wellbeing of child brides. A concerted effort, therefore, is warranted to mitigate the risk of the HBP and HBG disease burden among child brides.
Our research highlights that social determinants of health warrants further consideration and research. In India, the prevalence of hypertension and diabetes is typically higher in urbanized areas and among more affluent social groups [32, 36]. However, we found that the odds of having HBP and HBG comorbidity were significantly higher among child brides even after controlling for socioeconomic attributes such as household income and educational attainment as well as rural or urban residence and state fixed effects. This finding indicates that irrespective of sociodemographic background, the inherent nature of being married at a young-age places these young brides at risk of later life health outcomes as a result.
The study is subject to some limitations. First, the HBP and HBG conditions were not clinically diagnosed. Fasting glucose measurements, for instance, would have been more accurate in determining HBG condition. However, unlike other studies where respondent’s self-reported hypertension and/or diabetes conditions are used as prevalence measures [22], we used measured blood pressure and blood glucose levels to determine comorbidity. Second, age at first marriage data were self-reported. To mitigate any bias from probable misreporting, we used an alternative measure of child marriage that exploits the strong relationship between child marriage and adolescent motherhood. Lastly, because of the cross-sectional nature of data, we could not investigate the casual pathways of the relationship between child marriage and HBP and HBG comorbidity at young adult age. Future research using longitudinal data will shed more light on this issue.
## Conclusion
Our findings suggest that child brides at young adult age are more susceptible to HBP and HBG comorbidity. Our analyses can be linked to at least two of the United Nations Sustainable Development Goal (SDG) targets—target 3.4: reducing premature mortality from non-communicable diseases and target 5.4: eliminating child marriage [37]. Greater efforts are to eliminate the practice of child marriage as it has consequences for later life health outcomes. Orchestrated public health efforts to eliminate child marriage and to provide access to necessary healthcare, therefore, will improve the health and wellbeing of women in the developing countries as well as will facilitate attaining certain UN SDG targets. Further, targeted prevention and control initiatives are needed for the child brides who bear a disproportionately higher health risk than their peers married as adults.
## References
1. 1Mathur S, Greene M, Malhotra A. Too young to wed: the lives, rights and health of young married girls. Washington DC: International Center for Research on Women; 2003. Available from: https://www.issuelab.org/resources/11421/11421.pdf.
2. Parsons J, Edmeades J, Kes A, Petroni S, Sexton M, Wodon Q. **Economic impacts of child marriage: a review of the literature**. *The Review of Faith & International Affairs* (2015.0) **13** 12-22. DOI: 10.1080/15570274.2015.1075757
3. 3United Nation Children’s Fund (UNICEF). Looking Ahead Towards 2030: Eliminating child marriage through a decade of action; 2020. Available from: https://data.unicef.org/resources/looking-ahead-towards-2030-eliminating-child-marriage-through-a-decade-of-action/.
4. Patel PH, Sen B. **Teen motherhood and long-term health consequences**. *Maternal and child health journal* (2012.0) **16** 1063-1071. DOI: 10.1007/s10995-011-0829-2
5. Datta BK, Husain MJ, Kostova D. **Hypertension in women: the role of adolescent childbearing**. *BMC public health* (2021.0) **21** 1-14. DOI: 10.1186/s12889-021-11488-z
6. Rosendaal NT, Alvarado B, Wu YY, Velez MP, da Câmara SMA, Pirkle CM. **Adolescent childbirth is associated with greater Framingham risk scores for cardiovascular disease among participants of the IMIAS (international mobility in aging study)**. *Journal of the American Heart Association* (2017.0) **6** e007058. DOI: 10.1161/JAHA.117.007058
7. Datta B, Tiwari A. **Adding to her woes: child bride’s higher risk of hypertension at young adulthood**. *Journal of Public Health* (2022.0). DOI: 10.1093/pubmed/fdac026
8. Vikram K. **Early marriage and health among women at midlife: Evidence from India**. *Journal of Marriage and Family* (2021.0) **83** 1480-1501. DOI: 10.1111/jomf.12793
9. Emdin CA, Anderson SG, Woodward M, Rahimi K. **Usual blood pressure and risk of new-onset diabetes: evidence from 4.1 million adults and a meta-analysis of prospective studies**. *Journal of the American College of Cardiology* (2015.0) **66** 1552-1562. DOI: 10.1016/j.jacc.2015.07.059
10. Tsimihodimos V, Gonzalez-Villalpando C, Meigs JB, Ferrannini E. **Hypertension and diabetes mellitus: coprediction and time trajectories**. *Hypertension* (2018.0) **71** 422-428. DOI: 10.1161/HYPERTENSIONAHA.117.10546
11. Bower JK, Appel LJ, Matsushita K, Young JH, Alonso A, Brancati FL. **Glycated hemoglobin and risk of hypertension in the atherosclerosis risk in communities study**. *Diabetes care* (2012.0) **35** 1031-1037. DOI: 10.2337/dc11-2248
12. Iglay K, Hannachi H, Joseph Howie P, Xu J, Li X, Engel SS. **Prevalence and co-prevalence of comorbidities among patients with type 2 diabetes mellitus**. *Current medical research and opinion* (2016.0) **32** 1243-1252. DOI: 10.1185/03007995.2016.1168291
13. Bretzel RG. **Comorbidity of diabetes mellitus and hypertension in the clinical setting: a review of prevalence, pathophysiology, and treatment perspectives**. *Clinical therapeutics* (2007.0) **29** S35-S43. DOI: 10.1016/j.clinthera.2007.07.010
14. 14United Nation Children’s Fund (UNICEF). Ending Child Marriage: A profile of progress in India; 2019. Available from: https://data.unicef.org/resources/ending-child-marriage-a-profile-of-progress-in-india/.
15. 15International Institute for Population Sciences (IIPS) and ICF. National Family Health Survey (NFHS-4), 2015–16: India; 2017. Available from: https://dhsprogram.com/pubs/pdf/FR339/FR339.pdf
16. Mathur P, Kulothungan V, Leburu S, Krishnan A, Chaturvedi HK, Salve HR. **National noncommunicable disease monitoring survey (NNMS) in India: Estimating risk factor prevalence in adult population**. *PloS one* (2021.0) **16** e0246712. DOI: 10.1371/journal.pone.0246712
17. Tripathy JP, Thakur J, Jeet G, Jain S. **Prevalence and determinants of comorbid diabetes and hypertension: Evidence from non communicable disease risk factor STEPS survey, India**. *Diabetes & Metabolic Syndrome: Clinical Research & Reviews* (2017.0) **11** S459-S465. DOI: 10.1016/j.dsx.2017.03.036
18. Mishra T, Banerjee T. **Child marriage: some facts from selected Indian states”**. *Economics Bulletin* (2020.0) **40** 2093-2110
19. Corsi DJ, Subramanian S. **Socioeconomic gradients and distribution of diabetes, hypertension, and obesity in India**. *JAMA network open* (2019.0) **2** e190411-e190411. DOI: 10.1001/jamanetworkopen.2019.0411
20. Anjana RM, Deepa M, Pradeepa R, Mahanta J, Narain K, Das HK. **Prevalence of diabetes and prediabetes in 15 states of India: results from the ICMR–INDIAB population-based cross-sectional study**. *The lancet Diabetes & endocrinology* (2017.0) **5** 585-596. DOI: 10.1016/S2213-8587(17)30174-2
21. Britton LE, Berry DC, Hussey JM. **Comorbid hypertension and diabetes among US women of reproductive age: Prevalence and disparities**. *Journal of Diabetes and its Complications* (2018.0) **32** 1148-1152. DOI: 10.1016/j.jdiacomp.2018.09.014
22. Yaya S, El-Khatib Z, Ahinkorah BO, Budu E, Bishwajit G. **Prevalence and Socioeconomic Factors of Diabetes and High Blood Pressure Among Women in Kenya: A Cross-Sectional Study**. *Journal of Epidemiology and Global Health* (2021.0) **11** 397-404. DOI: 10.1007/s44197-021-00004-6
23. 23United Nations Population Fund (UNFPA). Girlhood, Not Motherhood: Preventing Adolescent Pregnancy; 2015. Available from: https://www.unfpa.org/publications/girlhood-not-motherhood
24. Ibarra-Nava I, Choudhry V, Agardh A. **Desire to delay the first childbirth among young, married women in India: a cross-sectional study based on national survey data**. *BMC Public Health* (2020.0) **20** 1-10. DOI: 10.1186/s12889-020-8402-9
25. Anita R, Saggurti N, Balaiah D, Silverman G. **Prevalence of Child Marriage and its Impact on the Fertility and Fertility Control Behaviors of Young Women in India**. *Lancet* (2010.0) **373** 1883-9
26. Jones NL, Gilman SE, Cheng TL, Drury SS, Hill CV, Geronimus AT. **Life course approaches to the causes of health disparities**. *American journal of public health* (2019.0) **109** S48-S55. DOI: 10.2105/AJPH.2018.304738
27. Lynch J, Smith GD. **A life course approach to chronic disease epidemiology**. *Annu Rev Public Health* (2005.0) **26** 1-35. DOI: 10.1146/annurev.publhealth.26.021304.144505
28. Sullivan SD, Umans JG, Ratner R. **Hypertension complicating diabetic pregnancies: pathophysiology, management, and controversies**. *The Journal of Clinical Hypertension* (2011.0) **13** 275-284. DOI: 10.1111/j.1751-7176.2011.00440.x
29. Kamal SM, Ulas E. **Child marriage and its impact on fertility and fertility-related outcomes in South Asian countries**. *International Sociology* (2021.0). DOI: 10.1177/0268580920961316
30. 30World Health Organization (WHO). Trends in Maternal Mortality: 1990 to 2015 – Estimates by WHO, UNICEF, UNFPA, World Bank Group and the United Nations Population Division; 2015. Available from: https://www.unfpa.org/publications/trends-maternal-mortality-1990-2015.
31. Devi P, Rao M, Sigamani A, Faruqui A, Jose M, Gupta R. **Prevalence, risk factors and awareness of hypertension in India: a systematic review**. *Journal of human hypertension* (2013.0) **27** 281-287. DOI: 10.1038/jhh.2012.33
32. Gupta R, Gaur K, Ram CVS. **Emerging trends in hypertension epidemiology in India**. *Journal of human hypertension* (2019.0) **33** 575-587. DOI: 10.1038/s41371-018-0117-3
33. Petrie JR, Guzik TJ, Touyz RM. **Diabetes, hypertension, and cardiovascular disease: clinical insights and vascular mechanisms**. *Canadian Journal of Cardiology* (2018.0) **34** 575-584. DOI: 10.1016/j.cjca.2017.12.005
34. Datta BK, Husain MJ, Asma S. **Assessing the relationship between out-of-pocket spending on blood pressure and diabetes medication and household catastrophic health expenditure: evidence from Pakistan**. *International journal for equity in health* (2019.0) **18** 1-12. DOI: 10.1186/s12939-018-0906-x
35. Datta BK, Husain MJ, Fatehin S. **The crowding out effect of out-of-pocket medication expenses of two major non-communicable diseases in Pakistan**. *International health* (2020.0) **12** 50-59. DOI: 10.1093/inthealth/ihz075
36. Tandon N, Anjana RM, Mohan V, Kaur T, Afshin A, Ong K. **The increasing burden of diabetes and variations among the states of India: the Global Burden of Disease Study 1990–2016**. *The Lancet Global Health* (2018.0) **6** e1352-e1362. DOI: 10.1016/S2214-109X(18)30387-5
37. 3724 United Nations (UN). Transforming our World: The 2030 Agenda for Sustainable Development; 2015. Available from: https://sdgs.un.org/2030agenda.
|
---
title: 'Prevalence and determinants of oral health conditions and treatment needs
among slum and non-slum urban residents: Evidence from Nigeria'
authors:
- Mary E. Osuh
- Gbemisola A. Oke
- Richard J. Lilford
- Eme Owoaje
- Bronwyn Harris
- Olalekan John Taiwo
- Godwin Yeboah
- Taiwo Abiona
- Samuel I. Watson
- Karla Hemming
- Laura Quinn
- Yen-Fu Chen
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021815
doi: 10.1371/journal.pgph.0000297
license: CC BY 4.0
---
# Prevalence and determinants of oral health conditions and treatment needs among slum and non-slum urban residents: Evidence from Nigeria
## Abstract
Oral diseases constitute a neglected epidemic in Low and Middle-Income Countries (LMICs). An understanding of its distribution and severity in different settings can aid the planning of preventive and therapeutic services. This study assessed the oral health conditions, risk factors, and treatment needs among adult residents in the slum and compared findings with non-slum urban residents in Ibadan, Nigeria. The Multistage sampling was used to select adult (≥18-years) residents from a slum and a non-slum urban sites. Information sought from participants included dietary habits, oral hygiene practices, and the use of dental services. Oral examinations were performed in line with WHO guidelines. Associations were examined using logistic regression. Mediation analysis was undertaken using generalized structural equation modeling. The sample comprised 678 slum and 679 non-slum residents. Median age in slum vs non-slum was 45 (IQR:32–50) versus 38 (IQR:29–50) years. Male: female ratio was 1:2 in both sites. Prevalence of oral diseases (slum vs non-slum sites): dental caries ($27\%$ vs $23\%$), gingival bleeding ($75\%$ vs $53\%$) and periodontal pocket ($23\%$ vs $16\%$). The odds of having dental caries were $21\%$ higher for the slum dwellers compared to non-slum residents (OR = 1.21, $95\%$ CI:0.94 to 1.56); and $50\%$ higher for periodontal pocket (OR = 1.50, $95\%$CI: 1.13 to 1.98), after adjusting for age and sex. There was little evidence that tooth cleaning frequency mediated the relationship between place of residence and caries (OR = 0.95, $95\%$CI: 0.87 to 1.03 [indirect effect], $38\%$ mediated) or periodontal pocket (OR = 0.95, $95\%$CI: 0.86 to 1.04, $15\%$ mediated). Thirty-five percent and $27\%$ of residents in the slum and non-slum sites respectively required the “prompt and urgent” levels of treatment need. Oral diseases prevalence in both settings are high and the prevalence was generally higher in the slum with correspondingly higher levels of prompt and urgent treatment needs. Participants may benefit from targeted therapeutic and health promotion intervention services.
## Introduction
Slum settlements provide homes for about 1 billion world population [1,2]. Slums are characterized by crowded, unhealthy places with a high risk of infection and injury [3,4]. Slum residents are often marginalized and have limited access to basic services [1,5]. Empirical evidence points to higher disease burdens among slum dwellers compared to their rural and urban counterparts [6–9]. This has led many to suggest that slums should be identified and studied separately from other types of urban settings lest the plight of slum dwellers gets lost in urban averages [1,7,10]. The health of slum dwellers has been a major concern to government and researchers as slum dwellers appeared to have comparatively greater health challenges compared to non-slum dwellers within the same urban continuum [6–8]. That said, the determinants of disease between slum and non-slum settings might be less clear-cut when it comes to non-communicable diseases. Whereas there is little doubt that children fare particularly badly in slums as compared to non-slum city areas, the situation for adults is more nuanced. Evidence points to a higher prevalence of smoking [11,12], obesity, and diabetes mellitus [13] among slum dwellers than residents of other parts of the city, although the prevalence of high blood pressure appeared similar in both settings [13].
Oral health is growing in significance as an integral part of health and exerts significant effects on individuals, communities, and the wider society [14–18]. For long, oral health is recognized to be an essential component of overall health and well-being [18–20]. Currently, oral diseases affect an estimated 3.5 billion people globally and untreated dental caries is one of the most prevalent non-communicable diseases [14,21]. The economic burden of oral diseases is considerable [22–24]. From the 2010 Global Burden of Disease Study, the global economic burden of dental diseases amounted to $442 billion yearly, of which $298 billion ($4.6\%$ of global health expenditure) was attributed to direct treatment costs and $144 billion to indirect costs in terms of productivity losses due to caries, periodontitis, and tooth loss [22]. We recently carried out a systematic review of population-based surveys of oral health among adult urban residents in LMICs and found only three eligible studies dealing with oral health in slums: one study in Ahmedabad, India compared oral health across slum versus non-slum urban locations [25] and the remaining two from Bangladesh and Indore city, central India examined slum populations exclusively [9,26]. None was found for sub-Saharan African countries. This is in spite of the fact that oral health issues are regarded as a substantial population health burden, especially for the disadvantaged and poor population groups in both the developing and developed countries [27,28]. We, therefore, undertook this study to compare oral health and the determinants of oral health across slum and non-slum city areas. It is worth noting that major risk factors for caries and periodontal disease which include unhealthy sugary foods, tobacco use, excessive alcohol consumption, and poor hygiene practices, are also risk factors for other non-communicable diseases such as cardiovascular diseases, cancer, chronic respiratory disease, and diabetes [14,15,29–31].
In this paper we compared oral health in a slum vs. a non-slum area with the following aims: 1) To find out whether and to what extent differences in oral conditions may occur across slum vs. non-slum areas. 2) To find out if any difference between slum and non-slum areas could simply be attributed to age and gender differences. 3)To examine known determinants (risk factors) of oral diseases by location. 4) To examine the effect of each risk factor on oral conditions net of age, gender, and location. 5) To examine the causal component of risk factors associated with the location identified through the above analyses.
We hypothesized that the determinants of oral health might pull in opposite directions with slum dwellers taking less cariogenic diet than non-slum city inhabitants while the non-slum residents exhibited better dental hygiene and had better access to dental health care [25,32–36].
## Design
The study was a descriptive cross-sectional survey with an analytical comparison between slum and non-slum sites in Ibadan, the capital city of Oyo State, South-Western Nigeria. A purposive sampling method was deployed in the selection of the slum (Idikan) and the non-slum urban (Okeado) sites. The sites were selected based on similar ethnic and religious profiles.
## Ethics
Ethical approvals for the research protocol were obtained from the Biomedical and Scientific Research Ethics Committee (BSREC: $\frac{37}{18}$-19) of the University of Warwick and the Oyo State Research Ethics Review Committee (AD $\frac{13}{479}$/1247) of Ibadan, Nigeria (S1 and S2 Files).
## Sample size calculation
Sample size calculations were made to ensure sufficient statistical power for assessing the prevalence of oral diseases and for comparing the proportions of oral disease between slum and non-slum sites [37,38]. The estimates of prevalence for oral disease outcomes used were $12.8\%$ for dental caries and $75.4\%$ for periodontal disease [39,40]. For estimating oral disease prevalence, the level of confidence was set at $95\%$ and the desired level of precision was 5 percentage points. For comparison between slum and non-slum sites, a difference of $10\%$ was assumed with an alpha of 0.05 and $80\%$ power. Based on the largest sample size suggested by the calculations, a minimum sample size of 650 people in each slum and non-slum setting and a total of 1,300 was required.
## Mapping of study sites for easy sampling
The boundaries of each site were determined together with the National Population Commission (NPopC) staff (a commission vested with the duty of publishing information or data on population for various purposes in Nigeria) who were familiar with the communities. The site boundaries were identified on Google Map from where structures were delineated and digitized using high-resolution satellite imagery from the DigitalGlobe. A building-footprint database was created for each site using this high-resolution satellite imagery. A total of 1,923 and 3711 structures were extracted from the slum and the non-slum areas, respectively.
## Sampling strategy
We used a multistage sampling technique. First, we selected 1000 building structures each for slum and non-slum sites from the structures identified from satellite imagery described above using the random selection algorithm in Quantum GIS software. Second, we located the buildings using the Geographic Positioning System (GPS) technology aided by the local internet network. Teams of researchers created an inventory of all households within each selected building, then selected one household per building at random using the OpenDataKit (ODK) client application. Lastly, we obtained an inventory of all adults in each selected household from which we selected one eligible adult participant at random, again using ODK software.
## Inclusion and exclusion criteria
To be eligible, participants had to be at least 18 years of age and must be residents in the area for at least one year. Visitors, guests, house staff, and chronically ill patients, were excluded from the study.
## Data collection exercise
Following the random selection of an eligible household member to be included in the study, a data collection team visited the selected household to meet the selected person. If the selected adult was available, he/she was enrolled in the survey and when not available, a second and up to a third visit (if needed) was made. During the meeting, the number of visits before a successful meeting was noted, then detailed information about the project was shared, and clarifications were provided, where needed. The participant was asked about their willingness to participate in the research and if yes, they were invited to sign the consent form. After obtaining the participant’s signature, a good location with adequate light and sitting arrangement was identified in the home environment with the assistance of the participant. The participant was then seated on a regular chair and the questionnaire was administered by the trained recording clerk while the dentist prepared for the oral examination (light source—headlamp, fresh tray of sterile and sealed set of examination instruments, wooden spatula, sterile gloves, and face mask). After the administration of the questionnaire, oral examination was performed by the dentist, and findings were entered into the prepared tablets by the recording clerk. Following the oral examination, if a participant was found to have dental problems that required urgent attention, they were referred to the University College Hospital with a unique identifier that entitled them to subsidized cost of care. Others who did not require urgent dental treatment benefitted from oral hygiene practice reinforcement and education.
On the few occasions (8 times) that the participants declined to partake in the research exercise, the reasons for declining were noted and the process was discontinued. Thereafter, another eligible member within the same household was randomly selected as a replacement for the person that declined.
At the end of the data collection exercise, each participant received a tube of toothpaste and a toothbrush as a token of appreciation for their time in participation in the study.
## Instrument validation
The questionnaire for this study was an adaptation of a questionnaire from the National Institute for Health Research—Global Health Research Unit (NIHR-GHRU) [41] and the WHO adult oral health questionnaire [42]. The face validity of the adapted questionnaire was checked by two experts in the field of dental public health.
## Training of data collection teams
Ten teams, each consisting of a qualified dentist and an experienced recording clerk, were trained to collect data. The activities of each team were supervised and monitored by the first author. All team members were trained in understanding and interpretation of the questionnaire as well as the application of criteria and codes for various oral diseases and appropriate recording in line with the WHO standards [42]. The dentists were fresh graduates of the University of Ibadan and were specially trained at the University College Hospital (UCH) dental clinic, to elicit and record findings on oral examination in line with the WHO set criteria [42]. The WHO’s Community Periodontal Index (CPI) probe (a specially designed, lightweight metallic probe) for the assessment of periodontal status and disposable materials [42] were used in the training of the research team members. Examiners’ interrater variability was tested for dental caries, periodontal pocket formation and other oral health conditions, in which 10 examiners interpreted 10 different scenarios for both oral health outcomes.
## Pre-test
A pre-test of the study was conducted at the Abadina community of the University of Ibadan using 50 randomly selected buildings and households. The exercise enabled verification of the veracity of information obtained from satellite imagery by a physical on-the-spot visit of the 50 buildings (5 buildings per team). The exercise also enabled the collection of other useful information on the QField, an Android-based GIS for collecting field data. The processes of sampling study participants using the ODK software were tested. The duration of data collection activities (average time and range; including both questionnaire and oral examination), as well as the recording of information into the tablet computer, were verified. Ambiguity in the questionnaire design was also identified and promptly reviewed.
## Questionnaire
Participants responded to a structured and standardized questionnaire which was interviewer-administered. Information collected included participants’ socio-demographic information, period of residence, the experience of a dental problem within the last 12-month period, self-assessment of oral health status, dental hygiene practices (frequency of tooth cleaning, use of aids for oral hygiene, use of toothpaste containing fluoride), utilization of dental services (dental visits, reasons for dental visits), consumption of sugary foods and drinks, use of tobacco, and consumption of alcohol. The questionnaire was translated into the local (Yoruba) language and back-translated in line with WHO recommendations.
## Oral examination
Oral examinations were performed by the dentist in each team. The WHO’s Community Periodontal Index (CPI) probe and disposable materials [42] were deployed for the oral examination. Mouth examination was carried out with the aid of visual and tactile senses. The teeth were examined systematically from the first to the fourth quadrant. Diagnosis of oral conditions was made in accordance with the WHO criteria for oral health surveys [42] and the findings were recorded using the WHO standardized oral health assessment format which was prepared into the tablet computers. The following information was obtained from all participants: dentition status (carious teeth, missing due to caries and filled teeth for the reason of caries), periodontal status (gingival bleeding and pocket formation); loss of attachment, and intervention urgency [42]. The lifetime caries experience—DMFT (sum of decayed, missing, and filled teeth) [42], was derived for each participant, and the presence and stages of periodontal diseases (gingival bleeding, periodontal pocket, or periodontal attachment loss) were recorded [43]. Sterilization of all clinical dental instruments was done at the end of each day’s exercise at the Primary Oral Health Care Centre, University College Hospital (UCH), Ibadan, based in Idikan.
## Data analysis
Demographic characteristics and oral health outcomes were described by place of residence (slum and non-slum). For categorical variables, counts and percentages were reported. For numeric variables, means and standard deviations or medians and interquartile ranges were reported where appropriate. The frequency of cariogenic food consumption was determined by response to 8 items whose scores were added to construct a continuous variable reporting a sum score for each individual (range 8–48). The mean score was derived and used as a cut-off to produce binary output of less and more frequent intake [44,45]. Alcohol intake within the preceding 30 days was also determined using the mean number of drinks by those who drank as the cut-off to produce a binary output into; moderate and excessive alcohol intake [46].
Our objective was to explore the possible reasons for any differences in oral health outcomes between people living in the slum vs. those in the non-slum area. To this end, we first examined the relationship between place of residence and each of the two main oral health outcomes (dental caries, and periodontal pocket formation).
We chose dental caries and periodontal pocket formation because of their well-known competing risk for tooth loss in the adult population [47–50].
Second, we determined whether particular risk factors such as alcohol use, tobacco use, diet, and oral hygiene habits had a relationship with the place of residence and oral health outcomes. This is because these risk factors independently or collectively play significant roles in the development of oral diseases [51–56]. To establish causal pathways, we predetermined whether certain risk factors (diet and oral hygiene habits) acted as mediators. For example, we explored whether a cariogenic diet mediates the relationship between place of residence and dental caries. These risk factors were chosen as potential mediators for analysis on the basis of our dental knowledge as well as their mediating roles in other situations for example tooth brushing as a mediator between cognitive impairment and oral health outcome [57] and nutrition as a mediator between oral and systemic disease [58].
Associations between oral health outcomes, place of residence, and risk factors were explored using logistic regression models, unadjusted and adjusted for age and sex. Risk factors considered were alcohol intake, tobacco use, cariogenic diet, and frequency of tooth cleaning. Odds ratios, $95\%$ confidence intervals, and p-values were reported. Mediation analysis was performed using generalized structural equation modeling (gsem command in Stata). The total effect, direct effect, and indirect effect were calculated using the nonlinear combination command (nlcom). The proportion of the relationship mediated was calculated by taking the proportion of the indirect effect divided by the total effect. The effect values from the non-linear combination command were then transformed to odds ratios by taking the exponential. All effects were reported with $95\%$ confidence intervals and p-values. Percentage agreement and Gwet’s AC1 index were used to measure interrater variability of researchers for diagnosing the oral health conditions. The statistical analysis was carried out using IBM SPSS version 26 and Stata version 16.1.
## Sample characteristics
A total of 1,357 participants were included, evenly distributed between the slum and non-slum residence. Overall, there were 245 ($36\%$) males in slum residences and 234 ($35\%$) males in non-slum residences (Table 1). The average age was higher in the slum residences; the median age was 45 years (interquartile range [IQR]: 32 to 50 years) compared to a median of 38 (IQR: 29–50) in non-slum residences. Oral health perceptions and hygiene practices, and utilization of dental services are summarized in S1 Table. A lower percentage of those in slum residences had seen a dentist ($17\%$ versus $24\%$), used a toothbrush for tooth cleaning ($82\%$ versus $98\%$), and brushed at least twice daily ($24\%$ versus $27\%$) compared to non-slum residents.
**Table 1**
| Characteristic | Slum(n = 678) | Non-slum(n = 679) |
| --- | --- | --- |
| Male | 245 (36.1) | 234 (34.5) |
| Age–median (IQR) | 45 (32 to 60) | 38 (29 to 50) |
| Age | | |
| <35 years | 192 (28.3) | 260 (38.3) |
| Adult age group (35–44 years) | 127 (18.7) | 186 (27.4) |
| 45–54 years | 114 (16.8) | 109 (16.1) |
| 55–64 years | 102 (15.0) | 61 (9.0) |
| Elderly age group (65–74 years) | 88 (13.0) | 40 (5.9) |
| 75 years and above | 55 (8.1) | 23 (3.4) |
| Current marital status | | |
| Married | 437 (64.5) | 458 (67.5) |
| Living together but not married | 9 (1.3) | 13 (1.9) |
| Divorced or separated | 34 (5.0) | 7 (1.0) |
| Widowed | 110 (16.2) | 50 (7.4) |
| Never married and never lived together | 88 (13.0) | 151 (22.2) |
| Highest educational level | | |
| No education | 175 (25.8) | 42 (6.2) |
| Primary school | 179 (26.4) | 61 (9.0) |
| Secondary school | 285 (42.0) | 301 (44.3) |
| Post-secondary school | 32 (4.7) | 183 (27.0) |
| University education | 7 (1.0) | 92 (13.5) |
| Currently working | 533 (78.6) | 552 (81.3) |
| Length of time of residence in the neighbourhood | | |
| 1–10 years | 102 (15.0) | 203 (29.9) |
| 11–20 years | 103 (15.2) | 80 (11.8) |
| >20years | 473 (69.8) | 396 (58.3) |
| Wealth quintile/status (SES) | | |
| 1st quintile (lowest) | 231 (34.1) | 40 (5.9) |
| 2nd quintile (lower) | 167 (24.6) | 105 (15.5) |
| 3rd quintile (middle) | 157 (23.2) | 114 (16.8) |
| 4th quintile (higher) | 97 (14.3) | 175 (25.8) |
| 5th quintile (highest) | 26 (3.8) | 245 (36.1) |
| Main drinking water source | | |
| Piped/ tap water | 31 (4.6) | 47 (6.9) |
| Borehole | 74 (10.9) | 283 (41.7) |
| Well/ spring/ tanker or cart supply | 5 (0.7) | 8 (1.2) |
| Rainwater | 64 (9.4) | 10 (1.5) |
| Sachet/ bottled water | 504 (74.3) | 331 (48.7) |
## Oral health conditions
Prevalence of the common oral health conditions among the overall study sample was: dental caries—$25\%$; gingival bleeding—$64\%$; pocket formation—$19\%$; and dental trauma—$26\%$. Other findings include: missing teeth—$14\%$; filled teeth– $2\%$; attachment loss—$10\%$ (S2 Table). Regarding the levels of treatment needed or intervention urgency: very few participants ($4\%$) had no need for dental treatment. Sixty-four percent of the overall study sample required the “preventive or routine” level of dental treatment, while $31\%$ had the “prompt” and “immediate/ urgent” levels of treatment need (S2 Table). Examiners’ assessment for oral health conditions revealed substantial, moderate, and almost perfect levels of interrater agreements for the three common oral health conditions respectively: Dental caries—$77\%$ agreement (Gwet’s index: 0.71), periodontal pocket formation—$66\%$ agreement (Gwet’s index: 0.49), dental trauma—$96\%$ agreement (Gwet’s index: 0.96). Other interrater agreement findings were: filled teeth—$94\%$ agreement (Gwet’s index: 0.94); gingival attachment loss- $67\%$ agreement (Gwet’s index: 0.50; enamel fluorosis- $80\%$ agreement (Gwet’s index: 0.74); dental erosion- $86\%$ agreement (Gwet’s index: 0.84).
## Associations between the place of residence and oral health conditions
The number and prevalence of oral health conditions among the participants, according to their residential settings, are shown in S2 Table: Dental caries ($27\%$ versus $23\%$) and periodontal pocket formation ($23\%$ vs $16\%$) were more prevalent in slum residences than non-slum residences (Table 2 and Fig 1). Even after adjusting for age and sex, the odds of having dental caries were estimated to be $21\%$ higher for people who lived in slums compared to non-slum residences (OR = 1.21, $95\%$ CI: 0.94 to 1.56); and $50\%$ higher (OR = 1.50, $95\%$ CI: 1.13 to 1.98) for periodontal pocket. The number of participants with no need for dental treatment was $3\%$ and $5\%$ in slum and non-slum sites respectively. The type of treatment required in both sites was mostly preventive/ routine dental treatment and this comprised $62\%$ versus $67\%$ of residents in the slum and the non-slum setting respectively. Prompt and urgent levels of treatment were required for $35\%$ (slum) versus $28\%$ (non-slum) of participants (S2 Table).
**Fig 1:** *Distribution of key survey data (demographic characteristics, oral health conditions and risk factors) by slum versus non-slum site.* TABLE_PLACEHOLDER:Table 2
## Associations between the place of residence and risk factors for oral health conditions
Alcohol intake for participants was low in both slum and non-slum residence: moderate ($10\%$ versus $13\%$) and excessive ($2\%$ versus $4\%$). Tobacco use was slightly higher in slum compared to non-slum residences ($13\%$ versus $9\%$), whereas higher consumption of cariogenic diet was slightly lower in slum compared to non-slum residences ($44\%$ versus $50\%$). After adjusting for age and sex, the odds of having a higher consumption of cariogenic diet was $10\%$ lower (OR = 0.90, $95\%$ CI:0.72 to 1.13) in slum residences compared to non-slum residences. A slightly lower percentage of participants cleaned their teeth at least twice daily in slum residences compared to non-slum residences ($24\%$ versus $27\%$), After adjusting for age and sex, the odds of tooth cleaning at least twice a day were $19\%$ lower (OR = 0.81, $95\%$ CI: 0.63 to 1.04) in slums compared to non-slum residences (Table 2 and Fig 1).
## Association between risk factors and oral health conditions
Associations between oral health conditions and potential risk factors, including cariogenic diet, alcohol use, tobacco use, and frequency of tooth cleaning were reported unadjusted and adjusted for age group and sex (S3 and S4 Tables). The odds of having dental caries were $28\%$ higher (OR = 1.28, $95\%$ CI:0.97 to 1.69) for people who cleaned their teeth at least twice daily compared to less than twice daily, adjusting for age and sex. The odds of having periodontal pocket formations was $103\%$ higher (OR = 2.03, $95\%$ CI:0.97 to 4.22) for people who had an excessive intake of alcohol compared to no alcohol intake and $28\%$ higher (OR = 1.28, $95\%$ CI:0.95 to 1.73) for people who cleaned their teeth at least twice daily compared to less, adjusting for age and sex.
## Validity check for logistic regression
Assumptions of logistic regression were checked for each of the models reported above. All models had binary outcomes; little or no multicollinearity was found between independent variables; the influence of outliers did not affect the results and the observations were independent.
## Mediation analysis
Mediation via tooth cleaning frequency. The odds of having dental caries were estimated to be $22\%$ higher for people who lived in slums compared to non-slum residents (OR = 1.22, $95\%$ CI:0.95 to 1.58); and $51\%$ higher (OR = 1.51, $95\%$ CI:1.14 to 2.01) for periodontal pocket formation, after adjusting for tooth cleaning frequency, age and sex (direct effects, Fig 2 and Table 3). The odds of tooth cleaning at least twice a day were $19\%$ lower (OR = 0.81, $95\%$ CI:0.63 to 1.04) in slums compared to non-slum residences, adjusting for age and sex (indirect effect). The odds of having dental caries were $30\%$ higher (OR = 1.30, $95\%$ CI:0.98 to 1.71) and periodontal pocket formation was $31\%$ higher (OR = 1.31, $95\%$ CI:0.96 to 1.77) for cleaning teeth at least twice daily compared to less, adjusting for place of residence, age, and sex (indirect effect).
**Fig 2:** *Casual pathways between place of residence and oral health conditions (dental caries and periodontal pocket formation) with frequency of tooth cleaning as a potential mediator, controlling for age group and sex.* TABLE_PLACEHOLDER:Table 3 *Mediation via* cariogenic diet. The odds of having dental caries were estimated to be $21\%$ higher for people who lived in slums compared to non-slum residents (OR = 1.21, $95\%$ CI:0.94 to 1.56); and $49\%$ higher (OR = 1.49, $95\%$ CI:1.12 to 1.97) for periodontal pocket formation, after adjusting for cariogenic diet, age and sex (direct effects). The odds of having a more frequent cariogenic diet were $10\%$ lower (OR = 0.90, $95\%$ CI:0.72 to 1.13) for people in slum compared to non-slum residents. The odds of having dental caries was $2\%$ higher (OR = 1.02, $95\%$ CI:0.79 to 1.32) and periodontal pocket formation was $41\%$ lower (OR = 0.59, $95\%$ CI:0.44 to 0.78) for having a more frequent cariogenic diet compared to less, adjusting for place of residence, age and sex (indirect effect).
There was no evidence cariogenic diet mediated the relationship between place of residence and dental caries (OR = 1.00, $95\%$ CI:0.97 to 1.02 [total indirect effect], estimated $1\%$ mediated) or periodontal pocket formation (OR = 1.06, $95\%$ CI:0.93 to 1.19 [total indirect effect], estimated $12\%$ mediated; see Fig 3 and Table 4).
**Fig 3:** *Causal pathways between place of residence and oral health outcomes (dental caries and periodontal pocket formation) with cariogenic diet as a potential mediator, controlling for age group and sex.* TABLE_PLACEHOLDER:Table 4
## Discussion
Our study has established that the prevalence of common dental diseases in the two communities studied ranged from moderate to very high for dental caries and periodontal disease respectively. Complete absence of oral diseases was detected in only about $4\%$ of participants who were adjudged to require no form of dental treatment. The high prevalence of some oral diseases such as caries and periodontal disease is comparable to reports from available literature worldwide [59–61]. The fact that these diseases are chronic, silent, and slowly progressive in nature [14–18], often renders them prone to neglect [62,63]. Given the economic effects of dental disease and their effect on general health and well-being [22–24], more attention should be paid to early detection and treatment of dental diseases [23].
Magnifying the already heavy oral disease burden is that the slum residents recorded higher prevalence rates relative to their non-slum counterparts, although the difference was modest. The observed pattern of distribution would seem to mimic the general distribution pattern of non-communicable diseases and risk factors associated with ill-health such as diabetes mellitus, smoking and obesity, which were reported as preponderant in disadvantaged communities in previous studies [6–8].
According to the WHO, dental caries is the most common non-communicable diseases worldwide and is a major public health problem, globally [64]. The global overall prevalence of dental caries for all ages combined as reported in the global burden of disease study 2010 is approximately $35\%$ [65]. In comparison, caries prevalence from our study population is $25\%$. This lower prevalence may be explained in part by the age of our study participants which is predominantly younger, since the disease is known to increase with age [42]. Other likely determinants of the lower prevalence rates recorded in our study include the habit of consuming less of free sugar [66], fluoride use [67], as well as indigenous caries prevention measures [68]. Further studies to investigate the role and importance of each factor in disease aetiology for our study population should enhance efforts at controlling the same. The difference in the prevalence of caries observed between the slum and the non-slum locations was not significant, even after adjusting for age and gender. However, when the estimated prevalence of dental caries from our study of $27\%$ and $23\%$ in the slum and non-slum urban settings respectively, were compared with that reported from a similar study in India [25], which reported $61\%$ and $47\%$ in the respective settings, the difference was marked. Whereas in our study, participants were selected at random information on how India’s study participants were selected was not recorded and this methodological factor may explain the marked difference in findings between our study and the Indian study.
Access to clean drinking water, especially with respect to its fluoride content, plays a significant role in caries prevention. The main source of fluoride is water, while other sources include food, industrial exposure, drugs, and cosmetics [69]. The WHO, specified a fluoride concentration level of 0.5 mg/L (equivalent to 0.5 parts per million [ppm]) in drinking water for its beneficial effects in the prevention of caries and a tolerance limit of 1.5 mg/L [28,70]. In a tropical environment like Nigeria, a fluoride level of 0.3–0.6 ppm in drinking water was recommended depending on fluoride ingestion from other sources [71]. Although our study did not assess the fluoride concentration in participants’ drinking water, available studies revealed that the majority of drinking water sources (58–$75\%$) in all but one LGA in Nigeria contained 0.3 mg/L or less of fluoride [71]. A recent study conducted in south-west Nigeria reported a smaller proportion ($48\%$) of drinking water with low fluoride from similar sources: ground water- well or borehole [71,72]. Access to clean drinking water is generally a challenge for a large proportion of the Nigerian population but the situation is worse in the slums [73,74]. The non-recognition of slums in official discourses often limit their consideration in the planning of essential public services such as water [74]. The aforementioned may explain why the commonest source of drinking water in both the slum and non-slum areas in our study is packaged water (sachet/ bottle water) followed by borehole. Available reports have confirmed these water sources as clean drinking water sources with adequate fluoride concentration [71,75]. Furthermore, fluoride toothpaste is recommended for the prevention of caries, and is usually advised in conjunction with good oral hygiene [76]. The majority of our study participants reported use of fluoridated toothpaste in cleaning their mouth. Most of the commercially available toothpastes in Nigeria are reported to have the standard recommended concentration of 1000–1500 ppm [72,77,78]. This range is considered both beneficial against caries and safe against dental fluorosis for any age if toothpaste was the individual’s sole source of fluoride. The combination of access to clean/ safe drinking water source and the use of fluoridated toothpaste by the majority of our study participants probably make up for the low fluoride content available in the regions’ ground water. Therefore, there may be justification to shift focus to other causes of caries such as dietary [79], cleaning habits and other risk factors [80].
More than two-thirds of the entire population in our study had periodontal disease. This comprised gingival bleeding of $64\%$; periodontal pocket formation of $19\%$ and an attachment loss of $10\%$. Previous studies in different settings, inclusive of slums have utilized various criteria in the assessment of periodontal disease prevalence. These include the Community Periodontal Index of Treatment Needs (CPITN), Clinical Attachment Level (CAL), and/or probing depth (PD). The reported prevalence rates varied but were equally high [25,26,81,82]. Our findings that periodontal disease affected close to three-quarters of the slum dwellers and about half of the non-slum residents further buttresses the heavy burden of periodontal disease being borne by populations in LMICs particularly those residing in slums. However, our findings were again dissimilar to those reported in the study in India, in which the non-slum dwellers had a higher prevalence of gingivitis than slum dwellers while having an equal prevalence for gingival bleeding [25]. This may be explained by their purposively selected small [300] population and that their participants were drawn from across the ages (children and adults). The difference in the prevalence of periodontal pocket formation observed between the slum and the non-slum locations in our study were statistically significant, and remained so, even after adjusting for age and sex. Overall, literature on population oral health surveys among adult residents of LMICs are sparse and more data/studies will be needed to make an adequate comparison.
In terms of the oral healthcare needs, participants in both settings had high needs for oral healthcare, and most of the needs related to preventive/routine treatments. The high treatment need observed in this study is similar to other studies which reported high prevalence of oral diseases [9,26]. However, when the level of treatment required was compared between the two residential locations, more of the slum dwellers required urgent level of care, while more of the non-slum residents required the preventive/routine dental treatments, relative to their counterparts respectively.
In tackling the oral health needs of the populace, access to basic publicly funded oral health services should be central [83,84]. In Nigeria as well as other LMICs, oral health is largely considered less important than the general health, and receives little attention in terms of health care planning and service delivery [85]. Oral health services are delivered in all health care levels (primary, secondary and tertiary) in Nigeria. The primary level of oral care is concerned with the prevention of dental diseases and management of dental emergencies, however, the services are available at only a few primary health care centres (PHCCs) nationwide [79,83,85]. General treatment of oral disorders is provided at the secondary care facilities: private dental clinics, general hospitals, institutions run by faith-based organisations, federal medical centres and armed forces hospitals. Specialised oral health services including treatment of oral diseases and patient rehabilitation are provided at the tertiary (teaching hospitals) [83,85]. Traditional oral health care providers are also available in most regions as well as few non-governmental organisations, all providing oral health services to the people [17,85,86]. The primary level of oral care, is considered as the veritable means to ensure easy access to oral health service to people in all locations, due to its widespread location in the country [83–85], but it has not succeeded in delivering on its service mandate to the people due to persistent challenges (manpower and other resources). Therefore, most oral health services are directed towards the provision of curative and rehabilitative care from the secondary and tertiary levels of care [85]. These levels of care are available mainly in the big cities and largely through out of pocket payment [17]. To date the inclusion of oral health into existing PHCs has continued to suffer setbacks in terms of facilities and capacity building [79,85]. Consequently, a number of independent initiatives targeted at charting a direction for Primary Oral Health Care were set up in Nigeria’s dental schools, but these are too few to meet the current oral health need, they lack adequate manpower and are largely funded from out of pocket [79,85]. The Nigeria’s National Health Insurance Scheme (NHIS) has a mandate, through various prepayment systems, to design and implement a social health insurance scheme that can facilitate easier access to affordable and available quality health care services and achieve a universal health coverage (UHC) [83]. But this equally has its own challenges which range from poor financing of the health system by governments, an absence of prepayment schemes, and a growing population of poorly paid people who reside majorly in slums [83,85]. As such, the NHIS billing system concerning oral health is largely considered only at the level of secondary care [85,87]. These challenges in accessing oral health care may also explain the low use of professional dental services observed among our study participants. A revision of the position of oral health services on the NHIS should advocate to include the coverage of preventive oral health care services as recommended by the WHO [85,88].
Tooth brushing, at least twice daily is the professionally recommended routine that should contribute to individual oral hygiene [89,90]. We found that only a small proportion of our study participants engaged in twice daily mouth cleaning routine across both settings. The indigenous norms, cultural practices, and habits native to the people as well as their limited exposure to correct information from school and through the media may have played a role [68,91]. The low prevalence of mouth cleaning was in stark contrast to findings from a national oral health survey conducted in Malawi which reported a much higher prevalence of $75\%$ for brushing teeth at least twice daily [92]. About $40\%$ in their study reported brushing their teeth three or more times a day. This however, did not seem to reduce the caries prevalence ($49\%$) reported among their adult populations.
In advance of our data analysis, we hypothesized that different determinants would be at play in slum vs. non-slum areas. Previous work had suggested that diets in slums may be less cariogenic than in non-slum urban areas [66,93–95], while at the same time we suspected that hygiene practice and access to dental care may be poorer in slum areas. In the mediation analysis of the association between residential places (slum vs non-slum) and oral health conditions (dental caries and periodontal pocket formation), we found little evidence of either the frequency of tooth cleaning or cariogenic diet being potential mediators. This was unexpected given the established influence of these factors on oral health. The lack of a mediation effect may partly reflect the relatively small observed differences in tooth cleaning practice, cariogenic diet consumption, and dental caries between slum and non-slum populations. Previous studies have linked the pattern of sugar consumption [96–98] and tooth brushing [99,100] with periodontal disease. The unexpected finding in our study may be due to the effect of unknown confounders which require further investigation. They may also indicate issues related to self-reported measures of cariogenic diet consumption and tooth cleaning practice, and the grouping of exposure for these variables. For example, how the respondents’ self-reported brushing frequencies approximate their actual tooth brushing behavior is not clear. In addition, the act of tooth brushing itself is complex as it combines many other variables such as the duration of brushing, the design and quality of the brush, the brushing method and the toothpaste used [99–102]. We have used brushing at least twice daily as the threshold for defining higher versus lower frequency of tooth cleaning, but a previous systematic review suggested the effect might be similar even with brushing at least once daily [99]. Maintenance of good oral hygiene is the foundation on which most oral diseases can be prevented [103]. However, achieving good oral hygiene involves more than brushing the teeth twice daily. A combination of quality in the act of tooth brushing itself [99–102], with the right cleaning frequency, is considered essential and plays a significant role in achieving the level of hygiene that can be effective against oral diseases [89,91]. Moreover, the role of professional tooth cleaning at regular intervals in inhibiting oral diseases should not be overlooked [54,104]. Data from our study suggested that the majority of both the slum and the non-slum residents had never visited a dentist similar to other studies [68,105]. Lastly, our study is cross-sectional and one may indeed require longitudinal studies to properly assess a causal effect. On the whole, our hypotheses regarding key determinants of oral health conditions in slum versus non-slum urban settings were not supported by the data that we observed.
## Study limitations
Our study has certain limitations. First, the study was carried out in only one slum and one non-slum area, and so the findings may not be generalizable to all slums. Second, the cross-sectional design precludes causal inferences, hence only associations can be drawn. However, our study has some major strengths: To the best of our knowledge, it is the first community oral health study in the LMICs to deploy the use of GIS software with GPS technology to select representative samples from randomly selected households and buildings in dense urban areas such as the slums. In the slums, it is rarely feasible to conduct a comprehensive door-to-door survey of every individual or household unit, that can ensure representativeness [41,106]. The adoption of this approach in our study is therefore an innovation among oral health studies.
Additionally, we utilized the latest in the manual series of WHO oral health survey methods [42] in the assessment of public oral health status. The methods were recommended for all oral health surveys to enhance easy comparison of results among populations globally. Lastly, our study is the second and currently, the largest of its kind in the LMICs to compare oral health survey findings related to the prevalence of oral diseases and their determinants among adults residing in the slum and non-slum settings, making a significant contribution to the currently sparse evidence base.
## Conclusion
Oral disease prevalence is high in both the slum and the non-slum urban locations, and the prevalence is generally higher in the slum relative to the non-slum setting. We did not find clear evidence of the mediation effect of cariogenic diet consumption and tooth brushing frequency on the relationship between place of residence and oral health conditions. More than a third and more than a quarter of the slum and the non-slum residents respectively required the prompt and urgent levels of dental treatments. The participants may benefit from interventions targeted at therapeutic, preventive, and oral health promotion services.
## References
1. Ezeh A, Oyebode O, Satterthwaite D, Chen Y-F, Ndugwa R, Sartori J. **The health of people who live in slums 1 The history, geography, and sociology of slums and the health problems of people who live in slums**. *The Lancet* (2016) **389** 547-58
2. Saglio-Yatzimirsky M-C. *Dharavi: from mega-slum to urban paradigm.* (2021)
3. Simon FR, Adeoke A, Adewale B. **Slum Settlements Regeneration in Lagos Mega-city: an Overview of a Waterfront Makoko Community**. *International Journal of Education and Research* (2013) **1** 1-16
4. Pat-Mbano E, Nwadiaro E. **The rise of urban slum in Nigeria: implications on the urban landscape**. *International Journal of Development and Management Review (INJODEMAR)* (2012) **7** 257-69
5. Riley LW, Ko AI, Unger A, Reis MG. **Slum health: diseases of neglected populations**. *BMC international health and human rights* (2007) **7** 2. DOI: 10.1186/1472-698X-7-2
6. Mberu BU, Haregu TN, Kyobutungi C, Ezeh AC. **Health and health-related indicators in slum, rural, and urban communities: a comparative analysis.**. *Global Health Action* (2016) **9** 33163. PMID: 27924741
7. Kyobutungi C, Ziraba AK, Ezeh A, Yé Y. **The burden of disease profile of residents of Nairobi’s slums: Results from a Demographic Surveillance System.**. *Population Health Metrics* (2008) **6** 1. DOI: 10.1186/1478-7954-6-1
8. Marlow MA, Maciel ELN, Sales CMM, Gomes T, Snyder RE, Daumas RP. **Tuberculosis DALY-Gap: Spatial and Quantitative Comparison of Disease Burden Across Urban Slum and Non-slum Census Tracts.**. *Journal of Urban Health* (2015) **92** 622-34. DOI: 10.1007/s11524-015-9957-0
9. Airen B, Dasar P, Nagarajappa S, Kumar S, Jain D, Warhekar S. **Dentition status and treatment need in urban slum dwellers in Indore city, Central India.**. *Journal of Indian Association of Public Health Dentistry* (2014) **12** 163-6
10. Lilford RJ, Oyebode O, Satterthwaite D, Melendez-Torres GJ, Chen YF, Mberu B. **Improving the health and welfare of people who live in slums**. *The Lancet* (2017) **389** 559-70. DOI: 10.1016/S0140-6736(16)31848-7
11. Khan MMH, Khan A, Kraemer A, Mori M. **Prevalence and correlates of smoking among urban adult men in Bangladesh: slum versus non-slum comparison**. *BMC Public Health* (2009) **9** 149. DOI: 10.1186/1471-2458-9-149
12. Gupta V, Yadav K, Anand K. **Patterns of tobacco use across rural, urban, and urban-slum populations in a north Indian community.**. *Indian journal of community medicine: official publication of Indian Association of Preventive & Social Medicine* (2010) **35** 245-51. PMID: 20922100
13. Snyder RE, Rajan JV, Costa F, Lima HC, Calcagno JI, Couto RD. **Differences in the prevalence of non-communicable disease between slum dwellers and the general population in a large urban area in Brazil**. *Tropical medicine and infectious disease* (2017) **2** 47. PMID: 30270904
14. 14WHO. Oral health: Key facts. The World Health Organization; 2020.. (2020)
15. Peres MA, Macpherson LMD, Weyant RJ, Daly B, Venturelli R, Mathur MR. **Oral diseases: a global public health challenge**. *The Lancet* (2019) **394** 249-60. DOI: 10.1016/S0140-6736(19)31146-8
16. Varenne B.. *WHO Oral Health Fact Sheet* (2012)
17. Akpata ES. **Oral health in Nigeria.**. *International Dental Journal* (2004) **54** 361-6. DOI: 10.1111/j.1875-595x.2004.tb00012.x
18. Petersen P.. **Improvement of oral health in Africa in the 21st century—the role of the WHO Global Oral Health Programme**. *Developing Dentistry* (2004) **5** 9-20
19. Glick M, Williams DM, Kleinman DV, Vujicic M, Watt RG, Weyant RJ. **A new definition for oral health developed by the FDI World Dental Federation opens the door to a universal definition of oral health**. *British dental journal* (2016) **221** 792. DOI: 10.1038/sj.bdj.2016.953
20. Hummel J, Philips K, Holt B, Hayes C, Health Q. *Qualis Health* (2015)
21. Disease GBD, Injury I, Prevalence C, Looker K. **Global, regional, and national indicence, prevalence, and years lived with disability for 355 diseases and injuries for 195 countries, 1990–2017. a systematic analysis for the Global Burden of Disease Study 2017**. (2018) **392** 1789-858
22. Listl S, Galloway J, Mossey PA, Marcenes W. **Global Economic Impact of Dental Diseases. Research Report: Clinical**. *Journal of Dental Research* (2015) **94** 1355-61. DOI: 10.1177/0022034515602879
23. Tonetti MS, Jepsen S, Jin L, Otomo‐Corgel J. **Impact of the global burden of periodontal diseases on health, nutrition and wellbeing of mankind: A call for global action**. *Journal of clinical periodontology* (2017) **44** 456-62. DOI: 10.1111/jcpe.12732
24. Botelho J, Machado V, Leira Y, Proença L, Chambrone L, Mendes JJ. **Economic burden of periodontitis in the United States of America and Europe–an updated estimation**. *Journal of Periodontology* (2021)
25. Patel AB, Shah RR, Ramanuj VB. **Comparative study of oral hygienic practices and oral health status among people residing in urban and urban slum of Ahmedabad municipal corporation.**. *Int J Community Med Public Health* (2017) **4** 2181-5
26. Hannan MA, Chowdhury MT, Khan MA, Chowdhury AF, Shahidullah KM, Saha AK. **Prevalence of Gingivitis, Plaque accumulation and Decayed, Missing and Filled Teeth among slum population in Bangladesh.**. *Bangladesh Medical Research Council Bulletin* (2014) **40** 47-51. DOI: 10.3329/bmrcb.v40i2.25182
27. Bernabe E, Marcenes W, Hernandez CR, Bailey J, Abreu LG, Alipour V. **Global, Regional, and National Levels and Trends in Burden of Oral Conditions from 1990 to 2017: A Systematic Analysis for the Global Burden of Disease 2017 Study**. *Journal of Dental Research* (2020) **99** 362-73. DOI: 10.1177/0022034520908533
28. Petersen PE, Bourgeois D, Ogawa H, Estupinan-Day S, Ndiaye C. **The global burden of oral diseases and risks to oral health**. *Bulletin of the World Health Organization* (2005) **83** 661-9. PMID: 16211157
29. Lee J-H, Oh J-Y, Youk T-M, Jeong S-N, Kim Y-T, Choi S-H. **Association between periodontal disease and non-communicable diseases: A 12-year longitudinal health-examinee cohort study in South Korea**. *Medicine* (2017) **96** e7398. DOI: 10.1097/MD.0000000000007398
30. Kantovitz KR, Pascon FM, Rontani RMP, Gaviao MBD, Pascon FM. **Obesity and dental caries—A systematic review**. *Oral Health & Preventive Dentistry* (2006) **4**. PMID: 16813143
31. Schulze MB, Manson JE, Ludwig DS, Colditz GA, Stampfer MJ, Willett WC. **Sugar-sweetened beverages, weight gain, and incidence of type 2 diabetes in young and middle-aged women**. *JAMA* (2004) **292** 927-34. DOI: 10.1001/jama.292.8.927
32. Deprivation Locker D.. **oral health: a review**. *Community Dent Oral Epidemiol* (2000) **28** 161-9. DOI: 10.1034/j.1600-0528.2000.280301.x
33. Gautam D, Vikas J, Amrinder T, Rambhika T, Bhanu K. **Evaluating dental awareness and periodontal health status in different socioeconomic groups in the population of Sundernagar, Himachal Pradesh, India.**. *Journal of International Society of Preventive & Community Dentistry* (2012) **2** 53. DOI: 10.4103/2231-0762.109367
34. Sanders AE, Slade GD, Turrell G, John Spencer A, Marcenes W. **The shape of the socioeconomic–oral health gradient: implications for theoretical explanations**. *Community Dentistry and Oral Epidemiology* (2006) **34** 310-9. DOI: 10.1111/j.1600-0528.2006.00286.x
35. Gautam DK, Vikas J, Amrinder T, Rambhika T, Bhanu K. **Evaluating dental awareness and periodontal health status in different socioeconomic groups in the population of Sundernagar, Himachal Pradesh, India.**. *J Int Soc Prev Community Dent* (2012) **2** 53-7. DOI: 10.4103/2231-0762.109367
36. Aikins E, Braimoh O. **Utilization of dental services among civil servants in Port Harcourt, Nigeria**. *Journal of Dental Research and Review* (2015) **2** 62-6
37. Taherdoost H.. **Determining Sample Size; How to Calculate Survey Sample Size.**. *International Journal of Economics and Management Systems* (2017) **2** 236-9
38. Bartlett JE, Kotrlik JW, Higgins CC. **Organizational Research: Determining Organizational Research: Determining Appropriate Sample Size in Survey Research.**. *Information Technology, Learning, and Performance Journal* (2001) **19** 43-50
39. Lawal F, Alade O. **Dental caries experience and treatment needs of an adult female population in Nigeria.**. *African Health Sciences* (2017) **17** 905. DOI: 10.4314/ahs.v17i3.34
40. Umoh AO, Azodo CC. **Prevalence of gingivitis and periodontitis in an adult male population in Nigeria.**. *Nigerian Journal of Basic and Clinical Sciences* (2012) **9** 65-9
41. Bakibinga P, Kabaria C, Kyobutungi C, Manyara A, Mbaya N, Mohammed S. **A protocol for a multi-site, spatially-referenced household survey in slum settings: methods for access, sampling frame construction, sampling, and field data collection**. *BMC Medical Research Methodology* (2019) **19** 109. DOI: 10.1186/s12874-019-0732-x
42. Petersen P, Baez RJ. *Oral Health Surveys Basic Methods* (2013) 1-137
43. Wiebe CB, Putnins EE. **The periodontal disease classification system of the American Academy of Periodontology-an update**. *Journal-Canadian Dental Association* (2000) **66** 594-9. PMID: 11253351
44. Nurelhuda N, Malde M, Ahmed M, TA T, Nurelhuda N. **Correlation between caries experience in Sudanese school children and dietary habits, according to a food frequency questionnaire and a modified 24-hr recall method**. *African Journal of Food, Agriculture, Nutrition and Development* (2013) **13**
45. Cade J, Thompson R, Burley V, Warm D. **Development, validation and utilisation of food-frequency questionnaires–a review**. *Public Health Nutrition.* (2002) **5** 567-87. DOI: 10.1079/PHN2001318
46. Tjønneland A, Grønbæk M, Stripp C, Overvad K. **Wine intake and diet in a random sample of 48763 Danish men and women**. *The American Journal of Clinical Nutrition* (1999) **69** 49-54. DOI: 10.1093/ajcn/69.1.49
47. Akhter R, Hassan NMM, Aida J, Zaman KU, Morita M. **Risk indicators for tooth loss due to caries and periodontal disease in recipients of free dental treatment in an adult population in Bangladesh**. *Oral Health & Preventive Dentistry* (2008) **6**
48. Susin C, Oppermann RV, Haugejorden O, Albandar JM. **Tooth loss and associated risk indicators in an adult urban population from south Brazil**. *Acta Odontologica Scandinavica* (2005) **63** 85-93. DOI: 10.1080/00016350510019694
49. Silva Junior MF, Batista MJ, de Sousa MdLR. **Risk factors for tooth loss in adults: A population-based prospective cohort study**. *Plos One* (2019) **14** e0219240. DOI: 10.1371/journal.pone.0219240
50. Gilbert GH, Shelton BJ, Chavers LS, Bradford Jr. EH. **Predicting Tooth Loss During a Population-Based Study: Role of Attachment Level in the Presence of Other Dental Conditions**. *Journal of Periodontology* (2002) **73** 1427-36. DOI: 10.1902/jop.2002.73.12.1427
51. Pitiphat W, Merchant AT, Rimm EB, Joshipura KJ. **Alcohol Consumption Increases Periodontitis Risk**. *Journal of Dental Research* (2003) **82** 509-13. DOI: 10.1177/154405910308200704
52. Peres MA, Sheiham A, Liu P, Demarco FF, Silva AER, Assunção MC. **Sugar Consumption and Changes in Dental Caries from Childhood to Adolescence**. *Journal of Dental Research* (2016) **95** 388-94. DOI: 10.1177/0022034515625907
53. Bernabé E, Vehkalahti MM, Sheiham A, Aromaa A, Suominen AL. **Sugar-sweetened beverages and dental caries in adults: A 4-year prospective study**. *Journal of Dentistry* (2014) **42** 952-8. DOI: 10.1016/j.jdent.2014.04.011
54. Bellini HT, Arneberg P, Von Der Fehr FR. **Oral hygiene and caries.**. *Acta Odontologica Scandinavica* (1981) **39** 257-65. DOI: 10.3109/00016358109162287
55. Andlaw RJ. **Oral hygiene and dental caries—a review**. *International dental journal* (1978) **28** 1-6. PMID: 346493
56. Winn DM. **Tobacco Use and Oral Disease.**. *Journal of Dental Education* (2001) **65** 306-12. PMID: 11336115
57. Lee KH, Plassman BL, Pan W, Wu B. **Mediation Effect of Oral Hygiene on the Relationship Between Cognitive Function and Oral Health in Older Adults**. *Journal of Gerontological Nursing* (2016) **42** 30-7. DOI: 10.3928/00989134-20151218-03
58. Ritchie CS, Joshipura K, Hung H-C, Douglass CW. **Nutrition as a mediator in the relation between oral and systemic disease: associations between specific measures of adult oral health and nutrition outcomes.**. *Critical reviews in oral biology & medicine* (2002) **13** 291-300. DOI: 10.1177/154411130201300306
59. Kassebaum N, Bernabé E, Dahiya M, Bhandari B, Murray C, Marcenes W. **Global burden of untreated caries: a systematic review and metaregression**. *Journal of Dental Research* (2015) **94** 650-8. DOI: 10.1177/0022034515573272
60. Watt RG, Petersen PE. **Periodontal health through public health–the case for oral health promotion**. *Periodontology 2000* (2012) **60** 147-55. DOI: 10.1111/j.1600-0757.2011.00426.x
61. Kaste L, Gift H, Bhat M, Swango P. **Prevalence of incisor trauma in persons 6 to 50 years of age: United States, 1988–1991**. *Journal of Dental Research* (1996) **75** 696-705. DOI: 10.1177/002203459607502S09
62. Varenne B, Petersen PE, Fournet F, Msellati P, Gary J, Ouattara S. **Illness-related behaviour and utilization of oral health services among adult city-dwellers in Burkina Faso: evidence from a household survey**. *BMC Health Services Research* (2006) **6** 164. DOI: 10.1186/1472-6963-6-164
63. Okunseri C, Born D, Chattopadhyay A. **Self-reported dental visits among adults in Benin City**. *Nigeria. International Dental Journal* (2004) **54** 450-6. DOI: 10.1111/j.1875-595x.2004.tb00303.x
64. 64World Health Organization. Sugars and dental caries. World Health Organization; 2017.. *Sugars and dental caries* (2017)
65. Marcenes W, Kassebaum NJ, Bernabé E, Flaxman A, Naghavi M, Lopez A. **Global burden of oral conditions in 1990–2010: a systematic analysis**. *J Dent Res* (2013) **92** 592-7. DOI: 10.1177/0022034513490168
66. Ismail AI, Tanzer JM, Dingle JL. **Current trends of sugar consumption in developing societies**. *Community Dentistry and Oral Epidemiology* (1997) **25** 438-43. DOI: 10.1111/j.1600-0528.1997.tb01735.x
67. Malago J, Makoba E, Muzuka AN. **Fluoride levels in surface and groundwater in Africa: a review.**. *Am J Water Sci Eng* (2017) **3** 1-17
68. Olusile AO, Adeniyi AA, Orebanjo O. **Self-rated oral health status, oral health service utilization, and oral hygiene practices among adult Nigerians**. *BMC Oral Health* (2014) **14** 140. DOI: 10.1186/1472-6831-14-140
69. Hardwick K, Barmes D, Writer S, Richardson LM. **International Collaborative Research on Fluoride**. *Journal of Dental Research* (2000) **79** 893-904. DOI: 10.1177/00220345000790040301
70. **Guidelines for drinking-water quality: second addendum**. *Recommendations* (2008) **1**
71. Akpata ES, Danfillo I, Otoh E, Mafeni J. **Geographical mapping of fluoride levels in drinking water sources in Nigeria**. *African health sciences* (2009) **9**. PMID: 21503173
72. Ibiyemi O, Ibiyemi S. **Fluoride concentrations and labeling information on adult toothpastes from Nigeria and the United Kingdom**. *African Journal of Medicine and Medical Sciences* (2021) **50** 107-15
73. Aliu IR, Akoteyon IS, Soladoye O. **Living on the margins: Socio-spatial characterization of residential and water deprivations in Lagos informal settlements, Nigeria**. *Habitat International* (2021) **107** 102293
74. Akpabio EM, Wilson N-AU, Essien KA, Ansa IE, Odum PN. **Slums, women and sanitary living in South-South Nigeria.**. *Journal of Housing and the Built Environment* (2021) **36** 1229-48. DOI: 10.1007/s10901-020-09802-z
75. Ani F, Akaji E, Uguru N, Ndiokwelu E. **Fluoride content of commercial drinking water and carbonated soft drinks available in Southeastern Nigeria: dental and public health implications**. *Niger J Clin Pract* (2020) **23** 65-70. DOI: 10.4103/njcp.njcp_248_19
76. Toumba KJ, Twetman S, Splieth C, Parnell C, van Loveren C, Lygidakis NΑ. **Guidelines on the use of fluoride for caries prevention in children: an updated EAPD policy document.**. *European Archives of Paediatric Dentistry.* (2019) **20** 507-16. DOI: 10.1007/s40368-019-00464-2
77. Cury JA, Tenuta LMA. **Evidence-based recommendation on toothpaste use.**. *Brazilian oral research* (2014) **28** 1-7. DOI: 10.1590/S1806-83242014.50000001
78. Walsh T, Worthington HV, Glenny AM, Appelbe P, Marinho VC, Shi X. **Fluoride toothpastes of different concentrations for preventing dental caries in children and adolescents**. *Cochrane database of systematic reviews* (2010). DOI: 10.1002/14651858.CD007868.pub2
79. Oke G.. *In Primary Mental, Oral, Eye, and Ear Care in Nigeria. Nigerian Health Review 2007: Primary Health Care in Nigeria: 30 Years afer Alma Ata. National Health Review* (2007) 226-34
80. Okeigbemen SA, Ibiyemi O. **Prevention of Dental Caries in Nigeria: A Narrative Review of Strategies and Recommendations from 1999 to 2019**. *J Int Soc Prev Community Dent.* (2020) **10** 240-5. DOI: 10.4103/jispcd.JISPCD_423_19
81. Jaafar N, Hakim H, Mohd Nor NA, Mohamed A, Saub R, Esa R. **Is the burden of oral diseases higher in urban disadvantaged community compared to the national prevalence?**. *BMC Public Health* (2014) **14** S2
82. Nazir M, Al-Ansari A, Al-Khalifa K, Alhareky M, Gaffar B, Almas K. **Global Prevalence of Periodontal Disease and Lack of Its Surveillance.**. *The Scientific World Journal* (2020) **2020** 2146160. DOI: 10.1155/2020/2146160
83. Gaines J, Malumfashi D, Omenka S, Oti N. *TAPI–Desk Study* (2020)
84. Northridge ME, Kumar A, Kaur R. **Disparities in Access to Oral Health Care.**. *Annual Review of Public Health* (2020) **41** 513-35. DOI: 10.1146/annurev-publhealth-040119-094318
85. Adeniyi A, Sofola O, Kalliecharan R. **An appraisal of the oral health care system in Nigeria.**. *International Dental Journal* (2012) **62** 292-300. DOI: 10.1111/j.1875-595X.2012.00122.x
86. Oke GA, Bankole OO, Denloye OO, Danfillo IS, Enwonwu CO. **Traditional and emerging oral health practices in parts of Nigeria**. *Odonto-stomatologie tropicale = Tropical dental journal* (2011) **34** 35-46. PMID: 22457991
87. Adeniyi A, Onajole A. **The National Health Insurance Scheme (NHIS): a survey of knowledge and opinions of Nigerian dentists’ in Lagos.**. *African Journal of Medicine and Medical Sciences* (2010) **39** 29-35. PMID: 20632669
88. Helderman WvP, Benzian H. **Implementation of a Basic Package of Oral Care: towards a reorientation of dental NGOs and their volunteers**. *International dental journal* (2006) **56** 44-8. DOI: 10.1111/j.1875-595x.2006.tb00073.x
89. Ganss C, Schlueter N, Preiss S, Klimek J. **Tooth brushing habits in uninstructed adults—frequency, technique, duration and force**. *Clinical Oral Investigations* (2009) **13** 203-8. DOI: 10.1007/s00784-008-0230-8
90. Attin T, Hornecker E. **Tooth brushing and oral health: how frequently and when should tooth brushing be performed?**. *Oral Health & Preventive Dentistry* (2005) **3**. PMID: 16355646
91. Bashiru BO, Anthony IN. **Oral self-care practices among university students in Port Harcourt, Rivers State**. *Nigerian medical journal: journal of the Nigeria Medical Association* (2014) **55** 486-9. DOI: 10.4103/0300-1652.144703
92. Msyamboza KP, Phale E, Namalika JM, Mwase Y, Samonte GC, Kajirime D. **Magnitude of dental caries, missing and filled teeth in Malawi: National Oral Health Survey.**. *BMC Oral Health* (2016) **16** 1-6. DOI: 10.1186/s12903-016-0190-3
93. Sheiham A.. **Changing Trends in Dental Caries**. *International Journal of Epidemiology* (1984) **13** 142-7. DOI: 10.1093/ije/13.2.142
94. Mazengo MC, Tenovuo J, Hausen H. **Dental caries in relation to diet, saliva and cariogenic microorganisms in Tanzanians ot selected age groups**. *Community Dentistry and Oral Epidemiology* (1996) **24** 169-74. DOI: 10.1111/j.1600-0528.1996.tb00836.x
95. Enwonwu CO. **Socio-economic factors in dental caries prevalence and frequency in Nigerians.**. *Caries Research* (1974) **8** 155-71. DOI: 10.1159/000260104
96. Moreira ARO, Batista RFL, Ladeira LLC, Thomaz EBAF, Alves CMC, Saraiva MC. **Higher sugar intake is associated with periodontal disease in adolescents.**. *Clinical Oral Investigations* (2021) **25** 983-91. DOI: 10.1007/s00784-020-03387-1
97. Lula EC, Ribeiro CC, Hugo FN, Alves CM, Silva AA. **Added sugars and periodontal disease in young adults: an analysis of NHANES III data**. *The American Journal of Clinical Nutrition* (2014) **100** 1182-7. DOI: 10.3945/ajcn.114.089656
98. Sidi AD, Ashley FP. **Influence of Frequent Sugar Intakes on Experimental Gingivitis**. *Journal of Periodontology* (1984) **55** 419-23. DOI: 10.1902/jop.1984.55.7.419
99. Kumar S, Tadakamadla J, Johnson N. **Effect of toothbrushing frequency on incidence and increment of dental caries: a systematic review and meta-analysis**. *Journal of Dental Research* (2016) **95** 1230-6. DOI: 10.1177/0022034516655315
100. Frandsen A.. **Mechanical oral hygine practices.**. *Dental Plaque Control Measures and Oral Hygine Practices* (1986) 93-116
101. Ashley P.. **Toothbrushing: why, when and how?**. *Dental Update* (2001) **28** 36-40. DOI: 10.12968/denu.2001.28.1.36
102. Asadoorian J.. **Tooth brushing**. *Canadian Journal of Dental Hygiene* (2006) **40** 1-14
103. Bakdash B.. **Current patterns of oral hygiene product use and practices**. *Periodontology 2000* (1995) **8** 11-4. DOI: 10.1111/j.1600-0757.1995.tb00041.x
104. Isidor F, Karring T. **Long‐term effect of surgical and non‐surgical periodontal treatment. A 5‐year clinical study**. *Journal of Periodontal Research* (1986) **21** 462-72. DOI: 10.1111/j.1600-0765.1986.tb01482.x
105. Adeniyi AA, Oyapero A. **Predisposing, enabling and need factors influencing dental service utilization among a sample of adult Nigerians**. *Population Medicine* (2020) **2** 1-9
106. Levy PS, Lemeshow S. *Sampling of populations: methods and applications* (2013)
|
---
title: 'Prevalence of microvascular and macrovascular complications of diabetes in
newly diagnosed type 2 diabetes in low-and-middle-income countries: A systematic
review and meta-analysis'
authors:
- Faith Aikaeli
- Tsi Njim
- Stefanie Gissing
- Faith Moyo
- Uazman Alam
- Sayoki G. Mfinanga
- Joseph Okebe
- Kaushik Ramaiya
- Emily L. Webb
- Shabbar Jaffar
- Anupam Garrib
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021817
doi: 10.1371/journal.pgph.0000599
license: CC BY 4.0
---
# Prevalence of microvascular and macrovascular complications of diabetes in newly diagnosed type 2 diabetes in low-and-middle-income countries: A systematic review and meta-analysis
## Abstract
There is an excessive burden of diabetes complications in low-resource settings. We conducted a systematic review to determine the nature and frequency of diabetes complications in newly diagnosed with type 2 diabetes. A systematic search was performed using Medline, CINAHL and Global Health online databases from inception to July 2020. Articles reporting prevalence of microvascular or macrovascular complications within six months of type 2 diabetes diagnosis and published in English or French from low- and middle-income countries (LMICs) were eligible for analysis. Data were extracted using a standardized data extraction tool. Descriptive statistics were used to describe the prevalence of micro and macrovascular complications in newly diagnosed type 2 diabetes. Assessment of heterogeneity was conducted using the inconsistency index (I2) and Cochran-Q chi2 statistical tests. Publication bias was assessed by the Funnel plot and Egger test. A total of 3 292 records underwent title or abstract screening and 95 articles underwent full text review. Thirty-three studies describing 13 283 participants (aged 20 years and older) met the inclusion criteria. The eligible studies were from Asia ($$n = 24$$), Africa ($$n = 4$$), Oceania ($$n = 2$$), South America ($$n = 2$$) and the Caribbean ($$n = 1$$). For microvascular complications, the median prevalence (interquartile range) of retinopathy, nephropathy and neuropathy were $12\%$ ($6\%$-$15\%$), $15\%$ ($7\%$-$35\%$) and $16\%$ ($10\%$$25\%$) respectively. For macrovascular complications, the median prevalence (interquartile range) was $10\%$ ($7\%$-$17\%$) for ischaemic heart disease, $6\%$ ($1\%$-$20\%$) for peripheral arterial disease and $2\%$ ($1\%$-$4\%$) for stroke. There was evidence of substantial heterogeneity between studies for all outcomes (I2 > $90\%$. We found a high prevalence of complications in newly diagnosed type 2 diabetes in LMICs. Findings suggest that many people live with diabetes and are only diagnosed when they present with complications in LMICs. Research is needed to guide timely and effective identification of people living with diabetes in these settings.
## Introduction
Globally an estimated 463 million adults aged 20–79 years are currently living with diabetes. This number is expected to increase by $51\%$ to 700 million by 2045 [1]. Type 2 diabetes mellitus is the most common form of diabetes, representing about $90\%$ of all diabetes cases worldwide [2]. It is characterised by a long asymptomatic period of five to seven years from onset to diagnosis, such that many patients present with complications at the time of diagnosis [3]. Within the first ten years from diagnosis, an estimated $27\%$ of people with type 2 diabetes die [4].
Microvascular and macrovascular complications are the major cause of morbidity and mortality in people with diabetes [2]. Macrovascular complications include myocardial infarction, stroke, peripheral vascular disease and diabetic foot. There is an increase in five-year mortality in patients diagnosed with macrovascular complications [5]. Microvascular complications of type 2 diabetes include retinopathy, nephropathy and neuropathy [6] of which there is an excess burden in persons newly diagnosed with type 2 diabetes [7]. Importantly, newly diagnosed diabetes is associated with substantial premature death not only from vascular disease but also other non-vascular causes of mortality [8].
Studies characterising recent and long-term trends in diabetes-related complications globally are based on data predominantly coming from high-income countries [9, 10]. Low- and middle- income countries (LMICs) are the focus of this study as they are home to $79\%$ of adults with diabetes, and there is a paucity of evidence synthesis detailing the burden of diabetes complications in these settings [1].
We sought to systematically review the literature on both microvascular and macrovascular complications at presentation among patients diagnosed recently with type 2 diabetes in LMICs.
## Materials and methods
This review followed a protocol which was registered in the PROSPERO database (CRD42019126762; available from https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42019126762. The findings have been reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (S1 Checklist).
## Search strategy and selection criteria
We searched the following online databases: Medline, CINAHL and Global Health using predefined search strategies for relevant abstracts. The main search terms included: “Diabetes mellitus” “newly diagnosed” “microvascular complications” and “macrovascular complications” (S1 Text).
All articles reporting complications at diagnosis (regardless of diagnostic method, definition or classification) published in English or French and from LMICs, up to July 2020 were included. Countries with a gross national income below US $12 376 defined LMICs according to the World Bank Country and Lending Groups classification [11].
Case series, studies with a sample size of less than 30 participants, letters to the editor, reviews, editorials, commentaries, conference abstracts of unpublished studies and studies including participants with gestational, or type 1 diabetes were excluded. Cross sectional, cohort and case control studies, case series with > 30 patients and randomised control studies were included. For multiple studies presenting results of the same population, the study with the most complete data was included. For studies with several publications of findings over time, the most recent was included.
## Data extraction
Articles returned by the search were saved to EndNote software which was used to remove duplicates. Titles and abstracts of the articles obtained were subsequently assessed for eligibility using the inclusion and exclusion criteria by two reviewers working independently.
Full text articles were then retrieved and assessed by two reviewers independently (TN, SG, FA and EW), and their references were also screened. Disagreements were resolved by discussion with a third author (OJ). Full texts of articles that could not be retrieved or articles where important information was missing were requested from corresponding authors through emails. Reminders were sent weekly, and the articles were excluded if no response was obtained after a month.
## Data management
An a priori data extraction tool was created on Microsoft Excel 2010 and pre-tested. Following full text screening, data were extracted into the tool by two independent reviewers (TN, SG, FA and EW). A third author (AG) checked that the data was correct and resolved discrepancies by discussion. The following information was extracted: surname of first author; date of publication; country; region; study design; definition used for diagnosis of type 2 diabetes; age range of participants; health facility type; various microvascular and macrovascular complications; their diagnostic criteria and their respective proportions in the participants.
## Assessment of risk of bias
Included studies were assessed for methodological quality and risk of bias using the Quality Assessment Tool for Observational, Cohort and Cross-Sectional Studies of the National Health Institute (S1 Table) by two independent reviewers (FA and EW). Studies were classified to have either “good”; “fair” or “poor” quality.
## Data synthesis and analysis
Preliminary checks for heterogeneity were performed to assess the possibility of combining evidence from primary studies in a meta-analytic approach. The Cochran-Q chi2 statistical test for heterogeneity was used to assess interstudy variability while the inconsistency statistic (I2) quantified the proportion of between study heterogeneity with values of $0\%$-$25\%$, $50\%$-$75\%$, >$75\%$ representing low, medium, and substantial heterogeneity, respectively [12]. Due to the substantial heterogeneity observed between studies (I2 >$90\%$), combining available evidence in a meta-analytic approach was not feasible. Therefore, descriptive statistics (median, range and interquartile range) were used to describe the prevalence of study outcomes from primary studies. Prevalence data for the individual studies were summarised in Forest plots.
Where possible, subgroup analyses were conducted to explore sources of heterogeneity, i.e. whether prevalence of a complication varied with the following prespecified study characteristics: region in which the study was conducted, gender, age, criteria for diagnosis, and type of health facility. For this subgroup analysis, random effects meta-analysis models were fitted to estimate the pooled prevalence and associated $95\%$ confidence interval for each chronic complication within each subgroup. The selection of these covariates was guided by their clinical or evidence-based relevance. A small number of characteristics was chosen to reduce the likelihood of false positive results. Evidence for publication bias was assessed graphically by creating funnel plots from the inverse variance of the proportion of newly diagnosed type 2 diabetes with various complications. Statistical tests for the funnel plot asymmetry were done using the Egger test while the non-parametric trim and fill tests were conducted to account for potentially missing studies.
## Search results
Database searches yielded 3 288 articles and four additional articles were identified from the reference list of which, 2 720 remained after duplicate removal. Titles and abstracts of candidate articles were screened to exclude 2 625 articles leaving 95 records for full text screening. Sixty-two articles were excluded for the following reasons: 18 studies did not identify newly diagnosed patients, 13 studies assessed complications after six months, 12 studies were in other languages, nine studies included patients with type 1 diabetes and in six studies the study population were patients with specific diseases. Two studies included data from the same population, one study only had an abstract published and one author did not respond to queries for full texts. A total of 33 studies met the inclusion criteria and were included in the final review (Fig 1).
**Fig 1:** *Prisma flow chart of the systematic review and article selection.*
## Study characteristics
The characteristics of the 33 studies included in the final review are described in Table 1 [2–4, 9, 10, 13–40]. The studies reported on a total of 13 283 (aged ≥20 years) participants from fourteen countries belonging to LMIC group as per the World Bank classification criteria [11]. Among these studies, 31 were cross sectional [2–4, 9, 10, 13–37, 40], one a cohort [38] and one a case-control study [39]. More than two thirds of the studies, 24 ($73\%$), were conducted in Asia [2, 3, 9, 14–16, 18, 19, 21, 23, 25–29, 31–35, 37–39] while only four were from Africa [10, 13, 30, 36], two from Oceania [17, 20], two from South America [22, 24] and one from the Caribbean [4]. A detailed assessment of the studies included in this review was carried out and the risk scores for bias are presented in S1 Table.
**Table 1**
| Complication | Number of studies | Location of studies | Number of participants | Median prevalence (IQR) |
| --- | --- | --- | --- | --- |
| Retinopathy | 22 | Asia– 20 | 10 427 | 12% (6%-15%) |
| Retinopathy | 22 | Africa– 1 | 10 427 | 12% (6%-15%) |
| Retinopathy | 22 | Oceania—1 | 10 427 | 12% (6%-15%) |
| Nephropathy | 22 | Asia– 17 | 10 409 | 15% (7%-35%) |
| Nephropathy | 22 | Africa– 2 | 10 409 | 15% (7%-35%) |
| Nephropathy | 22 | Oceania– 2 | 10 409 | 15% (7%-35%) |
| Nephropathy | 22 | South America—1 | 10 409 | 15% (7%-35%) |
| Microalbuminuria | 11 | Asia– 7 | 2 276 | 24% (12%-44%) |
| Microalbuminuria | 11 | Africa– 1 | 2 276 | 24% (12%-44%) |
| Microalbuminuria | 11 | Oceania– 2 | 2 276 | 24% (12%-44%) |
| Microalbuminuria | 11 | South America—1 | 2 276 | 24% (12%-44%) |
| Macroalbuminuria | 8 | Asia– 5 | 7 180 | 6% (4%-24%) |
| Macroalbuminuria | 8 | Oceania– 2 | 7 180 | 6% (4%-24%) |
| Macroalbuminuria | 8 | South America—1 | 7 180 | 6% (4%-24%) |
| Neuropathy | 17 | Asia– 14 | 9 701 | 16% (10%-25%) |
| Neuropathy | 17 | Africa– 2 | 9 701 | 16% (10%-25%) |
| Neuropathy | 17 | South America—1 | 9 701 | 16% (10%-25%) |
| Myocardial Infarction | 1 | Sri Lanka | 597 | 7%* (6% - 10%) |
| Ischaemic Heart Disease | 10 | Asia– 10 | 8 418 | 10% (7%-17%) |
| Peripheral Arterial Disease | 6 | Asia– 4 | 2 041 | 6% (1%-20%) |
| Peripheral Arterial Disease | 6 | Africa– 1 | 2 041 | 6% (1%-20%) |
| Peripheral Arterial Disease | 6 | South America– 1 | 2 041 | 6% (1%-20%) |
| Stroke | 4 | Asia– 4 | 2 332 | 2% (1% - 4%) |
| Diabetic Foot | 2 | Asia -2 | 105 | 1%* (0%-1%) |
Ten studies were rated to be of “good” quality [13–22], seventeen were given a “fair” quality rating [2–4, 9, 10, 23–34] and six studies were given a rating of “poor” quality [35–40]. Publication bias was assessed graphically by a funnel plot and the Egger test (Fig 2).
**Fig 2:** *Evidence for publication bias by type of complication from diabetes.*
## a) Retinopathy
This microvascular complication was reported in 22 studies which enrolled a total of 10 427 newly diagnosed type 2 diabetes. Characteristics of these studies and methods used for diagnosis are described in S2 Table. The majority of studies were from Asia ($$n = 20$$), one study was from Africa and another was from Oceania. The prevalence of diabetic retinopathy ranged from $2\%$-$33\%$ (Fig 3), with studies documenting the lowest and highest prevalence both from Asia.
**Fig 3:** *Forest plot illustrating the prevalence of retinopathy among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.*
When stratified by study region, the pooled prevalence of retinopathy in Asia was $11\%$ ($9\%$-$13\%$), $13\%$ ($10\%$-$16\%$) in Africa and $15\%$ ($8\%$-$25\%$) in Oceania (P value = 0.469) suggesting that regional differences did not account for the heterogeneity observed (Fig 4). Further analyses were impossible to conduct due to sparsity of data. Publication bias assessed by the Egger test for symmetry showed significant results ($$P \leq 0.002$$) suggesting possible publication bias (Fig 2).
**Fig 4:** *Forest plot illustrating the prevalence of retinopathy among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries, by region.*
## b) Nephropathy
Nephropathy was reported in 22 studies ($$n = 10$$ 409). The majority of studies were from Asia ($$n = 17$$), two studies were from Africa, two from Oceania and one study was from South America (Table 1). Characteristics of the studies and methods used for the diagnosis of nephropathy are described in S3 Table. The prevalence of nephropathy ranged from $1\%$ to $63\%$ among the eligible primary studies (Fig 5).
**Fig 5:** *Forest plot illustrating the prevalence of nephropathy among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.*
Eleven studies reported microalbuminuria with prevalence ranging from $10\%$ to over $50\%$ among eligible studies (Fig 6). Macroalbuminuria was reported in eight studies and the prevalence of this condition among primary studies ranged from $1\%$-$24\%$ (Fig 7). Further subgroup analyses were not feasible due to the different criteria used to define nephropathy and the paucity of studies in different categories. The funnel plot was asymmetrical, and this was supported by a statistically significant p value of the Egger test ($$P \leq 0.049$$) suggesting possible publication bias (Fig 2).
**Fig 6:** *Forest plot of the prevalence of microalbuminuria among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.* **Fig 7:** *Forest plot of the prevalence of macroalbuminuria among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.*
## c) Peripheral neuropathy
The prevalence of diabetic neuropathy was reported in 17 studies ($$n = 9$$ 701). Fourteen of the studies were from Asia, two were from Africa and one from the Caribbean. The characteristics of the studies and methods used to diagnose neuropathy are summarised in S4 Table.
The prevalence of neuropathy ranged from $3\%$ to $65\%$ (Fig 8). Sub-group analysis by region gave a pooled prevalence of $4\%$ ($4\%$-$5\%$) for the two studies conducted in Africa, $20\%$ ($14\%$-$28\%$) for a study from the Caribbean and $21\%$ ($17\%$-$25\%$) for 14 studies conducted in Asia. The p value for the difference between subgroups was statistically significant ($p \leq 0.01$), suggesting that the region in which the study was conducted accounted for some of the heterogeneity observed (Fig 9). A funnel plot did not demonstrate evidence of publication bias (Egger test, $$p \leq 0.11$$) (Fig 2).
**Fig 8:** *Forest plot illustrating the prevalence of neuropathy among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.* **Fig 9:** *Forest plot illustrating the prevalence of neuropathy among newly diagnosed type 2 diabetes mellitus participants in low- and middle-income countries, by region.*
## d) Myocardial Infarction (MI)
The proportion of participants with MI at diagnosis of type 2 diabetes was reported by Weerasuriya et al. [ 27]. This study was conducted in a Sri Lankan specialised diabetic clinic and the diagnostic criteria were symptoms suggestive of MI. A total of 597 participants were screened. Forty-four participants had symptoms suggestive of MI, representing a MI prevalence of $7\%$ ($5\%$-$10\%$).
## e) Ischaemic Heart Disease (IHD)
The proportion of participants presenting with IHD was reported in ten studies ($$n = 8$$ 418), S5 Table describes characteristics of these studies and criteria used for the diagnosis of IHD. The prevalence of IHD in newly diagnosed type 2 diabetes ranged from $1\%$-$27\%$ (Fig 10) with substantial heterogeneity. The funnel plot did not demonstrate evidence of publication bias (Egger test, $$P \leq 0.98$$) (Fig 2).
**Fig 10:** *Forest plot of the prevalence of IHD among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.*
## f) Peripheral Arterial Disease (PAD)
An estimate for newly diagnosed type 2 diabetic patients presenting with PAD at diagnosis were reported in six studies. Four of the studies were from Asia, one was from Africa and one from South America. A total of 2 041 participants were studied and characteristics and criteria for diagnosis are summarised in S6 Table.
The prevalence of PAD in newly diagnosed type 2 diabetic participants ranged from $1\%$-$40\%$ (Fig 11) with substantial heterogeneity. Funnel plot symmetry was used to assess the risk of publication bias for which there was no evidence (Egger P value = 0.09) (Fig 2).
**Fig 11:** *Forest plot illustrating the prevalence of peripheral arterial disease among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.*
## g) Stroke
Four studies reported the proportion of participants presenting with stroke at the time of diagnosis of type 2 diabetes. All studies were conducted in Asia ($$n = 2$$,332 patients). The study characteristics and criteria for diagnosis of stroke are summarized in S7 Table. The reported prevalence ranged from $0\%$ to $5\%$ (Fig 12), with no evidence of publication bias (Egger P value = 0.72) (Fig 2).
**Fig 12:** *Forest plot of the prevalence of stroke among newly diagnosed type 2 diabetes mellitus patients in low- and middle-income countries.*
## h) Diabetic foot
The proportion of newly diagnosed type 2 diabetes patients presenting with diabetic foot disease was reported by two studies carried out in Asia. Gupta et al. [ 3] studied 105 patients in a tertiary hospital in India and one patient was reported to have diabetic foot at first diagnosis. Chandrashekar et al. [ 38] studied 44 patients and found no patients presenting with diabetic foot at time of diagnosis. Characteristics of these studies are summarised in S8 Table.
## Summary of the findings
This systematic review reports on the burden of diabetes complications at the time of diagnosis of type 2 diabetes in LMICs.
We found that the prevalence of microvascular complications in newly diagnosed type 2 diabetes was estimated at $12\%$ (IQR: $6\%$-$15\%$) for retinopathy. Diabetic retinopathy is one of the leading causes of blindness among working age adults around the world [41] The median prevalence of retinopathy in our study was higher than that reported from studies done in high income countries like Denmark ($6.8\%$) [42] in the Netherlands ($0.7\%$) [43] and South Korea ($2.8\%$) [44] but was lower than the $18\%$ found in two studies in the UK [45, 46]. These differences could reflect retinopathy screening and diagnostic methods and definitions, early identification of people with type 2 diabetes and other health systems and individual level factors [46].
The median proportion of patients presenting with nephropathy was estimated at $15\%$ (IQR: $7\%$-$35\%$) in this study and is higher than that reported in high income countries. It is twice the reported estimate in the United Kingdom ($7.2\%$) [47] and Netherlands ($12.4\%$) [43] and six times that reported in Denmark ($3\%$) [41]. Similarly, in this review we found a prevalence of diabetic neuropathy of $16\%$. The prevalence reported in high-income settings is much lower for patients presenting at diagnosis ranging from 2.3 to $8.2\%$ in the United Kingdom [48, 49], $9.8\%$ in Northern Ireland [50], $8.7\%$ in Finland [51], 6.3 in Germany [52], $6\%$ in Australia [53], $4\%$ in Denmark [54] and 1.7 to $3.9\%$ in the Netherlands [43].
The median prevalence of macrovascular complications in newly diagnosed type 2 diabetes was lower compared to microvascular complications, with estimated prevalence ranging between $6\%$-$10\%$ for peripheral arterial disease, myocardial infarction, and ischemic heart disease. The median prevalence of stroke and the diabetic foot in the eligible studies was low at $2\%$ (IQR: $1\%$-$4\%$) and $1\%$ ($0\%$-$1\%$) respectively. Macrovascular complications such as cardiovascular diseases are the leading cause of disability and death among individuals with diabetes, accounting for approximately half the deaths globally [55]. However, recent data from high-income countries demonstrates that rates of myocardial infarction, stroke and amputation are decreasing among people with diabetes, in parallel with declining mortality [56], possibly as a result of earlier diagnosis and more aggressive management of diabetes and other co-morbidities that may contribute to the development of these complications. The prevalence of macrovascular complications in this review were higher than those reported in high-income settings. The prevalence of IHD in the US is $7.5\%$ and of stroke is $1.7\%$. In Europe the Discover study reported a prevalence of $3.1\%$ for peripheral arterial disease [57]. There are very few studies that explored macrovascular complications in LMICs, hence these complications are less well understood. The current understanding of the international burden of and variation in diabetes-related complications is poor [56], a knowledge gap which this study attempted to narrow.
Our findings demonstrate a considerable burden of diabetes complications at time of diagnosis in LMICs. The findings may be due to a lack of or inaccessibility of services, resulting in delays in seeking care among people living with diabetes in LMICs compared to individuals from high-income settings. In high-income countries, a steady decline in all-cause mortality rates and in the incidence of complications in persons with type 2 diabetes has been seen [58]. The lower prevalence observed in high-income countries is likely attributable to an enabling policy environment which facilitates early detection of type 2 diabetes, resulting in timely clinical management of the disease which then prevents and/or minimizes onset of complications [41]. There is however a lack of data to fully investigate these trends in diabetes complications in high income countries and almost no data from other high-risk areas of the world [58].
The socio-economic implications of a high burden of diabetes and related complications for LMICs are immense, as the management of these conditions can consume vast amounts of household and governments’ spending [59]. The younger age at which type 2 diabetes is occurring in people in LMICs means that complications from diabetes threatens economic productivity and livelihoods of families and communities [54]. The inability to work has a knock-on effect on access to care. Most people in LMICs access care largely by out-of-pocket payments due to weak health systems that cannot meet health care demands. Therefore, failure to earn an income becomes a barrier to accessing healthcare, increasing morbidity and mortality.
From a clinical perspective, the high prevalence of diabetes complications at time of diagnosis suggests a need to improve early identification of people with undiagnosed diabetes. Several studies have evaluated targeted screening for type 2 diabetes and demonstrated effectiveness in identifying undiagnosed people who had a considerable prevalence of microvascular complications [60], however evidence on the cost effectiveness of screening strategies remains unclear [61, 62]. In the context of multimorbidity and increasing service integration, clinic based opportunistic screening particularly targeting high risk groups, may be a feasible approach to increasing early diagnosis of people with type 2 diabetes. However further evidence is needed, particularly on the cost effectiveness of these approaches, in resource limited settings.
## Strengths and limitations
According to our knowledge, this systematic review is among the first to estimate the prevalence of micro and macrovascular complications of diabetes in LMICs. The majority of studies included in this review enrolled large numbers of participants, improving generalizability of findings to populations from which participants were drawn and improving precision of proportions we reported. We only included articles published in peer reviewed journals, with >$80\%$ of the articles receiving a methodological quality rating of “fair/good”, potentially making our findings more robust. However, our review has some important limitations to consider.
The substantial interstudy heterogeneity observed in our study required a formal assessment of potential sources of variability such as gender, age, and type of health facility. However, efforts to explore the source of this variability statistically, were challenging due to paucity of studies in subgroup analyses. Potential reasons for the heterogeneity observed in these findings may relate to different screening procedures for these conditions, differences in definitions and methods of diagnosis in each study, differences in subjects’ ethnicities and their genetic predisposition to diabetes, especially considering that most of our eligible articles ($73\%$) were from Asia, where the socio-economic profiles of countries within the region can differ widely. Even within the LMIC spectrum itself, geographic differences are anticipated with respect to quality of life, access to and quality of healthcare services. The prevalence of co-morbidities like hypertension and high cholesterol, and the extent to which these are diagnosed and treated may also have an impact on differences seen in prevalence of some microvascular and macrovascular complications [46]. This may be a contributing factor to some of the variability observed in this study.
Finally, the high interstudy variability observed may be due to differences in time frames in which the studies were carried, where the publication period for our eligible studies was from 1989–2020. We acknowledge that language restrictions in our inclusion criteria may have limited the scope of the search. Additionally, grey literature formed part of the exclusion criteria. Further work on this topic can include grey literature to increase the review’s comprehensiveness and timeliness to foster a balanced picture of the available evidence [63].
## Conclusion and recommendations
The prevalence of micro and macrovascular complications of diabetes at the time of diagnosis appears higher in LMICs compared to high income countries, however the high heterogeneity observed makes firm conclusions challenging. Our findings suggests that people in LMICs have a higher burden of undiagnosed diabetes complications. Further investigation of cost-effective ways for early identification and treatment of people with diabetes (or prediabetes) to reduce the associated morbidity and mortality is needed. Screening for complications at time of diagnosis should become routine practice to provide opportunity for timely intervention.
## References
1. 1International Diabetes Federation, Diabetes Atlas
2019. [Cited 2021 November 04]. Available from: https://diabetesatlas.org/.. *Diabetes Atlas* (2019.0)
2. Sosale A, Kumar P, Sadikot SM, Nigam A, Bajaj S, Zargar AH. **Chronic complications in newly diagnosed patients with Type 2 diabetes mellitus in India**. *Indian J Endocrinol Metab* (2014.0) **18** 355-360. DOI: 10.4103/2230-8210.131184
3. Gupta A, Singh TP. **Occurrence of complications in newly diagnosed type 2 diabetes patients: a hospital based study**. *Journal Of The Indian Medical Association* (2013.0) **111** 245-7. PMID: 24475555
4. Guízar JM, Kornhauser C, Malacara JM, Amador N, Barrera JA, Esparza R. **Renal Functional Reserve in Patients with Recently Diagnosed Type 2 Diabetes mellitus with and without Microalbuminuria**. *Nephron* (2001.0) **87** 223-230. DOI: 10.1159/000045919
5. Cusick M, Meleth AD, Agron E, Fisher MR, Feed GF, Knatterrud GL. **Associations of mortality and diabetes complications in patients with type 1 and type 2 diabetes: early treatment diabetic retinopathy study report no. 27**. *Diabetes Care* (2005.0) **28** 617-625. DOI: 10.2337/diacare.28.3.617
6. Kharroubi AT, Darwish HM. **Diabetes mellitus: The epidemic of the century**. *World J Diabetes* (2015.0) **6** 850-867. DOI: 10.4239/wjd.v6.i6.850
7. Palladino R, Tabak AG, Khunti K, Valabhji J, Majeed A, Millet C. **Association between pre-diabetes and microvascular and macrovascular disease in newly diagnosed type 2 diabetes**. *BMJ Open Diabetes Research & Care* (2020.0) **8** e001061. DOI: 10.1136/bmjdrc-2019-001061
8. Rao Kondapally Seshasai S, Kaptoge S, Thompson A, Di Angelantonio E, Gao P, Sarwar N. **Diabetes Mellitus, Fasting Glucose, and Risk of Cause-Specific Death**. *New Eng J Med.* (2011.0) **364** 829-841. DOI: 10.1056/NEJMoa1008862
9. Jammal H, Khader Y, Alkhatib S, Abujbara M, Alomari M, Ajlouni K. **Diabetic retinopathy in patients with newly diagnosed type 2 diabetes mellitus in Jordan: Prevalence and associated factors**. *Journal of Diabetes* (2013.0) **5** 172-179. DOI: 10.1111/1753-0407.12015
10. Khalil SA, Rohoma KH, Guindy MA, Zaki A, Hassanein M, Malaty AH. **Prevalence of Chronic Diabetic Complications in Newly Diagnosed versus Known Type 2 Diabetic Subjects in a Sample of Alexandria Population, Egypt**. *Curr Diabetes Rev* (2019.0) **15** 74-83. DOI: 10.2174/1573399814666180125100917
11. 11World Bank Country and Lending Groups. [Cited 2021 October 31]. Available from: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups.
12. Sedgwick P.. **Meta-analyses: what is heterogeneity?**. *BMJ: British Medical Journal* (2015.0) **350** h1435. DOI: 10.1136/bmj.h1435
13. Shaw JE, de Courten M, Dowse GK, Gareeboo H, Tuomilehto J, Alberti KG. **Diabetic neuropathy in Mauritius: prevalence and risk factors**. *Diabetes Res and Clin Prac* (1998.0) **42** 131-139. DOI: 10.1016/s0168-8227(98)00100-4
14. Tasnim F, Farkhanda M, Muneeza Z, Shagufta N. **Pakistan Prevalence of retinopathy detected by fundoscopy among newly diagnosed type 2 diabetic patients visiting a local hospital in Lahore**. *Journal of Zoology* (2017.0) **49** 367-372
15. Raman R, Gupta A, Krishna S, Kulothungan V, Sharma T. **Prevalence and risk factors for diabetic microvascular complications in newly diagnosed type II diabetes mellitus. Sankara Nethralaya Diabetic Retinopathy Epidemiology and Molecular Genetic Study (SN-DREAMS, report 27)**. *J Diabetes Complications* (2012.0) **26** 123-128. DOI: 10.1016/j.jdiacomp.2012.02.001
16. Hayat AS, Baloch GH, Shaikh N. **Journal Of Frequency and pattern of retinopathy in newly diagnosed type 2 diabetic patients at tertiary care settings in Abbottabad**. *JAMC* (2012.0) **24** 87-89
17. Collins VR, Finch CF, Zimmet PZ, Linnane AW. **Prevalence and risk factors for micro- and macroalbuminuria in diabetic subjects and entire population of Nauru**. *Diabetes* (1989.0) **38** 1602-1610. DOI: 10.2337/diab.38.12.1602
18. Kumar M, Rawat R, Verma V, Zafar K, Kumar G. **Chronic complications in newly diagnosed patients with type 2 diabetes mellitus in rural area of western Uttar Pradesh, India**. *Int J Res Med Sci.* (2016.0) 2292-2296
19. Wani FA, Raina AA, Nazir A, Maqbool M, Bhat MH, Sha PA. **Prevalence of Microvascular Complications in Newly Diagnosed Type-2 Diabetes Mellitus**. *Int J of Sci Study* (2016.0) **3** 899-902
20. Collins VR, Plehwe WE, Imo TT, Toelupe PM, Taylor HR, Zimmet PZ. **High prevalence of diabetic retinopathy and nephropathy in Polynesians of Western Samoa**. *Diabetes Care* (1995.0) **18** 1140-1149. DOI: 10.2337/diacare.18.8.1140
21. Wang XL, Lu JM, Pan CY, Tian H, Li CL. **A comparison of urinary albumin excretion rate and microalbuminuria in various glucose tolerance subjects**. *Diabet Med* (2005.0) **22** 332-335. DOI: 10.1111/j.1464-5491.2004.01408.x
22. Adams OP, Howitt C, Unwin N. **The prevalence of peripheral neuropathy severe enough to cause a loss of protective sensation in a population‐based sample of people with known and newly detected diabetes in Barbados: a cross‐sectional study**. *Diabetic Medicine* (2019.0) **36** 1629-1636. DOI: 10.1111/dme.13989
23. Raza SA, Bada F, Rasheed F, Meerza F, Azam S, Jawa A. **Cardiovascular disease risk factors in Pakistani population with newly diagnosed Type 2 diabetes mellitus: A cross-sectional study of selected family practitioner clinics in four provinces of Pakistan (CardiP Study)**. *J Pak Med Ass.* (2019.0) **69** 306-312. PMID: 30890819
24. Felicio JSK, Abdallah Zahalan N, de Souza Resende F, Nascimento de Lemos M, Jardim da Motta Correa Pinto R, Jorge Kzan de Souza Neto N. **An evaluation including a type 2 diabetes mellitus drug-naive patients cohort**. *Diabetes & Vascular Disease Research* (2019.0) **16** 344-350. PMID: 30786752
25. Liu F, Bao Y, Hu R, Zhang X, Li H, Zhu D. **Screening and prevalence of peripheral neuropathy in type 2 diabetic outpatients: a randomized multicentre survey in 12 city hospitals of China**. *Diabetes Metab Res Rev* (2010.0) **26** 481-489. DOI: 10.1002/dmrr.1107
26. Ramachandran AS, Vijay V, Viswanathan M. **Diabetic retinopathy at the time of diagnosis of NIDDM in south Indian subjects**. *Diabetes Research And Clinical Practice* (1996.0) **32** 111-114. DOI: 10.1016/0168-8227(96)01185-0
27. Weerasuriya NSS, Dissanayake A, Subasinghe Z, Wariyapola D, Fernando DJ. **Long-term complications in newly diagnosed Sri Lankan patients with type 2 diabetes mellitus**. *QJM: Monthly Journal Of The Association Of Physicians* (1998.0) **91** 439-443. DOI: 10.1093/qjmed/91.6.439
28. Bansal D, Gudala K, Esam HP, Nayakallu R, Vyamusani VR, Bansali A. **Microvascular Complications and Their Associated Risk Factors in Newly Diagnosed Type 2 Diabetes Mellitus Patients**. *Int J Chronic Dis* (2014.0) 201423. DOI: 10.1155/2014/201423
29. Ali A, Iqbal F, Taj A, Iqbal Z, Amin MJ, Iqbal QZ. **Prevalence of microvascular complications in newly diagnosed patients with type 2 diabetes**. *Pak J Med Sci* (2013.0) **29** 899-902. DOI: 10.12669/pjms.294.3704
30. Dowse GK, Collins VR, Plehwe W, Gareeboo H, Fareed D, Hemraj F. **Prevalence and risk factors for diabetic retinopathy in the multiethnic population of Mauritius**. *A J Epi* (1998.0) **14** 448-57. DOI: 10.1093/oxfordjournals.aje.a009470
31. Wahab S, Shaikh Z, Kazmi WH. **The Frequency of retinopathy in newly diagnosed type 2 diabetes patients**. *JPMA Journal Of The Pakistan Medical Association* (2008.0) **58** 557-561. PMID: 18998309
32. Sosale B, Sosale AR, Mohan AR, Kumar PM, Saboo B, Kandula S. **Cardiovascular risk factors, micro and macrovascular complications at diagnosis in patients with young onset type 2 diabetes in India: CINDI 2**. *Indian J Endocrinol Metab* (2016.0) **20** 114-118. DOI: 10.4103/2230-8210.172277
33. Tzeng TF, Hsieh MC, Shin SJ. **Association of nephropathy and retinopathy, blood pressure, age in newly diagnosed type 2 diabetes mellitus**. *The Kaohsiung Journal Of Medical Sciences* (2001.0) **17** 294-301. PMID: 11559967
34. Agarwal S, Raman R, Kumari RP, Deshmukh H, Paul PG, Gnanamoorthy P. **Diabetic retinopathy in type II diabetics detected by targeted screening versus newly diagnosed in general practice**. *Ann Acad Med Singap* (2006.0) **35** 531-535. PMID: 17006579
35. Dhobi GN, Masoodi SR, Bashir MI, Wani AI, Zargar AH. **Prevalence of hypertension in patients with new onset type 2 diabetes mellitus**. *J Indian Med Assoc.* (2008.0) **106** 92. PMID: 18705251
36. Agaba El, Puepet FH. **Prevalence of Microalbuminuria in newly diagnosed type 2 diabetic patients in Jos Nigeria**. *Afric J Med Scie* (2004.0) **33** 19-22. PMID: 15490788
37. Shukla VKR, Chandra A. **A study of newly diagnosed type 2 diabetes mellitus patients from rural areas**. *J Assoc Physicians India* (2014.0) **62** 682-684. PMID: 25856935
38. Chandrashekar N, Maity N, Kalra P. **Profile of Microvascular Complications in Newly Diagnosed Type 2 Diabetics and its Association with Correlates of Metabolic Syndrome in a Tertiary Hospital: An Observational Study**. *J Pharm Res* (2017.0) **16** 125
39. Bajaj S, Agarwal SK, Varma A, Singh VK. **Association of depression and its relation with complications in newly diagnosed type 2 diabetes**. *Indian J Endoc Metab* (2012.0) **16** 759-763. DOI: 10.4103/2230-8210.100670
40. Gill HK, Yadav SB, Ramesh V, Bhatia E. **A prospective study of prevalence and association of peripheral neuropathy in Indian patients with newly diagnosed type 2 diabetes mellitus**. *J Postgrad Med* (2014.0) **60** 270-275. DOI: 10.4103/0022-3859.138750
41. Gedebjerg A, Almdal TP, Berencsi K, Rungby J, Nielsen JS, Witte DR. **Prevalence of micro- and macrovascular diabetes complications at time of type 2 diabetes diagnosis and associated clinical characteristics: A cross-sectional baseline study of 6958 patients in the Danish DD2 cohort**. *J Diabetes Complications* (2018.0) **32** 34-40. DOI: 10.1016/j.jdiacomp.2017.09.010
42. Bek T, Lund-Andersen H, Hansen AB, Johnsen KB, Sandbaek A, Lauritzen T. **The prevalence of diabetic retinopathy in patients with screen-detected type 2 diabetes in Denmark: the ADDITION study**. *Acta Ophthalmol* (2009.0) **87** 270-274. DOI: 10.1111/j.1755-3768.2008.01207.x
43. Klein Woolthuis EP, de Grauw WJ, van Keeken SM, Akkermans RP, van de Lisdonk EH, Metsemakers JF. **Vascular Outcomes in Patients With Screen-Detected or Clinically Diagnosed Type 2 Diabetes: Diabscreen Study Follow-up**. *Ann Fam Med* (2013.0) **11** 20-27. DOI: 10.1370/afm.1460
44. Jee D, Lee WK, Kang S. **Prevalence and Risk Factors for Diabetic Retinopathy: The Korea National Health and Nutrition Examination Survey 2008–2011**. *Investigative ophthalmology & visual science* (2013.0) **54** 6827. DOI: 10.1167/iovs.13-12654
45. Looker HC, Nyangoma SO, Cromie D, Olson JA, Leese GP, Black M. **Diabetic retinopathy at diagnosis of type 2 diabetes in Scotland**. *Diabetologia* (2012.0) **55** 2335-42. DOI: 10.1007/s00125-012-2596-z
46. Shah S, Feher M, McGovern A, Sherlock J, Whyte MB, Munro N. **Diabetic retinopathy in newly diagnosed Type 2 diabetes mellitus: Prevalence and predictors of progression; a national primary network study**. *Diabetes Res Clin Pract* (2021.0) **175** 108776. DOI: 10.1016/j.diabres.2021.108776
47. Adler AI, Stevens RJ, Manley SE, Bilous RW, Cull CA, Holman RR. **Development and progression of nephropathy in type 2 diabetes: the United Kingdom Prospective Diabetes Study (UKPDS 64)**. *Kidney Int* (2003.0) **63** 225-32. DOI: 10.1046/j.1523-1755.2003.00712.x
48. Azam M.. **Diabetes Complications at Presentation and One Year by Glycated Haemoglobin at Diagnosis in a Multiethnic and Diverse Socioeconomic Population: Results from the South London Diabetes Study**. *J of Diabetes Res* (2015.0) **2015** 587673-587678. PMID: 26090473
49. Davis TM, Stratton IM, Fox CJ, Holman RR, Turner RC. **Prospective Diabetes Study 22. Effect of age at diagnosis on diabetic tissue damage during the first 6 years of NIDDM**. *Diabetes Care* (1997.0) **20** 1435-1441. PMID: 9283793
50. McDaid EA, Monaghan B, Parker AI, Hayes JR, Allen JA. **Peripheral Autonomic Impairment in Patients Newly Diagnosed With Type II Diabetes**. *Diabetes Care* (1994.0) **17** 1422. DOI: 10.2337/diacare.17.12.1422
51. Töyry JP, Partanen JV, Niskanen LK, Länsimies EA, Uusitupa MI. **Divergent development of autonomic and peripheral somatic neuropathies in NIDDM**. *Diabetologia* (1997.0) **40** 953-958. DOI: 10.1007/s001250050773
52. Ratzmann KP, Raschke M, Gander I, Schimke E. **Prevalence of peripheral and autonomic neuropathy in newly diagnosed type II (noninsulin-dependent) diabetes**. *The J Diab Comp* (1991.0) **5** 1-5. DOI: 10.1016/0891-6632(91)90002-7
53. Maple-Brown L, Cunningham J, Dunne K, Whitbread C, Howard D, Weeramanthri T. **Complications of diabetes in urban Indigenous Australians: The DRUID study**. *Diabetes research and clinical practice* (2008.0) **80** 455-462. DOI: 10.1016/j.diabres.2008.01.011
54. Mutyambizi C, Pavlova M, Chola L, Hongoro C, Groot W. **Cost of diabetes mellitus in Africa: a systematic review of existing literature**. *Globalization and Health* (2018.0) **14** 3. DOI: 10.1186/s12992-017-0318-5
55. Einarson TR, Acs A, Ludwig C, Panton U. H. **Prevalence of cardiovascular disease in type 2 diabetes: a systematic literature review of scientific evidence from across the world in 2007–2017**. *Cardiovasc Diabetol.* (2018.0) **17** 83. DOI: 10.1186/s12933-018-0728-6
56. Harding JL, Pavkov ME, Magliano DJ, Shaw JE, Gregg EW. **Global trends in diabetes complications: a review of current evidence**. *Diabetologia* (2019.0) **62** 3-16. DOI: 10.1007/s00125-018-4711-2
57. Kosiborod M, Gomes MB, Nicolucci A, Pocock S, Rathmann W, Shestakova MV. **Vascular complications in patients with type 2 diabetes: prevalence and associated factors in 38 countries (the DISCOVER study program)**. *Cardiovasc diabetol.* (2018.0) **17** 150. DOI: 10.1186/s12933-018-0787-8
58. Pearson-Stuttard J, Buckley J, Cicek M, Gregg EW. **The Changing nature of mortality and morbidity in patients with diabetes**. *Endocrinol Metab Clin North Am* (2021.0) **50** 357-368. DOI: 10.1016/j.ecl.2021.05.001
59. Azevedo M, Alla S. **Diabetes in sub-saharan Africa: Kenya, Mali, Mozambique, Nigeria, South Africa and zambia**. *Int J Diabetes Dev Ctries* (2008.0) **28** 101-108. DOI: 10.4103/0973-3930.45268
60. Spijkerman AM, Dekker JM, Nijpels G, Adriaanse MC, Kostense PJ, Ruwaard D. **Microvascular complications at time of diagnosis of type 2 diabetes are similar among diabetic patients detected by targeted screening and patients newly diagnosed in general practice: the hoorn screening study**. *Diabetes Care* (2003.0) **26** 2604-2608. DOI: 10.2337/diacare.26.9.2604
61. Icks A, Haastert B, Gandjour A, John J, Löwel H, Holle R. **Cost-effectiveness analysis of different screening procedures for type 2 diabetes: the KORA Survey 2000**. *Diabetes Care* (2004.0) **27** 2120-8. DOI: 10.2337/diacare.27.9.2120
62. Kaur G, Chauhan AS, Prinja S, Teerawattananon Y, Muniyandi M, Rastogi A. **Cost-effectiveness of population-based screening for diabetes and hypertension in India: an economic modelling study**. *Lancet Public Health* (2022.0) **7** e65-e73. DOI: 10.1016/S2468-2667(21)00199-7
63. Paez A.. **Grey literature: An important resource in systematic reviews**. *J Evid Based Med.* (2017.0). DOI: 10.1111/jebm.12265
|
---
title: The incremental cost of implementing the world health organization Package
of essential non-communicable (PEN) diseases interventions in Iran
authors:
- Mehrdad Azmin
- Farnam Mohebi
- Moein Yoosefi
- Naser Ahmadi
- Saeed Shirazi
- Mitra Modirian
- Farshad Farzadfar
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021820
doi: 10.1371/journal.pgph.0000449
license: CC BY 4.0
---
# The incremental cost of implementing the world health organization Package of essential non-communicable (PEN) diseases interventions in Iran
## Abstract
World-Health-Organization’s PEN package proposes a minimum set of cost-effective interventions for early diagnosis and management of Non-Communicable-Disease (NCD). IraPEN (the PEN package implemented in Iran), adopted from PEN and Iran National Action Plans for NCDs, addresses challenges regarding NCD prevention and control. IraPEN was piloted in four districts of Iran. In this research, we estimate incremental per-capita cost of IraPEN program implementation in two of the pilot districts. We utilized a bottom-up, ingredient-based costing approach. Institutional expenditure data was collected via information forms. Information pertaining to personnel costs was gathered by performing task time measurements using Direct Observation Method. An individual-level survey was conducted in under-study districts to determine program coverage and its users’ demographic information via systematic random cluster sampling. Sampling of districts was based on systematic random cluster sampling. In each district, 250 families in 25 clusters proportional to urban or rural populations were randomly selected by postal codes. All family members eligible for the program were interviewed. Interviews were organized and conducted in each district by NCD experts in provincial Universities of Medical Sciences. Costs were re-categorized into fixed and variable costs based on their dependency on the program’s coverage. Fixed and variable costs were, respectively, divided by total eligible populations and covered populations in each district to calculate cost per-capita for each protocol. Total per-capita cost per-service was then calculated for each protocol and whole program by adding these figures. All costs are reported in US$ 2015–2016. The incremental costs of IraPEN implementation per user, with and without introduction cost, were US$24.90 and US$25.32, respectively. Total incremental cost per-capita for each protocol ranged between US$1.05 to US$7.45. The human resources and supplies had the highest contribution in total program cost ($74.97\%$ and $15.76\%$, respectively). The present study shows that IraPEN program implementation to be a high-cost package within Iranian context, that necessitates cautions in other similar contexts for implementation. It is, however, difficult to make decisions on implementation of NCD prevention and control programs purely based on their cost. Informed decision making requires assessment of a programs’ effectiveness and justifications and alterations to the current package could reduce the costs, leading to increased efficiency of the program.
## Introduction
The 2030 Agenda for Sustainable Development *Goals is* to reduce premature mortality from Non-Communicable Diseases (NCDs) by one third [1]. World Health Organization’s (WHO) “Package of Essential Non-Communicable Diseases Intervention for Primary Healthcare in Low-Resource Settings” (PEN) proposes a minimum set of cost-effective interventions for early diagnosis and management of NCDs [2] and is introduced as a tool to reduce the NCD-attributable premature mortality. Although the feasibility and cost-effectiveness of PEN has been assessed in a number of low-resource countries, the capacity to implement the package, its cost of implementation and effectiveness are, however, still under question. [ 3–16]. In other words, there is not adequate information on how the cost of implementation differs from the predictions of WHO and in what extend. The detailed information on the costs of implementing PEN package is required for the countries that are willing to adopt this package because countries need accurate estimations of cost of implementation to assess whether they should adopt the package in terms of costs and efficiency and in consideration of their resources, what are the detailed costs of implementation to configurate the package according to their resources and needs, and how to insure the sustained implementation of the practice if they decide to adopt it. This study helps feel the existing gap on the aforementioned notions by studying the case of pilot implementation of PEN in Iran and estimating the detailed cost of PEN implementation in Iran.
Due to high burden of NCDs in Iran [17], with the proportion of deaths attributable to NCDs being $82\%$ in 2017 in Iran, Ministry of Health and Medical Education (MoHME) has developed a “National Action Plan for Prevention and Control of Non-Communicable Diseases and the related risk factors in Iran, 2015–2025” [18]. IraPEN (the PEN package implemented in Iran) is the name of the plan adopted from both WHO PEN package and National Action Plans for NCDs [2, 18]. Considering the high burden of NCDs and the financial limitations in addressing the burden of these diseases and prioritization concerns, Iranian healthcare policy makers are motivated to apply suggestions of expert organizations on efficient and effective policies and interventions. Thus, IraPEN is an attempt by healthcare policymakers and deputies to address challenges regarding NCD prevention and control in Iran by implementing a screening and referral system for the most common NCDs or with highest burden by healthcare providers at primary health care centers [19]. *In* general, healthcare policymakers considered IraPEN to be a comprehensive and effective program and exceeding the previously established type of practice in being comprehensive and integrated. Nevertheless, and like any other comprehensive policy, there are cost and structural concerns for upscaling the program.
IraPEN includes 5 protocols for: 1) prevention of myocardial infarction and stroke through integrated care of diabetes, hypertension, and dyslipidemia (MI & Stroke prevention); 2,3) screening of breast, cervical (female cancer prevention), and colorectal cancers (CRC prevention); 4) screening of asthma and other respiratory disorders including chronic obstructive pulmonary disease (Respiratory diseases screening); 5) Survey of NCD risk factors related to nutrition, alcohol/tobacco use, and low physical activity, followed by health education and counseling on health behaviors (NCD risk factors survey) [19]. This program was funded by MoHME and universities of medical sciences responsible for delivering healthcare programs at the district level.
There was a variety of managerial, political, and organizational challenges for the full-scale implementation of IraPEN [20]. Additionally, the upscaling and continued implementation the program required the long-term provision of financial resource. Therefore, MoHME started with piloting IraPEN in 4 districts to assess the feasibility and incremental cost of the program. In this research, we seek address the knowledge gap regarding the cost and resource use of IraPEN in the two pilot districts of Naqadeh and Shahreza from February 2016 to March 2017. The provided evidence can help with its future planning and may also be useful to other developing countries planning to implement cost-effective interventions to control the rapid emergence of NCDs and their associated complications.
## Materials and methods
IraPEN program pilot implementation was performed in four districts of Iran. The districts were chosen based on the assessment of the experts in MoHME on the following matters: 1) the set of locations being representative of all parts of the country in terms of the healthcare provision and population characteristics, 2) the respective Universities of Medical Sciences being representative of the types of practices that are performed in the medical universities of the country because medical universities are responsible for the provision of the type of the care that is prescribed by PEN package, and 3) the estimated cost of implementation based on experts’ previous experiences and opinions, and not mathematical estimations, do not exceed the budget of the pilot program. This resulted in choosing 4 districts in 4 provinces of the country.
In this study, we estimated the incremental per capita cost of IraPEN program pilot-implementation in Shahreza and Naqadeh districts of Iran using a bottom-up approach [21]. An overall view of analysis framework has been illustrated in Fig 1. Capital and recurrent expenditure data was initially gathered via information forms. Expenditures were then re-categorized into fixed and variable costs. Cost per capita calculated by dividing fixed and variable costs by total eligible populations and the population covered by the pilot programs respectively. Total per capita cost per service was then calculated for each protocol, by including specific and non-specific costs of the protocol, and the whole program by adding these figures. The protocols consisted of a combination of the interventions designed for: 1) prevention of myocardial infarction and stroke through integrated care of diabetes, hypertension, and dyslipidemia (MI & Stroke prevention); 2,3) screening of breast, cervical (female cancer prevention), and colorectal cancers (CRC prevention); 4) screening of asthma and other respiratory disorders including chronic obstructive pulmonary disease (Respiratory diseases screening); 5) Survey of NCD risk factors related to nutrition, alcohol/tobacco use, and low physical activity, followed by health education and counseling on health behaviors (NCD risk factors survey) [19].
**Fig 1:** *An overall view of analysis framework to calculate incremental cost of implementing the world health organization Package of essential non-communicable (PEN) diseases interventions in Iran.*
## A. Institutional data collection
Institutional costs comprised expenditure by MoHME, as well as the Universities of Medical Sciences (UMSs) and District Health Networks (DHNs) responsible for delivering IraPEN services. Institutional expenditure data was collected by submitting request for information (RFI) forms to Health and Treatment deputies at MoHME and UMSs of the districts under study. The relevant authorities were asked to provide the amount spent (in Iran Rials) on each item from February 2016 to March 2017, and provide official financial receipt numbers and reports where available. Expenditure on equipment and supplies was detailed by name of items, total units purchased, unit costs, and number of units used during the period under study. The validity of the reports was assessed and approved by Health and Treatment deputies at MoHME and UMSs and by the authors. In case of any ambiguity or suspect, the authors were provided the chance to contact the responsible person for data collection. Nevertheless, the authors and the deputies did not observe and report any case of potential invalid data entry.
Information pertaining to service providers’ costs attributable to delivering IraPEN services was gathered by performing task time measurements using the Direct Observation Method (DOM). We recorded the time spent by physicians, midwives, and community health workers (CHWs) working in IraPEN delivery sites in rural and urban areas. The gathered data included details on the attending patient, appointment date, service(s) provided, and per service. This data was collected from 2 urban and 2 rural sites in each district for 10–20 patients visited at each site. The time of observing was not announced to the provision centers in advance and the observer recorded data until he/she reached at least 10 patients, continued with the cap of 20 patients. To elaborate, if the observer started observing the site in a time that 15 patients visited till the end of the working hours of the center, it was considered adequate data gathering. The observers were suggested to attend the sites when they expect to see at least 10 patients. If they failed in their attempt to gather enough data, they needed to visit the center again to continue the data gathering. Average daily time spent on managing the program on UMS, urban, and rural levels was also collected. Data on number of service providers (191 physicians, midwives, and CHWs) assigned to the delivery of IraPEN and their salaries in urban and rural areas in each district was provided by their respective UMSs and DHNs.
## B. Individual-level data collection
An individual-level survey was conducted in the districts under study to determine program coverage and its users’ demographic information. We should not that the randomized individual level survey was only conducted in the two districts to determine coverage, effect on wider healthcare utilization, and out of pocket expenditure attributable to the program. The districts for the survey were a random choice between two options for each location. The districts of Shahreza (province of Isfahan) and Baft (province of Kerman) are very close to another both geographically and also in terms of the population characteristics, healthcare access and healthcare system attributes, and assessment of experts on the acceptability and process of IraPEN implementation and Shahreza was chosen between the two districts. This is also true for Maraqeh (province of East Azerbaijan) and Naqadeh (province of West Azarbaijan) districts and Naqadeh was chosen.
Sampling was based on systematic random cluster sampling. In each district, 250 families in 25 clusters proportional to urban or rural populations were randomly selected by postal codes. The cluster construction was as follows: first, all the recorded postal codes were two divided into two stratified clusters (urban and rural). For rural and urban areas, the required number of samples were determined proportional to the number of populations in each. Then, we needed to distribute this sample in the areas. So, we assigned numbers to the geographical blocks (as clusters) of each stratum and then randomly selected required number of blocks. Withing each block, 10 individual postal codes was chosen as each being a representation of one family. We started contacting the eligible families residing in each block and kept inviting until we reached 10 invited families in each block. All family members of a postal code eligible for IraPEN program were interviewed. Interviews were organized and conducted in each district by NCD experts in UMSs of West Azerbaijan and Isfahan. Data collection was initially carried out using paper forms and later entered into the study database by collaborators using a web portal specifically developed for this study.
## Cost estimation approach
*In* general, after calculation the Capital and Recurrent costs via of each protocol, costs were re-categorized into fixed and variable costs. To calculate the cost per capita, we divided fixed and variable costs were, respectively, by the total eligible populations and covered populations in each district and added up these figures to calculate the total cost of each protocol and the whole program. Therefore, we did not perform any inferential analysis for population-based estimations and we directly used and reported the collected data for cost calculations.
## C. Cost per Capita estimation
To make the cost of protocols comparable to one another, we adopted the strategy to standardize the costs with dividing the cost by the eligible population. Considering that we are not performing regression analysis, we did not transform the estimations to meet distribution requirements of regression. The eligible population for each protocol was extracted from the 2016 national census [23]. The costs were recategorized into variables and fixed costs; supplies and personnel were considered variable and other costs were considered fixed costs. It is important to note that some items included in the fixed costs category, particularly equipment should strictly be classified as ‘Mixed’ or “semi-variable” expenses. For the purposes of this paper, however, it was assumed that they would remain constant throughout the program’s implementation. Because the cost of program introduction are one-off costs and only applicable in the first year of implementation, the final costs are reported both incorporating and not incorporating them. Costs for delivery of IraPEN were estimated for individual protocols, carried out by first tagging each equipment and supply item with the service protocol it was used in. The expenditures on supplies and equipment used in multiple protocols, such as disposable gloves, were divided equally between those protocols. Laboratory costs incurred for analyzing cervical and colorectal cancer screening samples were also not included. The staffing costs for delivery of each service was estimated by calculating the cost of healthcare staffing per service using times per service as indicated in staffing time-sheets. Cost of physician visit was attributed to specific protocols accordingly. Non-specific costs, which are costs not related to a service (e.g., program introduction, supervision, and customization) were apportioned to each protocol equally. Fixed costs were divided by the total number of eligible individuals in each of the districts. Variable costs were divided by target population covered by the program. Total per capita cost per service was then calculated for each protocol by adding these figures.
All costs are reported in US$ using the official central bank of Iran exchange rate for 2015–2016 reported at US$1 = 36440 IRR [24]. The timeframe of analysis is annual.
In terms of technical limitations, institutional cost calculations relied fundamentally on the quality of data provided by entities responsible for planning and implementation of the program. Considerable efforts were made in accurate collection of expenditure data, such as requests for detailed cost breakdowns and financial document numbers. There may, however, be inaccuracies in the information provided, such as over-declaration of consumables used at the point of service. Additional efforts were made to address anomalous figures by reconfirmation with relevant sources. Determining the validity of supplied data beyond these efforts were outside the scope of this study. Several assumptions and simplifications, inevitable in research of this nature, were made in calculations. The number and length of visits per provider were based on timesheets collected specially for this study around the same dates. The primary simplification made here was to assume constant and continued demand for visits throughout the implementation period. The estimations, therefore, ignore variations in demand over time. The non-inclusion of laboratory costs for analyzing cervical and colorectal cancer screening samples was another limitation. This may significantly increase the costs associated with these two protocols.
## Summary
Taken together, and as depicted in Fig 1, the cost components were as follows: Then we divided the incremental cost by the population covered and reported them as final results.
## Ethical considerations
This study was assessed by multiple organizations for ethical considerations. This study was reviewed and approved by Deputy of Health, Ministry of Health and Medical Education of Iran before the study began. Tehran University of Medical Sciences IRB approved of this study. Informed written consent was received and documented from the participants, after describing the details of the study. The need for consent was waived by the ethics committee to gather administrative data.
## Results
A total of 637 individuals, from 551 families, who met IraPEN eligibility criteria were interviewed, out of whom 265 individuals had received IraPEN services at least once, corresponding to an overall coverage of $40.74\%$ in two districts. Naqadeh and Shahreza had a coverage of $46.1\%$, and $36.82\%$, respectively. Over $75\%$ of participants living in rural areas were covered by IraPEN.
Total expenditures are presented in Table 1. Introduction cost only contributed to $2.07\%$ of the total costs. The human resources and supplies had the highest contribution in the total program cost ($74.97\%$ and $15.76\%$, respectively). Expenditure on human resources also contributed the highest portion of cost of each protocol (S2 Table). The cost categories with lowest contribution were retraining followed by customization.
**Table 1**
| Unnamed: 0 | Variable costs | Variable costs.1 | Variable costs.2 | Variable costs.3 | Fixed costs | Fixed costs.1 | Fixed costs.2 | Fixed costs.3 | Fixed costs.4 | Fixed costs.5 | Fixed costs.6 | Fixed costs.7 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Protocol | Recurrent cost per protocol per year | Recurrent cost per protocol per year | Recurrent cost per protocol per year | Recurrent cost per protocol per year | Capital cost per protocol per year | Capital cost per protocol per year | Recurrent cost per protocol per year | Recurrent cost per protocol per year | Recurrent cost per protocol per year | Capital cost per protocol per year | Capital cost per protocol per year | Capital cost per protocol per year |
| | Supply | Share of supply cost in each protocol (%) | Personnel | Share of personnel cost in each protocol (%) | Equipment | Share of equipment cost in each protocol (%) | Consultancy | Retraining | Supervision | Introduction | Customization | Supervision |
| MI & Stroke prevention | 102859.24 | 72.26 | 186455.94 | 27.55 | 7389.52 | 82.96 | 4291.00 | 395.87 | 3348.26 | 3754.37 | 1482.81 | 1654.70 |
| Respiratory diseases screening | 671.95 | 0.47 | 37361.09 | 5.52 | 1513.00 | 16.99 | 4291.00 | 395.87 | 3348.26 | 3754.37 | 1482.81 | 1654.70 |
| CRC prevention | 7413.57 | 5.21 | 81284.81 | 12.01 | 0.00 | 0.00 | 4291.00 | 395.87 | 3348.26 | 3754.37 | 1482.81 | 1654.70 |
| Female cancer prevention | 31270.27 | 21.97 | 163802.66 | 24.20 | 5.04 | 0.06 | 4291.00 | 395.87 | 3348.26 | 3754.37 | 1482.81 | 1654.70 |
| NCD risk factors survey | 129.39 | 0.09 | 207893.64 | 30.72 | 0.00 | 0.00 | 4291.00 | 395.87 | 3348.26 | 3754.37 | 1482.81 | 1654.70 |
| Total | 142344.42 | | 676798.14 | | 8907.55 | | 21455.02 | 1979.39 | 16741.34 | 18771.88 | 7414.07 | 8273.52 |
| Share of each cost category from total cost of IraPEN implementation (%) | 15.77 | | 74.98 | | 0.99 | | 2.38 | 0.22 | 1.85 | 2.08 | 0.82 | 0.92 |
When focusing on the proportion of each program in equipment, personnel, and supplies MI & Stroke prevention had the highest share of supplies and equipment costs and NCD risk factors survey had the highest share of personnel costs. Interestingly, NCD risk factors survey was the protocol with the lowest share in supplies and equipment costs (Table 1).
The final incremental costs of the IraPEN implementation per user, with and without introduction cost, were US$24.90 and US$25.32, respectively. The total incremental cost per capita for each protocol ranged between US$1.05 to US$7.45. Among all protocols, MI & Stroke prevention protocol had the highest cost per user followed by NCD risk factors survey (Table 2). The protocol with the lowest cost was respiratory diseases screening.
**Table 2**
| Unnamed: 0 | Variable costs | Variable costs.1 | Fixed costs | Fixed costs.1 | Fixed costs.2 | Fixed costs.3 | Fixed costs.4 | Fixed costs.5 | Fixed costs.6 | Total | Total.1 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Protocol | Recurrent | Recurrent | Capital | Recurrent | Recurrent | Recurrent | Capital | Capital | Capital | Total | Total |
| | Supply | Personnel | Equipment | Consultancy | Retraining | Supervision | Introduction | Customization | Supervision | Without introduction cost | with introduction cost |
| MI & Stroke prevention | 2.33 | 9.47 | 0.16 | 0.10 | 0.008 | 0.07 | 0.09 | 0.03 | 0.04 | 7.45 | 7.53 |
| Respiratory diseases screening | 0.03 | 1.86 | 0.03 | 0.10 | 0.008 | 0.07 | 0.09 | 0.03 | 0.04 | 1.05 | 1.13 |
| CRC prevention | 0.63 | 6.89 | 0.00 | 0.10 | 0.008 | 0.07 | 0.09 | 0.03 | 0.04 | 3.95 | 4.03 |
| Female cancer prevention | 2.20 | 11.58 | 0.00 | 0.10 | 0.008 | 0.07 | 0.09 | 0.03 | 0.04 | 7.13 | 7.23 |
| NCD risk factors survey | 0.01 | 10.42 | 0.00 | 0.10 | 0.008 | 0.07 | 0.09 | 0.03 | 0.04 | 5.32 | 5.40 |
| Total | 5.21 | 40.22 | 0.19 | 0.52 | 0.04 | 0.37 | 0.49 | 0.19 | 0.23 | 24.91 | 25.33 |
## Discussion
This research aimed to estimate the incremental cost of PEN implementation in Iran from healthcare provider perspective. The results show human resources to have had the highest contribution, followed by supplies. The final incremental costs of the IraPEN implementation per capita, excluding the cost of program introduction, was US$24.90. Total incremental cost per capita for each protocol ranged between US$1.05 and US$7.45, with MI & Stroke prevention protocol being the highest and respiratory diseases screening protocol being the lowest. The highest cost of supplies and equipment were for MI & Stroke prevention protocol and the highest cost of personnel was for NCDs risk factors survey protocol.
Our study is one of the first to evaluate the incremental implementation cost of health interventions proposed by PEN program in primary health care settings in a middle-income country [16]. Original WHO cost estimates for implementing the PEN’s package of “best-buy” s, was less than US$3 per capita in a UMIC, much lower than our US$24.90 figure for IraPEN [2, 25–27]. Best-buys should constitute a small proportion of the overall health spending of a country, less than $1\%$ in upper middle-income countries [27]. Our results show the cost of IraPEN package to be $6.80\%$ of the annual per capita healthcare expenditure in Iran [28], and can thus can be considered high-cost. The observed differences could be generally due to two reasons of accuracy and context. First, WHO cost calculations are mainly based on the total expected cost of implementation and in many cases, there is lack of accurate data and the gap is filled by expert estimations [29]. Additionally, WHO calculations are based on predicted costs, whereas we assessed the realized costs of implementation. Therefore, we consider the approach of this study to be accurate and thus more valid. The differences could come from the particular context of the study. Nevertheless, we expect not this to be the case because WHO considers Iran an upper-middle income country and incorporates the characteristics of upper-middle income countries in recommendations. And finally, and as depicted later in the discussion, WHO do not recommend detailed implementation. Therefore, a part of the difference could come from lack of efficiency of the IraPEN implementation and this is why we believe with configurations in the proposed plan, we could get closer to WHO target.
Nevertheless, this costing report can be considered as best estimate currently available for implementing a PEN variant in an MIC. Our study was the first attempt at gathering and presenting empirical evidence regarding the cost of implementing a PEN-based program. We took ingredients approach of costing for supplies, human resources, and equipment which is considered to be more precise compared to other methods like reference costing [22, 30, 31]. The results may offer empirical inputs for macroeconomic costing tools, such as OneHealth, for more realistic cost projections of PEN program in the future and in different countries. The information presented here may also be used internationally by policymakers to estimate the budgets necessary for implementing PEN in settings with previously established primary health care system.
Our study faced some limitations which should be discussed in brief. On top of all, though cost matters, an important factor to consider in making the decision about implementation of an intervention or a policy is the perspective of the patients and healthcare deputies. Due to the financial limitations and being out of the scope of the main goal in the study design, we did not systematically record the perspectives of the stakeholders. Nevertheless, we could infer that a great part of the opinion about the program for healthcare deputies is formed by the cost of the program in our informal discussion policy makers and decision makers in MoHME, UMSs, and DHSs. Additionally, the process of IraPEN implementation did not change the type and configuration of the service provision to patients in the sense that they did notice any big change in the healthcare center and service provision very likely stayed the same in their eyes. This design helped us in two matters. First, we are confident that the measurements are not affected by the patients’ expected outcomes of the changes. Additionally, as patients probably did not notice any difference, their perspective is very likely to be the same before and after the implementation of IraPEN. Therefore, the decision about continuing or halting the program could put patients’ perspective aside in further decision-making steps.
Our estimations just covered incremental cost of implementation in the Iranian healthcare context. The total cost of the program is required to be calculated for further studies and considerations. Nevertheless, and considering all strengths and limitations, our findings could be utilized in low and upper-middle income countries with limited resources and funding to inform the requirements for development and implementation of low cost and effective interventions for prevention and control of NCDs. Based on our findings we suggest some strategies to reduce the cost of the PEN program implementation in the Iranian context as well as other settings: 1) Because the human resource training, preparation, wages, and their supervision contributed the most of the program implementation cost, we suggest alterations of the current program to reduce the necessity and role of humans. Inserting the low-cost innovative technology to the chain of care have shown to reduction and efficiency improvements [32–35]. A primary example of this would be to utilize eHealth strategies in CVD risk calculation and consequent education on associated behavioral risk factors. In addition, any future performance monitoring system and cost reduction strategy requires the relevant data to be collected. Therefore, further strengthening of electronic record keeping for the program will assist in gathering valuable intelligence regarding local needs besides cost reduction. Moreover, it allows for customizable resource planning and allocation of the program which would help furtherly with increasing the efficiency of the program and decreasing its cost. 2) Restriction of the program’s inclusion criteria to target individuals at higher risk of NCDs will lower cost of human resources and the required materials as well. Provision of services could be planned according to the disease’s prevalence and their relative burden in the population. For instance, the incidence rate of cervical and colorectal cancers in the Iranian population was, respectively, 3.2 and 7.65 in 100,000 in 2015 [36]. It therefore seems that the colorectal cancer poses a greater population risk than cervical cancer and should be thus prioritized. Additionally, the cost-effectiveness of these early screening interventions will need to be thoroughly considered and funds should be prioritized and allocated accordingly. 3) Besides direct interventions for patients from healthcare providers, there is a need for exploration of more innovative behavioral approaches, targeted at the reduction of NCDs risk factors at the public and individual levels [27, 37]. It should however be noted that such interventions, while relatively straightforward to initiate, are shown to have mixed records of success [38, 39] and require careful review of the available evidence prior to making and implementing policy decisions [40].
## Conclusions
Taken together, the present study showed that the IraPEN program implementation is considered a high-cost package in Iran. It is, however, difficult to make decisions on the implementation of NCD prevention and control programs purely based on their cost. Informed decision making, in this regard, requires an assessment of a programs’ effectiveness at national and subnational levels. Moreover, justifications and alterations to the current package could reduce the costs, leading to increased efficiency of the program, in the similar contexts comparing to Iran.
## References
1. 1NCD and the Sustainable Development Goals-WHO Global Coordination Mechanism on the Prevention and Control of NCDs. 2015 20190508]; Available from: https://www.who.int/global-coordination-mechanism/ncd-themes/sustainable-development-goals/en/.
2. Organization W.H.. *Package of essential noncommunicable (PEN) disease interventions for primary health care in low-resource settings* (2010.0)
3. Kontsevaya A., Farrington J.. *Implementation of a Package of Essential Noncommunicable (PEN) Disease Interventions in Kyrgyzstan: Evaluation of Effects and Costs in Bishkek After One Year* (2017.0)
4. Wangchuk D.. **Package of essential noncommunicable disease (PEN) interventions in primary health-care settings of Bhutan: a performance assessment study**. *WHO South-East Asia journal of public health* (2014.0) **3** 154. PMID: 28607301
5. Hyon C.S.. **Package of essential noncommunicable disease (Pen) interventions in primary health-care settings in the Democratic people’s Republic of Korea: a feasibility study**. *WHO South-East Asia journal of public health* (2017.0) **6** 69. PMID: 28857065
6. Latt T.S.. **Gaps and challenges to integrating diabetes care in Myanmar**. *WHO South-East Asia journal of public health* (2016.0) **5** 48. PMID: 28604398
7. Dorji T.. **An approach to diabetes prevention and management: the Bhutan experience**. *WHO South-East Asia journal of public health* (2016.0) **5** 44. PMID: 28604397
8. Mutale W.. **Assessing capacity and readiness to manage NCDs in primary care setting: Gaps and opportunities based on adapted WHO PEN tool in Zambia**. *PloS one* (2018.0) **13** e0200994. PMID: 30138318
9. Upreti S.R.. **Strengthening policy and governance to address the growing burden of diabetes in Nepal**. *WHO South-East Asia journal of public health* (2016.0) **5** 40. PMID: 28604396
10. Zhang X.H.. **Implementation of World Health Organization Package of Essential Noncommunicable Disease Interventions (WHO PEN) for Primary Health Care in Low‐Resource Settings: A Policy Statement From the World Hypertension League**. *The Journal of Clinical Hypertension* (2016.0) **18** 5-6. PMID: 26646424
11. Nyarko K.M.. **Capacity assessment of selected health care facilities for the pilot implementation of Package for Essential Non-communicable Diseases (PEN) intervention in Ghana**. *The Pan African medical journal* (2016.0) **25**
12. Dukpa W.. **Is diabetes and hypertension screening worthwhile in resource-limited settings? An economic evaluation based on a pilot of a Package of Essential Non-communicable disease interventions in Bhutan**. *Health policy and planning* (2014.0) **30** 1032-1043. PMID: 25296642
13. Rogers H.E.. **Capacity of Ugandan public sector health facilities to prevent and control non-communicable diseases: an assessment based upon WHO-PEN standards**. *BMC health services research* (2018.0) **18** 606. PMID: 30081898
14. Aryal B.K.. **Assesssment of Health Facilities for Implementation of Package of Essential Non-communicable Disease in Nepal: Baseline Study in Kailali and Ilam District**. *Journal of Nepal Health Research Council* (2018.0) **16** 149-155. PMID: 29983428
15. Basu S.. **Implications of scaling up cardiovascular disease treatment in South Africa: a microsimulation and cost-effectiveness analysis**. *The Lancet Global Health* (2019.0) **7** e270-e280. PMID: 30528531
16. Prabhakaran D.. **Cardiovascular, respiratory, and related disorders: key messages from Disease Control Priorities**. *The Lancet* (2018.0) **391** 1224-1236
17. 17Institute of Health Metrics and Evaluation (IHME). GBD Compare 2017. 2018-08-19]; Available from: http://vizhub.healthdata.org/gbd-compare
18. Peykari N.. **National action plan for non-communicable diseases prevention and control in Iran; a response to emerging epidemic**. *Journal of Diabetes & Metabolic Disorders* (2017.0) **16** 3. DOI: 10.1186/s40200-017-0288-4
19. Cronbach L.J.. **Coefficient alpha and the internal structure of tests**. *psychometrika* (1951.0) **16** 297-334
20. Etemad K.. **A Challenges in Implementing Package of Essential Noncommunicable Diseases Interventions in Iran’s Healthcare System**. *Journal of health research in community* (2016.0) **2** 32-43
21. Chapko M.K.. **Equivalence of two healthcare costing methods: bottom‐up and top‐down**. *Health economics* (2009.0) **18** 1188-1201. DOI: 10.1002/hec.1422
22. Creese A., Parker D.. *Cost Analysis in Primary Health Care* (1994.0)
23. 23Iran, S.C.o. Population and Household of the Country by Province and Sub-province (Shahrestan)-Iran-Census 2016. 2016 2018-11-26]; Available from: https://www.amar.org.ir/Portals/1/census/2016/Population-and-Household-by-Province-and-Shahrestan.xlsx.
24. 24Foreign Exchange Rates, Central Bank of the Islamic Republic of Iran. Available from: https://www.cbi.ir/ExRates/rates_en.aspx.
25. Organization W.H.. *Scaling up action against noncommunicable diseases: how much will it cost?* (2011.0)
26. Mendis S., Chestnov O.. **Costs, benefits, and effectiveness of interventions for the prevention, treatment, and control of cardiovascular diseases and diabetes in Africa**. *Progress in cardiovascular diseases* (2013.0) **56** 314-321. PMID: 24267438
27. 27Bloom, D.E., et al., From burden to" best buys": reducing the economic impact of non-communicable disease in low-and middle-income countries. 2011, Program on the Global Demography of Aging.
28. 28Current health expenditure per capita, PPP (current international $), Iran. 2015 20181230]; Available from: https://data.worldbank.org/indicator/SH.XPD.CHEX.PP.CD?locations=IR.
29. 29Economics of Non-Communicable Diseases in India—World Economic Forum. 2014 2018-11-25]; Available from: http://www3.weforum.org/docs/WEF_EconomicNonCommunicableDiseasesIndia_Report_2014.pdf.
30. UNAIDS A.. *Costing Guidelines for HIV/AIDS Intervention Strategies* (2004.0)
31. Johns B., Baltussen R., Hutubessy R.. **Programme costs in the economic evaluation of health interventions**. *Cost Effectiveness and Resource Allocation* (2003.0) **1** 1. PMID: 12773220
32. Saleh S.. **eHealth as a facilitator of equitable access to primary healthcare: the case of caring for non-communicable diseases in rural and refugee settings in Lebanon**. *International journal of public health* (2018.0) **63** 577-588. PMID: 29546440
33. House W.. *Follow-up to the Political Declaration of the High-level Meeting of the General Assembly on the Prevention and Control of Non-communicable Diseases* (2013.0)
34. Majumdar A.. **mHealth in the prevention and control of non-communicable diseases in India: current possibilities and the way forward**. *Journal of clinical and diagnostic research: JCDR* (2015.0) **9** LE06. PMID: 25859473
35. Peiris D.. **Use of mHealth systems and tools for non-communicable diseases in low-and middle-income countries: a systematic review**. *Journal of cardiovascular translational research* (2014.0) **7** 677-691. PMID: 25209729
36. Aday L.A., Andersen R., Fleming G.V.. *Health care in the US: Equitable for whom?* (1980.0)
37. 37Nugent, R., Benefits and Costs of the Noncommunicable Disease Targets for the Post-2015 Development Agenda. Copenhagen Consensus Center. Perspective Paper, 2015.
38. Brambila-Macias J.. **Policy interventions to promote healthy eating: a review of what works, what does not, and what is promising**. *Food and nutrition bulletin* (2011.0) **32** 365-375. PMID: 22590970
39. Bader P., Boisclair D., Ferrence R.. **Effects of tobacco taxation and pricing on smoking behavior in high risk populations: a knowledge synthesis**. *International journal of environmental research and public health* (2011.0) **8** 4118-4139. PMID: 22163198
40. Heydari G.. **Prevalence of smuggled and foreign cigarette use in Tehran, 2009**. *Tobacco control* (2010.0) **19** 380-382. PMID: 20876076
|
---
title: A socio-ecological framework examination of drivers of blood pressure control
among patients with comorbidities and on treatment in two Nairobi slums; a qualitative
study
authors:
- Shukri F. Mohamed
- Teresia Macharia
- Gershim Asiki
- Paramjit Gill
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021823
doi: 10.1371/journal.pgph.0001625
license: CC BY 4.0
---
# A socio-ecological framework examination of drivers of blood pressure control among patients with comorbidities and on treatment in two Nairobi slums; a qualitative study
## Abstract
Despite the known and effective treatments to control blood pressure, there is limited information on why there are high uncontrolled hypertension rates in urban slum settings. The aim of this paper is to explore the views of treated people with uncontrolled hypertension and other key stakeholders on the facilitators and barriers to blood pressure control among people with comorbid conditions in two Nairobi slums. The study was conducted in two Nairobi slums namely, Korogocho and Viwandani. This study used a qualitative methodology using interviews and focus group discussions. Barriers and facilitators to blood pressure control were explored using the Social Ecological Model (SEM) framework. A total of 57 participants were interviewed for this study. There were 31 in-depth interviews and two focus group discussions among participants with uncontrolled hypertension and with comorbidities. Additionally, 16 key informant interviews were conducted with healthcare providers and decision/policymakers. All interviews were audio-recorded, transcribed verbatim and analysed thematically. This study identified barriers and facilitators to blood pressure control among patients with uncontrolled hypertension at the patient/individual level, family and community level, health system level and at the policy level. High cost of hypertension medicines, the constant unavailability of medicines at the health facilities, unsupportive family and environment, poor medicines supply chain management, availability and use of guidelines were among the barriers reported. The results show that uncontrolled hypertension is a major public health issue in slums of Nairobi and they highlight barriers to blood pressure control at different levels of the socio-ecological model. These findings can be used to design holistic interventions to improve blood pressure control by addressing factors operating at multiple levels of the socio-ecological framework.
## Introduction
Uncontrolled hypertension (UHTN) is an important risk factor for cardiovascular diseases (CVDs) and a leading contributor to death [1,2]. Globally, an estimated 1.28 billion people had hypertension in 2019 and the global age-standardised prevalence of hypertension was $32\%$ in women and $34\%$ in men among adults aged 30–79 years [3]. The highest proportion (1.04 billion—$75\%$) of people with hypertension were from low-and-middle-income-countries (LMICs) while $25\%$ (349 million) were from high-income-countries (HICs) [4]. The highest prevalence ($30\%$) of hypertension is in the African region compared to $18\%$ in the Americas region among those aged 18 years and over [5]. Urbanization is thought to be a key driver for the rise in hypertension in SSA [6] and with urbanization, more and more people are moving to cities and living in slums or slum like conditions with limited health-services available. Currently more than half ($55\%$) of the global population live in urban areas. The UN predicts that this proportion will rise to $68\%$ by 2050 [7]. In 2017, a Lancet article reported intense urban growth over the last 50 years with more than half of city populations living in slums [8].
The increasing trend and high burden of hypertension in LMICs is worrisome particularly because most of the limited resources for healthcare spending is allocated to managing infectious disease burden in these countries. The burden of hypertension among slum populations specifically is high and rising [9]. Barriers to blood pressure control exist at various levels. The Social Ecological Model (SEM) provides different levels to assess barriers in blood pressure control.
Previous literature has shown that economic constraints have been cited as a major barrier to blood pressure control at the individual level while having knowledge and understanding of ones’ own condition are thought to be an important facilitator at this level [10]. Literature supports the importance of family and community in hypertension management. For instance, a study conducted by Flynn and colleagues [11] reported that family members of individuals with hypertension usually helped them with meal preparations, taking their medications and attending appointments. Similarly, a study conducted in Eritrea reported that patients with hypertension highly valued the support they received from their families and community in hypertension care [10]. Health systems in SSA are already overburdened with communicable diseases. Gaps in capacity for implementation of essential non-communicable disease (NCD) intervention have been identified in low resource settings [12]. A study conducted in the slums of Nairobi also found major gaps in staffing, equipment and drugs for handling chronic diseases [13]. Despite the known and effective treatments to control high blood pressure, there is a dearth of information on the drivers of large uncontrolled hypertension rates in urban slum settings [14–17]. The aim of this study was to enrich our understanding on the facilitators and barriers to blood pressure control among people with comorbidities that exist at different levels of the socio-ecological model in two Nairobi slums. The study also explored why treated patients with hypertension still have uncontrolled blood pressure.
## Conceptual framework
This study used the Social Ecological Model (SEM) framework adapted from the Centers for Disease Control and Prevention [18] to understand the multiple levels of factors associated with uncontrolled hypertension and the interactions between the different levels within this system. There are four levels in this adapted SEM: Individual; family and community; health system, and policy/enabling environment (Fig 1). This model takes into account the complex interplay between the different levels and allows for the understanding of the range of factors that put people at risk for uncontrolled hypertension or protects them from having uncontrolled blood pressure. The overlapping levels in the model show how factors at the different levels influence each other. The solutions and gaps in hypertension care can be investigated by assessing these factors at these levels. Uncontrolled hypertension (UHTN) in this study is defined as systolic blood pressure of ≥140 mmHg and/or diastolic blood pressure of ≥90 mmHg in a patient taking anti-hypertensive medication.
**Fig 1:** *Factors affecting uncontrolled hypertension at the different levels.*
## Study site and participants
The study was conducted in two informal settlements or slums in Nairobi (Kenya) namely, Korogocho and Viwandani (Fig 2). The two slums are located on the outskirts of Nairobi City about 10 km from the city center. Viwandani slum is located near the city’s industrial area and it is home to many low income young people working in the industries who are predominantly male. Due to the industrial nature of work that favors employment for men only, a high proportion of married men in Viwandani do not live with their spouses who have been left behind in their rural origins to either farm and or take care of the children. The Korogocho site by contrast is a more established slum settlement with a high proportion of married men living with their spouses and children [19]. The study sites were chosen as the African Population and Health Research Center (APHRC) has been running the Nairobi Urban Health and Demographic Surveillance System (NUHDSS) in these two slums since 2003. The NUHDSS captures routine information on births, deaths and migration from households three times a year. In 2018, the NUHDSS covered 88,798 individuals in 33,462 households (APHRC 2018). The NUHDSS [20] provides a sampling frame for many nested studies including AWI-Gen study [21] from which this current study drew its study participants. The AWI-Gen study collected data on sociodemographic, anthropometric, biomedical and genetic information from 2003 study participants between the ages of 40 and 60 years in the NUHDSS using a cross-sectional survey.
**Fig 2:** *Map with the location of the Nairobi Urban Health and Demographic Surveillance System (NUHDSS) sites (Korogocho and Viwandani).*
The AWI-Gen study sample was used to purposively sample residents of the two slums with uncontrolled hypertension and comorbid conditions using the participants’ most recent blood pressure measurements (collected in 2018). All participants from the community were adults aged 45 years and older, previously diagnosed with hypertension, and had at least one of the following comorbidities; diabetes, dyslipidemia and overweight/obesity in the 2014–2015 AWI-Gen survey, and were receiving care for uncontrolled hypertension. Healthcare providers in the community and relevant decision/policymakers at county and national level were also approached.
## Study design
This study used a phenomenology approach to gain an in-depth understanding of the facilitators and barriers in controlling blood pressure among patients who are on treatment for hypertension and have a comorbid condition. A phenomenological approach is a form of qualitative enquiry that emphasizes lived experiences of individuals by exploring the meaning of a phenomenon while gaining a deeper understanding of the phenomenon [22]. The main goal of this approach is to identify a phenomena by how it is perceived by those with the lived experiences [23]. The consolidated criteria for reporting qualitative research (COREQ) was adopted in this study.
Data were collected via focus group discussions (FGDs) and in-depth interviews (IDIs). In-depth interviews were used because they provide rich participant views and allow for issues to be explored in more depth [24,25]. FGDs were used to supplement the interviews because they give participants an opportunity to reflect on other participants’ views while building on their views and they give a good understanding about participants’ views on the topic of interest [24,26,27]. IDIs and FGDs were conducted among people with uncontrolled hypertension and comorbidities while on hypertension treatment. Key informant interviews (KIIs) were conducted among key actors shaping hypertension care (healthcare providers, and decision/policy makers).
## Characteristics of study participants
A total of 57 people participated in the study. Thirty-one IDIs (15 in Korogocho and 16 in Viwandani) were conducted. Two FGDs were also conducted; one in each of the two slums among participants with uncontrolled hypertension and comorbidities. Eleven KIIs were conducted among healthcare providers in service provision for hypertension in the two study communities. In addition, five key stakeholder interviews were conducted with representation from the ministry of health; two from the national level and three others were from the sub-county health levels (Table 1). The sample size was determined by theoretical saturation [28].
**Table 1**
| Interview type | Participants | Number |
| --- | --- | --- |
| Site 1: Korogocho | Site 1: Korogocho | Site 1: Korogocho |
| In-depth interviews (IDIs) | Males | 6 |
| In-depth interviews (IDIs) | Females | 9 |
| KIIs–Healthcare providers | Males | 4 |
| KIIs–Healthcare providers | Females | 1 |
| Focus group discussions (FGDs) | Males | 3 |
| Focus group discussions (FGDs) | Females | 2 |
| Site 2: Viwandani | Site 2: Viwandani | Site 2: Viwandani |
| In-depth interviews (IDIs) | Males | 5 |
| In-depth interviews (IDIs) | Females | 10 |
| KIIs–Healthcare providers | Males | 3 |
| KIIs–Healthcare providers | Females | 3 |
| Focus group discussions (FGDs) | Males | 3 |
| Focus group discussions (FGDs) | Females | 3 |
| Site 3: National/county level | | |
| KIIs–Decision and policy makers | Males | 3 |
| KIIs–Decision and policy makers | Females | 2 |
| Total | | 57 |
## In-depth interviews and focus group discussions
IDIs and FGDs were conducted among participants with uncontrolled hypertension and comorbidities to understand their experiences and views about their hypertension care. They were also asked about facilitators and barriers to blood pressure control and solutions to the barriers mentioned at each of the SEM levels. Once no new information was emerging from the interviews, data collection was stopped.
## Key informant interviews with key stakeholders
For the key stakeholders’ interviews (policy/decision-makers and healthcare providers), an initial list of purposively selected study participants was generated with varying representation in sectors. Their selection was based on their role in hypertension care provision in the community or their ability to influence policy and decision making for hypertension care. Snowballing [29] was also used to identify additional key informants during interviews with the initial key informants selected. Key informant interviews were conducted with decision/policymakers to get their views on the challenges in the access and uptake of hypertension care in the study community and what can be done to improve access and uptake of hypertension care while interviews with healthcare providers sought to ascertain the healthcare providers’ prescription practices, conformity and knowledge of national guidelines, and how they treat patients with comorbidities. Both categories of the key stakeholders further provided their views on factors associated with uncontrolled hypertension in the community using the different levels in the SEM framework as a guide.
## Guide development, training of interviewers and pilot study
The initial topic guides were developed from the literature review informed by the conceptual framework. All study participants were asked to describe facilitators, barriers and solutions to blood pressure control at different levels of the adapted SEM (patient, family or community, health system and policy level). The guides were further revised following the pilot study.
Two research assistants collected the data for this study. Both had previous experience working in the community and had prior experience conducting qualitative interviews. They were conversant with the local language (Swahili) and the cultural nuances in the community. The team were trained on the study rationale, objectives, study approach, data collection procedures, note-taking and phone based interviews. They were further trained on research ethics and the study’s informed consent process. The team also reviewed the guides to understand the purpose of each question and the objective it answered.
A virtual pilot study was undertaken outside the study areas to test all the study guides. A debriefing session was held following the pilot exercise to discuss questions that were unclear, had wrong instructions, questions respondents struggled with or questions that were difficult for the participant to understand in all the developed guides. The guides were then revised accordingly before the actual the data collection.
## Data collection
Due to the COVID-19 pandemic, most of the interviews were conducted by phone. The FGDs were conducted in a face-to-face set up that adhered to COVID-19 control measures. The 2018 blood pressure recordings were used for selection into the study. Before the interviews begun, all study participants were asked to complete a brief questionnaire to provide their demographic information. To enhance data quality, the research assistants submitted their summary notes daily and several debriefing sessions were held with them to assess the quality of data that were collected. Data collection took place from June to August 2020 and data collection was stopped when theoretical saturation was reached—no new information was being generated from the interviews [28].
## Analysis
All interviews were recorded. Audio-recordings were first transcribed verbatim by a professional then translated into English by an independent translator. Transcripts were reviewed daily by the research team in order to get a sense of theoretical saturation. All the transcripts were imported into NVivo software (version 12, QSR International) for coding and further analysis.
The analysis was guided by Braun and Clarke’s six steps to conduct thematic analysis [30]. Coding and identification of quotes to go with each theme was done by two independent researchers. All the data were analysed and integrated together in the presentation of the themes. Using the SEM framework, data were examined and four major themes were deductively developed; these were facilitators and barriers to blood pressure control at the 1) patient/individual level, 2) family and community level, 3) health system level and 4) policy level. The solutions to the above barriers are also presented under each theme (Table 2).
**Table 2**
| Factors | Facilitators | Barriers | Solutions |
| --- | --- | --- | --- |
| Individual Level Factors | ○ Good understanding of blood pressure management○ Adherence to medical and lifestyle advise○ Blood pressure monitoring○ Stress management○ Health insurance | ○ Poverty, low socio-economic status○ Inability to adhere to diet○ Patients comorbidities & age○ Patient’s perception of physical activity○ Limited knowledge on blood pressure○ Unfavourable medication side effects & pill burden○ Behavioural factors: smoking & alcohol use○ Denial of hypertension diagnosis○ Language barrier | ○ Resources: income generating activities○ Free medicine○ Hypertension knowledge |
| Family and community level factors | ○ Social /family support○ Supportive social environment○ Easy access to facilities | ○ Stigma and unsupportive family environment○ No support to older people○ Social, physical and stressful home environment○ Limited hypertension knowledge | ○ Use of community health workers○ Attending clinic with care providers |
| Health System factors | ○ Technical capacity of healthcare providers○ Relationships with care providers○ Quality of care○ Facility hours of operation○ Availability of medication○ Free medication | ○ Inadequate follow ups & lack of appointments○ Providers heavy workload and bad attitude○ Low quality of care○ Limited provider training opportunities○ Poor supply management system○ High cost of medication & unavailability of medication (stock outs)○ Facilities not accepting patient’s insurance○ Expired medication○ Facility hours of operation○ Shortage of healthcare providers & long wait times○ Limited and worn-out equipment and tests & delays in equipment maintenance and replacement | ○ Imparting hypertension knowledge to patients○ Training & increasing healthcare providers○ Setting up special hypertension clinics & follow up mechanisms for patients○ Availing medication, tests and equipment including free medication○ Removal of taxes on hypertension medicines & medication subsidies○ Reducing patient wait times |
| Policy level factors | ○ Availability of guidelines | ○ Lack of guidelines in some facilities○ Policies prohibiting hypertension medicines in lower-level facilities○ No specific budget allocation dedicated for hypertension care○ Limited evidence to support policy development | ○ Research and interventions to inform policies○ Policy that would allow free medication including policy removing taxation on hypertension medication○ Availability of care guidelines in each facility |
## Ethics
This research study was approved by AMREF (P$\frac{773}{2020}$) and University of Warwick ethical review boards (BSREC $\frac{54}{19}$-20). Informed consent was collected from all participants in accordance with approved ethical procedures and guidelines. For the phone interviews, verbal consent was sought and it was audio-recorded. For the face-to-face interviews, written consent was obtained and this was documented.
## Organisation of the results
The results from this research are organized to match the various levels of the study’s conceptual framework.
## Perception and experiences of patient level facilitators and barriers
Knowledge, behaviour, practices, and healthcare experiences of residents who had uncontrolled hypertension living in the selected study sites were explored. The findings showed that a patient’s blood pressure control was facilitated by having a fair understanding of what they needed to do. They mentioned adherence to medication, frequent monitoring of BP, salt reduction, physical activity, diet control, weight control and lowering of alcohol consumption as key facilitating components of hypertension management. Monitoring of blood pressure regularly was mentioned and the monitoring ranged from daily, every three days to every three months. A good number of respondents recorded their readings for future reference. Respondents also noted that they knew what their target blood pressure was supposed to be.
Respondents mentioned managing stress as a facilitator in controlling their blood pressure. They felt it improved their emotional and physical health, which ultimately lowered their high blood pressure. Stress management techniques mentioned included exercising, listening to music, focusing on something calm or peaceful and talking to their friends.
Despite the patients’ experiences of how to manage their blood pressure, study participants frequently reported poverty as a significant barrier to blood pressure control in their communities because it restricted their access to medication. Majority of the participants repeatedly reported not getting medicine at the health facilities they visited and that unaffordability of hypertension medications affected their adherence to blood pressure medications. The situation was worsened by the COVID-19 pandemic with many people losing their jobs. Given the reality of low financial resources in the study area where sources of income were limited and wealth status was low, high costs of the medicine were felt to constrain the success of medication adherence. Respondents noted that they would only buy the medications that they could afford and stay without medications on the days they could not afford. Concerns about the cost of hypertension care went beyond the cost of medication to include other associated costs including consultation, testing and transport.
Having some sort of health insurance coverage was mentioned as providing access to healthcare. Patients reported getting treatment in some health facilities was facilitated by having health insurance.
The few individuals that mentioned having health insurance, mentioned having the National Health Insurance Fund (NHIF). This was reported as the government-initiated health insurance that is mainly accepted at government (public) facilities as well as a few other private health facilities in the community. Even though the public facilities accepted the NHIF, study participants reported that they hardly had medicines in stock. Others also mentioned that the facilities did not accept their insurance because the government was not reimbursing the health facilities for the services they provided. Healthcare providers noted that not having health insurance was a barrier to blood pressure control because the alternative would be to buy the medication though many patients in the community were not able to afford this.
Respondents noted that they were conflicted on whether to spend the little money they had on food or medicine to control their blood pressure. There was a perception among the patients that they had to prioritise eating for survival over buying their medications.
Participants also reported that their low economic status inhibited their ability to follow the recommended diet stating that the diet recommendations did not consider their financial status since they could only afford one meal a day. In addition, there was a challenge when only one individual in the family had hypertension or another comorbidity, thus requiring that there are two budgets to cater for food in the family, which was considered unfeasible.
There were misconceptions and difficulties in implementing physical activity in this community. The most common exercises mentioned were house chores and walking. Some respondents stated that they were too old for exercise and they seemed to perceive that exercise was only vigorous activity such as running.
Apart from discouraging physical activity, age was also mentioned as a barrier as it led to forgetfulness that affected the patients’ care in terms of medication adherence and keeping appointments for follow up. Health care providers also noted that it was not easy to care for old patients especially those who lived alone. They mentioned that the older patients were stubborn and hardly followed instructions. In addition, older patients were described as vulnerable because they relied on other people for their care including food preparation and healthcare The stresses of daily life and responsibility of taking care of the family were also reported as barriers in blood pressure control. Participants stated that hypertensive patients were prone to stress especially due to their low economic status that hindered their capability to buy medication and provide for their family.
Knowledge on the causes and management of hypertension was limited among the majority of the respondents and this resulted in medication non-adherence. In addition, some of the interviewees were unaware of the asymptomatic nature of hypertension and the rationale for its lifelong treatment. The idea of having to take drugs continuously was also thought to be a burden and some respondents reported that they only took drugs when they felt unwell or when they experienced hypertension side effects. Medication side effects were a significant barrier to blood pressure control in the community. Some respondents reported having to stop taking their medications due to unfavourable side effects.
While patients possessed some general knowledge of their condition and hypertension, the level of knowledge was limited. Only a few patients were able to recall what their optimal BP was or could identify their target BP as informed by the health care provider and not all could remember their most recent blood pressure measurements.
Behavioural factors such as smoking and alcohol consumption were noted by health care providers to be very prevalent in the study community and this was a significant barrier to blood pressure control.
Most respondents reported that they also had other comorbidities complicating their management. Comorbidities were mentioned as a barrier to practicing lifestyle changes to control hypertension especially diet modifications. In addition, the comorbidities meant that the number of drugs increased which was a burden to the patient. The most common comorbidity mentioned by respondents were diabetes and dyslipidemia.
In addition, denial and use of herbs was mentioned as a barrier by both healthcare providers and participants with uncontrolled hypertension.
## Proposed patient-level solutions
Most of the solutions recommended at this level were around securing resources to enable the patients to access hypertension care. Patients recommended income generating activities to enable them buy medications and cater for their other expenses. Patients also reported that if they could get free medication, then that would reduce their stress and in turn their blood pressure would be controlled. Others thought a change in their environment would help control their blood pressures. Some knew they needed to take their medications in order to control their blood pressure and so they felt it was important to work hard in order to pay for their medications.
Policymakers felt it was important for patients to get more information about their condition and how they can manage themselves well. Patients with hypertension felt that getting advice from the health care providers was important for their care. Health care providers suggested using community health volunteers (CHVs) with similar characteristics as the patients (e.g., an elderly CHV to pass the message to the patients in a language they understood).
## Perception and experiences of family and community-level facilitators and barriers
The availability of CHVs who work in the community was seen as a facilitator for blood pressure control in the community. The CHVs were credited for supporting the elderly in taking medication, giving hypertension information to the community members and referring patients to the hospitals for further care. In addition, the CHVs supported the health providers in reaching those who did not come to the blood pressure clinics In addition, having social and physical support from family members to help the elderly clients was mentioned as an important facilitator. These family members were reported to provide health care assistance including accompanying the elderly to the health facilities for their regular appointments, supporting them in adhering to medication use as advised, and reminding them of the appointment dates.
Most participants reported having easy access to clinics where they went for screening and medication. They reported these clinics to be near their homes and therefore did not require a lot of resources to get to them. The community was reported to have both public and private health facilities. Some of the community members noted that they went outside their community for care when they needed specialised care and when referred.
Some patients reported their neighbourhoods to be supportive environments for blood pressure control. They noted getting support from neighbours with regards to food and money to buy medicines when they could not afford to buy for themselves. The community environment was also described by some respondents as an avenue to de-stress. Respondents noted that they would assist their friends and neighbours when they felt stressed for social support. The friends and community members also tried not to stress the patients.
Despite the aforementioned facilitators, many barriers were mentioned at this level. Even though many patients reported accessing care in the community, other patients had to go far from their homes when they needed specialised care, which was a barrier as they needed to have transport. In addition, patients who wanted to avoid the stigma associated with having hypertension opted to seek care far from their homes.
An unsupportive family environment was also noted as a barrier in the community. Participants with hypertension reported not being able to eat different foods as required for their condition from those consumed in the house. Some respondents also noted that there was no one to accompany the older patients for their appointments and language was sometimes an issue. Others mentioned the noisy community environment also meant that it was not conducive for blood pressure control. Since family members are involved in influencing the lifestyle of the patients, including the food they eat and the money and time given to seeking health care and adherence to medication, lack of information among them led to lack of support. Patients also reported a stressful home environment as a barrier. Uncooperative spouses, alcoholic and drug abusing children were among the stresses mentioned.
Participants also reported stigma that they experienced as patients with hypertension. This led to the patients having to travel far to seek medical attention just to conceal their HIV status. Another barrier identified at the family/community level was the lack of knowledge within the community which has led to myths and misconceptions in the community about hypertension. For instance, some people in the community believed that being diagnosed with hypertension was a death sentence hence there was no need to seek medical care while others thought that once one has hypertension, it is not treatable or manageable. Myths and misconceptions in the community were also reported as barriers.
Participants noted an increase in number of traditional healers in the community and it was thought that they may be affecting the patients with hypertension adversely.
Finally, the environment where the study was carried out was described as a barrier for a number of reasons. The limited space in the urban informal settlements discouraged exercise and walking due to congestion. In addition, the environment impacted on the foods the patients consumed.
## Proposed family and community level solutions to the barriers
Several solutions were suggested at the family and community level. The most mentioned solution was making use of the community health strategy to support continuous monitoring and screening as well adherence to medication and clinic appointments.
Another solution was the creation of general awareness in the community about their condition. This involved where patients needed to seek treatment.
The health care providers who mentioned language barrier as an issue suggested that the patients could come with their care providers for appointments. Health care providers felt that this would help during the clinic appointment and it was hoped that the care provider would support the patient with adherence to the medication due to being present when the information is given to the patients
## Perception and experiences of health system level facilitators and barriers
Health professionals were mentioned to be the key source of information for hypertensive patients. Patients reported having good communication with their doctors and the service providers were generally described to be cooperative. The capacity of the health providers was also reported as a facilitator in that the healthcare providers were able to treat patients who had comorbidities and were able to change the prescription when needed. Some of the facilitators mentioned at the provider level included provider training received and the providers’ ability to follow the guidelines for hypertension care in the facility.
According to the study participants, their doctors changed their medication according to their blood pressure measurements. Most patients with uncontrolled hypertension mentioned that their medications had changed over time. Some reported taking higher doses while others noted that the number of drugs for hypertension also increased.
Health care providers offering continuous care and following up with their clients was also noted to be a facilitator. Some health care providers were also working with CHVs to support in follow-up of patients The patients reported that they were happy with the care they were receiving and there seemed to be good rapport between the patients and the health care providers. Some even got financial support from the health care providers who were also described as being friendly. One respondent noted that their blood pressure was stabilized when it was really high while others mentioned that they were also advised on how to manage their blood pressure.
Further, some respondents described that the wait times were appropriate and they were able to return to their daily activities after their appointments.
Respondents further noted many facilitators in regards to health systems. Healthcare providers noted that there were systems in place for following up patients particularly those whose blood pressure have not been controlled. Other respondents also mentioned that systems were in place in their facilities to quantify their needs for medicines for hypertension care even though the shipment for medicines mostly never arrived on time due to various reasons.
Some of the respondents noted they were able to find medications in the facilities they visited and that the medications were offered at no cost. They also mentioned they rarely found stock outs of the medications they needed and in the instances that the facility they used was out of medication, they were given a date to come back for their medication.
Very few participants indicated that the operating hours of the facilities they visit for hypertension care was fine. A few of the respondents mentioned that the facilities they visited provided 24 hour services while others mentioned specific hours in which hypertensive patients were seen.
Despite the provider-level facilitators reported, many barriers were reported about the healthcare providers in the community. Inability to regularly follow-up and closely monitor patients was described as a significant barrier. Healthcare providers in the community felt their message would have more impact if they were able to follow-up with their patients more frequently. The main barrier to regular follow-up and monitoring was lack of appointments in the facilities they worked in. Other barriers mentioned about the care the providers gave were lack of time due to the providers’ heavy workload, lack of training and knowledge or expertise to treat hypertensive patients that had comorbidities.
The high workload of doctors was noted as a challenge as there was no time for a detailed discussion between the patient and the service providers. From the interviews, it was clear that the respondents were desperate for information on hypertension as they were not getting answers from their health care providers.
The expertise of the healthcare providers was also mentioned as a barrier. Some patients did not trust the care they received so they either stopped seeking care and only went to buy medication from chemists or changed hospitals.
Another barrier mentioned by patients with hypertension was that providers were noted to offer the patients only what was in stock, which was particularly challenging for those patients on multi-drug regimens. Health facilities were also described to have a poor supply management system as medication stock outs were mentioned repeatedly as a major barrier to blood pressure control. Patients and health care providers alike mentioned this to be a problem. Patients reported that even if the medications were free in public facilities, these facilities often lacked the medications and patients had to purchase the medication elsewhere. This was particularly challenging for those patients who were on multi-drug regimens for their hypertension treatment. In addition, due to the inefficient supply system, it was noted that when some medicines arrived, some were expired and patients were using them before they realized they were expired.
A surprising finding in regards to out of stock medicines at the public facilities is that patients reported that they were always asked to purchase the medications from elsewhere and that the medications were always available in private pharmacies yet they were unavailable at the public facilities in which they sought care where they expected to receive the medications for free. It also emerged that facilities purposely did not stock all types of medications for hypertension—they only stocked the cheaper medications.
Medication stock out was a common feature in most public facilities regardless of where the facility was located. In addition to not stocking medications, a near absence of public facilities was mentioned in the community. In each of the two study communities, respondents mentioned that there was only one public level facility in each of the study sites and that they almost always didn’t not have medications. Healthcare providers at the public facilities also advised the patients to buy medications that were out of stock in other private facilities.
Facility hours of operation was cited by many as a barrier to blood pressure control. The government owned facilities did not operate beyond the normal working hours (8am to 5pm) or on the weekend thus limiting care to those who have to work and need care outside those working hours. The clinic days for hypertension care was also reported to be once a week and the hours to be seen were also short.
Another major barrier mentioned was that the facilities were short staffed; this shortage of staff was the norm across all the facilities thus leading to high workload to the current staff and this also led to long wait times for patients to be seen. Healthcare providers reported that if they were not available on the clinic day, then the patients would have to come back on another day and so the patients would have to do without their medications if they were completely out. Patients with hypertension also noted that they had to wait months to be seen. Others reported that sometimes they would spend their whole day to be seen, while some patients mentioned not booking other activities on the same day as their clinic days. They also noted that they had to arrive very early if they were to be seen early. Patients noted that they were attended to faster in private facilities. Policymakers for the area also noted that there was a shortage of staff in the public community facilities thus the staff are overwhelmed.
The patients also mentioned the attitude of the healthcare providers as a barrier. Some patients felt that they couldn’t consult the health providers and some mentioned that the healthcare providers shouted at them.
Quality of care at the facilities was reported to be compromised because of the high workload thus the doctors were not able to give proper attention to patient care.
Lack of equipment, reagents for tests and lab tests in general were cited as barriers to hypertension care in the community because it meant patients needed to go and get the tests in places that required them to pay for the tests. Study participants also reported that sometimes the equipment broke down or they needed maintenance. Delays in replacing and repairing equipment were also cited.
## Proposed solutions at the health-system level
Various solutions at the health providers’ level were proposed by the patients, the health care providers themselves as well as the policy/decision makers. The respondents noted that it was important to ensure the patients and the community in general be knowledgeable on hypertension. However, the health providers noted that the heavy workload was a hindrance to achieving this. A healthcare provider mentioned increasing the number of healthcare providers’ as a solution to the heavy workload and increasing providers’ capacity through trainings and seminars.
To counter the heavy workload, setting up special clinics for hypertensive patients was proposed as a possible solution. In addition, the need to set up a follow-up mechanism for patients was proposed. The option to provide continuous support was also mentioned as a possible solution.
The main solutions mentioned at the health system mainly revolved around availability of medications, tests and equipment for hypertension. Solutions particularly for the public facilities were to ensure that they have continuous medications in stock. A consistent supply of medications for hypertension was suggested by patients and healthcare providers who acknowledged the medication stock out was a major challenge affecting blood pressure control. They also suggested that people with hypertension be treated that same way people with HIV are treated because people with HIV are never without medication.
Majority of the respondents felt that medications for hypertension should be provided for free. Almost all respondents felt the cost of hypertension medications was prohibitive and the government needed to intervene and provide these medications for free especially to the elderly.
Participants also suggested that taxes needed to be removed from hypertensive medications to make the medications cheaper for patients to buy. They also suggested that a donor can be approached to help with medications while another suggestion was to engage a non-governmental organization to help with the provision of subsidized medications for hypertension.
Since healthcare provider staffing was a major barrier, most of the respondents suggested that the number of doctors attending to patients with hypertension needed to be increased so that the long wait times are reduced for patients.
Some respondents felt that providers needed to provide hypertension care counselling on issues relating to what they could do to improve their blood pressure control and information on what would be the ideal nutrition for them.
Participants noted the need for better linkage between facilities to improve referrals. They also recommended special clinics in the community to reduce referrals to outside facilities which some patients find difficult to access. Strengthening linkages between the community and the facility was also suggested as was the integration of programs to avoid missing other conditions that the patients have that need attention.
## Policy-level facilitators and barriers
Few facilitators at the policy level were mentioned. Healthcare providers mentioned that policies/guidelines were available for the management of hypertensive patients.
Barriers mentioned at the policy level included the lack of guidelines, not having up to date guidelines on hypertension, guidelines not being cascaded to lower level facilities and not having a budget line or specific allocation for hypertension care within the healthcare budget.
It further emerged that to have access to certain hypertension medications and products, health facilities needed to be designated at a higher level and this may have contributed to the medication stock outs experienced in the lower level facilities.
Allocation of resources was cited also as a barrier to blood pressure management. It was noted that there is no specific allocation or budget line for hypertension care, which falls under non-communicable diseases. It was further noted that much of the emphasis in regards to resource allocation was more geared towards communicable diseases yet there was a growing burden of non-communicable diseases.
Healthcare providers felt there was an absence of healthcare providers in the policymaking arena. They felt that health care representatives needed be present in decision making.
It was suggested that data on hypertension needed to be strengthened. More data was needed to know the number of people dying from hypertension as it is known for Malaria. Data was mentioned to be important because it informs policies needed. Increased research on hypertension was suggested to inform interventions and policies.
## Proposed policy-level solutions
At the policy level, several strategies were suggested by the study participants that would help with blood pressure control. Respondents emphasized repeatedly that the government needed to provide hypertension medications and tests at no cost. Respondents felt that patients with hypertension needed to be accorded the same benefits as patients with HIV and TB and be given medications and testing at no cost.
Study respondents suggested that medications for hypertension should not be taxed in order to make them more affordable for patients, leading to better adherence. Additionally strategies suggested that donors who can support the provision of hypertension drugs should be approached. Some hypertensive patients felt they needed monetary support from the government so that they could buy their medications and foods in line with hypertension care. Another strategy suggested to help with reducing the medication burden was the implementation of universal health coverage.
In regards to guidelines, there were suggestions that it should be available at all facilities and the guidelines needed to be the most updated versions. It was further suggested that there should be policies for non-communicable diseases as they are for communicable diseases.
Healthcare providers suggested that there be more awareness and advocacy activities around NCDs in general hopefully to garner the same attention that HIV has. Policy and decision makers noted more effort needs to be put on non-communicable diseases as has been put on communicable diseases.
## Discussion
This study used the SEM framework [31] to assess the facilitators, barriers and solutions to blood pressure control among participants with uncontrolled hypertension and comorbidities in two Nairobi slums. This study provides key insights collated from patients’, healthcare providers’ and policy/decision makers’ perspective on the facilitators and barriers encountered by people with uncontrolled hypertension in the slums and solutions to identified barriers. The application of the SEM framework to analyse the data collected demonstrates that there is a need to intervene at multiple levels of the SEM framework.
Our findings reveal that access to medication is a major barrier to blood pressure control among patients with uncontrolled hypertension and comorbidities in Korogocho and Viwandani. High prices and the poor socio-economic capabilities of the urban slum residents in this study have limited access to treatment thus affecting compliance to hypertension medications. High cost of medicines has been confirmed in slums across Africa and other low-and-middle income countries [32,33]. While access to medicines is a barrier to blood pressure control in some low-middle-income countries, it seems other countries have started embracing universal health coverage thus providing healthcare and medications for free. For instance a study conducted in Eritrea looking at barriers and facilitators of hypertension management reported that patients appreciated their governments support in providing free medication for hypertension thus improving adherence to medication [10]. A study in Malaysia also found that patients with hypertension had no problem with accessing medications because they were provided free of charge at public facilities [34].
At the individual level, adherence to medication regimens by patients is also affected by the regular unavailability of drugs in the facilities that are expected to provide them for free. Adherence was further worsened by the cost of the medicines thus patients were buying inadequate doses and missing or skipping doses due to inability to buy the medicines. This was evident in a previous study conducted among low income earners in five regions in Kenya [35]. The study found that $38\%$ of households forwent healthcare needs due to lack of money and at times bought less than their required treatment regimen. While compliance to medications is important in hypertension management, in this study compliance is affected due to inaccessibility of medications in the community. Unavailability of medication and the cost of medications were barriers mentioned by the majority of study participants and this adversely affected blood pressure control and adherence to treatment. One of the sustainable development goals’ (SDG) target, “access to quality essential health-care services and access to safe, effective, quality and affordable essential medicines and vaccines for all” [36] is not being met in these two communities.
Other barriers at the patient level were poverty, lack of formal employment, lack of information and knowledge, misconceptions, not having medical cover, being older in age and depending on others for hypertension care. These barriers are similar to those reported in similar settings across sub-Saharan Africa [10,37]. Knowledge about hypertension plays an important role in blood pressure control. In the current study patients’ knowledge was limited and the study participants with uncontrolled hypertension did not seem to understand the rationale for the lifelong treatment of hypertension. Previous research has shown lack of knowledge coupled with misperceptions about the disease can affect adherence to treatment [38,39]. A study by Meinema et al. [ 39] conducted among African Surinamese and Ghanaians with uncontrolled hypertension in the Netherlands showed that using a culturally adapted hypertension education program gave the patients a better understanding of hypertension and improved their understanding about the chronic nature of hypertension thus improving medication adherence.
In the current study, all the patients were older adults with other comorbidities in addition to having uncontrolled hypertension. Comorbidities can have an effect on blood pressure control, and this may explain the inadequacy of blood pressure control in this population. Studies in high income countries have shown that people with comorbidities have higher risks for uncontrolled hypertension [40,41]. A review of the literature estimated that more than $50\%$ of the older adults have multimorbidity and the prevalence of multimorbidity increases with age [42]. A recent study conducted in the current patient population also found that close to a third ($28.7\%$) of the study participants had multimorbidity (defined as two or more chronic conditions) and the commonest identified chronic conditions were hypertension and obesity in this population [43]. There is also literature from similar settings showing that hypertension co-exists with other comorbidities [16,44–46]. Excess weight is a known risk factor for high blood pressure. An earlier study also in the same study population suggested lifestyle changes in this community has led to a rise in overweight and or obesity [47]. The prevalence of overweight/obesity coupled with hypertension may explain the inadequacy of the blood pressure control. A recent study in China has found a positive association between BMI and blood pressure [48]. Therefore, in addition to clinical management of patient comorbidities, this population will require lifestyle changes to address modifiable risk factors to enable them to get their blood pressures under control.
Staffing for healthcare providers was noted to be low in this study. Previous research has shown that increased health care personnel staffing has a positive effect on health outcomes [49,50]. However, most low- and middle-income countries including Kenya lack the needed number of healthcare personnel to provide essential services. Likely reasons for this shortfall include migration of health personnel in search of greener pastures and the reduced capacity of countries or institutional bodies graduating people with the needed qualification for healthcare. It has been estimated that countries with fewer than 23 physicians, nurses and midwives per 10,000 population generally fail to achieve adequate coverage rates for selected primary health-care interventions [51]. The pervasive lack of skilled care is likely the reason for reduced communication among physicians with patients as has been reported in the current study. Good communication between patients and their physicians has many benefits. Benefits include improved compliance to prescribed treatments [52]. A recent article by Zulman and colleagues [53] looking at practices to foster patient physician connection in clinical encounters pointed out how impersonal patient and physician encounters have become.
A health service provision assessment conducted in the above two slums about a decade ago revealed that the majority of public health facilities did not have the required staff, equipment, drugs or the mandate to handle chronic diseases [13]. These barriers continue to persist in 2020 and as a result, the majority of healthcare visits continue to occur in private facilities which are also typical in the slum areas. Drug and equipment stock out in public facilities force the poor urban populations to visit private pharmacies for care thus increasing their out of pocket expenditure [35]. A study conducted in a rural part of Western Kenya also found that lack of drugs at the facilities patients visited was among the health system barriers to blood pressure control [37]. An earlier study by Buigut et al [54] in the study area showed that seeking care in a public health facility was associated with increased odds of experiencing catastrophic health expenditure and for this reason many informal slum residents would forgo health service utilization. A more recent collaborative study looking to improve slum health in the study area also found that many households in the study area spent a significant proportion of their money on healthcare [55].
Out of pocket payments for health services and lack of health insurance coverage was identified in this study as a barrier to accessing health care coverage. A study conducted in the above two slums in 2012 revealed nearly $90\%$ of the slum residents did not have access to any type of health insurance [56]. A more recent study in 2018 in the *Viwandani slum* revealed only $43\%$ of the sampled population had health insurance [57]. While this is an improvement from the study conducted in 2012 by Kimani and colleagues, the coverage is still very low. A study by The Improving Health in Slums Collaborative [58] examining inequalities of healthcare need, access, use and expenditure within slums including the current study areas revealed that there is a very high degree of inequality of household budgets in slums and that this translates into inequities in the access to and use of healthcare services in slums.
Decision making requires reliable information. Hypertension and NCDs in general have not received the same type of support in funding as communicable diseases as evidenced by global initiatives such as the Global Fund to Fight AIDS, Tuberculosis and Malaria (GFATM) and the United States President’s Emergency Plan for AIDS Relief (PEPFAR) which have significant resources and good reporting systems. In this study, participants reported lack of and limited mortality data on hypertension. This type of information is readily available for other conditions such as Malaria, Tuberculosis and HIV due to the funding attached to these programs. Funding of research and good information systems are therefore important in hypertension care in order to provide early warning, basis for planning, analysis of health data among other functions [36].
Hypertension has been identified to be a major contributor [59] to the observed rising deaths due to NCDs in Kenya. These deaths have risen from $35\%$ to $45\%$ in a span of seven years (2003 to 2010) [60]. Results from the current study showed that there were no clear guidelines for hypertension care in some of the facilities which is likely to contribute to some of the above deaths. It also emerged that even though the guidelines are disseminated to all levels of the health facilities, access to medication is limited to facilities in the higher levels thus also contributing to the gaps in access in medication observed in the two slum communities.
This study showed that majority of patients lacked health insurance and this has been previously described in the 2013 Kenya budget and utilisation survey. The survey reported that approximately $83\%$ of the Kenyan population lack financial protection from health care costs and about 1.5 million Kenyans are pushed into poverty each year as a result of paying for health care [61]. The $\frac{2015}{16}$ Kenya Integrated Household Budget Survey revealed $19\%$ of the Kenyan population had some form of health insurance which is a slight improvement from the 2013 estimate [62]. Even though the provision of public health care in *Kenya is* subsidized, it is inadequate due to the dense population in urban areas. Furthermore, the public health care system suffers from inadequate infrastructure and workforce, long queues, and shortage of drugs [63]. In 2018, the Government of Kenya committed to achieving Universal Health Coverage (UHC) by the year 2022. This is a bold initiative and a major step in the right direction for many Kenyans who lack financial protection. The National Hospital Insurance Fund (NHIF) is the main insurance scheme in Kenya and it is expected to improve the provision of healthcare services in Kenya. The scheme covers both the formal and the informal sector. Coverage is high in the formal sector due to the mandatory nature of contributions from employers while the coverage from the informal sector is very low due to the voluntary nature of contributions. The majority of the study respondents in this study have strongly articulated the need to have free healthcare and the full implementation of the UHC in Kenya can make this a reality. Otherwise, the provision of care will be inequitable and more biased towards those who can afford the premium contributions.
## Strengths and limitations
To our knowledge, this is one of the first studies to ask healthcare providers, policymakers and patients with uncontrolled hypertension and comorbidities in an urban slum setting to explore their experiences and views about facilitators, barriers and solutions to hypertension management at the different levels of the SEM framework. This study captured an integrated and diverse range of perspectives on facilitators and barriers to blood pressure control in slum communities by purposively engaging with patients with uncontrolled hypertension, health care providers and policy/decision makers. Examining uncontrolled hypertension through the socio-ecological model has increased our understanding of how to tackle blood pressure control while highlighting potential strategies at the different levels of the SEM. However, some limitations should be noted. The results from this study are from two *Nairobi slum* communities and even though it may be applicable to other similar slum settings; it may not be generalizable beyond slum communities. While a strength of this study is that views from different study participants were sought for each of the SEM levels including the families/community members who did not participate in the study, it is thus possible that this particular groups’ perspectives may not have been accurately captured by the current study respondents. Another limitation to consider is the timing of data collection which corresponded with the current pandemic spread of SARS-CoV-2 which posed disruptions in the delivery of health care services. Thus it may be that some of the patients’ challenges discussed could have been a result of the measures put in place to curb further spread of the virus. Similarly, the switch to phone interviews for most of the interviews rather than the traditional face-to-face interviews due to the pandemic may have reduced or limited the level of detail needed to capture the non-verbal cues that are important in guiding further discussions. Nonetheless, findings from this study can help inform efforts to develop multi-level interventions to improve hypertension control among similar urban slum residents.
## Recommendations
At the patient level, barriers affecting patients’ access to hypertension medication need to be removed and or alleviated through the provision of free medications or subsidized medicines. Also, more frequent educational sessions should be conducted with patients so that they are well informed about their conditions and what they need to do to control their blood pressure. At the community level, hypertension care awareness is critical in ensuring a good understanding among the community and family members on hypertension care. Approaches at this level should also consider more involvement of community health workers/volunteers. At the health system level, approaches should focus on improvements at various levels within the health system structures such as; human resources, health management, health systems and governance. Lastly, at the policy level there is need for policies and directives that ensure equitable care is received by all including those in the slum communities or those seeking care at lower level health facilities. To address this, policy makers could consider expanding the healthcare mandate at lower level facilities by extending treatment for hypertension care in lower level facilities.
## Conclusion
This study presents the findings from a qualitative study of multiple levels of factors associated with uncontrolled hypertension. The findings demonstrate that uncontrolled hypertension is a major public health issue in slums of Nairobi and it is associated with barriers at different levels of the socio-ecological framework. The findings from the present study can be used to design interventions to address the interplay of factors operating at multiple levels of the SEM, from the patient level all the way to the policy level. Importantly there is a need for policies that facilitate equitable care in slums through increased access to subsidized or free medication.
## References
1. Stanaway JD, Afshin A, Gakidou E, Lim SS, Abate D, Abate KH. **Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *The Lancet* (2018.0) **392** 1923-94. DOI: 10.1016/S0140-6736(18)32225-6
2. Roth GA, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N. **Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980–2017: a systematic analysis for the Global Burden of Disease Study 2017**. *The Lancet* (2018.0) **392** 1736-88. DOI: 10.1016/S0140-6736(18)32203-7
3. Zhou B, Carrillo-Larco RM, Danaei G, Riley LM, Paciorek CJ, Stevens GA. **Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants**. *The Lancet* (2021.0) **398** 957-80. DOI: 10.1016/S0140-6736(21)01330-1
4. Mills KT, Bundy JD, Kelly TN, Reed JE, Kearney PM, Reynolds K. **Global Disparities of Hypertension Prevalence and Control: A Systematic Analysis of Population-Based Studies From 90 Countries.**. *Circulation* (2016.0) **134** 441-50. DOI: 10.1161/CIRCULATIONAHA.115.018912
5. 5World Health Organization. Global status report on noncommunicable diseases 2014: World Health Organization; 2014.. *Global status report on noncommunicable diseases 2014* (2014.0)
6. Poulter N, Khaw K, Hopwood B, Mugambi M, Peart W, Sever P. **Determinants of blood pressure changes due to urbanization: a longitudinal study. Journal of hypertension Supplement: official journal of the**. *International Society of Hypertension* (1985.0) **3** S375-7
7. 768% of the world population projected to live in urban areas by 2050, says UN [Internet]. 2018 [cited March, 2019]. Available from: https://www.un.org/development/desa/en/news/population/2018-revision-of-world-urbanization-prospects.html.
8. Ezeh A, Oyebode O, Satterthwaite D, Chen Y-F, Ndugwa R, Sartori J. **The history, geography, and sociology of slums and the health problems of people who live in slums**. *The lancet* (2017.0) **389** 547-58. DOI: 10.1016/S0140-6736(16)31650-6
9. Uthman OA, Ayorinde A, Oyebode O, Sartori J, Gill P, Lilford RJ. **Global prevalence and trends in hypertension and type 2 diabetes mellitus among slum residents: a systematic review and meta-analysis.**. *BMJ Open.* (2022.0) **12** e052393. DOI: 10.1136/bmjopen-2021-052393
10. Gebrezgi MT, Trepka MJ, Kidane EA. **Barriers to and facilitators of hypertension management in Asmara, Eritrea: patients’ perspectives.**. *Journal of Health, Population, and Nutrition,.* (2017.0) **36** 11. DOI: 10.1186/s41043-017-0090-4
11. Flynn SJ, Ameling JM, Hill-Briggs F, Wolff JL, Bone LR, Levine DM. **Facilitators and barriers to hypertension self-management in urban African Americans: perspectives of patients and family members**. *Patient Preferrence and Adherence* (2013.0) **7** 741-9. DOI: 10.2147/PPA.S46517
12. Mendis S, Al Bashir I, Dissanayake L, Varghese C, Fadhil I, Marhe E. **Gaps in capacity in primary care in low-resource settings for implementation of essential noncommunicable disease interventions**. *International Journal of Hypertension* (2012.0) 2012. DOI: 10.1155/2012/584041
13. Kyobutungi C, Ezeh A, Vlahov D, Boufford J, Pearson C, Norris L. **Chronic disease care in Nairobi’s urban informal settlements.**. *Urban Health: Global Perspectives,.* (2010.0) 139-56
14. Hulzebosch A, van de Vijver S, Oti SO, Egondi T, Kyobutungi C. **Profile of people with hypertension in Nairobi’s slums: a descriptive study.**. *Globalization and health* (2015.0) **11** 26. DOI: 10.1186/s12992-015-0112-1
15. Olack B, Wabwire-Mangen F, Smeeth L, Montgomery JM, Kiwanuka N, Breiman RF. **Risk factors of hypertension among adults aged 35–64 years living in an urban slum Nairobi, Kenya.**. *BMC public health.* (2015.0) **15** 1251. DOI: 10.1186/s12889-015-2610-8
16. Joshi MD, Ayah R, Njau EK, Wanjiru R, Kayima JK, Njeru EK. **Prevalence of hypertension and associated cardiovascular risk factors in an urban slum in Nairobi, Kenya: a population-based survey.**. *BMC public health.* (2014.0) **14** 1. PMID: 24383435
17. Van de Vijver SJ, Oti SO, Agyemang C, Gomez GB, Kyobutungi C. **Prevalence, awareness, treatment and control of hypertension among slum dwellers in Nairobi, Kenya**. *Journal of Hypertension* (2013.0) **31** 1018-24. DOI: 10.1097/HJH.0b013e32835e3a56
18. 18Centers for Disease Control Prevention. The social-ecological model: A framework for prevention 2015.
19. Emina J, Beguy D, Zulu EM, Ezeh AC, Muindi K, Elung’ata P. **Monitoring of health and demographic outcomes in poor urban settlements: evidence from the Nairobi Urban Health and Demographic Surveillance System**. *Journal of Urban Health* (2011.0) **88** 200-18. DOI: 10.1007/s11524-011-9594-1
20. Beguy D, Elung’ata P, Mberu B, Oduor C, Wamukoya M, Nganyi B. **HDSS Profile: the Nairobi Urban Health and Demographic Surveillance System (NUHDSS).**. *International Journal of Epidemiology* (2015.0)
21. Ali SA, Soo C, Agongo G, Alberts M, Amenga-Etego L, Boua RP. **Genomic and environmental risk factors for cardiometabolic diseases in Africa: methods used for Phase 1 of the AWI-Gen population cross-sectional study.**. *Global health action.* (2018.0) **11** 1507133. DOI: 10.1080/16549716.2018.1507133
22. Petty NJ, Thomson OP, Stew G. **Ready for a paradigm shift? Part 2: Introducing qualitative research methodologies and methods.**. *Manual Therapy* (2012.0) **17** 378-84. DOI: 10.1016/j.math.2012.03.004
23. Lester S.. *An introduction to phenomenological research* (1999.0)
24. Gill P, Stewart K, Treasure E, Chadwick B. **Methods of data collection in qualitative research: interviews and focus groups**. *British Dental Journal* (2008.0) **204** 291. DOI: 10.1038/bdj.2008.192
25. Kvale S.. *Methods of analysis.* (1996.0)
26. Kitzinger J.. **Focus group research: using group dynamics**. *Qualitative research in health care* (2005.0) **56** 70
27. Petty NJ, Thomson OP, Stew G. **Ready for a paradigm shift? Part 1: Introducing the philosophy of qualitative research.**. *Manual therapy.* (2012.0) **17** 267-74. DOI: 10.1016/j.math.2012.03.006
28. Sebele-Mpofu FY. **Saturation controversy in qualitative research: Complexities and underlying assumptions. A literature revie**. *Cogent Social Sciences* (2020.0) **6** 1838706. DOI: 10.1080/23311886.2020.1838706
29. Boyatzis RE. *Transforming qualitative information: Thematic analysis and code development* (1998.0)
30. Braun V, Clarke V. **Using thematic analysis in psychology.**. *Qualitative Research in Psychology* (2006.0) **3** 77-101. DOI: 10.1191/1478088706qp063oa
31. 31Centers for Disease Control Prevention. The social-ecological model: A framework for prevention. Atlanta, GA: CDC Retrieved from https://wwwcdcgov/violenceprevention/overview/social-ecologicalmodelhtml. 2015.. *The social-ecological model: A framework for prevention* (2015.0)
32. Cameron A, Ewen M, Ross-Degnan D, Ball D, Laing R. **Medicine prices, availability, and affordability in 36 developing and middle-income countries: a secondary analysis**. *Lancet* (2009.0) **373** 240-9. DOI: 10.1016/S0140-6736(08)61762-6
33. Ahmed SAKS, Ajisola M, Azeem K, Bakibinga P, Chen Y-F, Choudhury NN. **Impact of the societal response to COVID-19 on access to healthcare for non-COVID-19 health issues in slum communities of Bangladesh, Kenya, Nigeria and Pakistan: results of pre-COVID and COVID-19 lockdown stakeholder engagements**. *BMJ Global Health* (2020.0) **5** e003042. DOI: 10.1136/bmjgh-2020-003042
34. Tan CS, Hassali MA, Neoh CF, Saleem F. **A qualitative exploration of hypertensive patients’ perception towards quality use of medication and hypertension management at the community level**. *Pharmacy practice* (2017.0) **15** 1074. DOI: 10.18549/PharmPract.2017.04.1074
35. Zollmann Julie, Ravishankar Nirmala. *Struggling to thrive: How Kenya’s low-income families (try to) pay for healthcare.* (2016.0)
36. 36World Health Organization. Monitoring the building blocks of health systems: a handbook of indicators and their measurement strategies: World Health Organization; 2010.. *Monitoring the building blocks of health systems: a handbook of indicators and their measurement strategies* (2010.0)
37. Naanyu V, Vedanthan R, Kamano JH, Rotich JK, Lagat KK, Kiptoo P. **Barriers Influencing Linkage to Hypertension Care in Kenya: Qualitative Analysis from the LARK Hypertension Study.**. *J Gen Intern Med* (2016.0) **31** 304-14. DOI: 10.1007/s11606-015-3566-1
38. Rajpura J, Nayak R. **Medication adherence in a sample of elderly suffering from hypertension: evaluating the influence of illness perceptions, treatment beliefs, and illness burden.**. *Journal of Managed Care Pharmacy.* (2014.0) **20** 58-65. DOI: 10.18553/jmcp.2014.20.1.58
39. Meinema JG, van Dijk N, Beune EJ, Jaarsma DA, van Weert HC, Haafkens JA. **Determinants of adherence to treatment in hypertensive patients of African descent and the role of culturally appropriate education.**. *PLoS One.* (2015.0) **10** e0133560. DOI: 10.1371/journal.pone.0133560
40. Degli Esposti E, Di Martino M, Sturani A, Russo P, Dradi C, Falcinelli S. **Risk factors for uncontrolled hypertension in Italy.**. *Journal of Human Hypertension* (2004.0) **18** 207. DOI: 10.1038/sj.jhh.1001656
41. Liu X, Song P. **Is the Association of Diabetes With Uncontrolled Blood Pressure Stronger in Mexican Americans and Blacks Than in Whites Among Diagnosed Hypertensive Patients**. *American Journal of Hypertension* (2013.0) **26** 1328-34. DOI: 10.1093/ajh/hpt109
42. Marengoni A, Angleman S, Melis R, Mangialasche F, Karp A, Garmen A. **Aging with multimorbidity: a systematic review of the literature.**. *Ageing Research Reviews* (2011.0) **10** 430-9. DOI: 10.1016/j.arr.2011.03.003
43. Mohamed SF, Haregu TN, Uthman OA, Khayeka-Wandabwa C, Muthuri SK, Asiki G. **Multimorbidity from Chronic Conditions among Adults in Urban Slums: The AWI-Gen Nairobi Site Study Findings.**. *Global Heart* (2021.0) **16** 6. DOI: 10.5334/gh.771
44. Hendriks ME, Wit FW, Roos MT, Brewster LM, Akande TM, de Beer IH. **Hypertension in sub-Saharan Africa: cross-sectional surveys in four rural and urban communities**. *PloS one* (2012.0) **7** e32638. DOI: 10.1371/journal.pone.0032638
45. Jenson A, Omar AL, Omar MA, Rishad AS, Khoshnood K. **Assessment of hypertension control in a district of Mombasa, Kenya.**. *Global Public Health.* (2011.0) **6** 293-306. DOI: 10.1080/17441692.2010.510478
46. Mathenge W, Foster A, Kuper H. **Urbanization, ethnicity and cardiovascular risk in a population in transition in Nakuru, Kenya: a population-based survey.**. *BMC public health* (2010.0) **10** 569. DOI: 10.1186/1471-2458-10-569
47. Ziraba AK, Fotso JC, Ochako R. **Overweight and obesity in urban Africa: A problem of the rich or the poor?**. *BMC public health.* (2009.0) **9** 465. DOI: 10.1186/1471-2458-9-465
48. Linderman GC, Lu J, Lu Y, Sun X, Xu W, Nasir K. **Association of Body Mass Index With Blood Pressure Among 1.7 Million Chinese Adults.**. *JAMA Network Open.* (2018.0) **1** e181271. DOI: 10.1001/jamanetworkopen.2018.1271
49. Anand S, Bärnighausen T. **Health workers and vaccination coverage in developing countries: an econometric analysis**. *The Lancet* (2007.0) **369** 1277-85. DOI: 10.1016/S0140-6736(07)60599-6
50. Speybroeck N, Kinfu Y, Dal Poz MR, Evans DB. *Reassessing the relationship between human resources for health, intervention coverage and health outcomes.* (2006.0)
51. 51World Health Organization. The world health report 2006: working together for health: World Health Organization; 2006.. *The world health report 2006: working together for health* (2006.0)
52. Harmon G, Lefante J, Krousel-Wood M. **Overcoming barriers: the role of providers in improving patient adherence to antihypertensive medications.**. *Current Opinion in Cardiology* (2006.0) **21** 310-5. DOI: 10.1097/01.hco.0000231400.10104.e2
53. Zulman DM, Haverfield MC, Shaw JG, Brown-Johnson CG, Schwartz R, Tierney AA. **Practices to Foster Physician Presence and Connection With Patients in the Clinical Encounter**. *JAMA* (2020.0) **323** 70-81. DOI: 10.1001/jama.2019.19003
54. Buigut S, Ettarh R, Amendah DD. **Catastrophic health expenditure and its determinants in Kenya slum communities**. *International Journal for Equity in Health* (2015.0) **14** 46. DOI: 10.1186/s12939-015-0168-9
55. **Primary care doctor and nurse consultations among people who live in slums: a retrospective, cross-sectional survey in four countries**. *BMJ Open* (2022.0) **12** e054142. DOI: 10.1136/bmjopen-2021-054142
56. Kimani JK, Ettarh R, Kyobutungi C, Mberu B, Muindi K. **Determinants for participation in a public health insurance program among residents of urban slums in Nairobi, Kenya: results from a cross-sectional survey.**. *BMC Health Services Research* (2012.0) **12** 1-11. DOI: 10.1186/1472-6963-12-66
57. Otieno PO, Wambiya EOA, Mohamed SF, Donfouet HPP, Mutua MK. **Prevalence and factors associated with health insurance coverage in resource-poor urban settings in Nairobi, Kenya: a cross-sectional study**. *BMJ Open* (2019.0) **9** e031543. DOI: 10.1136/bmjopen-2019-031543
58. **Inequity of healthcare access and use and catastrophic health spending in slum communities: a retrospective, cross-sectional survey in four countries.**. *BMJ Global Health* (2021.0) **6** e007265. DOI: 10.1136/bmjgh-2021-007265
59. 59Institute for Health Metrics and Evaluation and the International Centre for Humanitarian Affairs. The Global Burden of Disease: Generating Evidence, Guiding Policy in Kenya. Nairobi, Kenya: 2016.. *Institute for Health Metrics and Evaluation and the International Centre for Humanitarian Affairs. The Global Burden of Disease: Generating Evidence* 2016
60. Phillips-Howard PA, Laserson KF, Amek N, Beynon CM, Angell SY, Khagayi S. **Deaths Ascribed to Non-Communicable Diseases among Rural Kenyan Adults Are Proportionately Increasing: Evidence from a Health and Demographic Surveillance System, 2003–2010.**. *PloS one.* (2014.0) **9** e114010. DOI: 10.1371/journal.pone.0114010
61. 61Ministry of Health. 2013 Kenya Household Health Expenditure and Utilisation Survey. Nairobi: Government of Kenya; 2014.. *2013 Kenya Household Health Expenditure and Utilisation Survey* (2014.0)
62. 62Kenya National Bureau of Statistic (KNBS). 2015/16 Kenya Integrated Household Budget Survey (KIHBS). 2018 March 2018. Report No.
63. Muriithi MK. **The determinants of health-seeking behavior in a Nairobi slum**. *Kenya. European Scientific Journal* (2013.0) **9**
|
---
title: 'Prevalence, determinants, and association of overweight/obesity with non-communicable
disease-related biomedical indicators: A cross-sectional study in schoolteachers
in Kabul, Afghanistan'
authors:
- Sharifullah Alemi
- Keiko Nakamura
- Ahmad Shekib Arab
- Mohammad Omar Mashal
- Yuri Tashiro
- Kaoruko Seino
- Shafiqullah Hemat
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021827
doi: 10.1371/journal.pgph.0001676
license: CC BY 4.0
---
# Prevalence, determinants, and association of overweight/obesity with non-communicable disease-related biomedical indicators: A cross-sectional study in schoolteachers in Kabul, Afghanistan
## Abstract
Overweight/obesity constitutes a major risk factor for non-communicable diseases (NCDs), whose global prevalence is growing rapidly, including in Afghanistan. However, the effects of risk factors on NCDs have rarely been studied in the educator workforce. Therefore, the objective of this study is to determine the prevalence, determinants, and association of overweight/obesity with NCD-related biomedical indicators among schoolteachers in Afghanistan. The sample comprised 600 schoolteachers aged 18 years and above. We conducted questionnaire interviews, anthropometric measurements, and blood biochemistry tests. The main explanatory variable was overweight/obesity (body mass index ≥ 25.0 kg/m2). NCD-related biomedical indicators were the outcome variables. Poisson regression models were applied to investigate the association between overweight/obesity and outcome variables. The prevalence of overweight/obesity was $58.2\%$, which was significantly higher in women, those aged 41–50 years, married participants, and those with 10–20 years of working experience than in their counterparts. After adjusting for sociodemographic variables and lifestyle behaviors, overweight/obesity was significantly associated with hypertension (adjusted prevalence ratio [aPR] = 1.83, $95\%$ confidence interval [CI]: 1.33–2.51); elevated levels of glycosylated hemoglobin (HbA1c) (aPR = 1.35, $95\%$ CI: 1.01–1.79), total cholesterol (aPR = 1.67, $95\%$ CI:1.14–2.44), low-density lipoprotein cholesterol (LDL-C) (aPR = 1.29, $95\%$ CI: 1.10–1.50), and triglycerides (aPR = 1.98, $95\%$ CI: 1.57–2.50), and having three or more comorbidities (aPR = 1.90, $95\%$ CI: 1.47–2.47). Our findings demonstrated a high prevalence of overweight/obesity among schoolteachers. In addition, we found significant associations of overweight/obesity with a higher prevalence of hypertension; elevated serum levels of HbA1c, total cholesterol, LDL-C, and triglycerides; and comorbid conditions in schoolteachers. The findings highlight the need for worksite interventions that promote weight control among schoolteachers with overweight/obesity to reduce the burden of NCDs.
## Introduction
Overweight and obesity are growing public health concerns accounting for at least 4.7 million deaths globally in 2017 [1]. Despite being traditionally considered a concern of high-income countries, overweight and obesity rates among adults have continued to increase in low- and middle-income countries [2]. According to the recent non-communicable disease (NCD) risk factors survey in Afghanistan, the prevalence of hypertension, overweight, obesity, and elevated levels of fasting blood glucose and total cholesterol was $23.5\%$, $25.8\%$, $17.0\%$, $9.2\%$, and $16.9\%$, respectively [3]. Previous research analyzing data from a national survey in Afghanistan found that higher age (30 years and over), hypertension, and type 2 diabetes mellitus were among factors positively associated with overweight/obesity [4]. Individuals with overweight/obesity have a greater risk of developing adverse health outcomes, including diabetes mellitus, cardiovascular diseases, cancer, hypertension, and dyslipidemia [5–7]. However, there is a limited understanding of the association between overweight/obesity and NCD-related biomedical indicators. Among the risk factors for NCDs, overweight and obesity are particularly concerning as they potentially reverse many health benefits that would lead to improved life expectancy. The early detection and control of NCD risk factors are regarded as an effective strategy for tackling NCDs. Halting overweight and obesity by promoting healthy lifestyle behaviors, including a balanced diet and regular physical activity, may substantially contribute to achieving the target of reducing NCD-related deaths and disabilities.
Schoolteachers constitute one of the largest and high-risk occupational groups more exposed to the most frequent predictors of overweight and obesity, including poor dietary habits, insufficient physical activity, and spending long working hours on sedentary activities [8–10]. Excess body weight has become more common among employment groups, having negative consequences, including sick leave, more frequent absenteeism and doctor visits, and increased healthcare costs [11–13]. Similarly, participants with overweight/obesity are at risk of functional impairment, early retirement, and reduced health-related quality of life [14–16]. Teachers’ teaching quality and productivity may significantly improve when they are healthy, thus having a beneficial impact on students’ learning outcomes [17, 18]. Addressing NCDs and their risk factors in schoolteachers contributes to their health outcomes and schoolchildren’s educational development and learning outcomes [19, 20]. Thus, investigating health risks among the occupational group of schoolteachers and supporting them to adopt healthy lifestyle behaviors is vital in public health research. Given that schoolteachers spend approximately half of their waking time at school, the school environment and teaching conditions should be reformed to promote and reinforce healthy lifestyle behaviors, such as focusing on food quality, physical activity facilities, and health literacy, which contribute to the prevention of weight gain.
The transition to more urban life and changes in lifestyle behaviors, such as consuming more energy-dense diets and foods high in fat and sugars as well as increases in physical inactivity due to sedentary work/life and modern modes of transportation, have all contributed to increased body weight [21]. Although some research has been carried out on the adverse health outcomes of overweight/obesity among individuals, little is known about the influence of overweight/obesity on increasing the risk of NCDs among schoolteachers. In addition, generating evidence on NCD risk factors among schoolteachers is an important step in designing and developing school-based interventions for the prevention and control of NCDs. This study builds on previous research carried out on the adverse health outcomes of overweight/obesity but differs in that it is the first study to investigate the association between overweight/obesity and a wide range of objectively measured NCD-related biomedical indicators among an important, but rarely-studied occupation group of schoolteachers. Therefore, our study aimed to determine the prevalence, determinants, and association of overweight/obesity with NCD-related biomedical indicators among schoolteachers in Afghanistan. This study highlights the burden of overweight/obesity as a risk factor for NCD development among adults in Afghanistan. The findings should encourage schoolteachers to modify their lifestyle behaviors to prevent overweight and obesity and reduce the burden of NCDs.
## Study design and setting
This was a cross-sectional study conducted in February 2017 that involved 600 schoolteachers from 210 primary, middle, and high public schools across all municipal districts in Kabul city. All permanent male and female schoolteachers were eligible for recruitment. Schoolteachers with short-term contracts and those hired for teachers’ training programs were excluded. Based on the formal invitation letter from the Ministry of Education in partnership with the Ministry of Public Health, principals of individual schools were requested to select and introduce one to four schoolteachers who met the eligibility criteria. School principals selected eligible schoolteachers from the list and introduced them to participate in the study. The ratio of male to female schoolteachers in our study is comparable to the sex ratio of schoolteachers in Kabul city. The sample size calculation is described in detail elsewhere [22].
## Data collection and measurements
Data were collected in three phases. First, a face-to-face interview was conducted using a questionnaire that included questions about participants’ sociodemographic characteristics, health status, medication history, lifestyle behaviors, and NCD-related knowledge. Second, trained male and female medical staff conducted anthropometric measurements, including height, weight, and blood pressure (BP) measurements. Height (cm) was measured using a stadiometer and weight (kg) using a calibrated weighing scale. Body mass index (BMI) was calculated as follows: weight (kg)/height squared (m2). OMRON monitors (OMRON Healthcare, Kyoto, Japan) were used for BP measurement. The average of two different systolic and diastolic BP readings measured at 3–5-minute intervals was used. After blood pressure measurement, blood samples were drawn by laboratory technicians from all the participants for the blood biochemistry tests, which included measurements of glycosylated hemoglobin (HbA1c), total cholesterol, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and triglyceride levels. HbA1c was measured using a fully automated HbA1c analyzer (Clover A1c), and lipid measurements were performed using a Micro-lab 300 semi-automated clinical chemistry analyzer.
## Study variables
The main explanatory variable was overweight/obesity, defined as a BMI ≥ 25.0 kg/m2. Six NCD-related biomedical indicators were considered dependent variables: BP, HbA1c, total cholesterol, LDL-C, HDL-C, and triglycerides. BP was assessed as a dichotomous, categorical variable (<$\frac{130}{85}$ mmHg/≥$\frac{130}{85}$ mmHg). Other binary variables were HbA1c (<$5.5\%$/≥$5.5\%$), total cholesterol (<200 mg/dL/≥200 mg/dL), LDL-C (<100 mg/dL/≥100 mg/dL), HDL-C (≥40 mg/dL/<40 mg/dL), and triglycerides (<150 mg/dL/≥150 mg/dL). Comorbidities included hypertension, elevated HbA1c, high total cholesterol, high LDL-C, low HDL-C, and high triglycerides. The presence of multiple biomedical indicators was categorized into less than three and three or more comorbidities. The cut-off values for normal and elevated blood pressure were in compliance with the categories reported in the 2017 American College of Cardiology/American Heart Association guidelines for the prevention, detection, evaluation, and management of high blood pressure in adults [23]. The cut-off levels for NCD-related biomedical indicators were set according to clinical practice and guidelines, including the Adult Treatment Panel III (ATP-III) guidelines, systematic reviews, and original studies conducted in countries with similar contexts [24–27]. Sociodemographic variables included sex, age, education attainment, marital status, working experience, and monthly income. Lifestyle-behavior variables included physical exercise/walking, fruit/vegetable consumption, and tobacco use.
## Statistical analysis
Data analyses were performed using Stata software (version 15.1; Stata Corp). The chi-squared test was used to compare the characteristics of the weight-status groups. Considering the high prevalence (>$10\%$) of the binary outcome variables, Poisson regression models were employed [28, 29]. We estimated prevalence ratios (PRs) using Poisson regression models with robust variance to identify correlates of overweight/obesity and investigate the effects of sociodemographic variables and lifestyle behaviors on the relationship between overweight/obesity and NCD-related biomedical indicators. The multivariate models were adjusted for sex, age, education attainment, marital status, working experience, monthly income, physical exercise/walking, consumption of fruits/vegetables, and tobacco use. To investigate the effect modification of sex on the association between the explanatory variable and measured outcomes, a sex-stratified analysis was performed. The sex-stratified models were also adjusted for sociodemographic variables and lifestyle behaviors. The statistical assumptions for the Poisson regression model were checked prior to model fitting. The variance inflation factor (VIF) was computed for the set of independent variables, and only variables with VIF less than 5 were included in the model; multicollinearity between the set of included variables was not observed. The deviance goodness of fit and Pearson goodness of fit tests were also performed using the Stata command “estat gof” to assess the overall goodness of fit and adequacy of the Poisson regression model. The test results indicated that the Poisson regression models fit our data well. Statistical significance was set at P ≤ 0.05.
## Ethical considerations
Ethical approval was obtained from the Tokyo Medical and Dental University Research Ethics Committee and the Afghanistan Ministry of Public Health Institutional Review Board. This research complied with the ethical principles set by the Declaration of Helsinki. All participants were provided with information about the study protocol along with written informed consent forms and the right to not participate or withdraw.
## Results
Table 1 shows the sociodemographic characteristics by weight status of study participants. Two-thirds of the participants ($69.3\%$) were women. Most participants were aged between 41 and 50 years and were married. In the total sample ($$n = 600$$), $58.2\%$ were classified as overweight/obese, which was significantly higher among females than in males ($64.7\%$ vs. $43.5\%$). The chi-squared test also revealed that a significantly larger proportion of participants aged 41–50 years old, those currently married, and those with 10–20 years of working experience had overweight/obesity compared to their counterparts.
**Table 1**
| Variables | n (%) | Non-overweight/obese (BMI<25.0 kg/m2) | Overweight/obese (BMI≥25.0 kg/m2) | P |
| --- | --- | --- | --- | --- |
| Variables | n (%) | (n = 251; 41.8%) | (n = 349; 58.2%) | P |
| Sex | | | | |
| Male | 184 (30.7) | 104 (56.5) | 80 (43.5) | <0.001 |
| Female | 416 (69.3) | 147 (35.3) | 269 (64.7) | |
| Age (years) | | | | |
| 18–30 | 142 (23.7) | 90 (63.4) | 52 (36.6) | <0.001 |
| 31–40 | 148 (24.7) | 59 (39.9) | 89 (60.1) | |
| 41–50 | 191 (31.8) | 56 (29.3) | 135 (70.7) | |
| ≥51 | 119 (19.8) | 46 (38.7) | 73 (61.3) | |
| Education attainment | | | | |
| 12th grade (high school) graduate | 32 (5.3) | 12 (37.5) | 20 (62.5) | 0.470 |
| 14th grade (2-year college) graduate | 362 (60.3) | 146 (40.3) | 216 (59.7) | |
| College/university graduate or higher | 206 (34.4) | 93 (45.1) | 113 (54.9) | |
| Marital status | | | | |
| Never married | 122 (20.3) | 70 (57.4) | 52 (42.6) | <0.001 |
| Currently married | 478 (79.7) | 181 (37.9) | 297 (62.1) | |
| Working experience (years) | | | | |
| <10 | 191 (31.8) | 109 (57.1) | 82 (42.9) | <0.001 |
| 10–20 | 197 (32.8) | 68 (34.5) | 129 (65.5) | |
| ≥21 | 212 (35.4) | 74 (34.9) | 138 (65.1) | |
| Monthly income, Afghanis† | | | | |
| ≤10,000 | 190 (31.7) | 91 (47.9) | 99 (52.1) | 0.102 |
| 10,001–20,000 | 272 (45.3) | 109 (40.1) | 163 (59.9) | |
| >20,001 | 138 (23.0) | 51 (37.0) | 87 (63.0) | |
## Prevalence of weight status by NCD-related biomedical indicators and lifestyle behaviors
The prevalence of hypertension, elevated HbA1c, and high triglyceride levels were $25.7\%$, $29.7\%$, and $42.7\%$, respectively. Of the participants, $20.2\%$, $58.7\%$, and $28.8\%$ had high total cholesterol, high LDL-C, and low HDL-C levels, respectively. Overweight/obesity was significantly more prevalent among participants with hypertension; increased serum levels of HbA1c, total cholesterol, LDL-C, and triglycerides; and low HDL-C levels (Table 2).
**Table 2**
| Variables | n (%) | Non-overweight/obese (BMI<25.0 kg/m2) | Overweight/obese (BMI≥25.0 kg/m2) | P |
| --- | --- | --- | --- | --- |
| Variables | n (%) | (n = 251; 41.8%) | (n = 349; 58.2%) | P |
| NCD-related biomedical indicators | | | | |
| Blood pressure (mmHg) | | | | |
| Normal (<130/85) | 446 (74.3) | 213 (47.8) | 233 (52.2) | <0.001 |
| Elevated (≥130/85) | 154 (25.7) | 38 (24.7) | 116 (75.3) | |
| HbA1c (%) | | | | |
| Normal (<5.5) | 422 (70.3) | 196 (46.4) | 226 (53.6) | <0.001 |
| Elevated (≥5.5) | 178 (29.7) | 55 (30.9) | 123 (69.1) | |
| Total cholesterol (mg/dL) | | | | |
| Normal (<200) | 479 (79.8) | 219 (45.7) | 260 (54.3) | <0.001 |
| Elevated (≥200) | 121 (20.2) | 32 (26.5) | 89 (73.5) | |
| LDL-C (mg/dL) | | | | |
| Normal (<100) | 248 (41.3) | 133 (53.6) | 115 (46.4) | <0.001 |
| Elevated (≥100) | 352 (58.7) | 118 (33.5) | 234 (66.5) | |
| HDL-C (mg/dL) | | | | |
| Normal (≥40) | 427 (71.2) | 193 (45.2) | 234 (54.8) | 0.009 |
| Low (<40) | 173 (28.8) | 58 (33.5) | 115 (66.5) | |
| Triglyceride level (mg/dL) | | | | |
| Normal (<150) | 344 (57.3) | 183 (53.2) | 161 (46.8) | <0.001 |
| Elevated (≥150) | 256 (42.7) | 68 (26.6) | 188 (73.4) | |
| Lifestyle behaviors | | | | |
| Physical exercise/walking (per day) | | | | |
| <1 hour | 330 (55.0) | 132 (40.0) | 198 (60.0) | 0.314 |
| ≥1 hour | 270 (45.0) | 119 (44.1) | 151 (55.9) | |
| Consumption of fruits/vegetables (per week) | | | | |
| <4 times | 204 (34.0) | 93 (45.6) | 111 (54.4) | 0.181 |
| ≥4 times | 396 (66.0) | 158 (39.9) | 238 (60.1) | |
| Tobacco use | | | | |
| No | 568 (94.7) | 233 (41.0) | 335 (59.0) | 0.089 |
| Yes | 32 (5.3) | 18 (56.2) | 14 (43.8) | |
## Overweight/obesity and sociodemographic and lifestyle-behaviors
After adjusting for sociodemographic and lifestyle-behavior factors, sex, age, and marital status remained significant correlates of overweight/obesity. Female sex, 31 years of age or over, and being married increased the likelihood of being overweight/obese by 1.48, 1.51, and 1.29 times, respectively (Table 3).
**Table 3**
| Variables | Crude prevalence ratio (95% CI) | P | Adjusted prevalence ratio (95% CI) | P.1 |
| --- | --- | --- | --- | --- |
| Sex | | | | |
| Male | 1.00 | | 1.00 | |
| Female | 1.49 (1.24–1.78) | <0.001 | 1.48 (1.22–1.81) | <0.001 |
| Age (years) | | | | |
| 18–30 | 1.00 | | 1.00 | |
| 31–40 | 1.64 (1.27–2.11) | <0.001 | 1.51 (1.13–2.02) | 0.005 |
| 41–50 | 1.93 (1.52–2.44) | <0.001 | 1.58 (1.16–2.15) | 0.003 |
| ≥51 | 1.67 (1.29–2.17) | <0.001 | 1.51 (1.08–2.11) | 0.015 |
| Education attainment | | | | |
| 12th grade (high school) graduate | 1.00 | | 1.00 | |
| 14th grade (2-year college) graduate | 0.95 (0.72–1.26) | 0.747 | 0.89 (0.69–1.16) | 0.402 |
| College/university graduate or higher | 0.88 (0.65–1.18) | 0.387 | 0.84 (0.64–1.11) | 0.230 |
| Marital status | | | | |
| Never married | 1.00 | | 1.00 | |
| Currently married | 1.46 (1.17–1.81) | 0.001 | 1.29 (1.01–1.64) | 0.043 |
| Working experience (years) | | | | |
| <10 | 1.00 | | 1.00 | |
| 10–20 | 1.53 (1.26–1.85) | <0.001 | 1.14 (0.91–1.43) | 0.252 |
| ≥21 | 1.52 (1.25–1.83) | <0.001 | 1.05 (0.82–1.34) | 0.697 |
| Monthly income, Afghanis † | | | | |
| ≤10,000 | 1.00 | | 1.00 | |
| 10,001–20,000 | 1.15 (0.97–1.36) | 0.102 | 1.12 (0.95–1.31) | 0.173 |
| >20,001 | 1.21 (1.00–1.46) | 0.046 | 1.18 (0.98–1.42) | 0.075 |
| Lifestyle behaviors | | | | |
| Physical exercise/walking (per day) | | | | |
| <1 hour | 1.00 | | 1.00 | |
| ≥1 hour | 0.93 (0.81–1.07) | 0.317 | 0.97 (0.85–1.11) | 0.706 |
| Consumption of fruits/vegetables (per week) | | | | |
| <4 times | 1.00 | | 1.00 | |
| ≥4 times | 1.10 (0.95–1.28) | 0.191 | 1.09 (0.94–1.26) | 0.251 |
| Tobacco use | | | | |
| No | 1.00 | | 1.00 | |
| Yes | 0.74 (0.50–1.11) | 0.142 | 0.91 (0.59–1.39) | 0.662 |
## Overweight/obesity and non-communicable disease-related biomedical indicators
Table 4 shows the results of multivariate Poisson regression analyses. After adjusting for sociodemographic and lifestyle-behavior factors, participants with overweight/obesity had a 1.83 times higher likelihood of having hypertension than non-overweight/obese participants (adjusted prevalence ratio [aPR] = 1.83, $95\%$ confidence interval [CI]: 1.33–2.51). Age was a factor that positively influenced the rate of hypertension. Participants with overweight/obesity aged 41 years and over were more likely to have hypertension. On the other hand, participants with overweight/obesity earning a monthly income of more than 20 thousand Afghanis were less likely to have hypertension than their counterparts. Overweight/obesity was significantly associated with high HbA1c levels, with a 1.35 times higher likelihood for participants with overweight/obesity than for non-overweight/obese participants (aPR = 1.35, $95\%$ CI: 1.01–1.79). Participants with overweight/obesity aged 31 years or over and those consuming fruits/vegetables more frequently were more likely to have high HbA1c levels than those in other categories. On the other hand, participants with overweight/obesity earning a monthly income of 10–20 thousand Afghanis were less likely to have higher HbA1c levels than their counterparts. Overweight/obesity markedly increased the rate of high total cholesterol (aPR = 1.67, $95\%$ CI: 1.14–2.44). Participants with overweight/obesity aged 41 years or over and those with a 14th grade/2-year college or higher education were more likely to have high total cholesterol. Overweight/obesity was associated with elevated LDL-C levels (aPR = 1.29, $95\%$ CI: 1.10–1.50). The likelihood of elevated LDL-C levels was higher in female participants with overweight/obesity than in male participants with overweight/obesity. Multivariate analysis revealed no association between overweight/obesity and low HDL-C levels. Overweight/obesity was also associated with high triglyceride levels (aPR = 1.98, $95\%$ CI: 1.57–2.50). Female participants with overweight/obesity were less likely to have higher triglyceride levels than male participants with overweight/obesity. Overweight/obesity was associated with a markedly higher likelihood of having three or more comorbidities (aPR = 1.90, $95\%$ CI: 1.47–2.47). Participants with overweight/obesity aged 41 years or older and those using tobacco were more likely to have three or more comorbidities.
**Table 4**
| Variables | Adjusted prevalence ratio (95% CI) | Adjusted prevalence ratio (95% CI).1 | Adjusted prevalence ratio (95% CI).2 | Adjusted prevalence ratio (95% CI).3 | Adjusted prevalence ratio (95% CI).4 | Adjusted prevalence ratio (95% CI).5 | Adjusted prevalence ratio (95% CI).6 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Variables | Hypertension (≥130/85 mmHg) | Elevated HbA1c (≥5.5%) | Elevated total cholesterol (≥200 mg/dL) | Elevated LDL-C (≥100 mg/dL) | Low HDL-C (<40 mg/dL) | Elevated triglyceride level (≥150 mg/dL) | Comorbidity (≥3 or more) |
| Overweight /obesity (BMI≥25.0 kg/m2) | | | | | | | |
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 1.83 (1.33–2.51) *** | 1.35 (1.01–1.79) * | 1.67 (1.14–2.44) ** | 1.29 (1.10–1.50) ** | 1.23 (0.92–1.63) | 1.98 (1.57–2.50) *** | 1.90 (1.47–2.47) *** |
| Sex | | | | | | | |
| Male | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Female | 0.75 (0.56–1.02) | 1.13 (0.82–1.56) | 1.35 (0.88–2.07) | 1.22 (1.01–1.46) * | 1.47 (1.04–2.09) * | 0.76 (0.61–0.96) * | 1.02 (0.78–1.32) |
| Age (years) | | | | | | | |
| 18–30 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| 31–40 | 1.88 (0.73–4.87) | 1.68 (1.00–2.82) * | 1.47 (0.66–3.27) | 1.34 (1.00–1.78) * | 1.38 (0.79–2.38) | 0.95 (0.65–1.39) | 1.54 (0.92–2.60) |
| 41–50 | 4.07 (1.54–10.75) ** | 2.53 (1.43–4.48) ** | 2.44 (1.02–5.87) * | 1.38 (1.01–1.88) * | 1.89 (1.04–3.43) * | 1.16 (0.76–1.75) | 2.26 (1.29–3.95) * |
| ≥51 | 4.89 (1.78–13.46) ** | 3.66 (2.00–6.70) *** | 3.17 (1.27–7.90) * | 1.40 (1.00–1.95) * | 2.49 (1.33–4.65) ** | 1.00 (0.64–1.58) | 2.61 (1.45–4.68) ** |
| Education attainment | | | | | | | |
| 12th grade (high school) graduate | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| 14th grade (2-year college) graduate | 0.72 (0.48–1.08) | 0.89 (0.58–1.37) | 4.30 (1.09–16.92) * | 1.23 (0.88–1.72) | 1.92 (0.90–4.11) | 1.21 (0.78–1.88) | 1.22 (0.80–1.87) |
| College/university graduate or higher | 0.75 (0.48–1.18) | 0.82 (0.52–1.31) | 3.84 (0.97–15.19) | 1.21 (0.85–1.70) | 1.65 (0.76–3.60) | 1.19 (0.76–1.87) | 1.15 (0.74–1.79) |
| Marital status | | | | | | | |
| Never married | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Currently married | 1.06 (0.57–1.98) | 0.80 (0.53–1.21) | 1.17 (0.63–2.16) | 1.11 (0.87–1.41) | 1.33 (0.83–2.14) | 1.19 (0.84–1.68) | 0.92 (0.62–1.37) |
| Working experience (years) | | | | | | | |
| <10 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| 10–20 | 1.74 (0.93–3.24) | 1.05 (0.69–1.58) | 0.82 (0.42–1.59) | 0.91 (0.72–1.15) | 0.77 (0.50–1.19) | 1.12 (0.80–1.56) | 1.14 (0.77–1.69) |
| ≥21 | 1.52 (0.77–2.99) | 0.97 (0.60–1.57) | 0.94 (0.47–1.89) | 1.11 (0.87–1.42) | 0.83 (0.53–1.31) | 1.18 (0.81–1.71) | 1.12 (0.74–1.71) |
| Monthly income, Afghanis † | | | | | | | |
| ≤10,000 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| 10,001–20,000 | 0.76 (0.58–1.01) | 0.70 (0.53–0.92) * | 1.04 (0.72–1.50) | 0.91 (0.78–1.05) | 0.86 (0.64–1.14) | 1.03 (0.84–1.26) | 0.90 (0.72–1.13) |
| >20,001 | 0.70 (0.49–0.99) * | 0.86 (0.63–1.19) | 1.09 (0.70–1.69) | 1.01 (0.85–1.21) | 1.02 (0.72–1.42) | 0.97 (0.75–1.26) | 0.95 (0.72–1.24) |
| Lifestyle behaviors | | | | | | | |
| Physical exercise/walking (per day) | | | | | | | |
| <1 hour | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| ≥1 hour | 0.84 (0.64–1.10) | 0.84 (0.65–1.07) | 0.97 (0.71–1.34) | 0.92 (0.81–1.05) | 1.07 (0.84–1.37) | 0.94 (0.78–1.13) | 0.89 (0.73–1.09) |
| Consumption of fruits/vegetables (per week) | | | | | | | |
| <4 times | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| ≥4 times | 1.03 (0.79–1.33) | 1.33 (1.01–1.74) * | 0.94 (0.68–1.31) | 0.95 (0.83–1.09) | 0.92 (0.71–1.20) | 0.95 (0.79–1.14) | 0.91 (0.74–1.12) |
| Tobacco use | | | | | | | |
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 1.22 (0.79–1.86) | 1.28 (0.77–2.13) | 1.71 (0.87–3.33) | 1.28 (0.96–1.71) | 1.73 (1.05–2.85) * | 1.21 (0.89–1.65) | 1.57 (1.10–2.22) * |
## Results of sex-stratified multivariate analysis
The results of sex-stratified multivariate analyses are presented in Table 5. After adjusting for sociodemographic and lifestyle-behavior variables, male participants with overweight/obesity were more likely to have elevated BP; high levels of HbA1c, LDL-C, and triglycerides; low HDL-C levels; and three or more comorbidities than their non-overweight/obese counterparts. On the other hand, female participants with overweight/obesity were more likely to have elevated BP; high levels of total cholesterol and triglycerides; and three or more comorbidities than their non-overweight/obese counterparts. Moreover, multivariate models were applied to check the effect modification in subgroups for other socioeconomic variables, including age, education, and income, and no statistically significant subgroup differences were found.
**Table 5**
| Variables | Adjusted prevalence ratio (95% CI) | Adjusted prevalence ratio (95% CI).1 | Adjusted prevalence ratio (95% CI).2 | Adjusted prevalence ratio (95% CI).3 | Adjusted prevalence ratio (95% CI).4 | Adjusted prevalence ratio (95% CI).5 | Adjusted prevalence ratio (95% CI).6 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Variables | Hypertension (≥130/85 mmHg) | Elevated HbA1c (≥5.5%) | Elevated total cholesterol (≥200 mg/dL) | Elevated LDL-C (≥100 mg/dL) | Low HDL-C (<40 mg/dL) | Elevated triglyceride level (≥150 mg/dL) | Comorbidity (≥3 or more) |
| Male stratum | Male stratum | Male stratum | Male stratum | Male stratum | Male stratum | Male stratum | Male stratum |
| Overweight /obesity (BMI≥25.0 kg/m2) | | | | | | | |
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 1.97 (1.30–2.97) ** | 1.89 (1.17–3.07) * | 1.57 (0.79–3.15) | 1.82 (1.35–2.44) *** | 2.02 (1.18–3.46) * | 2.44 (1.76–3.39) *** | 2.86 (1.83–4.46) *** |
| Female stratum | Female stratum | Female stratum | Female stratum | Female stratum | Female stratum | Female stratum | Female stratum |
| Overweight /obesity (BMI≥25.0 kg/m2) | | | | | | | |
| No | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 |
| Yes | 1.83 (1.12–2.99) * | 1.14 (0.81–1.61) | 1.60 (1.02–2.51) * | 1.08 (0.91–1.29) | 0.97 (0.71–1.33) | 1.72 (1.26–2.36) ** | 1.49 (1.09–2.02) * |
## Discussion
Our results demonstrated that overweight/obesity is independently associated with hypertension; higher serum levels of HbA1c, total cholesterol, LDL-C, and triglycerides; and having three or more comorbidities. These findings are particularly concerning, given that over half of the schoolteachers in this study had overweight or obesity. The current prevalence of overweight in Afghan adults is estimated at $25.8\%$, and that of overweight and obesity combined is $42.8\%$ [3]. The preliminary results of a population-based cross-sectional study in Kandahar province of Afghanistan indicated that the prevalence of overweight and obesity was $32.8\%$ and $31.0\%$, respectively, and that of central obesity was $63.7\%$, which was higher in females than males [30]. Overweight/obesity is increasing faster in Afghanistan due to rapid urbanization, changes in dietary patterns, and the tendency of adults to adopt a more sedentary lifestyle.
Several epidemiological studies have documented the association of overweight/obesity with hypertension and its pathological effects on BP [4, 7, 31, 32]. Weight gain, particularly among adults, appears to be a significant risk factor for developing hypertension [33]. The gradual and moderate body weight reduction achieved by regular physical exercise and consumption of low-calorie diets is recommended to normalize BP in hypertensive and normotensive individuals [34]. A modest weight loss, also defined as weight loss of $5\%$–$10\%$ of baseline weight, is regarded as an effective strategy to lose weight and lower BP in individuals with hypertension [34]. A study of adolescents with obesity showed that a weight-loss program comprising diet, behavior change, and exercise resulted in a greater reduction in BP than a program that only included diet and behavior change [35]. Therefore, it is imperative to educate schoolteachers with overweight/obesity about the effects of weight loss on hypertension and to guide them towards weight reduction. Furthermore, non-overweight/obese schoolteachers must be encouraged to maintain healthy body weight.
Our study also found a significant relationship between overweight/obesity and HbA1c levels. Excess body weight is a leading risk factor for diabetes [5], and weight gain is significantly associated with the risk of diabetes [36]. A previous study in Afghanistan analyzing data from a national survey found a positive association between overweight/obesity and diabetes [4]. Another study conducted in urban areas in Kabul province found that obesity was positively associated with diabetes [37]. Individuals with overweight/obesity have a considerably higher lifetime diabetes risk than healthy individuals [38]. The mechanism is partially explained by the metabolic changes that occur as adipocytes make the body cells less sensitive to insulin, thus altering glucose production in the body. A cohort study found that patients with type 2 diabetes who lost $10\%$ of their body weight after diagnosis were more likely to achieve glycemic control, despite weight regain after four years, than those who had stable weight or weight gain [39]. Lifestyle changes, such as limiting fat intake combined with exercise, tend to be effective in weight loss, thus potentially helping to delay or prevent the onset of diabetes.
A significant association between overweight/obesity and high serum levels of blood lipids was observed in our study. Multiple factors contribute to the pathophysiology of dyslipidemia in obesity that includes increased production of very low-density lipoprotein (VLDL) by the liver, decreased release of triglycerides into the circulation, and failure to trap free fatty acids increased flux of free fatty acids from fat cells to the liver and the formation of small dense low-density lipoprotein particles [40]. A comprehensive lifestyle modification program that includes diet, exercise, and behavior change has been recommended for adults with overweight/obesity to lose weight and lower blood lipid levels, particularly for women, as they have exhibited a higher prevalence of overweight/obesity [41, 42]. Health experts also recommend weight loss to lower BP, hyperglycemia, and elevated levels of blood lipids in individuals with overweight and obesity complicated by hypertension, diabetes mellitus, or dyslipidemia [43].
We observed that female teachers had an increased likelihood of being overweight/obese than their male counterparts. These results are consistent with those of studies conducted in Tanzania, Ghana, and Ethiopia [8, 44, 45]. The sex-stratified analysis revealed that male teachers with overweight/obesity had an increased likelihood of having abnormal levels of all NCD-related biomedical indicators except elevated levels of total cholesterol, whereas female teachers with overweight/obesity had an increased likelihood of having abnormal levels of three of the six indicators than their non-overweight/obese counterparts. Although speculative, a higher prevalence of overweight/obesity in female teachers may predict an increased likelihood of NCD-related biomedical indicators in this group. However, the sex-stratified analysis indicated that male teachers with overweight/obesity are more susceptible to NCD risk. Insights from such modeling can be used to inform current clinical practice and healthcare professional medical advice for male and female patients with overweight/obesity.
Sex differences in overweight/obesity conditions may be explained by physiological and sociocultural mechanisms [46]. The distribution of adipose tissues, their metabolism, and the levels of sex hormones are key physiological mechanisms that vary depending on sex and contribute to variations in body weight and shape [47]. In Afghanistan’s context, traditional beliefs and personal attitudes toward body weight may also contribute to the extent of overweight/obesity. Traditionally, fatness has been considered a sign of beauty, superior health, and strength, an attitude that may lead to the consumption of high-energy foods and reduced physical exercise. In addition, the cultural and environmental barriers, such as insufficient single-sex facilities, that prevent women from engaging in physical activity outside the house also contribute to the tendency of females to have more sedentary lifestyles than males. Further research is needed to assess the attitudes and perceptions of people toward excess body weight and barriers to the reduction of weight or maintaining a healthy weight in Afghanistan. The perception of beauty and body size appears to have been changing in recent years, with global influences and comparisons to models and celebrities supporting the notion that female thinness (i.e., being healthy) tends to be valued in the marriage market [48]. Therefore, traditional and personal beliefs and attitudes should not be overlooked when developing health programs to combat overweight/obesity and its adverse health consequences. These findings indicate further exploration of factors contributing to weight gain in males and females, including social determinants of health, genetics, and environmental factors. Furthermore, the prevalence of overweight/obesity was higher among married adults than unmarried adults. Previous studies have reported similar findings [8, 44]. Married life has been associated with weight-related behaviors. The association between marriage and excess body weight could be due to several factors. According to a cohort study, an increase in body weight among married adults is associated with increased social eating behaviors and consumption of denser foods due to social obligations, which increase the risk of becoming overweight or obese [48].
The results of this study provide a clear picture of the overweight/obesity burden and its role as a major risk factor in a well-educated population. Therefore, effective measures are required to address this public health problem. The Ministry of Public *Health is* urged to develop and implement well-grounded risk-factor control programs targeting high-risk populations, including schoolteachers. Furthermore, promoting healthy lifestyle behaviors that reduce the prevalence of overweight and obesity, such as consuming a balanced diet and engaging in regular physical exercise, may help to minimize the future burden of NCDs. In the school setting, there is a need to design and implement school-based interventions that incorporate nutrition, physical activity, and sedentary behavior modification. Raising schoolteachers’ awareness and knowledge about NCDs and healthy lifestyles through training and counseling will help improve the prevention and control of NCDs among schoolteachers and their students. In addition, creating enabling school environment for schoolteachers and students could provide physical activity opportunities that support weight loss or maintaining a healthy weight.
The strengths of this study included the objective measurement of NCD-related biomedical indicators and BMI, which provide actual and useful data. Furthermore, we measured HbA1c, which is a reliable biomarker for assessing cumulative glycemic history over the past 2–3 months and does not require fasting for several hours before measurement. A relatively homogenous sample of schoolteachers of similar socioeconomic backgrounds minimized variations in education and income levels. This study focused on a topic that has been scarcely addressed among schoolteachers in Afghanistan, and it provided the necessary information for health policy planning and clinical practice. Notwithstanding, the study also had certain limitations. First, the cross-sectional study design did not allow for the determination of a causal relationship. In addition, the self-reported assessment of lifestyle behaviors, including fruit/vegetable intake, physical exercise/walking, and tobacco use, might have been subjected to social-desirability bias and misreporting. Participants are health education volunteers at schools who voluntarily consented to participate in the study among schoolteachers. Targeting all available and eligible teachers at schools during sample selection would thereby minimize selection bias. Therefore, participants in our study may not represent schoolteachers at-large. Finally, caution should be taken when generalizing the study findings; however, we leveraged a large sample of schoolteachers from all districts of Kabul, which includes citizens from various provinces and ethnic groups.
In conclusion, this study demonstrated a relatively high prevalence of overweight/obesity among schoolteachers. We also found statistically significant associations of overweight/obesity with higher prevalence of hypertension; elevated serum levels of HbA1c, total cholesterol, LDL-C, and triglycerides; and comorbid conditions in schoolteachers. These findings highlight that overweight/obesity is a major predictor of NCD burden in schoolteachers in Afghanistan, implying that effective awareness and behavior change interventions are warranted to promote a healthy body weight and lower the risk of NCDs and their complications.
## References
1. Stanaway JD, Afshin A, Gakidou E, Lim SS, Abate D, Abate KH. **Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of Disease Study 2017**. *Lancet* (2018.0) **392** 1923-1994. DOI: 10.1016/S0140-6736(18)32225-6
2. Ng M, Fleming T, Robinson M, Thomson B, Graetz N, Margono C. **Global, regional and national prevalence of overweight and obesity in children and adults 1980–2013: A systematic analysis**. *Lancet* (2014.0) **384** 766-781. DOI: 10.1016/S0140-6736(14)60460-8
3. 3World Health Organization and Ministry of Public Health. Afghanistan National Non-Communicable Disease Risk Factors Survey 2018. In: World Health Organization [Internet]. Geneva: World Health Organization; 2020. [cited 18 Oct 2022]. https://extranet.who.int/ncdsmicrodata/index.php/catalog/782
4. Pengpid S, Peltzer K. **Underweight and overweight/obesity among adults in Afghanistan: prevalence and correlates from a national survey in 2018**. *J Health Popul Nutr* (2021.0) **40** 25. DOI: 10.1186/s41043-021-00251-0
5. Al-Goblan AS, Al-Alfi MA, Khan MZ. **Mechanism linking diabetes mellitus and obesity**. *Diabetes Metab Syndr Obes* (2014.0) **7** 587-591. DOI: 10.2147/DMSO.S67400
6. Calle EE, Rodriguez C, Walker-Thurmond K, Thun MJ. **Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults**. *N Engl J Med* (2003.0) **348** 1625-1638. DOI: 10.1056/NEJMoa021423
7. Brown CD, Higgins M, Donato KA, Rohde FC, Garrison R, Obarzanek E. **Body mass index and the prevalence of hypertension and dyslipidemia**. *Obes Res* (2000.0) **8** 605-619. DOI: 10.1038/oby.2000.79
8. Zubery D, Kimiywe J, Martin HD. **Prevalence of overweight and obesity, and its associated factors among health-care workers, teachers, and bankers in Arusha City, Tanzania**. *Diabetes Metab Syndr Obes* (2021.0) **14** 455-465. DOI: 10.2147/DMSO.S283595
9. Delfino LD, Tebar WR, Gil FC, de Souza JM, Romanzini M, Fernandes RA. **Association of sedentary behaviour patterns with dietary and lifestyle habits among public school teachers: a cross-sectional study**. *BMJ Open* (2020.0) **10** e034322. DOI: 10.1136/bmjopen-2019-034322
10. Rocha SV, Cardoso JP, dos Santos CA, Munaro HLR, Vasconcelos LRC, Petroski EL. **Overweight/obesity in teachers: prevalence and associated factors**. *Rev Bras Cineantropom Desempenho Hum* (2015.0) **17** 450-459. DOI: 10.5007/1980-0037.2015V17N4P450
11. Neovius K, Johansson K, Kark M, Neovius M. **Obesity status and sick leave: a systematic review**. *Obes Rev* (2009.0) **10** 17-27. DOI: 10.1111/j.1467-789X.2008.00521.x
12. Jans MP, van den Heuvel SG, Hildebrandt VH, Bongers PM. **Overweight and obesity as predictors of absenteeism in the working population of the Netherlands**. *J Occup Environ Med* (2007.0) **49** 975-980. DOI: 10.1097/JOM.0b013e31814b2eb7
13. Goetzel RZ, Gibson TB, Short ME, Chu BC, Waddell J, Bowen J. **A multi-worksite analysis of the relationships among body mass index, medical utilization, and worker productivity**. *J Occup Environ Med* (2010.0) **52: Suppl 1** S52-S58. DOI: 10.1097/JOM.0b013e3181c95b84
14. Jenkins KR. **Obesity’s effects on the onset of functional impairment among older adults**. *Gerontologist* (2004.0) **44** 206-216. DOI: 10.1093/geront/44.2.206
15. Houston DK, Cai J, Stevens J. **Overweight and obesity in young and middle age and early retirement: the ARIC study**. *Obesity* (2009.0) **17** 143-149. DOI: 10.1038/oby.2008.464
16. Mar J, Karlsson J, Arrospide A, Mar B, Martínez De Aragón G, Martinez-Blazquez C. **Two-year changes in generic and obesity-specific quality of life after gastric bypass**. *Eat Weight Disord* (2013.0) **18** 305-310. DOI: 10.1007/s40519-013-0039-6
17. Scheuch K, Haufe E, Seibt R. **Teachers’ Health**. *Dtsch Arztebl Int* (2015.0) **112** 347-356. DOI: 10.3238/arztebl.2015.0347
18. Merrill RM, Aldana SG, Pope JE, Anderson DR, Coberley CR, Whitmer RW. **Presenteeism according to healthy behaviors, physical health, and work environment**. *Popul Health Manag* (2012.0) **15** 293-301. DOI: 10.1089/pop.2012.0003
19. 19United Nations Development Program. Addressing the social determinants of noncommunicable diseases. In: United Nations Development Program [Internet]. New York: United Nations Development Program; 2013. [cited 02 Feb 2023]. https://www.undp.org/publications/discussion-paper-addressing-social-determinants-noncommunicable-diseases
20. 20World Health Organization and United Nations Development Program. What ministries of education need to know—Noncommunicable diseases. In: World Health Organization [Internet]. Geneva: World Health Organization; 2016. [cited 02 Feb 2023]. https://www.paho.org/en/node/59019
21. 21World Health Organization. Global NCD target: halt the rise in diabetes. In: World Health Organization [Internet]. Geneva: World Health Organization; 2016. [cited 16 Nov 2021]. https://apps.who.int/iris/handle/10665/312280
22. Arab AS, Nakamura K, Seino K, Hemat S, Mashal MO, Tashiro Y. **Lipid and diabetic profiles of school teachers in Afghanistan facing food insecurity and their association with knowledge relating to healthy lifestyle**. *Food Nutr Sci* (2019.0) **10** 678-693. DOI: 10.4236/FNS.2019.106050
23. Reboussin DM, Allen NB, Griswold ME, Guallar E, Hong Y, Lackland DT. **Systematic Review for the 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines [published correction appears in Hypertension. 2018;71: e145]**. *Hypertension* (2018.0) **71** E116-E135. DOI: 10.1161/HYP.0000000000000067
24. Razi F, Khashayar P, Ghodssi-Ghassemabadi R, Mehrabzadeh M, Peimani M, Bandarian F. **Optimal glycated hemoglobin cutoff point for diagnosis of type 2 diabetes in Iranian adults**. *Can J Diabetes* (2018.0) **42** 582-587. DOI: 10.1016/j.jcjd.2018.03.005
25. Zhang X, Gregg EW, Williamson DF, Barker LE, Thomas W, Bullard KMK. **A1C level and future risk of diabetes: a systematic review**. *Diabetes Care* (2010.0) **33** 1665-1673. DOI: 10.2337/dc09-1939
26. **Executive summary of the third report of the National Cholesterol Education Program (NCEP) (Adult Treatment Panel III)**. *JAMA* (2001.0) **285** 2486-2497. DOI: 10.1001/JAMA.285.19.2486
27. Tabatabaei-Malazy O, Qorbani M, Samavat T, Sharifi F, Larijani B, Fakhrzadeh H. **Prevalence of dyslipidemia in Iran: A systematic review and meta-analysis study**. *Int J Prev Med* (2014.0) **5** 373-393. PMID: 24829725
28. Martinez BAF, Leotti VB, Silva GSE, Nunes LN, Machado G, Corbellini LG. **Odds ratio or prevalence ratio? An overview of reported statistical methods and appropriateness of interpretations in cross-sectional studies with dichotomous outcomes in veterinary medicine**. *Front Vet Sci* (2017.0) **4** 193. DOI: 10.3389/fvets.2017.00193
29. Thompson ML, Myers JE, Kriebel D. **Prevalence odds ratio or prevalence ratio in the analysis of cross sectional data: what is to be done?**. *Occup Environ Med* (1998.0) **55** 272-277. DOI: 10.1136/oem.55.4.272
30. Sahrai MS, Huybrechts I, Biessy C, Rinaldi S, Ferrari P, Wasiq AW. **Determinants of obesity and metabolic health in the Afghan population: protocol, methodology, and preliminary results**. *J Epidemiol Glob Health* (2022.0) **12** 113-123. DOI: 10.1007/s44197-021-00026-0
31. Choukem SP, Kengne AP, Nguefack ML, Mboue-Djieka Y, Nebongo D, Guimezap JT. **Four-year trends in adiposity and its association with hypertension in serial groups of young adult university students in urban Cameroon: a time-series study**. *BMC Public Health* (2017.0) **17** 499. DOI: 10.1186/s12889-017-4449-7
32. Leggio M, Lombardi M, Caldarone E, Severi P, D’Emidio S, Armeni M. **The relationship between obesity and hypertension: an updated comprehensive overview on vicious twins**. *Hypertens Res* (2017.0) **40** 947-963. DOI: 10.1038/hr.2017.75
33. Huang Z, Willett WC, Manson JE, Rosner B, Stampfer MJ, Speizer FE. **Body weight, weight change, and risk for hypertension in women**. *Ann Intern Med* (1998.0) **128** 81-88. DOI: 10.7326/0003-4819-128-2-199801150-00001
34. Mertens IL, van Gaal LF. **Overweight, obesity, and blood pressure: the effects of modest weight reduction**. *Obes Res* (2000.0) **8** 270-278. DOI: 10.1038/oby.2000.32
35. Rocchini AP, Katch V, Anderson J, Hinderliter J, Becque D, Martin M. **Blood pressure in obese adolescents: effect of weight loss**. *Pediatrics* (1988.0) **82** 16-23. DOI: 10.1542/PEDS.82.1.16
36. Koh-Banerjee P, Wang Y, Hu FB, Spiegelman D, Willett WC, Rimm EB. **Changes in body weight and body fat distribution as risk factors for clinical diabetes in US men**. *Am J Epidemiol* (2004.0) **159** 1150-1159. DOI: 10.1093/aje/kwh167
37. Saeed KMI. **Prevalence of risk factors for non-communicable diseases in the adult population of urban areas in Kabul City, Afghanistan**. *Cent Asian J Glob Health* (2013.0) **2**. DOI: 10.5195/cajgh.2013.69
38. Narayan KMV, Boyle JP, Thompson TJ, Gregg EW, Williamson DF. **Effect of BMI on lifetime risk for diabetes in the U.S**. *Diabetes Care* (2007.0) **30** 1562-1566. DOI: 10.2337/dc06-2544
39. Feldstein AC, Nichols GA, Smith DH, Stevens VJ, Bachman K, Rosales AG. **Weight change in diabetes and glycemic and blood pressure control**. *Diabetes Care* (2008.0) **31** 1960-1965. DOI: 10.2337/dc08-0426
40. Klop B, Elte JWF, Cabezas MC. **Dyslipidemia in obesity: mechanisms and potential targets**. *Nutrients* (2013.0) **5** 1218-1240. DOI: 10.3390/nu5041218
41. Wadden TA, Webb VL, Moran CH, Bailer BA. **Lifestyle modification for obesity: new developments in diet, physical activity, and behavior therapy**. *Circulation* (2012.0) **125** 1157-1170. DOI: 10.1161/CIRCULATIONAHA.111.039453
42. Alemi S, Nakamura K, Arab AS, Mashal MO, Tashiro Y, Seino K. **Gender-specific prevalence of risk factors for non-communicable diseases by health service use among schoolteachers in Afghanistan**. *Int J Environ Res Public Health* (2021.0) **18** 5729. DOI: 10.3390/ijerph18115729
43. **Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults: the evidence report [published correction appears in Obes Res 1998;6: 464]**. *Obes Res* (1998.0) **6** 51S-209S. DOI: 10.1002/J.1550-8528.1998.TB00690.X
44. Addo PNO, Nyarko KM, Sackey SO, Akweongo P, Sarfo B. **Prevalence of obesity and overweight and associated factors among financial institution workers in Accra Metropolis, Ghana: a cross sectional study**. *BMC Res Notes* (2015.0) **8** 599. DOI: 10.1186/s13104-015-1590-1
45. Darebo T, Mesfin A, Gebremedhin S. **Prevalence and factors associated with overweight and obesity among adults in Hawassa city, southern Ethiopia: a community based cross-sectional study**. *BMC Obes* (2019.0) **6** 8. DOI: 10.1186/s40608-019-0227-7
46. Cooper AJ, Gupta SR, Moustafa AF, Chao AM. **Sex/gender differences in obesity prevalence, comorbidities, and treatment**. *Curr Obes Rep* (2021.0) **10** 458-466. DOI: 10.1007/s13679-021-00453-x
47. Chang E, Varghese M, Singer K. **Gender and sex differences in adipose tissue**. *Curr Diab Rep* (2018.0) **18** 69. DOI: 10.1007/s11892-018-1031-3
48. Averett SL, Sikora A, Argys LM. **For better or worse: Relationship status and body mass index**. *Econ Hum Biol* (2008.0) **6** 330-349. DOI: 10.1016/j.ehb.2008.07.003
|
---
title: 'Investments for effective functionality of health systems towards Universal
Health Coverage in Africa: A scoping review'
authors:
- Humphrey Cyprian Karamagi
- Ali Ben Charif
- Solyana Ngusbrhan Kidane
- Tewelde Yohanes
- David Kariuki
- Maritza Titus
- Charles Batungwanayo
- Aminata Binetou-Wahebine Seydi
- Araia Berhane
- Jacinta Nzinga
- David Njuguna
- Hillary Kipchumba Kipruto
- Edith Andrews Annan
- Benson Droti
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021830
doi: 10.1371/journal.pgph.0001076
license: CC BY 4.0
---
# Investments for effective functionality of health systems towards Universal Health Coverage in Africa: A scoping review
## Abstract
The health challenges in Africa underscore the importance of effectively investing in health systems. Unfortunately, there is no information on systems investments adequate for an effective functional health system. We aimed to address this by conducting a scoping review of existing evidence following the Joanna Briggs Institute Manual for Evidence Synthesis and preregistered with the Open Science Framework (https://osf.io/bvg4z). We included any empirical research describing interventions that contributed to the functionality of health systems in Africa or any low-income or lower-middle-income regions. We searched Web of Science, MEDLINE, Embase, PsycINFO, Cochrane Library, CINAHL, and ERIC from their inception, and hand-searched other relevant sources. We summarized data using a narrative approach involving thematic syntheses and descriptive statistics. We identified 554 unique reports describing 575 interventions, of which 495 reported evidence of effectiveness. Most interventions were undertaken in Africa ($80.9\%$), covered multiple elements of health systems (median: 3), and focused on service delivery ($77.4\%$) and health workforce ($65.6\%$). Effective interventions contributed to improving single ($35.6\%$) or multiple ($64.4\%$) capacities of health systems: access to essential services ($75.6\%$), quality of care ($70.5\%$), demand for essential services ($38.6\%$), or health systems resilience ($13.5\%$). For example, telemedicine models which covered software (technologies) and hardware (health workers) elements were used as a strategy to address issues of access to essential services. We inventoried these effective interventions for improving health systems functionality in Africa. Further analyses could deepen understanding of how such interventions differ in their incorporation of evidence for potential scale across African countries.
## Introduction
A health system “consists of all organizations, people, and actions whose primary intent is to promote, restore, or maintain health” [1]. Many health systems in countries across the world often do not meet the service requirements of their populations. Low- and middle-income countries confront the world’s most dramatic health and social services crises. In Africa, home to 54 low- and middle-income countries, health systems almost totally collapsed with the outbreak of the 2013 Ebola epidemic in West Africa for example [2]. The experience of previous disease outbreaks in Africa suggests that the health impact of coronavirus disease 2019 (COVID-19) could be challenging [3,4]. That is in addition to the sharp rise in non-communicable diseases [5,6], which is adding another burden to African health systems. Many of the fastest growing populations are observed in Africa [7], further necessitating robust systems that are fit-for-purpose, aimed to achieve greater equality, combat hunger and malnutrition, and strengthen the coverage and quality of health and education systems [8]. A quarter of the world’s population will live in Africa by 2050 [7], creating a need to push for stronger health systems, able to effectively absorb shocks created by growing epidemics of communicable and non-communicable diseases and deliver the required essential services [9].
Health systems are the foundation to ensure healthy lives and promote well-being for all at all age, the third Sustainable Development Goal established by the United Nations in 2015 [10]. However, attainment of this goal is only possible through attainment of multiple targets brought together in three interconnected themes around which results need to be attained: universal health coverage, health security, and coverage of health determinants [11]. Thus, understanding of health systems has progressed from six building blocks [1] to a focus on having complex, dynamic systems that allow for an interplay across 13 elements or investments of health systems [11]. This includes: i) three tangible hardware elements (health workforce, health products, and health infrastructure); ii) four tangible software elements (service delivery processes, health governance processes, health information systems, and financial management systems); and iii) six intangible software elements (values and norms, beliefs, practices, organizational culture, interests and networks, and relationships and power) [11]. These elements are conceptualized in the framework for health systems strengthening towards universal health coverage in the context of the third Sustainable Development Goal in Africa (also known as the Framework of Actions) [11,12]. The Framework of *Actions is* a reference results chain depicting the flow of action from the health system pillar areas to specific outputs, to health service outcomes, all the way to impact in the form of the achievement of the third Sustainable Development Goal. A good health system is therefore one where the interplay amongst these elements allows the operationalization of the shifts needed, to attain the expected results [11].
Strengthening a health system implies the need for improving its functionality, which can be viewed from the status of four capacities [11,13]: 1) access to essential services − capacity to overcome barriers the population may face when accessing essential services that they need; 2) quality of care − capacity to ensure the process of care provision is person-oriented and effective; 3) demand for essential services − capacity to engage with the beneficiaries, to ensure what the systems provide is aligned to their own needs, and 4) resilience of the health system − capacity to anticipate, absorb, adapt, and transform itself when facing a shock event, minimizing its impact. Due to the complex set of issues hidden behind those capacities of health systems, each capacity is deconstructed into vital signs, representing a group of sub-capacities that, taken together, constitute the overall capacities [11,13]. Use of vital signs allows a more targeted group of indicators to be selected and monitored, and also provides more granular information on where a country needs to focus within each capacity. Access to essential services includes three vital signs: physical access, financial access, and sociocultural access. Quality of care includes three vital signs: user experiences, patient safety, and effectiveness of care. Demand for essential services includes two vital signs: individual healthy actions and health-seeking behaviours. Resilience includes two vital signs: specific resilience (emergency preparedness and response capacity) and the non-specific resilience (inherent capacity of the health system). By improving in these vital signs, we assure that the delivery of essential services to the population where and when they are needed [14]. Unfortunately, health systems in the World Health Organization (WHO) African Region are performing at an average of $52.9\%$ of what they can feasibly do. Comparing the contribution of the four capacities, all countries in the WHO African Region are underperforming, with access to essential services doing worst ($46.3\%$ of what is feasible), followed by health system resilience ($48.4\%$), demand for essential services ($51.4\%$), and quality of care ($61.6\%$) [13]. Thus, there is a need for implementing, sustaining, or scaling interventions to improve the functioning of health systems for attainment of universal health coverage and the other health-related Sustainable Development Goal targets in Africa.
Unfortunately, the contribution of interventions to the effective functioning of African health systems is unclear or unknown. Thus, the WHO Regional Office for Africa sought to consolidate a menu of effective interventions that could be used to strengthen the functionality of health systems at national, sub-national, and facility (i.e., organization) levels. Here, an effective intervention refers to practices, procedures, methods, strategies, or products that have been proven effective in improving the functionality of a health system through outcome evaluations under ideal (efficacy) or real-world (effectiveness) circumstances [15].
In this knowledge synthesis, we aimed to identify existing effective interventions that contribute to strengthening the functionality of health systems at national, sub-national, or facility (organizational) levels in Africa. We hypothesized that by inventorying effective interventions that contribute to improving the four capacities of health systems, we will help to generate analysis of health systems functionality, and develop guidance for African countries on “how to” re-pivot their health system development efforts to attain expected results.
## Design
We conducted a scoping review following the methodology recommended in the Joanna Briggs Institute (JBI) Manual for Evidence Synthesis [16]. This methodology is based on the Arksey and O’Malley framework [17] and an enhanced version developed by Levac and colleagues [18]. Scoping reviews are defined as “a type of evidence synthesis that aims to systematically identify and map the breadth of evidence available on a particular topic, field, concept, or issue, often irrespective of source (i.e., primary research, reviews, non-empirical evidence) within or across particular contexts” [19]. We used an online tool to identify relevant resources for designing this review [20]. We pre-registered the protocol of this review with the Open Science Framework (OSF) on May 19, 2022 (registration identifier: https://osf.io/bvg4z). We reported this review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) extension for Scoping Reviews (PRISMA-ScR) checklist (S1 Appendix) [21]. In this manuscript, the noun “report” refers to a document (paper or electronic) supplying information about a study and the noun “record” refers to the title or abstract of a report indexed in a database or website [22].
Integrated knowledge translation approach [23]: our review involved extensive participation of female and male international experts as equal members of the core team. Team members represented key stakeholder groups across African countries (e.g., Republic of Congo, Comoros, Eritrea, Kenya, Burundi, Namibia, Uganda, and South Africa), including decision makers, clinicians, and researchers. No patient or member of the public was involved in this work. Team members met weekly to discuss the progress of work during the current week. We used a virtual teamwork space using Google Drive and communicated using Microsoft Teams and WhatsApp. We also conducted three multidisciplinary consultations with knowledge users from the WHO Regional Office for Africa to receive theoretical, conceptual, or practical insights for guiding the interpretation and dissemination of findings.
## Eligibility criteria
Following the “PICOS” (participants, intervention, comparator, outcome, and setting) framework [24], we used the following inclusion criteria: In other words, we included any empirical research describing an intervention that contributed to improving any capacity of health systems in Africa or in any other low-income or lower-middle-income regions outside Africa (Table 1).
**Table 1**
| Criteria | Inclusion | Exclusion |
| --- | --- | --- |
| Type of record or report | ◻ Any empirical research: ◻ Original study ◻ Evaluation study ◻ Knowledge synthesis ◻ Government document | ◻ Editorial (commentary, letter, note)◻ Conference abstract◻ Protocol◻ Retraction◻ Non-English and non-French reports |
| Participants | ◻ Any organization (e.g., regions, clinical sites, communities) or system (e.g., district health system) involved in the delivery or receipt of health care or services | ◻ The intervention is not intended to be used for a health organization or a health system◻ The intervention is not intended to be used in the field of health |
| Intervention | ◻ Any intervention (e.g., practice, product, procedure, strategy, method) covering at least one of the 13 elements of health systems | ◻ The study did not describe an intervention◻ The intervention did not cover any of the 13 elements of health systems |
| Outcomes | ◻ Any metrics or indicators related to at least one of the four capacities of health systems: ◻ Access to essential services ◻ Quality of care ◻ Demand for essential services ◻ Resilience of a health system | ◻ The intervention is not intended to be used to improve capacities of health systems in terms of access to essential services, quality of care, demand for essential services, or resilience of a health system |
| Setting | ◻ Any African country◻ Any low-income region◻ Any lower-middle-middle region | ◻ Only high-income or upper-middle-income regions outside Africa were targeted for the intervention |
## Literature search
We performed a comprehensive search to identify records through both electronic databases of peer-reviewed literature and secondary searches using other relevant sources. No restrictions regarding date of publication, language, place of publication, or type of reports were applied to our search strategy.
First, we searched the following seven electronic databases from their dates of inception to the final search date of May 3, 2022 (MEDLINE via Ovid, PsycINFO via Ovid, Cochrane Library, CINAHL via EBSCOhost, and ERIC via Ovid) or May 9, 2022 (Web of Science and *Embase via* Elsevier). ABC drafted the preliminary version of the search strategy for Ovid MEDLINE. The search terms were based on previous works to reflect three concepts: 1) intervention [27,28], 2) health system functionality [11,29], and 3) low- and middle-income regions [30,31]. For the latter, we used a search filter developed by the Cochrane Effective Practice and Organisation of Care (EPOC) in collaboration with the WHO and Campbell Collaboration [30,31]. The preliminary search strategy was reviewed by our core team of experts in health information systems, public health, implementation science, or knowledge syntheses from Africa (TY, DK, MT, CB, SNK, and MMR). The search terms were adapted to the above-mentioned databases. Details of the search strategy in each electronic database can be found in can be found in S2 Appendix.
Second, we identified other relevant reports by searching the WHO Library, Google Scholar, Google, and Hinari in English or French. We used multiple combinations of search terms related to the concepts of intervention, health system functionality, and low- and middle-income regions (S3 Appendix). We screened at least the first 30 results for each search, a threshold often used to analyse medical content available on websites. Indeed, results lower down the relevancy lists are often duplications of earlier results and it is rare for users to click past the third page of ten search results per page [32,33].
## Study selection process
We operationalized our inclusion criteria based on our PICOS elements. After removal of duplicates, ABC performed a calibration exercise and discussed with team members (SNK, TY, DK, MT, and CB) to ensure the criteria captured relevant studies. One reviewer (ABC, TY, DK, MT, CB, AB, JN, or DN) screened records and reports for relevance and selected eligible reports using the eligibility criteria. Each record or report was screened by one of those reviewers and checked by another (ABC, TY, or AB). All disagreements were resolved through discussion between the reviewer and the checker. For each ineligible report, we documented a reason for the exclusion. Records that referred to the same report were considered duplicates, but records that referred to reports that were merely similar were considered unique [22]. We used EndNote 20 software to identify and removed duplicates as well as standardized forms in Google Sheets for the selection process.
## Data collection process
We developed a form in Google Sheets to guide extraction of variables based on the Framework of Actions [12]. ABC performed a calibration exercise and discussed with team members (SNK, TY, DK, MT, and CB) to ensure the form captures relevant data. Eight reviewers (ABC, TY, DK, MT, CB, AB, JN, and DN) independently extracted data using this standardized form. Each included report was independently extracted by two of those reviewers. All disagreements were resolved through discussion with a third party (HCK, ABC, SNK, TY, AB, JN, or DN). Information extracted from each included report was:
## Risk of bias
Due to the nature of our research question, we did not perform an appraisal for risk of bias. This is consistent with the Joanna Briggs Institute (JBI) Manual for Evidence Synthesis [16].
## Data analysis
We summarized data using a narrative approach involving framework and content analysis. A unique report could include more than one intervention, but the individual intervention was the unit of analysis. We used framework analysis to classify data according to pre-defined categories and content analysis to organize qualitative data by counting and reporting the frequency of categories found. We analysed both elements and capacities of health systems through a meticulous examination of qualitative data to identify patterns, themes, or inferences. We used the PRISMA 2020 flowchart to describe the process of report selection [22]. We summarized the main characteristics of interventions, including elements of health systems covered by the intervention, capacities of health systems targeted by the intervention, and levels of intervention scope (national, sub-national, or facility levels) in a tabular display using SAS 9.4 software [34].
## Selection of interventions
Our electronic search identified 10,628 potentially relevant records. Of these, 4196 were duplicates, leaving 8494 records. Of these, 4719 did not meet the review criteria. Thus, we reviewed a total of 1713 reports, retained 517, and excluded 1196. Reasons for exclusion included: wrong type of report, no eligible intervention, no health organization or system targeted by the intervention, no metric related to a capacity of a health system, and wrong setting (S4 Appendix). In addition, our secondary searches led to the inclusion of 37 additional reports. Overall, we included a total of 554 unique reports from all sources, which described a total of 575 interventions or packages of interventions, of which 495 were reported effective in improving the functionality of health systems (Fig 1). Descriptions of each intervention can be found in the appendix (S5 Appendix).
**Fig 1:** *PRISMA 2020 flow diagram of the intervention inclusion process.*
## Characteristics of included interventions
Main characteristics of included interventions are outlined in Table 2. Most interventions were developed or implemented in Africa ($$n = 465$$, $80.9\%$), especially in South Africa ($$n = 52$$, $11.2\%$), followed by Ethiopia ($$n = 47$$, $10.1\%$), Uganda ($$n = 47$$, $10.1\%$), Tanzania ($$n = 45$$, $9.7\%$), Kenya ($$n = 40$$, $8.6\%$), and Nigeria ($$n = 38$$, $8.2\%$) (S5 Appendix). Most interventions were developed or implemented on a sub-national level ($$n = 317$$, $55.1\%$), followed by national ($$n = 287$$, $49.9\%$) and facility ($$n = 266$$, $46.3\%$) levels.
**Table 2**
| Characteristic | Characteristic.1 | Characteristic.2 | Total(n = 575) | Levels of intervention scope | Levels of intervention scope.1 | Levels of intervention scope.2 |
| --- | --- | --- | --- | --- | --- | --- |
| Characteristic | Characteristic | Characteristic | Total(n = 575) | National(n = 287)2 | Sub-national(n = 317)2 | Facility(n = 266)2 |
| Region 1 | Region 1 | Region 1 | n (% column) | n (% column) | n (% column) | n (% column) |
| | Africa | Africa | 371 (64.5) | 175 (61.0) | 235 (74.1) | 196 (73.7) |
| | Outside Africa | Outside Africa | 110 (19.1) | 47 (16.4) | 58 (18.3) | 38 (14.3) |
| | Africa and outside | Africa and outside | 94 (16.3) | 65 (22.6) | 24 (7.6) | 32 (12.0) |
| Evidence of efficacy or effectiveness 1 | Evidence of efficacy or effectiveness 1 | Evidence of efficacy or effectiveness 1 | n (% column) | n (% column) | n (% column) | n (% column) |
| | Established | Established | 495 (86.1) | 246 (85.7) | 273 (86.1) | 238 (89.5) |
| | Unknown | Unknown | 80 (13.9) | 41 (14.3) | 44 (13.9) | 28 (10.5) |
| Health system element covered 2 | Health system element covered 2 | Health system element covered 2 | n (% row) 2 | n (% row) 2 | n (% row) 2 | n (% row) 2 |
| | Tangible software | Tangible software | 512 (89.0) | 259 (90.2) | 285 (89.9) | 251 (94.4) |
| | | Service delivery processes | 445 (77.4) | 225 (78.4) | 245 (77.3) | 226 (85.0) |
| | | Health governance processes | 214 (37.2) | 126 (43.9) | 113 (35.6) | 115 (43.2) |
| | | Health information systems | 208 (36.2) | 108 (37.6) | 124 (39.1) | 129 (48.5) |
| | | Financial management systems | 107 (18.6) | 61 (21.3) | 61 (19.2) | 60 (22.6) |
| | Tangible hardware | Tangible hardware | 450 (78.3) | 229 (79.8) | 253 (79.8) | 230 (86.5) |
| | | Health workforce | 377 (65.6) | 192 (66.9) | 222 (70.0) | 189 (71.1) |
| | | Health infrastructure | 175 (30.4) | 82 (28.6) | 96 (30.3) | 104 (39.1) |
| | | Health products | 173 (30.1) | 84 (29.3) | 96 (30.3) | 112 (42.1) |
| | Intangible software | Intangible software | 269 (46.8) | 130 (45.3) | 153 (48.3) | 134 (50.4) |
| | | Practices | 202 (35.1) | 94 (32.8) | 117 (36.9) | 113 (42.5) |
| | | Values and norms | 113 (19.7) | 55 (19.2) | 69 (21.8) | 74 (27.8) |
| | | Relationships and power | 104 (18.1) | 59 (20.6) | 54 (17.0) | 49 (18.4) |
| | | Organizational culture | 100 (17.4) | 49 (17.1) | 55 (17.4) | 64 (24.1) |
| | | Interests and networks | 84 (14.6) | 48 (16.7) | 48 (15.1) | 44 (16.5) |
| | | Beliefs | 84 (14.6) | 35 (12.2) | 55 (17.4) | 64 (24.1) |
| | Number of elements covered | Number of elements covered | | | | |
| | | Median ± interquartile range | 3.0 ± 4.0 | 3.0 ± 4.0 | 3.0 ± 4.0 | 3.0 ± 4.0 |
## Elements of health systems covered by the included interventions
Most interventions covered three or more elements of health systems (median: 3 elements, interquartile range: 4): tangible software ($$n = 512$$, $89.0\%$), tangible hardware ($$n = 450$$, $78.3\%$), or intangible software ($$n = 269$$, $46.8\%$) (Table 2). For example, telemedicine models covered tangible software (e.g., technologies) and tangible hardware (e.g., health workers) elements and were used as a strategy to address issues of access and availability of speciality care in India [35].
For those that covered tangible software elements, 445 ($77.4\%$) were related to service delivery processes, 214 ($37.2\%$) were related to health governance processes, 208 ($36.2\%$) were related to health information systems, and 107 ($18.6\%$) were related to financial management systems. Predominantly, these interventions focused on identifying, understanding, or overcoming barriers that affect the effectiveness of health service delivery with emphasis on maternal and child health, communicable diseases, and the emergence of noncommunicable diseases (e.g., [36–39]) (S5 Appendix).
For those that covered tangible hardware elements, 377 ($65.6\%$) were related to the health workforce, 175 ($30.4\%$) were related to health infrastructure, and 173 ($30.1\%$) were related to health products. Predominantly, these interventions focused on overcoming barriers the population may face when accessing essential services that they need. This included increasing the numbers and skills of the available human resources for improving the capacity of the health system to respond to population needs. For example, community health workers were trained in multiple countries through the Structured Operational Research and Training IniTiative (SORT IT), building operational research capacity for improving public health and skills to mitigate the health system effects of the COVID-19 pandemic [40] (S5 Appendix).
With regards to the intangible software elements, 202 ($35.1\%$) were related to practices, 113 ($19.7\%$) were related to values and norms, 104 ($18.1\%$) were related to relationships and power, 100 ($17.4\%$) were related to organization culture, 84 ($14.6\%$) were related to interests and networks, and 84 ($14.6\%$) were related to beliefs. Predominantly, those interventions focused on: 1) preparedness, response, coverage, and programs (e.g., [41,42]); 2) quality improvement, benchmarking, and accreditation (e.g., [43–46]); 3) health worker management and supervision including availability and distribution, workload, capacity, salaries, benefits, and motivation (e.g., [47–50]); 4) evidence-based decision-making to enhance governance and policy-making for resilience including the use of health informatics, mobile health, and performance-based financing (e.g., [51–54]); 5) behaviour change communication (e.g., [55,56]); and 6) programs and reforms such as family medicine, sub-national health systems, task-shifting, and capacity building (e.g., [57–59]) (S5 Appendix).
## Contribution of effective interventions to the functionality of health systems
There were 495 interventions ($86.1\%$) for which evidence of efficacy or effectiveness for functionality of health systems has been established (Table 1). Those effective interventions contributed to improving either single ($$n = 176$$, $35.6\%$) or multiple ($$n = 319$$, $64.4\%$) capacities of health systems, including access to essential services ($$n = 374$$, $75.6\%$), quality of care ($$n = 349$$, $70.5\%$), demand for essential services ($$n = 191$$, $38.6\%$), and resilience of health systems ($$n = 67$$, $13.5\%$) (Table 3). Effective interventions that contributed to improving each capacity were mostly related to health service delivery ($$n = 401$$, $81.0\%$) and health workforce ($$n = 329$$, $66.5\%$). Through the examination of qualitative data, we found that effective interventions commonly focused on the following cross-cutting themes: mobilization of health workers, community involvement, capacity building, digital health systems, and financial systems (S5 Appendix).
**Table 3**
| Unnamed: 0 | Unnamed: 1 | Effective interventions that contributed to improving: | Effective interventions that contributed to improving:.1 | Effective interventions that contributed to improving:.2 | Effective interventions that contributed to improving:.3 | Effective interventions that contributed to improving:.4 | Effective interventions that contributed to improving:.5 | Effective interventions that contributed to improving:.6 | Effective interventions that contributed to improving:.7 | Effective interventions that contributed to improving:.8 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | | All2 | Access to services | Access to services | Quality of care | Quality of care | Demand for services | Demand for services | Resilience to shocks | Resilience to shocks |
| | | All2 | Mainly3 | All2 | Mainly3 | All2 | Mainly3 | All2 | Mainly3 | All2 |
| Effective interventions | Effective interventions | n = 495 | n = 193 | n = 374 | n = 193 | n = 349 | n = 39 | n = 191 | n = 30 | n = 67 |
| Element of health systems | Element of health systems | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 |
| Tangible software | Tangible software | 451 (91.1) | 185 (95.9) | 343 (91.7) | 18 (93.3) | 327 (93.7) | 37 (94.9) | 181 (94.8) | 27 (90.0) | 62 (92.5) |
| | Service delivery processes | 401 (81.0) | 166 (86.0) | 310 (82.9) | 169 (87.6) | 303 (86.8) | 31 (79.5) | 167 (87.4) | 24 (80.0) | 51 (76.1) |
| | Health governance processes | 188 (38.0) | 76 (39.4) | 156 (41.7) | 76 (39.4) | 144 (41.3) | 16 (41.0) | 97 (50.8) | 15 (50.0) | 39 (58.2) |
| | Health information systems | 185 (37.4) | 70 (36.3) | 134 (35.8) | 81 (42.0) | 136 (39.0) | 16 (41.0) | 81 (42.4) | 17 (56.7) | 34 (50.7) |
| | Financial management systems | 95 (19.2) | 45 (23.3) | 81 (21.7) | 35 (18.1) | 70 (20.1) | 11 (28.2) | 55 (28.8) | 10 (33.3) | 20 (29.9) |
| Tangible hardware | Tangible hardware | 391 (79.0) | 163 (84.5) | 305 (81.6) | 161 (83.4) | 282 (80.8) | 28 (71.8) | 155 (81.2) | 26 (86.7) | 53 (79.1) |
| | Health workforce | 329 (66.5) | 130 (67.4) | 253 (67.6) | 144 (74.6) | 243 (69.6) | 24 (61.5) | 133 (69.6) | 22 (73.3) | 45 (67.2) |
| | Health products | 160 (32.3) | 72 (37.3) | 126 (33.7) | 73 (37.8) | 130 (37.2) | 12 (30.8) | 83 (43.5) | 12 (40.0) | 26 (38.8) |
| | Health infrastructure | 156 (31.5) | 68 (35.2) | 126 (33.7) | 59 (30.6) | 117 (33.5) | 12 (30.8) | 71 (37.2) | 15 (50.0) | 29 (43.3) |
| Intangible software | Intangible software | 241 (48.7) | 19 (47.2) | 185 (49.5) | 97 (50.3) | 177 (50.7) | 24 (61.5) | 134 (70.2) | 19 (63.3) | 43 (64.2) |
| | Practices | 184 (37.2) | 70 (36.3) | 145 (38.8) | 71 (36.8) | 136 (39.0) | 18 (46.2) | 105 (55.0) | 17 (56.7) | 34 (50.7) |
| | Values and norms | 101 (20.4) | 36 (18.7) | 80 (21.4) | 43 (22.3) | 79 (22.6) | 13 (33.3) | 72 (37.7) | 7 (23.3) | 16 (23.9) |
| | Relationships and power | 91 (18.4) | 33 (17.1) | 68 (18.2) | 34 (17.6) | 61 (17.5) | 13 (33.3) | 51 (26.7) | 9 (30.0) | 20 (29.9) |
| | Organizational culture | 90 (18.2) | 28 (14.5) | 66 (17.6) | 45 (23.3) | 73 (20.9) | 8 (20.5) | 57 (29.8) | 7 (23.3 | 19 (28.4) |
| | Beliefs | 80 (16.2) | 30 (15.5) | 64 (17.1) | 32 (16.6) | 58 (16.6) | 14 (35.9) | 64 (33.5) | 5 (16.7) | 11 (16.4) |
| | Interests and networks | 75 (15.2) | 26 (13.5) | 56 (15.0) | 30 (15.5) | 53 (15.2) | 12 (30.8) | 44 (23.0) | 6 (20.0) | 15 (22.4) |
| Number of elements | Number of elements | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range | Median ± interquartile range |
| | Score ranging from 1 to 13 | 4.0 ± 4.0 | 4.0 ± 4.0 | 4.0 ± 4.0 | 4.0 ± 3.0 | 4.0 ± 4.0 | 5.0 ± 4.0 | 5.0 ± 5.0 | 5.5 ± 5.0 | 5.0 ± 5.0 |
| Number of capacities | Number of capacities | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 | n (% column)1 |
| | 1 | 176 (35.6) | 52 (26.9) | 76 (20.3) | 60 (31.1) | 78 (22.3) | 6 (35.9) | 7 (3.7) | 11 (36.7) | 15 (22.4) |
| | 2 | 174 (35.2) | 72 (37.3) | 153 (40.9) | 74 (38.3) | 128 (36.7) | 14 (35.9) | 50 (26.2) | 8 (26.7) | 17 (25.4) |
| | 3 | 123 (24.8) | 64 (33.2) | 123 (32.9) | 49 (25.4) | 121 (34.7) | 18 (46.2) | 112 (58.6) | 7 (23.3) | 13 (19.4) |
| | 4 | 22 (4.4) | 5 (2.6) | 22 (5.9) | 10 (5.2) | 22 (6.3) | 1 (2.6) | 22 (11.5) | 4 (13.3) | 22 (32.8) |
| | Median ± interquartile range | 2.0 ± 2.0 | 2.0 ± 2.0 | 2.0 ± 1.0 | 2.0 ± 2.0 | 2.0 ± 1.0 | 2.0 ± 1.0 | 3.0 ± 1.0 | 2.0 ± 2.0 | 3.0 ± 2.0 |
## Contribution of effective interventions to each capacity of health systems
Fig 2 provides examples of effective interventions that contributed to improving capacities of health systems in Africa or other low-income or lower-middle-income regions.
**Fig 2:** *Examples of effective interventions that contributed to improving the functionality of health systems.*
## Improving access to essential services
For effective access to essential services, interventions mainly focused on integration of community health workers into the primary healthcare system, digital health systems, financial systems based on vouchers, vehicle support, and involvement of high-level leadership. For example:
## Improving quality of care
For effective quality of care, interventions mainly focused on improving training of health workers, user involvement, accreditation approaches, case detection and referrals through community and outreach services, ambulance drivers on first aid, management and oversight at district level, equipping health facilities with basic emergency obstetric care gadgets, introduction of clinical guidelines and protocols for diagnosing and managing most common obstetric emergencies, monitoring systems to identify antenatal complications, reminding for antenatal care attendance, and mobile health applications. For example:
## Improving demand for essential services
For effective demand for essential services, interventions mainly focused on mobilization of health worker models, capacity building, performance-based financing programs, and leadership and governance practices. For example:
## Improving resilience of health systems
Resilience to external shocks has rarely been the domain of focus, except in the cases of Ebola and the pandemic of coronavirus disease 2019 (COVID-19). Beyond service delivery processes and health work force mobilization, practices ($$n = 17$$, $56.7\%$), health information systems ($$n = 17$$, $56.7\%$), health infrastructure ($$n = 15$$, $50\%$), and health governance processes ($$n = 15$$, $50.0\%$) were covered largely by effective interventions that mainly contributed to resilience of health systems (Table 3). For effective resilience of health systems, interventions mainly focused on government support, system preparedness, response and service coverage programs, health worker availability and distribution, and evidence-based decision making to enhance governance and policy making for resilience including the use of health informatics, mobile health, and performance-based financing. For example:
## Discussion
We identified existing interventions that could contribute to capacities of health systems in Africa and other low-income or lower-middle-income regions. Most interventions were developed or implemented in Africa, covered multiple elements of health systems, and focused on service delivery at sub-national levels. The interventions that showed efficacy or effectiveness in improving the four capacities were mostly related to service delivery processes and health workforce mobilization. These findings lead us to make the following observations.
First, most included interventions covered multiple elements with a focus on identifying, understanding, or overcoming service delivery barriers at sub-national level. This result suggests that service delivery touch as many as elements of the health system to bring substantial improvements in strengthening capacities of health systems at sub-national level. In fact, inefficient service delivery results from a combination of multiple causes including inadequate health products, inefficient human resources, corruption, and under-utilization of infrastructure [85]; whereas adequate service delivery results from a successful combination of the other elements of health systems. Generally, the 13 elements of health systems should not be seen as silos because they are integral parts of the same puzzle, and if one is left out, the whole arrangement may not function properly. For example, health financing and its management is crucial to ensure availability of health products to be delivered, purchase necessary equipment, and fund payments to staff [84]. The key considerations to health service provision in low- and middle-income countries are often linked to resource management and the use of health infrastructure such as technologies, electricity, and voucher schemes. Service delivery is considered good when equitable access to a comprehensive range of high-quality health services is ensured within an integrated and person-oriented continuum of care [86]. Health systems are only as effective as the services they provide [87]. As all other elements support the delivery of health services, disruption in these other elements will indirectly impact on the quality of delivery. Thus, there is a need for multi-pronged interventions involving tangible and intangible elements for improving each capacity of health systems. For example, we found that service delivery processes, health workforce mobilization, practice changes, health information systems, health infrastructure, and modes of governance were all elements capable of producing both intended and unintended consequences for health systems resilience. Thus, the cross-cutting nature and dynamic relationship between service delivery and the other elements of health systems created challenges in isolating a particular intervention to describe overall implications and recommendations for each element. The development of validated definitions and filters for these elements would be of value for future works in the field.
Second, the effective interventions related to service delivery processes and health workforce mobilization contributed to improving the four capacities of health systems. This suggests that service delivery and health workforce were the critical tangible elements of health systems in improving the functionality of health systems in low- and middle-income countries. Given the centrality of the health workers for delivering high-quality of essential services (quality and access) that respond to community needs (demand) when facing a shock event or not (resilience) in underserved areas, service delivery and health workforce will remain critical elements in improving overall health systems performance in low- and middle-income countries. The four capacities of health systems could mainly differ in accordance with elements related to health service delivery (e.g., delayed test results or appointment, difficulty navigating referral system, long wait for transfer or treatment, geographical inaccessibility, long travel distance, and lack of primary care service) or health workforce (e.g., poor attitude, mismanagement, misdiagnosis, lack of knowledge, strike, and unavailability of doctors) [88]. A competent health workforce is a vital resource for health services delivery and health systems can only function with health workers; the improvement of the four capacities of health systems is therefore dependent on their availability, accessibility, acceptability, and quality. We found that most effective interventions were undertaken to increase health workers numbers, performance, knowledge, and skills in order to address to population needs and reduce disparities across urban and rural areas [40,50,63,67,81,89,90]. Community health workers may be faced with inadequate access to training and reference materials, poor-quality communication systems for feedback from experts or supervisors in the diagnosis and management of complex cases, and difficulty maintaining patients within the continuum of care through follow-up visits or referrals, thereby impacting the access to high-quality of essential services they can deliver. There is need to align their curricula and build operational research capacity for strengthening service delivery systems and improving health system performance. For example, for effective resilience of health systems, some countries engaged, trained, and mobilized both female and male local health workers who are trusted by the community and understand the social and cultural complexities of the population, to fill the gap created by conventional health workers in a climate of distrust, where the latter were reluctant to treat patients, sick people were afraid to self-identify and caregivers were afraid to take children to the clinic [40,81]. Notably, we found that digital health has been leveraged to mitigate some of these challenges community health workers may face in low- and middle-income countries. An initiative of mobile consulting in Bangladesh, Kenya, Nigeria, Pakistan, and Tanzania, was explored to assess whether it is a viable option for communities with minimal resources, showing that there are indications of local readiness for mobile consulting in communities with minimal access to essential services, and that mobile consulting had the potential to strengthen health systems during and beyond the COVID-19 global pandemic [89]. The introduction of Millennium Villages Global Network (MVG-Net) in multiple African countries such as Ghana, Rwanda, Tanzania, and Uganda has facilitated point-of-care decision support through mobile phone systems based on a short message service [90]. This program assisted community health workers in real-time monitoring of their pre-determined community-based services, hence enhancing quality service delivery at community level. The Mobile Midwife, a mobile application implemented in Ghana in 2010, sends timely messages in local languages to register expectant mothers and new parents [63]. Thus, the use of mobile phone systems in low- and middle-income countries has taken place faster than any other infrastructural development. However, the field of digital health is severely under-researched in Africa, yet it can be an alternative for strengthening health systems and the ways in which health services are delivered. There is growing evidence that strengthening health workforce through digital health strategies can contribute to ensuring that populations have access to high quality, responsive, and sustainable health system for otherwise under-served populations in low- and middle-income countries [86,91,92]. African health systems are underperforming in all capacities, and countries need to prioritize on effective interventions (e.g., digital health models and training programs) related to health workforce mobilization and service delivery processes, which are critical tangible elements for improving functionality of health systems in low- and middle-income countries. The WHO Regional Office for Africa has proposed the adoption of a digital health platform to streamline different solutions to a cohesive whole [93].
Third, there is widespread enthusiasm for scaling effective interventions to improve functionality of African health systems [11,27], but we know little about the scalability of identified interventions. By scalability, we mean the ability of an effective intervention to change in size, while retaining effectiveness [94,95]. Pitfalls, problems, and difficulties with scaling even proven interventions suggest that it might be beneficial to identify phases as well as components of scaling of our included interventions [96]. Indeed, a previous knowledge synthesis found that $40\%$ of scaled public health interventions had not been trialled in real world [97]. Other studies showed that coverage of effective interventions (i.e., extent to which people, organizations, or systems adopt the intervention with fidelity) was least likely to be assessed [28,98]. Yet, for interventions to have a substantial impact, they need to be adopted by a large enough population over a sustained period and one of the potential drawbacks is that some of our included effective interventions might have ceased to exist since they might have been project or donor based. Also, interventions often need considerable adaptation to enable implementation at scale, a process that can reduce or remove the effects of interventions [99]. Assessing the scalability of identified interventions is important to determine whether significant investments in their scaling will achieve worthwhile benefits to the community. Thus, there is a need for a tool to help rate and rank our included interventions for their scalability across African countries. A systematic review published a useful inventory of tools, components, and items for assessing the scalability of interventions in health as well as interpretability criteria that could be used to weed out poor items [27]. This will help to propose a scalability assessment tool aligned with the Framework of Actions and to select effective interventions that could be successfully utilized to improve the functionality of health systems in African countries.
Finally, we acknowledged that caution should be exercised when interpreting the implications of our findings. First, the criteria for what constitutes an evidence-based intervention are not met by some interventions listed in our inventory [15]. However, at this early stage of this inventory, our interest is in creating an intervention pool and mapping the available evidence for strengthening functionality of health systems in Africa. Thus, we aimed to be as inclusive as possible as nothing can be done after the fact to compensate for interventions we neglected to include. Second, in the same implementation project, several interventions may be employed across the life of the project and these interventions could cover multiple elements of health systems. Thus, it may be difficult to isolate a particular intervention as it may relate to tangible and intangible elements. However, in our analyses, we looked only at frequency of occurrence of individual interventions as in some cases multiple interventions were highlighted in the same report. Third, the selection of eligible reports was performed by only one reviewer and checked by another. However, this pragmatic approach was in agreement with methodological guides on knowledge syntheses due to the low level of complexity regarding our eligibility criteria [100].
## Conclusions
We identified existing effective interventions which could contribute to strengthening the four capacities of the health systems in Africa. Across the 13 elements of health systems, service delivery and health workforce were critical elements, mostly covered by effective interventions for improving the four capacities of health systems. However, service delivery processes and health workforce mobilization can be challenging in underserved settings or where human and health system resources are scarce. In low- and middle-income countries, digital health models and capacity building have been utilized extensively to alleviate these challenges. Thus, to have a significant impact on the functionality of health systems, more attention and investment need to be directed towards scaling of effective interventions covering multiple elements that includes service delivery and health workforce. For example, the focus on health workforce will enhance access to essential services (by ensuring adequate number and skills mix are present where needed), quality of care (by ensuring people-centered care is streamlined), demand for essential services (by providing quality care for services that are requirement by the local population), and resilience (through means of agile, capacitated, and mobile workforce in place). However, effective investment in all 13 elements is a pre-requisite for the ambitious health goals, and it has attempted to provide the first set of effective compendia across the health systems elements and functionality dimensions. Aligned to the Framework of Actions, the set of effective interventions are expected to enhance the functionality of the health systems, which in return builds the foundation required for attainment of universal health coverage, health security, and determinants of health. Finally, further analyses could also deepen understanding of how included interventions differ in their incorporation of evidence about potential for scaling across African countries.
## Declarations
We wish to acknowledge the following persons for their assistance with various aspects of this knowledge synthesis: Mr. Christian Stéphane TOUNTA and Ms. Marie Claudine SAMBA. We would also like to acknowledge technical consultation and inputs from Jean Baptiste NIKIEMA, Adam AHMAT, Serge Marcial BATALIACK, Thandekile Ntombikayis MOYO, Gertrude AVORTRI, Moussa TRAORE, James Avoka ASAMANI and Juliet NABYONGA.
## References
1. 1World Health Organization (WHO). Everybody’s business–strengthening health systems to improve health outcomes: WHO’s framework for action. World Health Organization; 2007.. *Everybody’s business–strengthening health systems to improve health outcomes: WHO’s framework for action* (2007.0)
2. Coltart CEM, Lindsey B, Ghinai I, Johnson AM, Heymann DL. **The Ebola outbreak, 2013–2016: old lessons for new epidemics**. *Philos Trans R Soc B Biol Sci* (2017.0) **372** 20160297. DOI: 10.1098/rstb.2016.0297
3. Mboussou F, Ndumbi P, Ngom R, Kamassali Z, Ogundiran O, Van Beek J. **Infectious disease outbreaks in the African region: overview of events reported to the World Health Organization in 2018**. *Epidemiol Infect* (2019.0) **147** e299. DOI: 10.1017/S0950268819001912
4. Tessema GA, Kinfu Y, Dachew BA, Tesema AG, Assefa Y, Alene KA. **The COVID-19 pandemic and healthcare systems in Africa: a scoping review of preparedness, impact and response**. *BMJ Glob Health* (2021.0) **6** e007179. DOI: 10.1136/bmjgh-2021-007179
5. Chikowore T, Kamiza AB, Oduaran OH, Machipisa T, Fatumo S. **Non-communicable diseases pandemic and precision medicine: Is Africa ready?**. *EBioMedicine* (2021.0) **65** 103260. DOI: 10.1016/j.ebiom.2021.103260
6. Countdown NCD. **2030 collaborators. NCD Countdown 2030: worldwide trends in non-communicable disease mortality and progress towards Sustainable Development Goal target 3.4**. *Lancet Lond Engl* (2018.0) **392** 1072-1088. DOI: 10.1016/S0140-6736(18)31992-5
7. 7United Nations. World Population Prospects: The 2015 Revision. 2015. Available: https://population.un.org/wpp/Publications/.. *World Population Prospects: The 2015 Revision* (2015.0)
8. 8United Nations. World Population Prospects 2019. New York: Department of Economic and Social Affairs; 2019. Available: https://population.un.org/wpp.. *World Population Prospects 2019* (2019.0)
9. Gebremeskel AT, Otu A, Abimbola S, Yaya S. **Building resilient health systems in Africa beyond the COVID-19 pandemic response**. *BMJ Glob Health* (2021.0) **6** e006108. DOI: 10.1136/bmjgh-2021-006108
10. 10General Assembly. Resolution adopted by the General Assembly on 11 September 2015. New York: United Nations; 2015.. *Resolution adopted by the General Assembly on 11 September 2015* (2015.0)
11. Karamagi HC, Tumusiime P, Titi-Ofei R, Droti B, Kipruto H, Nabyonga-Orem J. **Towards universal health coverage in the WHO African Region: assessing health system functionality, incorporating lessons from COVID-19**. *BMJ Glob Health* (2021.0) **6** e004618. DOI: 10.1136/bmjgh-2020-004618
12. 12Regional Committee for Africa 67. Framework for health systems development towards universal health coverage in the context of the sustainable development goals in the African Region: report of the Secretariat. 2017
Aug. Report No.: AFR/RC67/10. Available: https://apps.who.int/iris/handle/10665/260237.. *Framework for health systems development towards universal health coverage in the context of the sustainable development goals in the African Region: report of the Secretariat* (2017.0)
13. 13Regional Committee for Africa 70. Report on the performance of health systems in the WHO African Region. World Health Organization. Regional Office for Africa; 2020. Report No.: AFR/RC70/13. Available: https://apps.who.int/iris/handle/10665/333713.. *Report on the performance of health systems in the WHO African Region* (2020.0)
14. 14World Health Organization. Regional Office for Africa. The state of health in the WHO African Region: an analysis of the status of health, health services and health systems in the context of the Sustainable Development Goals. World Health Organization. Regional Office for Africa; 2018. Available: https://apps.who.int/iris/handle/10665/275292.. *Regional Office for Africa. The state of health in the WHO African Region: an analysis of the status of health, health services and health systems in the context of the Sustainable Development Goals* (2018.0)
15. Gottfredson DC, Cook TD, Gardner FEM, Gorman-Smith D, Howe GW, Sandler IN. **Standards of Evidence for Efficacy, Effectiveness, and Scale-up Research in Prevention Science**. *Next Generation. Prev Sci* (2015.0) **16** 893. DOI: 10.1007/s11121-015-0555-x
16. Peters M, Godfrey C, McInerney P, Munn Z, Trico A, Khalil H, Aromataris E, Munn Z. *JBI Manual for Evidence Synthesis* (2020.0). DOI: 10.46658/JBIMES-20-12
17. Arksey H O, ’Malley L. **Scoping studies: towards a methodological framework**. *Int J Soc Res Methodol* (2005.0) **8** 19-32. DOI: 10.1080/1364557032000119616
18. Levac D, Colquhoun H, O’Brien KK. **Scoping studies: advancing the methodology**. *Implement Sci* (2010.0) **5** 69. DOI: 10.1186/1748-5908-5-69
19. Munn Z, Pollock D, Khalil H, Alexander L, Mclnerney P, Godfrey CM. **What are scoping reviews? Providing a formal definition of scoping reviews as a type of evidence synthesis**. *JBI Evid Synth* (2022.0) **20** 950-952. DOI: 10.11124/JBIES-21-00483
20. **Toronto, Canada: Knowledge Synthesis Team, Knowledge Translation Program, St**. *Michael’s Hospital* (2019.0)
21. Tricco AC, Lillie E, Zarin W, O’Brien KK, Colquhoun H, Levac D. **PRISMA Extension for Scoping Reviews (PRISMA-ScR): Checklist and Explanation**. *Ann Intern Med* (2018.0) **169** 467-473. DOI: 10.7326/M18-0850
22. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD. **The PRISMA 2020 statement: an updated guideline for reporting systematic reviews**. *BMJ* (2021.0) **372** n71. DOI: 10.1136/bmj.n71
23. 23Canadian Institutes of Health Research (CIHR). Guide to Knowledge Translation Planning at CIHR: Integrated and End-of-Grant Approaches. 6
Dec
2012 [cited 28 Feb 2018]. Available: http://www.cihr-irsc.gc.ca/e/45321.html#a5.. *Guide to Knowledge Translation Planning at CIHR: Integrated and End-of-Grant Approaches* (2012.0)
24. Huang X, Lin J, Demner-Fushman D. **Evaluation of PICO as a Knowledge Representation for Clinical Questions**. *AMIA Annu Symp Proc* (2006.0) **2006** 359-363. PMID: 17238363
25. Lane-Fall MB, Curran GM, Beidas RS. **Scoping implementation science for the beginner: locating yourself on the “subway line” of translational research**. *BMC Med Res Methodol* (2019.0) **19** 133. DOI: 10.1186/s12874-019-0783-z
26. 26World Bank. World Bank Country and Lending Groups. 2022 [cited 9 May 2022]. Available: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups.. *World Bank Country and Lending Groups* (2022.0)
27. Ben Charif A, Zomahoun HTV, Gogovor A, Abdoulaye Samri M, Massougbodji J, Wolfenden L. **Tools for assessing the scalability of innovations in health: a systematic review**. *Health Res Policy Syst* (2022.0) **20** 34. DOI: 10.1186/s12961-022-00830-5
28. Ben Charif A, Hassani K, Wong ST, Zomahoun HTV, Fortin M, Freitas A. **Assessment of scalability of evidence-based innovations in community-based primary health care: a cross-sectional study**. *CMAJ Open* (2018.0) **6** E520-E527. DOI: 10.9778/cmajo.20180143
29. Berman P, Bitran R. *Health systems analysis for better health system strengthening* (2011.0)
30. 30Cochrane Effective Practice and Organisation of Care (EPOC). Low- and middle-income country (LMIC) filters. 2020 [cited 25 Nov 2021]. Available: https://epoc.cochrane.org/lmic-filters.. *Low- and middle-income country (LMIC) filters* (2020.0)
31. Sutton A, Campbell F. **The ScHARR LMIC filter: Adapting a low- and middle-income countries geographic search filter to identify studies on preterm birth prevention and management**. *Res Synth Methods*. DOI: 10.1002/jrsm.1552
32. Hargrave DR, Hargrave UA, Bouffet E. **Quality of health information on the Internet in pediatric neuro-oncology**. *Neuro-Oncol* (2006.0) **8** 175-182. DOI: 10.1215/15228517-2005-008
33. van der Marel S, Duijvestein M, Hardwick JC, van den Brink GR, Veenendaal R, Hommes DW. **Quality of web-based information on inflammatory bowel diseases**. *Inflamm Bowel Dis* (2009.0) **15** 1891-1896. DOI: 10.1002/ibd.20976
34. 34SAS Institute Inc. SAS® 9.4 Statements: Reference. Cary, NC: SAS Institute Inc.; 2013.. *SAS® 9.4 Statements: Reference* (2013.0)
35. Agarwal D, Roy N, Panwar V, Basil A, Agarwal PM. **Bringing Health Care Closer to People—A Review of Various Telemedicine Models under the National Health Mission in India**. *Indian J Community Med Off Publ Indian Assoc Prev Soc Med* (2020.0) **45** 274-277. DOI: 10.4103/ijcm.IJCM_334_19
36. Lindtjorn B, Mitiku D, Zidda Z, Yaya Y. **Reducing Maternal Deaths in Ethiopia: Results of an Intervention Programme in Southwest Ethiopia**. *PloS One* (2017.0) **12** e0169304. DOI: 10.1371/journal.pone.0169304
37. Rawal LB, Kharel C, Yadav UN, Kanda K, Biswas T, Vandelanotte C. **Community health workers for non-communicable disease prevention and control in Nepal: a qualitative study**. *BMJ Open* (2020.0) **10** e040350. DOI: 10.1136/bmjopen-2020-040350
38. Lund S, Nielsen BB, Hemed M, Boas IM, Said A, Said K. **Mobile phones improve antenatal care attendance in Zanzibar: a cluster randomized controlled trial**. *BMC Pregnancy Childbirth* (2014.0) **14** 29. DOI: 10.1186/1471-2393-14-29
39. Serbanescu F, Goldberg HI, Danel I, Wuhib T, Marum L, Obiero W. **Rapid reduction of maternal mortality in Uganda and Zambia through the saving mothers, giving life initiative: results of year 1 evaluation**. *BMC Pregnancy Childbirth* (2017.0) **17** 42. DOI: 10.1186/s12884-017-1222-y
40. Zachariah R, Dar Berger S, Thekkur P, Khogali M, Davtyan K, Kumar AMV. **Investing in Operational Research Capacity Building for Front-Line Health Workers Strengthens Countries’ Resilience to Tackling the COVID-19 Pandemic**. *Trop Med Infect Dis* (2020.0) **5**. DOI: 10.3390/tropicalmed5030118
41. Mustafa S, Zhang Y, Zibwowa Z, Seifeldin R, Ako-Egbe L, McDarby G. **COVID-19 Preparedness and Response Plans from 106 countries: a review from a health systems resilience perspective**. *Health Policy Plan* (2021.0) czab089. DOI: 10.1093/heapol/czab089
42. Roman E, Wallon M, Brieger W, Dickerson A, Rawlins B, Agarwal K. **Moving malaria in pregnancy programs from neglect to priority: experience from Malawi, Senegal, and Zambia**. *Glob Health Sci Pract* (2014.0) **2** 55-71. DOI: 10.9745/GHSP-D-13-00136
43. Roberts DA, Ng M, Ikilezi G, Gasasira A, Dwyer-Lindgren L, Fullman N. **Benchmarking health system performance across regions in Uganda: a systematic analysis of levels and trends in key maternal and child health interventions, 1990–2011**. *BMC Med* (2015.0) **13** 285. DOI: 10.1186/s12916-015-0518-x
44. Mbah H, Ojo E, Ameh J, Musuluma H, Negedu-Momoh OR, Jegede F. **Piloting Laboratory Quality System Management in Six Health Facilities in Nigeria**. *PLOS ONE* (2014.0) **9** e116185. DOI: 10.1371/journal.pone.0116185
45. Sensalire S, Isabirye P, Karamagi E, Byabagambi J, Rahimzai M, Calnan J. **Saving Mothers, Giving Life Approach for Strengthening Health Systems to Reduce Maternal and Newborn Deaths in 7 Scale-up Districts in Northern Uganda**. *Glob Health Sci Pract* (2019.0) **7** S168-S187. DOI: 10.9745/GHSP-D-18-00263
46. Hirschhorn LR, Baynes C, Sherr K, Chintu N, Awoonor-Williams JK, Finnegan K. **Approaches to ensuring and improving quality in the context of health system strengthening: a cross-site analysis of the five African Health Initiative Partnership programs**. *BMC Health Serv Res* (2013.0) **13** S8. DOI: 10.1186/1472-6963-13-S2-S8
47. Schneider H, George A, Mukinda F, Tabana H. **District Governance and Improved Maternal, Neonatal and Child Health in South Africa: Pathways of Change**. *Health Syst Reform* (2020.0) **6** e1669943. DOI: 10.1080/23288604.2019.1669943
48. Seims LRK, Alegre JC, Murei L, Bragar J, Thatte N, Kibunga P. **Strengthening management and leadership practices to increase health-service delivery in Kenya: an evidence-based approach**. *Hum Resour Health* (2012.0) **10** 25. DOI: 10.1186/1478-4491-10-25
49. Teklehaimanot HD, Teklehaimanot A. **Human resource development for a community-based health extension program: a case study from Ethiopia**. *Hum Resour Health* (2013.0) **11** 39. DOI: 10.1186/1478-4491-11-39
50. Francetic I, Tediosi F, Salari P, de Savigny D. **Going operational with health systems governance: supervision and incentives to health workers for increased quality of care in Tanzania**. *Health Policy Plan* (2019.0) **34** ii77-ii92. DOI: 10.1093/heapol/czz104
51. Robin TA, Khan MA, Kabir N, Rahaman ST, Karim A, Mannan II. **Using spatial analysis and GIS to improve planning and resource allocation in a rural district of Bangladesh**. *BMJ Glob Health* (2019.0) **4** e000832. DOI: 10.1136/bmjgh-2018-000832
52. Siribaddana P, Hewapathirana R, Jayatilleke AU, Sahay S, Dissanayake VH. **Strengthening health systems through informatics capacity development among doctors in low-resource contexts: the Sri Lankan experience**. *WHO South-East Asia J Public Health* (2019.0) **8** 87-94. DOI: 10.4103/2224-3151.264852
53. Achoki T, Lesego A. **Implementing Health Financing Reforms in Africa: Perspectives of Health System Stewards**. *Ann Glob Health* (2016.0) **82** 903-911. DOI: 10.1016/j.aogh.2016.09.008
54. Hotchkiss DR, Aqil A, Lippeveld T, Mukooyo E. **Evaluation of the Performance of Routine Information System Management (PRISM) framework: evidence from Uganda**. *BMC Health Serv Res* (2010.0) **10** 188. DOI: 10.1186/1472-6963-10-188
55. Tran PD, Vu LN, Nguyen HT, Phan LT, Lowe W, McConnell MS. **Strengthening global health security capacity—Vietnam demonstration project, 2013**. *MMWR Morb Mortal Wkly Rep* (2014.0) **63** 77-80. PMID: 24476979
56. Wekesah FM, Mbada CE, Muula AS, Kabiru CW, Muthuri SK, Izugbara CO. **Effective non-drug interventions for improving outcomes and quality of maternal health care in sub-Saharan Africa: a systematic review**. *Syst Rev* (2016.0) **5** 137. DOI: 10.1186/s13643-016-0305-6
57. Seidman G, Atun R. **Does task shifting yield cost savings and improve efficiency for health systems? A systematic review of evidence from low-income and middle-income countries**. *Hum Resour Health* (2017.0) **15** 29. DOI: 10.1186/s12960-017-0200-9
58. Swanepoel M, Mash B, Naledi T. **Assessment of the impact of family physicians in the district health system of the Western Cape, South Africa**. *Afr J Prim Health Care Fam Med* (2014.0) **6** E1-8. DOI: 10.4102/phcfm.v6i1.695
59. Zwanikken PAC, Alexander L, Scherpbier A. **Impact of MPH programs: contributing to health system strengthening in low- and middle-income countries?**. *Hum Resour Health* (2016.0) **14** 52. DOI: 10.1186/s12960-016-0150-7
60. Wangmo S, Suphanchaimat R, Htun WMM, Tun Aung T, Khitdee C, Patcharanarumol W. **Auxiliary midwives in hard to reach rural areas of Myanmar: filling MCH gaps**. *BMC Public Health* (2016.0) **16** 914. DOI: 10.1186/s12889-016-3584-x
61. 61World Health Organization. What do we know about community health workers?
A systematic review of existing reviews. World Health Organization; 2020. Available: https://apps.who.int/iris/handle/10665/340717.. *A systematic review of existing reviews* (2020.0)
62. 62World Health Organization. Enhancing the role of community health nursing for universal health coverage. World Health Organization; 2017. Available: https://apps.who.int/iris/handle/10665/255047.. *Enhancing the role of community health nursing for universal health coverage* (2017.0)
63. Entsieh AA, Emmelin M, Pettersson KO. **Learning the ABCs of pregnancy and newborn care through mobile technology**. *Glob Health Action* (2015.0) **8** 29340. DOI: 10.3402/gha.v8.29340
64. Balakrishnan R, Gopichandran V, Chaturvedi S, Chatterjee R, Mahapatra T, Chaudhuri I. **Continuum of Care Services for Maternal and Child Health using mobile technology—a health system strengthening strategy in low and middle income countries**. *BMC Med Inform Decis Mak* (2016.0) **16** 84. DOI: 10.1186/s12911-016-0326-z
65. Witter S, Brikci N, Harris T, Williams R, Keen S, Mujica A. **The free healthcare initiative in Sierra Leone: Evaluating a health system reform, 2010–2015**. *Int J Health Plann Manage* (2018.0) **33** 434-448. DOI: 10.1002/hpm.2484
66. Souteyrand YP, Collard V, Moatti JP, Grubb I, Guerma T. **Free care at the point of service delivery: a key component for reaching universal access to HIV/AIDS treatment in developing countries**. *AIDS Lond Engl* (2008.0) **22** S161-8. DOI: 10.1097/01.aids.0000327637.59672.02
67. Lusambili AM, Muriuki P, Wisofschi S, Shumba CS, Mantel M, Obure J. **Male Involvement in Reproductive and Maternal and New Child Health: An Evaluative Qualitative Study on Facilitators and Barriers From Rural Kenya**. *Front Public Health* (2021.0) **9** 644293. DOI: 10.3389/fpubh.2021.644293
68. Dumont A, Gueye M, Sow A, Diop I, Konate MK, Dambe P. **[Using routine information system data to assess maternal and perinatal care services in Mali and Senegal (QUARITE trial)]**. *Util Donnees Recueillies En Routine Pour Evaluer Act Matern Au Mali Au Senegal Essai QUARITE* (2012.0) **60** 489-96. DOI: 10.1016/j.respe.2012.05.005
69. Sequeira Dmello B, Sellah Z, Magembe G, Housseine N, Maaloe N, van den Akker T. **Learning from changes concurrent with implementing a complex and dynamic intervention to improve urban maternal and perinatal health in Dar es Salaam, Tanzania, 2011–2019**. *BMJ Glob Health* (2021.0) **6**. DOI: 10.1136/bmjgh-2020-004022
70. Serbanescu F, Clark TA, Goodwin MM, Nelson LJ, Boyd MA, Kekitiinwa AR. **Impact of the Saving Mothers, Giving Life Approach on Decreasing Maternal and Perinatal Deaths in Uganda and Zambia**. *Glob Health Sci Pract* (2019.0) **7** S27-S47. DOI: 10.9745/GHSP-D-18-00428
71. Conlon CM, Serbanescu F, Marum L, Healey J, LaBrecque J, Hobson R. **Saving Mothers, Giving Life: It Takes a System to Save a Mother (Republication)**. *Glob Health Sci Pract* (2019.0) **7** 20-40. DOI: 10.9745/GHSP-D-19-00092
72. Acharya A.. *Environmental Health and Child Survival in Nepal: Health Equity, Cost-Effectiveness, and Priority-Setting* (2015.0)
73. Werdenberg J, Biziyaremye F, Nyishime M, Nahimana E, Mutaganzwa C, Tugizimana D. **Successful implementation of a combined learning collaborative and mentoring intervention to improve neonatal quality of care in rural Rwanda**. *BMC Health Serv Res* (2018.0) **18** 941. DOI: 10.1186/s12913-018-3752-z
74. Youngleson MS, Nkurunziza P, Jennings K, Arendse J, Mate KS, Barker P. **Improving a mother to child HIV transmission programme through health system redesign: quality improvement, protocol adjustment and resource addition**. *PloS One* (2010.0) **5** e13891. DOI: 10.1371/journal.pone.0013891
75. Rawal L, Jubayer S, Choudhury SR, Islam SMS, Abdullah AS. **Community health workers for non-communicable diseases prevention and control in Bangladesh: a qualitative study**. *Glob Health Res Policy* (2020.0) **6** 1. DOI: 10.1186/s41256-020-00182-z
76. Susanti H, James K, Utomo B, Keliat B-A, Lovell K, Irmansyah I. **Exploring the potential use of patient and public involvement to strengthen Indonesian mental health care for people with psychosis: A qualitative exploration of the views of service users and carers**. *Health Expect Int J Public Particip Health Care Health Policy* (2020.0) **23** 377-387. DOI: 10.1111/hex.13007
77. Zulu JM, Maritim P, Silumbwe A, Halwiindi H, Mubita P, Sichone G. **Unlocking Trust in Community Health Systems: Lessons From the Lymphatic Filariasis Morbidity Management and Disability Prevention Pilot Project in Luangwa District, Zambia**. *Int J Health Policy Manag* (2021.0). DOI: 10.34172/ijhpm.2021.133
78. Shaikh BT, Haq ZU, Tran N, Hafeez A. **Health system barriers and levers in implementation of the Expanded Program on Immunization (EPI) in Pakistan: an evidence informed situation analysis**. *Public Health Rev* (2018.0) **39** 24. DOI: 10.1186/s40985-018-0103-x
79. Evans-Lacko S, Hanlon C, Alem A, Ayuso-Mateos JL, Chisholm D, Gureje O. **Evaluation of capacity-building strategies for mental health system strengthening in low- and middle-income countries for service users and caregivers, policymakers and planners, and researchers**. *BJPsych Open* (2019.0) **5** e67. DOI: 10.1192/bjo.2019.14
80. Sacks E, Swanson RC, Schensul JJ, Gleave A, Shelley KD, Were MK. **Community Involvement in Health Systems Strengthening to Improve Global Health Outcomes: A Review of Guidelines and Potential Roles**. *Int Q Community Health Educ* (2017.0) **37** 139-149. DOI: 10.1177/0272684X17738089
81. Siekmans K, Sohani S, Boima T, Koffa F, Basil L, Laaziz S. **Community-based health care is an essential component of a resilient health system: evidence from Ebola outbreak in Liberia**. *BMC Public Health* (2017.0) **17** 84. DOI: 10.1186/s12889-016-4012-y
82. Simen-Kapeu A, Lewycka S, Ibe O, Yeakpalah A, Horace JM, Ehounou G. **Strengthening the community health program in Liberia: Lessons learned from a health system approach to inform program design and better prepare for future shocks**. *J Glob Health* (2021.0) **11** 07002. DOI: 10.7189/jogh.11.07002
83. Meyer D, Bishai D, Ravi SJ, Rashid H, Mahmood SS, Toner E. **A checklist to improve health system resilience to infectious disease outbreaks and natural hazards**. *BMJ Glob Health* (2020.0) **5** e002429. DOI: 10.1136/bmjgh-2020-002429
84. Shoman H, Karafillakis E, Rawaf S. **The link between the West African Ebola outbreak and health systems in Guinea, Liberia and Sierra Leone: a systematic review**. *Glob Health* (2017.0) **13** 1. DOI: 10.1186/s12992-016-0224-2
85. 85World Health Organization, Development O for EC and, Development IB for R and. Delivering quality health services: a global imperative for universal health coverage. World Health Organization; 2018. Available: https://apps.who.int/iris/handle/10665/272465.. *Delivering quality health services: a global imperative for universal health coverage* (2018.0)
86. Orton M, Agarwal S, Muhoza P, Vasudevan L, Vu A. **Strengthening Delivery of Health Services Using Digital Devices**. *Glob Health Sci Pract* (2018.0) **6** S61-S71. DOI: 10.9745/GHSP-D-18-00229
87. 87World Health Organization (WHO). Key components of a well functioning health system. 2010 [cited 3 Feb 2022]. Available: http://bibalex.org/baifa/en/resources/document/450818.. *Key components of a well functioning health system* (2010.0)
88. Gbenonsi G, Boucham M, Belrhiti Z, Nejjari C, Huybrechts I, Khalis M. **Health system factors that influence diagnostic and treatment intervals in women with breast cancer in sub-Saharan Africa: a systematic review**. *BMC Public Health* (2021.0) **21** 1325. DOI: 10.1186/s12889-021-11296-5
89. Harris B, Ajisola M, Alam RM, Watkins JA, Arvanitis TN, Bakibinga P. **Mobile consulting as an option for delivering healthcare services in low-resource settings in low- and middle-income countries: A mixed-methods study**. *Digit Health* (2021.0) **7** 20552076211033424. DOI: 10.1177/20552076211033425
90. Mechael P, Nemser B, Cosmaciuc R, Cole-Lewis H, Ohemeng-Dapaah S, Dusabe S. **Capitalizing on the characteristics of mHealth to evaluate its impact**. *J Health Commun* (2012.0) **17** 62-6. DOI: 10.1080/10810730.2012.679847
91. Long L-A, Pariyo G, Kallander K. **Digital Technologies for Health Workforce Development in Low- and Middle-Income Countries**. *A Scoping Review. Glob Health Sci Pract* (2018.0) **6** S41-S48. DOI: 10.9745/GHSP-D-18-00167
92. Carter H, Araya R, Anjur K, Deng D, Naslund JA. **The emergence of digital mental health in low-income and middle-income countries: A review of recent advances and implications for the treatment and prevention of mental disorders**. *J Psychiatr Res* (2021.0) **133** 223-246. DOI: 10.1016/j.jpsychires.2020.12.016
93. Ibeneme S, Karamagi H, Muneene D, Goswami K, Chisaka N, Okeibunor J. **Strengthening Health Systems Using Innovative Digital Health Technologies in Africa**. *Front Digit Health* (2022.0) **4**. DOI: 10.3389/fdgth.2022.854339
94. Milat AJ, King L, Bauman AE, Redman S. **The concept of scalability: increasing the scale and potential adoption of health promotion interventions into policy and practice**. *Health Promot Int* (2013.0) **28** 285-298. DOI: 10.1093/heapro/dar097
95. 95World Health Organization (WHO). Nine steps for developing a scaling-up strategy. 2010. Available: http://www.who.int/reproductivehealth/publications/strategic_approach/9789241500319/en/.. *Nine steps for developing a scaling-up strategy* (2010.0)
96. Zomahoun HTV, Ben Charif A, Freitas A, Garvelink MM, Menear M, Dugas M. **The pitfalls of scaling up evidence-based interventions in health**. *Glob Health Action* (2019.0) **12** 1670449. DOI: 10.1080/16549716.2019.1670449
97. Indig D, Lee K, Grunseit A, Milat A, Bauman A. **Pathways for scaling up public health interventions**. *BMC Public Health* (2017.0) **18** 68. DOI: 10.1186/s12889-017-4572-5
98. Ben Charif A, Zomahoun HTV, Massougbodji J, Khadhraoui L, Pilon MD, Boulanger E. **Assessing the scalability of innovations in primary care: a cross-sectional study**. *CMAJ Open* (2020.0) **8** E613-E618. DOI: 10.9778/cmajo.20200030
99. McCrabb S, Lane C, Hall A, Milat A, Bauman A, Sutherland R. **Scaling-up evidence-based obesity interventions: A systematic review assessing intervention adaptations and effectiveness and quantifying the scale-up penalty**. *Obes Rev* (2019.0) **20** 964-982. DOI: 10.1111/obr.12845
100. Johnston A, Kelly SE, Hsieh S-C, Skidmore B, Wells GA. **Systematic reviews of clinical practice guidelines: a methodological guide**. *J Clin Epidemiol* (2019.0) **108** 64-76. DOI: 10.1016/j.jclinepi.2018.11.030
|
---
title: Heterogeneity in the guidelines for the management of diabetic foot disease
in the Caribbean
authors:
- Bauer E. Sumpio
- Simone McConnie
- Dale Maharaj
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021831
doi: 10.1371/journal.pgph.0000446
license: CC BY 4.0
---
# Heterogeneity in the guidelines for the management of diabetic foot disease in the Caribbean
## Abstract
The prevalence of diabetes mellitus, diabetic foot (DF) disease and, as a result, lower extremity amputation rates remain high in the Caribbean. This study was undertaken to determine whether Caribbean countries have designated individuals that monitor DF disease and whether there are DF protocols consistent with the International Working Group on the Diabetic Foot (IWGDF) guidance documents. Relevant DF health care personnel(s) from the CARICOM and Dutch Caribbean countries were called or sent questionnaires regarding the presence of structured programs to monitor and manage DF problems in the population. All 25 countries ($100\%$) responded. $81\%$ of respondents could not identify any Ministry, Hospital or individual initiatives that monitored the DF. Only 9 ($36\%$) countries had any guidelines in place. Only 3 countries with guidelines in place utilized IWGDF guidelines. Only 6 ($24\%$) countries had podiatrists and 10 ($40\%$) had vascular surgery availability. 7 ($28\%$) countries had the components for a multidisciplinary team. The presence or the appointment of a designated individual and/or a multidisciplinary approach within the countries for DF disease was absent in the majority of respondent countries. Only a minority of countries implemented DF guidelines or had expertise available to organize a DF multidisciplinary team. Vascular surgery and podiatric care were noticeably deficient. These may be critical factors in the variability and reduced success in implementation of strategies for managing DF problems and subsequent amputations amongst these Caribbean countries.
## Introduction
Lower extremity (LE) ulceration is prevalent throughout the world and poses a major threat to limb integrity and life. Foot ulcers occur in up to 25 percent of patients with diabetes and precede more than 8 in 10 non-traumatic amputations [1]. In 2014, the World Health Organization (WHO) estimated that 422 million people were diagnosed with diabetes with worldwide prevalence rates reaching nearly $9.3\%$ [2]. This is a major global public health problem and it is estimated that a major amputation occurs every 20 secs world-wide [3]. In the Caribbean, the overall prevalence of diabetes mellitus is estimated to be approximately $9\%$ and is responsible for $13.8\%$ of all deaths among adults in the region [4]. The prevalence of diabetic foot (DF) disease is high in the Caribbean and unfortunately, some countries have been labeled as “amputation capitals of the world” because of their high LE amputation rates [5]. The loss of limb as a result of diabetes is especially harsh in the Caribbean because lower limb prosthetics are not routinely available [6–8].
The socio-economic impact of diabetes is profound. In 2015 the estimated annual global direct and indirect costs of diabetes was approximately US$1.3 trillion, with one in five diabetes dollars spent on lower extremity care [9]. In Latin America and the Caribbean, it was estimated that the cost of diabetes was US$135 billion in 2015 [10]. This burden included loss of productivity due to mortality and disability, as well as direct medical costs caused by treating diabetes and its long–term complications. The indirect cost of diabetes was US$826 million for the Caribbean [11]. In Trinidad and Tobago alone it was US$85 million [11]. Therefore, preventing foot ulcerations and/or LE amputations is critical from both medical, economical and socio-economical standpoints.
The pathophysiologic mechanisms underlying DF disease are multi-factorial and include neuropathy, infection, ischemia, abnormal foot structure and biomechanics [1, 12]. It is, therefore, not surprising that the management of the DF is a complex clinical problem requiring an interdisciplinary approach [13]. Implementation of evidence‐based management of DF disease has been shown to significantly reduce hospitalization, LE amputation, disability, mortality, and cost burdens [13, 14]. Despite public health and care-giver DF initiatives, there does not appear to be any significant improvement in the number of LE amputations in most of the Caribbean [6, 8, 15].
Many developed nations create their own DF disease guidelines, or adapt those issued by the International Working Group on the Diabetic Foot (IWGDF) that has developed and distributed evidence‐based, guidance documents developed through consensus of experts in clinical and research DF disease [16]. The 2015 IWGDF guidance documents provide recommendations on prevention, appropriate footwear and offloading, management of vascular disease, infections, wound healing and the need for use of a multidisciplinary approach. DFoot *International is* the implementation group of the IWGDF and is organized around seven regions of which North America and the Caribbean (NAC) is one [17]. D-Foot International promotes the global profile of DF prevention and care through awareness, guidance, education, research, and professional development. They promote training of healthcare professionals, in the implementation of appropriate strategies and building of teams in the prevention (through early screening) and management of DF problems effectively and provide strategies to develop foot services. Given the historical accounts of the dire state of DF management in the Caribbean, this study aimed to determine whether the Caribbean countries have protocols in place to monitor and manage DF disease consistent with the IWGDF guidance documents. We specifically queried whether there were responsible institutions or individuals in the country that were designated with this responsibility.
## Materials and methods
A questionnaire modeled from a previous study comparing diabetic foot guideline utilization in Western Pacific nations [18] was sent to 15 CARICOM member countries (Antigua and Barbuda, Bahamas, Barbados, Belize, Dominica, Grenada, Guyana, Haiti, Jamaica, Montserrat, St. Lucia, St. Kitts and Nevis, St. Vincent and the Grenadines, Suriname and Trinidad and Tobago), 5 CARICOM Associate member countries (Anguilla, Bermuda, British Virgin Island, Cayman Island and Turks and Caicos) [2], the Dominican Republic as well as the Dutch Caribbean countries (Aruba, Bonaire, Curacao and Sint Marteen). The known national representatives of the International Diabetic Foot (IDF) and DFoot organizations of these countries were contacted by email and/or phone and invited to participate in the survey. In countries that did not have any representation, we contacted the Ministry of Health (MOH), the national diabetes association or the national medical association.
The survey (Table 1) asked whether DF guidelines existed and whether the MOH, National Diabetes Association, Public Hospital, Medical University or Clinical Departments were responsible for implementation or involved in their enforcement. If guidelines existed, we sought to determine what they were and whether they followed IWGDF protocols. Information on the numbers of the relevant healthcare professionals known to be essential for the DF multidisciplinary team (general surgeons, orthopedic surgeons, vascular surgeons, podiatrists, infectious disease specialists, endocrinologists, wound care specialists and wound care nurses) was also queried.
**Table 1**
| Questionaire |
| --- |
| Do you know if your country has guidelines for the management of the Diabetic Foot?—Y/NCould you provide it?—Y/N |
| If you do not have it, we would like to determine if one exists in your country Does your country have a Diabetes Association- Y/NDo they have guidelines for the management of the Diabetic Foot?—Y/NDoes your country have a representative on the DFoot International?—Y/N Who is that person if not you?Does your country have a Medical (or other Paramedical) Association- Y/N Do they have guidelines for the management of the Diabetic Foot?—Y/NDoes your country have a University Department of Surgery/Medicine/Primary Care- Y/N Do they teach guidelines for the management of the Diabetic Foot?—Y/NDoes your country have a Ministry of Health- Y/N Is there a designated individual that deals with diabetes and/or diabetic foot- Y/NDo they have guidelines for the management of the Diabetic Foot?—Y/N |
| Please provide basic information about your country and provide the date and reference or basis for your answers PopulationIncidence of diabetesIncidence of diabetic foot (infections/Incidence of foot amputations? Incidence of diabetic foot amputations? Incidence of minor foot amputations (toe/forefoot)? Incidence of major amputations (BKA/AKA)Number of general surgeonsNumber of orthopedic surgeonsNumber of podiatristsNumber of endocrinologistsNumber of infectious disease specialistsNumber of vascular surgeonsNumber of wound care nursesNumber of Wound care specialists or others with diabetic foot interest. If not listed identify their profession. |
The responses were compiled in a Microsoft Excel database. Descriptive statistics were used to report the numbers and percentages (%) of responses. Since no individual or identifiable patient information was requested, no consent was required. This multi-national survey was waivered from Ethic Board approval.
## Results
Responses from 25 countries ($100\%$) were obtained (Table 2). $81\%$ of respondents could not identify any MOH, Hospital or individual initiatives that monitored DF. Only 9 ($36\%$) countries had guidelines for the management of the DF and they were distributed by different agencies. The main source appeared to be the National Diabetes Association or Medical Association. The MOH was only rarely responsible for disseminating the guidelines. The protocols were generally adapted from existing international guidelines, especially by the International Diabetes Foundation (IDF) but some countries developed their own. Only 3 countries utilized IWGDF guidelines.
**Table 2**
| Questions | YES |
| --- | --- |
| Do you know if your country has guidelines for the management of the diabetic foot? | 9 (36%) |
| Could you provide it? | 9 (36%) |
| Does your country have a Diabetes Association | 11 (44%) |
| Do they have guidelines for the management of the diabetic foot? | 3 (13%) |
| Does your country have a representative on the DFoot International? | 5 (20%) |
| Does your country have a Medical (or other Paramedical) Association? | 11 (44%) |
| Do they have guidelines for the management of the diabetic foot? | 0 |
| Does your country have a University Department of Surgery/Medicine /Primary Care? | 10 (40%) |
| Do they have guidelines for the management of the diabetic foot? | 5 (20%) |
| Does your country have a Ministry of Health? | 11 (44%) |
| Is there a designated individual that deals with diabetes and/or diabetic foot? | 5 (20%) |
| Do they have guidelines for the management of the diabetic foot? | 1 (4%) |
| Please provide basic information about your country and provide the date and reference or basis for your answerse. Number of general surgeonsf. Number of orthopedic surgeonsg. Number of podiatristsh. Number of endocrinologistsi. Number of infectious disease specialistsj. Number of vascular surgeonsk. Number of wound care nursesl. Number of Wound care specialists or others with diabetic foot interest. | 25 (100%)10 (40%)6 (24%)5 (20%)2 (8%)10 (40%)16 (64%)4 (16%) |
There appeared to be a fair number of general and orthopedic surgeons on the islands, but there were scarce or absent podiatrists, endocrinologists, vascular surgeons, or infectious disease specialists. Only 6 ($24\%$) countries had podiatrists and 10 ($40\%$) had vascular surgery availability but only three countries with specialist vascular training. In 4 out of the 10 countries, although vascular surgery was available, it was not lower limb focused. 7 ($28\%$) countries had the components for a multidisciplinary clinical team, but we could not identify any that had a functional unit.
Out of those countries that participated only 2 respondents had definitive data on the incidence of diabetic foot infection or ulceration; 3 knew the incidence of diabetic foot amputations, none knew the incidence of minor foot amputations and only 6 knew the incidence of major amputations.
## Discussion
The high prevalence of DF disease and LE amputations in the Caribbean has been long recognized [5, 19] but, unfortunately, very little progress has been made despite reports of improved outcomes around the world [13, 14]. The reasons underlying this lack of progress are not clear. This study was undertaken to determine whether the Caribbean countries have designated individuals or organizations that monitor DF disease and whether there are DF protocols consistent with the IWGDF guidance documents, and whether they are implemented by the health institutions, MOH or medical training facilities. Our study revealed some interesting observations. 20 of the 25 countries ($80\%$) were members of the IDF and yet, only 5 countries ($20\%$) had any guidance documents, and these were primarily managed by either interested professionals, the Diabetes Association or the Medical Association. Only 2 countries used the IWGDF guidance documents. Surprisingly, it did not appear that the MOH was involved. In most cases the MOH was unaware of guidelines for the DF. Second, there was a dearth of information regarding the incidence of DF disease and the number of major and minor amputations. There was little awareness of the scope of the DF problem. This is despite published data available (Table 3) [20].
**Table 3**
| COUNTRY | Population | DM Incidence |
| --- | --- | --- |
| Anguilla ** | 9670 | 13.3 |
| Antigua | 69700 | 13.3 |
| Aruba (D) | 77800 | 15.0 |
| Bahamas | 393244 | 9.2 |
| Barbados | 204100 | 17.8 |
| Belize | 229300 | 14.9 |
| Bermuda ** | 43400 | 15.8 |
| Bonaire (D) | 20104 | 17.8 |
| British Virgin Islands | 21100 | 14.7 |
| Cayman Islands ** | 41400 | 14.2 |
| Curacao (D) | 116000 | 17.0 |
| Dominica | 48900 | 12.9 |
| Dominican Republic | 6656000 | 8.7 |
| Grenada | 69500 | 9.7 |
| Guyana | 479100 | 10.5 |
| Haiti | 6396200 | 5.7 |
| Jamaica | 1936500 | 11.7 |
| Montserrat | 3400 | 13.9 |
| St. Kits and Nevis | 36900 | 14.2 |
| St. Lucia | 129300 | 11.5 |
| St. Maarten (D) | 26800 | 14.2 |
| St. Vincent’s | 110940 | 11.6 |
| Surinam | 367800 | 13.0 |
| Trinidad & Tobago | 982600 | 12.3 |
| Turks & Caicos** | 38718 | 11.0 |
The survey demonstrated that although the respondents acknowledged there was a high incidence of diabetes, DF infections or LE amputations in their country, there was a general lack of specific information that correlated with published data. Regarding the presence of a multidisciplinary DF team, there was also a wide disparity among the countries. Most countries have mainly general surgeons and a few orthopedic surgeons. Only a quarter have a podiatrist and under half had availability of vascular surgery, most of whom were not specialty trained. Only 3 countries reported specialty trained plastic surgeons that might have the skills of performing foot reconstructions for limb salvage. In contrast, the Dominican Republic reported over 200 general surgeons and orthopedic surgeons, 150 endocrinologists, 70 infectious disease specialists and 10 vascular surgeons but no podiatrist. Based on our study, it is clear that there are scant DF interdisciplinary teams in the Caribbean.
Obviously, no clear answer exists regarding which providers should be involved in this team approach or the extent of involvement provided by each member. It is well-established that an aggressive interdisciplinary approach to DF disease is required to provide optimal medical and surgical care for improved outcomes [13]. The presence of multiple practitioners caring for the same patient increases the opportunity for life-long follow-up surveillance of vascular and podiatric disease [13]. Numerous centers around the world have reported significant reductions in amputations and ulcer recurrence when limb assessment protocols have been established and an interdisciplinary team assembled [14, 21]. It is understandable that there are many barriers to forming a multidisciplinary team and establishing the right support structure for it to become successful. Our survey confirms that there are no regional centers in the Caribbean, to work towards this process of IWGDF implementation strategies which are foundational for the success of DF disease management. It is possible to assume that there is an inert disunity, within the countries that stave away team building capacity, for limb salvage.
Regardless, it is clear that limb preservation requires a series of steps including re-establishing adequate perfusion through adequate investigations inclusive of the microvascular status which is often overlooked, serial wound debridements, appropriate wound managed care and accessibility to materials to encourage prompt wound healing, aggressive infection management, and correction of underlying biomechanical abnormalities [1, 12, 13]. At a minimum, vascular surgeons with specialty in the lower limb, podiatrists and podiatric surgeons are essential components of the team [13]. Optimized wound care is then critical after required medical and surgical interventions have been accomplished. Whilst preventative DF care has a key role in managing the DF, it is essentially as important that when assaults to the DF occur the entire team is engaged whether under the same roof, within the same country or across the waters based on limited human specialty resources, as whatever is deemed necessary for saving a limb. Primary care physicians and podiatrists play important gate-keeper roles in monitoring DF and managing early foot trauma and infections.
Unfortunately, not all critical components of an interdisciplinary team are available in either general hospitals or wound care facilities in the Caribbean [11]. Some individual physicians and surgeons with experience and training across a broad spectrum of disciplines may appropriately treat conditions in areas that lack dedicated limb preservation centers, but for complex cases, the limb salvage results will likely be inferior to the team approach. Therefore, while the constituents of teams may differ in various locales based on myriad factors, there are certain critical elements in the management of a DF that constitutes an essential, professional skill set required of a dedicated DF care team. There are some bright spots in the Caribbean. Guyana reported a dramatic improvement in LE amputations with utilization of a complex interprofessional team and foot care based protocol [21–23]. Health care workers were trained with Canadian based programs including the International Interprofessional Wound Care Course (IIWCC) Michener Institute Diabetes Educator course, regionalized to cover approximately $90\%$ of the population. Over a period of 5 years, there was an approximately $70\%$ rate reduction in LE amputations. This government-initiated project, backed by dedicated medical staff has continued to this day and has been touted as a vision for all other Caribbean territories to emulate. Unfortunately, this country is the exception and most of the governments of the Caribbean and their local surgical communities do not have the capacity to establish and sustain the kinds of teams that are so desirable.
Given that diabetes is noted not only as the most common non-communicable disease (NCDs), but also as a leading cause of death due to its myriad of complications which can also lead to disability [27], it is perplexing that greater resources are not allocated to proactive mechanisms to stamp out such suffering. For a dire condition of over 50 years of academia and research on the diabetic foot, it was disappointing to note lack of knowledge of guidance protocols. Does this show that the disease, because it is so common, has made our health professionals become immune to this condition? Or does it exhibit the perpetual spiral of clinical inertia to implement programs and multidisciplinary teams? Or does it show the general lack of understanding of what a multidisciplinary team requires within the region? In either case it is a low hanging fruit given the expertise available to this region. It could signify that there was little priority placed on the DF, protocols for DF or LE amputation prevalence. It also possibly highlights the frustration at the complexity of the management of DF disease itself and/or general apathy towards the DF problem. Much of this may be due to the public’s perception of doctors and hospitals, in what might be termed the “amputation cycle”. As citizens are made aware of a countries high-rate of LE amputations they become hesitant to see their physicians at the early-stages of their lower limb disease believing that they too might be at risk of an amputation. Moreover, there are also potential national and physician factors relevant to this situation. The existence of a “substitution culture” [15] and compliance issues [24] transcends the Caribbean.
Our study has some nuances and limitations. First, although we were primarily interested in the Caribbean nations, we queried the CARICOM countries which includes some South and Central American countries (Guyana, Suriname, and Belize) because of their strong economic, political, and social linkages [2]. We also included the Dutch Caribbean islands (Aruba, Bonaire, Curacao and St. Marteen) and the Dominican Republic because of their strong ties to the neighboring island nations. The Caribbean countries are not only geographically diverse, being separated by the Caribbean Sea, but are a mix of races, with predominant African ancestry and a heavy mix of Indian, Chinese, and European backgrounds. The racial admixture varies between countries and may account for some of the demographic disparities within the Caribbean nations [25]. For example, the rates of amputation are higher in Afro-Trinidadians compared to Indo-Trinidadians [26]. Therefore, Caribbean countries may place different priorities on this problem.
Second, we did not include the US Virgin Islands or Cuba because of their distinct health economies. Cuba is one of the 20 countries of the IDF SACA region and the prevalence of diabetes in adults is $13.2\%$. The Cuban government operates a national health system and assumes fiscal and administrative responsibility for the health care of all its citizens and is unlike the neighboring Caribbean countries. The US territories also have a distinct health economics. The Virgin Islands have universal healthcare and have a law that states that the hospitals cannot deny benefits or services because of a person’s inability to pay. Therefore, if it’s unavailable on the Virgin Islands, the hospitals must be accommodated on the mainland of the USA. This makes the management of the diabetic foot unparallel to the other Caribbean territories.
Third, it is also interesting to note that only 3 countries (Haiti, Dominican Republic and Grenada) had reported diabetes mellitus incidence rates less than $10\%$ (Table 3). However, these statistics may reflect lack of population testing, reporting or poor access to healthcare by the population. For instance, a study of Haitians living in Miami revealed a diabetes prevalence rate of $33\%$ [27].
Lastly, the respondents to our questionnaire were DF health care personnel who represented different sectors of the nations. This underscores one of the primary problems of the island nations in the lack of any consistent group that was responsible for overseeing DF education and management. We made an extensive effort to contact individuals starting with a hierarchy of the MOH, the Diabetic Association, the Medical Association, medical schools, government and private hospitals and clinics along with aligning their published memberships or affiliation to IWGDF. The MOH was, for the most part, almost oblivious to the impact of the DF to the citizen’s health and so reliance on the management of the DF was left to the national diabetes associations or national medical associations. Unfortunately, as demonstrated by our study, even these organizations had inconsistent implementation of national or IWGDF guidelines.
Taken as a whole, our study demonstrates what appears to be a lack of a conscientious systematic approach by Caribbean developing countries to DF disease. DF disease seems to be managed ad hoc, and based on past experiences, memories and perceptions as opposed to scientifically known evidence-based practices. The lack of implementation of such approaches can be seen to have caused both a delayed response to seek help for the DF patient and an impression of a frustrated approach on the part of local treating physicians. Based on evidence-based approaches in the international arena, there are confirmations of the value of interdisciplinary approaches to managing the diabetic foot and promotion of avoidable amputations [13, 14, 21]. Government and MOH support are crucial for success and to help ease some of the barriers mentioned. In addition to health care interaction, dedicated infrastructure such as dedicated community foot clinics, vascular laboratory, endovascular operating rooms are necessary along with the budget for consumables and wound products.
In 1989, the St. Vincent Declaration proposed an aggressive approach to diabetes-related complications to reduce DF complications and LE amputations [28]. The recommendation for the establishment of vascular units was advised to reduce the amputation rate. However, our study indicated that only 10 countries had vascular specialist availability and only 4 countries had trained specialists with a dedicated unit. Another initiative in 2009 was the “Step-by-step” program of the World Diabetes Foundation spearheaded by and funded jointly through the Rotary Clubs [29]. Although the mission was for the health care teams to educate diabetes patients and the general population about preventive measures for DF problems and facilitate development of algorithms for foot care to enable and encourage multidisciplinary teamwork and unify diabetes care services, there has been no continuity or guidelines put in place for a teamed approach.
The role of primary care physicians and podiatrists in performing annual foot examinations to identify high-risk foot conditions such as neuropathy, vascular disease and foot deformities cannot be over-emphasized. Collaborative interaction among other diabetes care givers is optimal to provide patients with glycemic control, smoking cessation and patient education on daily foot care and use of proper footwear. In countries without a multidisciplinary team or with no input from the vascular or podiatric team., a significant number of proximal LE amputations are done as primary procedures [6, 13]. Appropriate, timely patient referral and dedicated service for the management of foot wounds and DF infection is crucial for improved outcomes. Input from the interdisciplinary team is critical. The need for DF guidelines and programs in the *Caribbean is* critical and mandatory at this stage.
In conclusion, it is expected that with reported improved outcomes of interdisciplinary approaches to the DF, there will be a motivation to have changed behavior of patients presenting early and physicians gaining knowledge of how such referrals can be appropriately guided to ensure preservation of a functional foot. Given the persistent trends of over 50 years of LE amputations, it is highly recommended, using developed country baseline results for successes with limb salvage, that the MOH, and relevant institutions consider implementation of multidisciplinary DF teams, DF guidance protocols and/or programs through policies which will enhance the streamlining of the at-risk DF, and screening programs to prevent DF ulcerations. Despite previous efforts for assisting MOH, there has been a lack of continuity. Therefore, there is also a need for MOH to actively facilitate a gatekeeper for continuity of these programs. Without the framework in place to facilitate implementation, it is expected that the revolving door feeding the frustrations and apathetic approach to the DF and the ensuing high rate of LE amputations will continue.
## References
1. Sumpio BE. **Foot ulcers**. *N Engl J Med* (2000.0) **343** 787-93. DOI: 10.1056/NEJM200009143431107
2. 2https://www.who.int/news-room/fact-sheets/detail/diabetes.
3. 3International Diabetes Federation and IWGDF. Time to Act. (Bakker K, Foster AVM, van Houtoum WH, Riley P, eds) 2005.
4. Bennett NR, Francis D.K., Ferguson T.S. **Disparities in diabetes mellitus among Caribbean populations: a scoping review**. *Int J Equity Health* (2015.0) **14** 23. DOI: 10.1186/s12939-015-0149-z
5. Hambleton IR, J R, Davis CR, Fraser HS, Chaturvedi N, Hennis AJ. **All-cause mortality after diabetes-related amputation in Barbados: a prospective case-control study**. *Diabetes Care* (2009.0) **32** 306-7. DOI: 10.2337/dc08-1504
6. 6Sumpio BJ SB, R Jonnalagadda, D Mahler, A Hennis, O Jordan, BE Sumpio Lower Extremity Amputations in Barbados: 1999 and 2009 –Has the Situation Changed? West Indian Med J 2019.
7. Kurup R, A A, Singh J. **A review on diabetic foot challenges in Guyanese perspective**. *J Diabetes Metab Syndr* (2019.0) **13** 905-12. DOI: 10.1016/j.dsx.2018.12.010
8. Pran L, B S, Cave C, Slim H, Harnanan D, Maharaj R, Naraynsingh V. **Quality of Life Experienced by Major Lower Extremity Amputees**. *Cureus* (2021.0) **13** e17440. DOI: 10.7759/cureus.17440
9. Bommer C, Sagalova V, Heesemann E, Manne-Goehler J, Atun R, Barnighausen T. **Global Economic Burden of Diabetes in Adults: Projections From 2015 to 2030**. *Diabetes Care* (2018.0) **41** 963-70. DOI: 10.2337/dc17-1962
10. Barcelo A AA, Gordillo-Tobar A, Segovia J, Qiang A. **The cost of diabetes in Latin America and the Caribbean in 2015: Evidence for decision and policy makers**. *J Glob Health* (2017.0) **7** 020410. DOI: 10.7189/jogh.07.020410
11. Cawich SO, Islam S, Hariharan S, Harnarayan P, Budhooram S, Ramsewak S. **The economic impact of hospitalization for diabetic foot infections in a Caribbean nation**. *Perm J* (2014.0) **18** e101-4. DOI: 10.7812/TPP/13-096
12. Edmonds M, Sumpio B. *Limb Salvage of the Diabetic Foot- An Interdisciplinary Approach* (2019.0)
13. Sumpio BE, Armstrong DG, Lavery LA, Andros G. **The role of interdisciplinary team approach in the management of the diabetic foot: a joint statement from the Society for Vascular Surgery and the American Podiatric Medical Association**. *J Vasc Surg* (2010.0) **51** 1504-6. DOI: 10.1016/j.jvs.2010.04.010
14. Anichini R, Z F, Cerretini I. **Improvement of diabetic foot care after the implementation of the International Consensus on the Diabetic Foot (ICDF): results of a 5‐year prospective study**. *Diabetes Res Clin Pract* (2007.0) **75** 153-8. DOI: 10.1016/j.diabres.2006.05.014
15. Cawich SO, N V, Jonallagadda R, Wilkinson C. **Caribbean “substitution culture” is a barrier to effective treatment of persons with diabetic foot infections**. *World J Surg Proced* (2019.0) **9** 12-8
16. Bakker K, A J, Lipsky BA, Van Netten JJ, Schaper NC. **The 2015 IWGDF guidance documents on prevention and management of foot problems in diabetes: development of an evidence‐based global consensus**. *Diabetes Metab Res Rev* (2016.0) **32** 2-6
17. 17https://d-foot.org.
18. Parker CN, Van Netten JJ, Parker TJ, Jia L, Corcoran H, Garrett M. **Differences between national and international guidelines for the management of diabetic foot disease**. *Diabetes Metab Res Rev* (2019.0) **35** e3101. DOI: 10.1002/dmrr.3101
19. Pran L, H D, Baijoo S, Cave C, Short A, Maharaj R, Cawich SO. **Major Lower Limb Amputations: Recognizing Pitfalls**. *Cureus* (2021.0) **13** e16972. DOI: 10.7759/cureus.16972
20. 20https://idf.org/our-network/regions-members/north-america-and-caribbean/members.html.
21. Lowe J, S R, Taha NY, Lebovic G, Rambaran M, Martin C, Bhoj I. **The Guyana Diabetes and Foot Care Project: Improved Diabetic Foot Evaluation Reduces Amputation Rates by Two-Thirds in a Lower Middle Income Country**. *J Endocrinol* (2015.0) **2015** 920124
22. Lowe J, S R, Taha NY, Lebovic G, Martin C, Bhoj I, Kirton R. **The Guyana Diabetes and Foot Care Project: a complex quality improvement intervention to decrease diabetes-related major lower extremity amputations**. *PLoS Med* (2015.0) **12** e1001814. DOI: 10.1371/journal.pmed.1001814
23. Woodbury MG, Sibbald RG, Ostrow B, Persaud R, Lowe JM. **Tool for Rapid & Easy Identification of High Risk Diabetic Foot: Validation & Clinical Pilot of the Simplified 60 Second Diabetic Foot Screening Tool**. *PLoS One* (2015.0) **10** e0125578. DOI: 10.1371/journal.pone.0125578
24. Adams OP, C A. **Are primary care practitioners in Barbados following diabetes guidelines?–a chart audit with comparison between public and private care sectors**. *BMC Res Notes* (2011.0) **4** 199. DOI: 10.1186/1756-0500-4-199
25. Islam S, H P, Cawich SO, Budhooram S, Bheem V. **Epidemiology of diabetic foot infections in an eastern Caribbean population: a prospective study**. *Perm J* (2013.0) **17** 37-40. DOI: 10.7812/TPP/12-126
26. Gulliford MC, M D. **Social inequalities in morbidity from diabetes mellitus in public primary care clinics in Trinidad and Tobago**. *Soc Sci Med* (1998.0) **46** 137-44. DOI: 10.1016/s0277-9536(97)00155-x
27. Rosen A, Sharpe I, Rosen J, Doddard M, Abad M. **The Prevalence of Type 2 Diabetes in the Miami-Haitian Community**. *Ethnicity & Disease* (2007.0) **17** S5-3
28. **Diabetes Mellitus in Europe: a Problem at all Ages in all Countries. A Model for Prevention and Self Care**. *Acta Diabetol* (1990.0) **27** 181-3
29. 29Abbas ZG BN, Morbach S, Urbancic-Rovan V, VanAcker K. The worldwide implementation of the ‘Train the Foot Trainer’ programme 2017 [https://www.woundsinternational.com/resources/details/worldwide-implementation-train-foot-trainer-programme.
|
---
title: The effect of socioeconomic factors on quality of life of elderly in Jaffna
district of Sri Lanka
authors:
- Sathees Santhalingam
- Sivayogan Sivagurunathan
- Shamini Prathapan
- Sivapalan Kanagasabai
- Luxmi Kamalarupan
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021859
doi: 10.1371/journal.pgph.0000916
license: CC BY 4.0
---
# The effect of socioeconomic factors on quality of life of elderly in Jaffna district of Sri Lanka
## Abstract
Globally, the proportion of the elderly is increasing. In comparison to other Southeast Asian countries, Sri Lanka’s population is rapidly aging. The elderly are a vulnerable age group that requires special attention to live a long and healthy life. As, there was a scarcity of data on the elderly’s quality of life, studying the level of quality of life and the associated factors of the elderly in the Jaffna district will provide insight into how to plan interventions to improve the elderly’s overall well-being in Jaffna District and Sri Lanka as well. The study aimed to determine the quality of life of the elderly in the Jaffna district of Sri Lanka and to study the association of socioeconomic factors with the quality of life. This cross-sectional study was conducted among 813 community-dwelling elderly in the Jaffna District of Sri Lanka. Socio-economic characteristics were recorded by way of a structured questionnaire. The WHOQOL-Bref questionnaire was used to assess quality of life in four domains: physical health, psychological, social participation and the environment. The statistical Package of Social Science Software (SPSS) version 21 was used to analyse the data. Univariate, bivariate, and multivariate analyses were applied, p-value less than 0.05 was considered statistically significant. Among the four QOL domains, the mean (SD) score for an environmental domain was (12.1±2.1), (12.0±2.8) for the psychological domain, (11.8±2.3) for the physical health domain, and (10.1±3.0) for the social relationship domain. Factors significantly associated with all domains of QOL included marital status, level of education, living arrangement, employment, level of income, income adequacy and ownership of the house. Furthermore, age, sex, religion, number of children, and presence of monthly income, were significantly associated with at least one domain of QOL of the elderly. According to these findings, the QOL of the elderly in the Jaffna district of Sri Lanka seems low. And it was associated with multiple socio-economic factors. Interventions to improve the QOL of the elderly are anticipated with a higher emphasis on social relationship for the elderly.
## Introduction
Nowadays the elderly population is increasing globally, and it is a cause of concern. Longer life expectancy, low fertility rates, remarkable public health programs, and breakthroughs in medicine and health care are credited to this occurrence. In 1950, only $8\%$ of the world’s population was over the age of 65; by 2000, that number had risen to $10\%$. [ 1]. Furthermore, it was predicted that the global elderly population would reach $22\%$ in 2050. Therefore, one-fifth of the world’s population will be 60 or older in the near future [2]. The optimal Quality of life of the elderly is vital to enjoying the longevity of humankind.
The World Health Organization defines the quality of life as "an individual’s perception of their position in life in the context of the culture and value system in which they live and in relation to their goals, expectations, standards, and concerns." [ 3]. It is a multidimensional evaluation of a person’s ability concerning their physical, mental, social, and environmental health that can reflect a person’s overall well-being and is the most important indicator of a healthy life.
Health refers to an individual’s ongoing physical, emotional, mental, and social ability to cope with his or her surroundings. Therefore, the socioeconomic status of an individual plays a major role in their health and overall QOL. Numerous studies have been undertaken to examine the relationships between socio-economic characteristics (gender, age, marital status, education, and income) and quality of life in older people.
The global share of the elderly population is highest among developing countries [4]. Developing countries face special challenges with population aging and demographic shifts as their economic growth is low compared to the developed countries.
Sri *Lanka is* one of the developing countries with a rapidly increasing elderly population [5, 6]. Also, Sri Lanka has the highest proportion of elderly people among the Southeast Asian countries both today and in the future predictions [7]. In the year 2000, one in ten Sri Lankans was reported to be elderly. Furthermore, by 2010, this percentage had risen to 1 in 5, and it is anticipated to rise to one-fourth by 2030 [6, 8]. In addition, between 2000 and 2030, Sri Lanka’s median age is predicted to rise from 26.9 to 39.2 years [9]. The demographic shift should be managed by implementing policies to enhance the well-being of the elderly. The policies should give special emphasis to the districts that have the highest proportion of elderly people.
Apart from the effect of population aging on the country, aging and age-related changes pose greater challenges to the wellbeing of the elderly. Aging is the multifaceted, ongoing degradation of a person’s organ systems and tissue that is complex, inexorable, and unavoidable [10]. Individuals’ natural functionality is affected by aging. As a result, when compared to other age groups, the elderly are at a higher risk of a variety of physical and psychological difficulties. Musculoskeletal difficulties, respiratory disorders, and gastrointestinal and cardiovascular problems are the most common difficulties [3]. Further, the most prevalent health concerns reported by the aged in Sri Lanka include hypertension, diabetes mellitus, vision and hearing impairments, arthritis, and asthma [11]. Aside from the health problems mentioned above, many elderly people are also dealing with various socio-economic problems. Inevitably Physical health problems in the elderly with socio-economic problems together can have a significant impact on their wellbeing. Therefore, it is well understood that aging has a direct and indirect impact on the quality of life (QOL) of the elderly. Determining the QOL of the elderly and studying the association of socio-economic factors will be the initial steps to improving the wellbeing of the elderly population.
Only a small number of studies have been undertaken in Sri Lanka to examine the elderly’s quality of life in Sri Lanka. Jaffna *District is* one of Sri Lanka’s 25 administrative districts. It has a greater proportion of the elderly population ($14.1\%$) than the rest of Sri Lanka ($12.4\%$) [8, 12]. The majority of the QOL studies have focused on the southern area of the country and the Sinhala Buddhist community. As a result, there is a scarcity of information specific to the elderly in the Jaffna community. But Thanujanan et al [2016] demonstrated that the quality of life of the elderly of Jaffna District was determined to be moderate by using the OPQOL scale [13].
According to previous studies, there are many significant connections between quality of life and the socio-demographic and economic characteristics of the elderly. The goal of the study is to determine the quality of life and the association of socioeconomic factors among the elderly population residing in the Jaffna district of Sri Lanka and to make recommendations to improve the quality of life of the elderly who live in Jaffna district and Sri Lanka as well.
## Study samples and procedures
People over the age of 60 who lived in the Jaffna District were included in this cross-sectional descriptive study. The study excludes institutionalized elderly people who make up about $0.45\%$ of Jaffna’s elderly population [14] as their living arrangements and care patterns differ from those of non-institutionalized elderly people.
In a recent study, the prevalence of self-reported QOL was $48\%$ [15]. Based on a prevalence of $48\%$, a precision of 0.05, and a confidence level of $95\%$, the sample size required was calculated to be 384 using the *Daniel formula* [16]. It was increased to roughly 880 to account for the $15\%$ nonresponse rate after being doubled to overcome the design effect [17]. Two-stage geographical cluster sampling was used to select the participants. For the first stage, all Grama Niladari (GN) areas [440] in Jaffna District were utilized as the sample framework, whereas people older than 60 years in each GN area were utilized as the sampling framework for the second stage. A total of 44 GN areas ($10\%$) were selected from the first stage. Twenty elderly people from each of the selected GN areas were selected randomly in the second stage. Altogether, 880 elderly people were included in the study.
## Measures
The socio-demographic characteristics and economic factors were assessed using an interviewer-administered questionnaire. The QOL of the elderly was determined by the previously validated Tamil version of the WHOQOL-Bref questionnaire. This questionnaire consists of 26 questions. The questions were divided into four categories to assess the perceived QOL of the elderly in the following four domains; physical health domain, psychological domain, social participation domain, and environmental domain. Responses were obtained on a 5-point Likert scale. Individual scores for all four domains of quality of life were calculated using the transformed score table. A higher score indicates better QOL in all four domains in elderly individuals [16]. Data collection was conducted by face-to-face interviews at the participants’ homes. Informed written consent was obtained from the participants before the interviews. Recall bias was considered minimal, as the information gathered was highly personal.
## Analysis
The Statistical Package for Social Sciences (SPSS) for Windows, version 16, was used to analyses all the data collected. The results for sociodemographic and economic variables, and quality of life were presented using both descriptive and inferential statistics. For categorical variables, the findings are expressed as proportions and frequencies. The continuous variables’ standard deviation and mean are presented. The mean variations in the quality of life with the socio-economic factors were identified using one-way ANOVA and independent t-tests. Variables with a p-value less than 0.05 in the univariate analysis were included in the multiple linear regression models, and variable multicollinearity was examined before inclusion to the model. This model was used to examine the independent relationship of socio-economic characteristics with physical health, psychological, social participation, and environmental domains. A p-value less than 0.05 was considered statistically significant.
## Ethical consideration
The Ethics Review Committee of the Faculty of Medicine at the University of Jaffna in Sri Lanka granted ethical approval for this study (Ref. No-J/ERC/$\frac{18}{92}$/DR/0059). Informed written consent was obtained from all the participants in the study. Data was anonymised to protect confidentiality during analysis.
## Characteristics of participants
The mean age of the participants was 71.1years (range 60–100 years, SD 7.7). As shown in Table 1, of the 813 participants, $53.5\%$ were men and $47\%$ were in the 60–69 years age category. The higher proportion of participants were Hindus ($85.9\%$). Most of the participants ($61.6\%$) were having a spouse, and $74.7\%$ had up to the secondary level of education. Less than half ($47\%$) of the participants never went for work outside the home while others were currently working or retired from their employment. Nearly half ($54.9\%$) of the participants had a monthly income below the national poverty line (food and non-food expenditures per person per month) of the region.
**Table 1**
| Sociodemographic Factors | Sociodemographic Factors.1 | F | % |
| --- | --- | --- | --- |
| Age (Years) | | | |
| | 60–69 years | 382.0 | 47.0 |
| | ≥70 years | 431.0 | 53.0 |
| Sex | | | |
| | Male | 435.0 | 53.5 |
| | Female | 378.0 | 46.5 |
| Religion (n = 797) | | | |
| | Hindu | 685.0 | 85.9 |
| | Christian | 107.0 | 13.5 |
| | Islam | 5.0 | 0.6 |
| Marital status (n = 811) | | | |
| | Unmarried | 20.0 | 2.5 |
| | With spouse | 500.0 | 61.6 |
| | widow | 261.0 | 32.2 |
| | Divorced/separated | 30.0 | 3.7 |
| Number of children | | | |
| | No | 58.0 | 7.1 |
| | 1 | 49.0 | 6.0 |
| | 2–3 | 424.0 | 52.2 |
| | More than 4 | 282.0 | 34.7 |
| Education | | | |
| | No schooling | 14.0 | 1.7 |
| | Primary | 98.0 | 12.1 |
| | Secondary | 607.0 | 74.7 |
| | Collegial | 58.0 | 7.1 |
| | Tertiary | 36.0 | 4.4 |
| Resides with (n = 803) | | | |
| | Spouse only | 254.0 | 31.6 |
| | Children only | 218.0 | 27.1 |
| | Spouse and children | 211.0 | 26.3 |
| | Alone | 95.0 | 11.8 |
| | Others | 25.0 | 3.1 |
| Occupation | | | |
| | Retired | 136.0 | 16.7 |
| | Currently working | 295.0 | 36.3 |
| | Never worked | 382.0 | 47.0 |
| Monthly income (n = 699) | | | |
| | Below national poverty line | 384.0 | 54.9 |
| | Above national poverty line | 315.0 | 45.1 |
## Univariate analysis
Univariate analysis revealed the participants’ levels of QOL in of different domains. Among the four QOL domains, the mean score for the environmental domain was higher (12.1±2.1) followed by psychological (2.0±2.8), physical health (11.8±2.3), and social participation domain (10.1±3.0).
## Bivariate analysis
Table 2 shows the individual associations of socio-economic characteristics with the four domains of QOL. All four domains of QOL were significantly associated with marital status, level of education, living arrangement, employment, source of income, level of income, income adequacy, and homeownership ($P \leq 0.05$). Physical health, psychological, and social participation domains were significantly associated with age and gender ($P \leq 0.05$). Religion was associated with the environmental domain ($$P \leq 0.001$$). The number of children was associated with the psychological, social relationship, and environmental domains ($P \leq 0.05$). The presence of monthly income was associated with the psychological and environmental domains of the QOL ($P \leq 0.05$).
**Table 2**
| Socio-demographic factors | Socio-demographic factors.1 | N | % | Physical domain Mean (SD) | Psychological domain Mean (SD) | Social relationship domain Mean(SD) | Environmental domain Mean (SD) |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Age (in years) | Age (in years) | | | | | | |
| | 60–69 years | 382.0 | 47.0 | 12.5 (2.0) | 12.3 (2.7) | 10.9 (3.0) | 12.2 (2.1) |
| | ≥70 years | 431.0 | 53.0 | 11.1 (2.3) | 11.7 (2.8) | 9.4 (2.9) | 12.0 (2.1) |
| | | | | P<0.001 | P = 0.001 | P<0.001 | P = 0.279 |
| Sex | Sex | | | | | | |
| | Male | 435.0 | 53.5 | 12.3 (2.2) | 12.3 (2.7) | 11.0 (2.9) | 12.2 (2.0) |
| | Female | 378.0 | 46.5 | 11.2 (2.3) | 11.6 (2.8) | 9.1 (2.9) | 12.0 (2.1) |
| | | | | P<0.001 | P = 0.002 | P<0.150 | P<0.001 |
| Religion | Religion | | | | | | |
| | Hindu | 685.0 | 85.9 | 11.8 (2.3) | 12.0 (2.8) | 10.1 (3.1) | 12.2 (2.1) |
| | Christian | 107.0 | 13.5 | 11.6 (2.4) | 11.5 (2.8) | 10.2 (2.6) | 11.4 (2.2) |
| | Islam | 5.0 | 0.6 | | | | |
| | | | | P = 0.334 | P = 0.053 | P = 0.937 | P = 0.001 |
| Marital status | Marital status | | | | | | |
| | Unmarried | 20.0 | 205.0 | 10.8 (2.7) | 9.2 (2.7) | 6.7 (1.9) | 10.6 (2.2) |
| | Married/living with spouse | 500.0 | 61.6 | 12.4 (2.1) | 12.7 (2.6) | 11.7 (2.6) | 12.5 (2.0) |
| | Widowed | 261.0 | 32.2 | 10.8 (2.1) | 11.2 (2.6) | 7.8 (1.8) | 11.6 (2.1) |
| | Divorced/Separated | 30.0 | 3.7 | 10.7 (1.8) | 9.0 (2.3) | 7.1 (1.0) | 10.1 (2.1) |
| | | | | P<0.001 | P<0.001 | P<0.001 | P<0.001 |
| Number of children | Number of children | | | | | | |
| | No | 58.0 | 7.1 | 11.1 (2.0) | 9.7 (2.4) | 8.9 (2.8) | 11.1 (1.9) |
| | 1 | 49.0 | 6.0 | 11.6 (1.9) | 11.2 (2.9) | 10.2 (2.8) | 12.0 (1.9) |
| | 2–3 | 424.0 | 52.2 | 11.9 (2.1) | 12.0 (2.6) | 10.2 (2.8) | 12.0 (1.9) |
| | More than 4 | 282.0 | 34.7 | 11.8 (2.7) | 12.5 (2.9) | 10.4 (3.4) | 12.5 (2.3) |
| | | | | P = 0.124 | P<0.001 | P<0.001 | P<0.001 |
| Education | Education | | | | | | |
| | No schooling | 14.0 | 1.7 | 10.0 (3.3) | 9.7 (2.8) | 8.4 (4.9) | 10.9 (3.5) |
| | Primary | 98.0 | 12.1 | 11.0 (2.6) | 11.2 (2.8) | 8.7 (3.1) | 11.3 (2.2) |
| | Secondary | 607.0 | 74.7 | 11.7 (2.1) | 11.8 (2.7) | 10.1 (2.9) | 12.0 (2.0) |
| | Collegiate | 58.0 | 7.1 | 13.1 (1.7) | 13.8 (2.2) | 12.2 (2.6) | 13.4 (1.8) |
| | Tertiary | 36.0 | 4.4 | 13.6 (1.6) | 14.2 (2.0) | 11.8 (3.0) | 14.4 (1.4) |
| | | | | P<0.001 | P<0.001 | P<0.001 | P<0.001 |
| Resides with | Resides with | | | | | | |
| | Spouse only | 254.0 | 31.6 | 12.3(2.1) | 12.6(2.6) | 11.8(2.1) | 12.6(1.9) |
| | Children only | 218.0 | 27.1 | 10.8(2.3) | 11.5(2.6) | 7.9(2.0) | 11.8(2.0) |
| | Spouse and children | 211.0 | 26.3 | 12.7(2.1) | 12.9(2.4) | 12.0(2.8) | 12.6(1.9) |
| | Alone | 95.0 | 11.8 | 11.2(2.1) | 10.5(2.8) | 7.4(1.7) | 11.2(2.2) |
| | Others | 25.0 | 3.1 | 10.2(2.1) | 8.6(1.8) | 6.9(1.9) | 9.6(2.3) |
| | | | | P<0.001 | P<0.001 | P<0.001 | P<0.001 |
| Occupation | Occupation | | | | | | |
| | Retired | 136.0 | 16.7 | 13.1 (1.9) | 13.7 (2.3) | 11.9 (2.6) | 13.8 (1.7) |
| | Currently working outside home | 295.0 | 36.3 | 12.2 (2.0) | 11.9 (2.7) | 11.0 (2.8) | 11.8 (1.8) |
| | Never worked outside home | 382.0 | 47.0 | 11.0 (2.3) | 11.4 (2.7) | 8.9 (2.8) | 11.7 (2.2) |
| | | | | P<0.001 | P<0.001 | P<0.001 | P<0.001 |
| Presence of monthly income | Presence of monthly income | | | | | | |
| | Yes | 676.0 | 83.8 | 11.8(2.3) | 12.1(2.6) | 10.2(3.0) | 12.2(2.0) |
| | No | 131.0 | 16.2 | 11.8(2.6) | 11.3(3.2) | 9.9(3.2) | 11.3()2.6 |
| | | | | 0.976 | 0.002 | 0.349 | 0.000 |
| Level of monthly income | Level of monthly income | | | | | | |
| | Below national poverty line | 384.0 | 54.9 | 11.0(2.2) | 10.9(2.4) | 9.2(2.6) | 11.3(1.7) |
| | Above national poverty line | 315.0 | 45.1 | 12.8(2.2) | 13.4(2.5) | 11.5(3.1) | 13.2(2.0) |
| | | | | P <0.001 | P <0.001 | P <0.001 | P <0.001 |
| Income adequacy | Income adequacy | | | | | | |
| | Adequate | 300.0 | 39.8 | 12.5(2.2) | 13.4(2.5) | 11.2(2.9) | 13.2(1.7) |
| | Inadequate | 454.0 | 60.2 | 11.3(2.2) | 11.2(2.6) | 9.5(2.9) | 11.5(2.0) |
| | | | | P<0.001 | P<0.001 | P<0.001 | P<0.001 |
| Living place | Living place | | | | | | |
| | Own house | 526.0 | 65.1 | 12.3(2.2) | 12.5(2.6) | 11.2(2.7) | 12.5(1.8) |
| | Others house | 265.0 | 32.8 | 10.7(2.2) | 11.0(2.8) | 8.1(2.4) | 11.4(2.3) |
| | Self-rented house | 17.0 | 2.1 | 11.0(1.6) | 10.8(2.9) | 8.5(2.9) | 11.5(3.1) |
| | | | | P <0.001 | P <0.001 | P <0.001 | P <0.001 |
## Multivariate analysis
Tables 3–6 show the effect of various socio-economic factors on physical health, psychological, social participation, and environmental domains. Factors such as male sex, higher monthly income, income adequacy, and home ownership all contributed positively to the physical health domain. While being older and not having formal education had a negative impact on the physical health domain of QOL.
In regards to the psychological domain of QOL, having more children, living with a spouse and or children, higher monthly income perceived income adequacy and ownership of house were positively contributed to the psychological domain of the QOL. The following factors negatively contributed to the psychological domain of QOL: higher age, not having formal education.
The social relationship domain was positively contributed by the factors such as male sex, having a spouse, living with spouse and or children, higher monthly income, income adequacy and ownership of a house, while negatively contributed by higher age, never going to work outside the home.
The environmental health domain of QOL was positively contributed by a higher number of children, living with spouses and or children, higher monthly income, income adequacy, and ownership of a house, while not having formal education was negatively contributed.
## Discussion
The scores of all four domains of QOL of the elderly in Jaffna district were lower than the global average of WHOQOL-bref scores of the elderly age group [18]. Sri Lanka faced three decades of civil war, which ended in 2009, Jaffna district was one of the most affected places by the war. The war-damaged the Jaffna community in all physical, psychological, social, and environmental dimensions. The lowest QOL of the elderly in the current study might be the residual effect of the said war. Further, the proportion of chronic diseases is high in the Jaffna district [19]. Chronic diseases might be a reason for the reduced physical health domains of the QOL.
Results show that the social participation domain had the lowest mean score in this study. The traditional social support system is being overwhelmed by the rapid expansion of the elderly population in developing countries [20]. As a result, the elderly have become a neglected population category. This could be one of the reasons for the lowest social participation among the elderly in the current study. Furthermore, deaths of close ones, as well as property losses due to the war [21] might lead to the social participation domain of QOL being the lowest of the four domains. The current finding is supported by the studies conducted in India [20] and Bangladesh [22], where the lowest domain score of QOL was reported for the social participation domain.
In the current study, age was found to be strongly associated with all domains of QOL except the elderly’s environmental domain. Similar findings have been observed in research undertaken in Sri Lanka [12] and other countries [23–26]. Dependency of the elderly increases as they get older [26], which may contribute to the lower QOL with the increase in age. When compared to the two age categories, the elderly who are young-elderly (age below 70 years) had the highest QOL in all domains. Even though physical health can be impaired with aging [27], the young elderly are still in the early stages of physical deterioration. They can carry out their daily activities more independently than elderly with higher age. This might increases their QOL. Furthermore, as one gets older, the risks of losing a spouse and being separated from one’s children rise. In the current study, the risk of living alone is higher among the elderly aged over 70 years compared to the elderly aged below 70 years of age (ODD 1.788 CL: 1.194–2.678), which can have a detrimental impact on one’s quality of life. These factors might contribute to the Higher QOL of the young elderly than their counterparts.
In the current study, males perceived higher QOL than females in all domains except the environmental domain. A prior study conducted in Jaffna supported the present finding [13]. Further research was carried out in Qatar [28], Lebanon [29], India [20, 30], Bangladesh [31], Vietnam [32] Indonesia [29] Malaysia [33], stating that gender has a substantial impact on QOL, with females experiencing lower QOL than males. The countries indicated above in Southeast Asia and the Middle East adhere to some traditional gender ideals, which influence male and female role performance in society which might contribute to the difference in QOL with gender. It can be noted that there was no gender difference in QOL was reported in the studies conducted in Japan [34] and Thailand [35]. These inconsistencies may be associate with the differences in cultural values surrounding gender in different countries. Jaffna society was known as a culturally enriched society with a higher ideological linage which may explain the gender differences in the perception of QOL of the elderly in the current study and reflect the similar pattern of its associated neighbouring countries.
Only the environmental domain scores showed significant differences between the religious groups in the current study. The elderly engage in more religious activities than other age groups [36] to cope with the challenges associated with aging [37]. Religion was known to play a significant role in the lives of the elderly Sri Lankans [12]. Practices and accompanying culture were significant among the three religious communities of this study, but they were not reflected in their QOL. This can be explained by the fact that the level of religious belief and faith may be linked to QOL, but religion itself does not alter the QOL of the elderly.
All four QOL domains of elderly QOL were found to be significantly associated with the marital status of the elderly. The majority of the elderly in the current study had spouses ($61.6\%$). They reported higher QOL when compared to the unmarried, widowed, or divorced/separate elderly. Prior studies in Sri Lanka [13, 38] and other countries had found that married status was linked to elderly people’s quality of life [20, 23, 26, 30]. It was found that people with spouses had better mental and physical health and lived longer than the people without spouses [39, 40, 41]. Also the elderly with a spouse can share and alleviate their partner’s distress. Furthermore, having a spouse ensures the receipt of care and avoids loneliness. In addition to that, having a healthy relationship with one’s spouse will improve one’s mental health [42]. These factors might contribute to the higher QOL of the elderly who have a spouse than those who don’t have a spouse.
The number of children of the elderly was significantly associated with all three QOL domains except for the physical health domain in the present study. QOL was highest among the elderly with more than four children. It was reported in a previous study that the elderly look at the children as resources for their old age [43]. Having more than four children [43], and living in a large family [32] also contributed to the elderly’s better QOL. A Sri Lankan study reported that the elderly are respected by their children as a good traditional practice [5]. Also, elderly people with more children were found to have enjoyed better living conditions, received more monetary assistance, and had more providers for in-kind and emotional care in Sri Lanka [5]. This may have contributed positively to pursuing a good QOL for the elderly. Jaffna society is also a traditional society with a higher emotional bond with their children. Having more children is associated with better QOL among the elderly in the current study.
The level of education was found to be significantly related to all four domains of QOL of the elderly in the current study. Several previous studies have also found a positive relationship between education level and on the perception of QOL among the elderly [12, 31, 41, 44] *Education is* a determinant of many other elements in an individual’s life, including occupation, income, living arrangement, property ownership, and social capital. It can have both a direct and indirect effect on an individual’s QOL. According to other researchers, people with a higher level of education are more likely to live a healthy lifestyle [45, 46], have better problem-solving abilities [22], are happier, have strong social relationships, and have better self-assessed health [47, 48]. This explains why the elderly with higher educational levels perceive higher QOL compared to those who have a lower level of education.
The majority of the elderly live with either their spouse or children or both in the current study. The elderly who reside with their spouse and or children perceive significantly higher QOL in all the domains, compared to the elderly who live alone or with other living arrangements. This finding was supported by several researchers in Sri Lanka [12], and other countries [20, 23, 26, 31, 32, 49]. Elderly people who live with their spouse and children are more likely to receive better care and emotional support, which could explain the current study’s findings.
Around one-third of the elderly ($36.3\%$) were currently working, and around half of the elderly ($47.0\%$) were unemployed while $16.7\%$ of them were retired from their work. Further employment status was significantly associated with all the domains of QOL of the elderly in the current study. The highest mean QOL was perceived by the elderly who were retired from previous employment. This finding was supported by several studies [13, 30, 31, 50, 51]. Even though the elderly who are currently working can have good physical functioning and social contact compared to the elderly who are retired, they perceive comparably low QOL. Most of the elderly ($94.8\%$) who were retired from their previous employment were pensioners in the present study. This can ensure financial independence which may contribute to their higher QOL. Also, the elderly who were never going to work scored the lowest QOL in all domains compared to the currently working and retired elderly. This may be associated with financial dependency and less interaction with their society.
A considerable proportion of participants ($83.8\%$) had a consistent monthly income. A very higher proportion ($83.8\%$) of respondents had a regular monthly income. It was significantly associated only with the psychological and environmental domains of QOL of the elderly. In other studies [52], the presence of a monthly income was identified as an associated factor of QOL for the elderly. While some researchers stated that it was not associated with the QOL of the elderly [53], as the level of income, income source, and wealth management abilities were important factors that contributed to the perception of QOL more than having a regular monthly income [53, 54].
Previous studies [55] have proven that economic independence was associated with higher QOL among the elderly. A higher level of income ensures that basic needs are met, that social participation and social respect are maintained, and that the elderly are not concerned about unexpected healthcare expenses, thereby improving their QOL. In the current study, income adequacy was significantly associated with all domains of the QOL of the elderly. It was reported as a direct contributing factor of QOL in the studies conducted in Sri Lanka [12, 13] and other countries [24, 25, 31, 50]. Income adequacy ensures a sense of financial security and a sense of independence from the needs of others. Perceived adequateness of income can reduce the fear of health care costs which may increase the QOL of the elderly.
The elderly who own a house had significantly higher QOL in all domains than those who live in other houses. This finding was supported by other researchers [15, 31, 32, 53]. The elderly prefer to live in their own houses to preserve their authority in the family and to maintain the provider role. It was reported that ownership of the house had a marked impact on the receipt of care of the elderly from their family members [15]. Thus, ownership of a house may ensure dignity and increase the probability of receiving better care for the elderly. These together may be contribute to the higher QOL of the elderly who live in their own house than the elderly who live in other houses.
## Conclusion and recommendation
The QOL of the elderly in the Jaffna district of Sri *Lanka is* low. The environment domain had the higher score, while the social health domain had the lowest score in this study. Marital status, level of education, living arrangement, employment status, source of income, level of income, income adequacy, and ownership of a house were associated with all the domains of QOL of the elderly.
Male gender, having spouse, number of children, living with spouse and or children, presence of monthly income, income adequacy, and ownership of house were positively contributed at least one domain of QOL of elderly. While, being older, in the current study of primary education, never went for work outside home, had a negative impact on at least one domain of QOL of elderly. Interventions to improve the QOL of the elderly are expected. When planning such interventions, unique factors such as the number of children, living arrangements, and ownership of house should be considered in addition to the known contributing factors to the QOL of the elderly. Also Interventions that place great emphasis on social support and social participation of elderly are recommended.
## References
1. **World Population Ageing.**. *Department of Economic and Social Affairs, Population Division.* (2013.0)
2. McNicoll G.. **World Population Ageing 1950–2050**. *Population and Development Review* (2002.0) **28** 814
3. Khaje-Bishak Y, Payahoo L, Pourghasem B, Asghari Jafarabadi M. **Assessing the quality of life in elderly people and related factors in Tabriz, Iran.**. *J Caring Sci* (2014.0) **3** 257-263. DOI: 10.5681/jcs.2014.028
4. 4United Nations. World Population Ageing 2017—Highlights, Department of Economic and Social Affairs, Population Division. 2017. https://www.un.org/en/development/desa/population/publications/pdf/ageing/WPA2017_Highlights.pdf Accessed 19 July 2022
5. 5World Bank. Sri Lanka addressing the needs of an aging population. Human Development Unit: South Asia Region. 2008. https://openknowledge.worldbank.org/bitstream/handle/10986/8105/433960ESW0P09410gray0cover01PUBLIC1.pdf?sequence=1&isAllowed=y Accessed 19 July 2022
6. Siddhisena KA. *Socio-economic implications of ageing in Sri Lanka: an overview. Oxford Institute of Ageing Working Papers.* (2005.0) **1** 27
7. 7United Nations. World Population and Future Resources, Population Division. 2009
8. 8Department of Census and Statistics. Census of Population and Housing 2012, provisional results, Department of Census and Statistics, Sri Lanka. 2013 http://www.statistics.gov.lk/pophousat/cph2011/pages/activities/Reports/SriLanka.pdf Accessed 19 July 2022
9. 9Department of Census and Statistics. Sri Lanka demographic and health survey 2016. 2016 http://www.statistics.gov.lk/Resource/en/Health/DemographicAndHealthSurveyReport-2016-Contents.pdf
10. 10Perera ELSJ. Ageing Population in Sri Lanka: Emerging Issues, Needs, and Policy Implications, United Nations Population Fund, Sri Lanka, Technical Report October 2017. 2017. https://srilanka.unfpa.org/sites/default/files/pub-pdf/UNFPA%20Ageing%20Monograph%20Report_0.pdf
11. Nigam Y, Knight J, Bhattacharya S, Bayer A. **Physiological changes associated with aging and immobility**. *Journal of aging research* (2012.0) **2**. DOI: 10.1155/2012/468469
12. Rathnayake S, Siop S. **Quality of life and its determinants among older people living in the rural community in Sri Lanka.**. *Indian Journal of Gerontology* (2015.0) **29** 131-153
13. Kanagaretnam T, Kamalarupan L, Thabotharan D, Coonghe PAD. **Quality of life and its selected determinants among elderly people living in Nallur.**. *Proceedings of Jaffna university international research conference-2018.* (2018.0) **198**
14. 14Department of census and Statistics. District Statistical Hand Book Jaffna. 2016
15. Perera R.. **A sociological study on elderly care in an urban community in Sri Lanka.**. *Proceedings of the second academic sessions –2004.* (2004.0) 142-146
16. Daniel WW. *Biostatistics: A Foundation for Analysis in the Health Sciences* (1999.0)
17. Lwanga SK, Lemeshow S. *Sample size determination in health studies A practical manual* (1991.0) **38**
18. Skevington SM, Lotfy M, O’Connell KA. **The World Health Organization’s WHOQOL-BREF quality of life assessment: psychometric properties and results of the international field trial.**. *Qual Life Res.* (2004.0) **13** 299-310. DOI: 10.1023/B:QURE.0000018486.91360.00
19. 19Department of census and statistics. National Survey on Self-reported Health in Sri Lanka. 2014 http://www.statistics.gov.lk/Resource/en/Health/NationalSurveyonSelf-reportedHealthinSriLanka2014.pdf Accessed on 17 June 2022
20. Kumar S G, Majumdar AGP. **Quality of Life (QOL) and Its Associated Factors Using WHOQOL-BREF among Elderly in Urban Puducherry, India.**. *Journal of Clinical and Diagnostic Research* (2014.0) **8** 54-57. DOI: 10.7860/JCDR/2014/6996.3917
21. 21International Crisis Group. Sri Lanka’s Authoritarian Turn: The Need for International Action, Belgium, Asia Report. 2013;(243). https://www.refworld.org/pdfid/5124deb32.pdf
22. Uddin MA, Soivong P, Lasuka D, Juntasopeepun P. **Factors related to quality of life among older adults in Bangladesh: A cross sectional survey.**. *Nurs Health Sci* (2017.0) **19** 518-524. DOI: 10.1111/nhs.12385
23. Sowmiya KR. **A Study on Quality of Life of Elderly Population in Mettupalayam, A Rural Area of Tamilnadu.**. *National journal of Research in Community Medicine* (2012.0) 123-177. DOI: 10.26727/NJRCM.2012.1.3.139–143
24. Nilsson J, Rana A. K, Kabir Z N. **Social capital and quality of life in old age: results from a cross-sectional study in rural Bangladesh**. *Journal of Aging and Health* (2006.0) **18** 419-434. DOI: 10.1177/0898264306286198
25. Rouhani S, Zoleikani P. **Socioeconomic status and quality of life in elderly people in rural area of Sari-Iran**. *Life Science Journal* (2013.0) **10** 74-78. DOI: 10.7537/marsaaj100718.09
26. Minh VH, Ng N, Bypsaa P, Wall S. **Patterns of subjective quality of life among older adults in rural Vietnam and Indonesia**. *Geriatrics & gerontology international* (2012.0) **12** 397-404. DOI: 10.1111/j.1447-0594.2011.00777.x
27. Tajvar M, Arab M, Montazeri A. **Determinants of health-related quality of life in elderly in Tehran, Iran.**. *BMC Public Health* (2008.0) **8** 323. DOI: 10.1186/1471-2458-8-323
28. Alipour F, Sajadi H, Fruzan A, Biglarian A. **The role of social support on quality of life of elderly.**. *Social Welfare Quarterly.* (2009.0) **33** 147-165
29. Sabbah I, Drouby N, Sabbah S, Retel- Rude N, Mercier M. **Quality of life in rural and urban populations in Lebanon using SF- 36 health survey.**. *Health Quality Life Outcomes* (2003.0) **1** 30. DOI: 10.1186/1477-7525-1-30
30. Kamra D.. **A community based epidemiological study on quality of life among rural elderly population of Punjab**. *International Journal of Recent Trends in Science and Technology* (2014.0) **11** 192-197
31. Khan M N, Mondal MNI, Hoque N, Islam M S, Shahiduzzaman M. **A study on quality of life of elderly population in Bangladesh.**. *American Journal of Health Research* (2014.0) **2** 152-157. DOI: 10.11648/j.ajhr.20140204.18
32. Hoi LV, Chuc NT, Lindholm L. **Health-related quality of life, and its determinants, among older people in rural Vietnam.**. *BMC public health.* (2010.0) **10** 1. DOI: 10.1186/1471-2458-10-549
33. Khan A R, Tahir I. **Influence of social factors to the quality of life of the elderly in Malaysia.**. *Open Medicine Journal* (2004.0) **1** 29-35. DOI: 10.2174/1874220301401010029
34. Lee Y, Shinkai SA. **Comparison of correlates of self- rated health and functional disability of older persons on the Far East: Japan and Korea.**. *Archive of Gerontology of Geriatrics* (2003.0) **37** 63-76. DOI: 10.1016/s0167-4943(03)00021-9
35. Assantachai P, Maranetra N. **Nationwide of the health and quality of life of elderly Thais attending clubs for the elderly**. *Journal of Thai Medical Association* (2003.0) **86** 938-946
36. Greenfield EA, Vaillant GE, Marks NF. **Do formal religious participation and spiritual perceptions have independent linkages with diverse dimensions of psychological well-being?.**. *Journal of Health and Social Behavior* (2009.0) **50** 196-212. DOI: 10.1177/002214650905000206
37. 37Kaplan DB, and Berkman BJ. Religion and spirituality in older adults. Merck Manual Professional version. 2019. Retrieved from https://www.merckmanuals.com/professional/geriatrics/social-issues-in-older-adults/religion-and-spirituality-in-older-adults on 19th July 2022
38. 38Department of Census and Statistics. Census of Population and Housing 2012. Department of Census and Statistics, Sri Lanka. 2012.
39. Hughes ME, Waite L J. **Marital biography and health at mid-life**. *Journal of Health and Social Behavior* (2009.0) **50** 344. DOI: 10.1177/002214650905000307
40. Simon R W.. **Revisiting the relationships among gender, marital status, and mental health**. *The American Journal of Sociology* (2002.0) **107** 1065-1096. DOI: 10.1086/339225
41. Shah VR, Christian DS, Prajapati AC, Patel MM, Sonaliya K N. **Quality of life among elderly population residing in urban field practice area of a tertiary care institute of Ahmedabad city, Gujarat.**. *J Family Med Prim Care* (2017.0) **6** 101-105. DOI: 10.4103/2249-4863.214965
42. Bierman A.. **Marital Status as Contingency for the Effects of Neighbourhood Disorder on Older Adults’ Mental Health. The journals of gerontology.**. *Series B, Psychological sciences and social sciences* (2009.0) **64** 425-34. DOI: 10.1093/geronb/gbp010
43. Knodel J, Chayovan N. **Inter-generational Family Care for and by Older People in Thailand**. *International Journal of Sociology and Social Policy* (2012.0) **32** 682-694. DOI: 10.1108/01443331211280719
44. Kumar D, Shankar H. **Prevalence of Chronic Diseases and Quality of Life among Elderly People of Rural Varanasi**. *International Journal of Contemporary Medical Research* (2018.0) **5** 2454-7379. DOI: 10.21276/ijcmr.2018.5.7.16
45. Van-Oort FVA, Van-Lenthe F, Mackenbach JP. **Co-occurrence of lifestyle risk factors and the explanation of education inequalities in mortality: results from the GLOBE study.**. *Preventive Medicine.* (2004.0) **39** 1126-1134. DOI: 10.1016/j.ypmed.2004.04.025
46. McDaid O, Hanly MJ, Richardson K, Kee F, Kenny RA, Savva GM. **The effect of multiple chronic conditions on self-rated health, disability and quality of life among the older populations of Northern Ireland and the Republic of Ireland: a comparison of two nationally representative cross-sectional surveys**. *BMJ Open* (2013.0) **3**. DOI: 10.1136/bmjopen-2013-002571
47. Poljicanin T, Ajdukovic D, Sekerija M, Pibernik-Okanovic M, Metelko Z, VuletiMavrinac G. **Diabetes mellitus and hypertension have comparable adverse effects on health-related quality of life**. *BMC Public Health* (2010.0) **10** 12. DOI: 10.1186/1471-2458-10-12
48. Lasheras C, Patterson AM, Casado C, Fernandez S. **Effects of Education on the Quality of Life, Diet, and Cardiovascular Risk Factors in an Elderly Spanish Community Population, Experimental Aging Research.**. (2001.0) **27** 257-270. DOI: 10.1080/036107301300208691
49. Deluga A, Kosicka B, Dobrowolska B, Chrzan-Rodak A, Jurek K, Wrońska I. **Lifestyle of the elderly living in rural and urban areas measured by the fantastic Life Inventory.**. *Annals Agricultural Environmental Medicine* (2018.0) **25** 562-567. DOI: 10.26444/aaem/86459
50. Naing MM, Nanthamongkolchai S, Munsawaengsub C. **Quality of Life of the Elderly People in Einme Township Irrawaddy Division, Myanmar.**. *Asia Journal of Public Health* (2010.0) 4-10
51. Kumari R, Dewan D, Langeri B, Gupta RK, Singh P. **Quality of Life and Its Associated Factors: A Comparative Study among Rural and Urban Elderly Population of North India**. *National Journal of Community Medicine* (2018.0) **9** 420-425
52. Kar B.. **Factors affecting quality of life of older- persons- a qualitative study from Bhubaneswar, India.**. *Journal of Geriatric Care and Research* (2017.0) **4** 47-54
53. Eliasi LG, Rasi H A. **Factors affecting quality of life among elderly population in Iran.**. *Humanities and Social Sciences* (2017.0) **5** 26-30
54. Akinyemi AI. **Assessment of the influence of socio- economic status on aging males’ symptoms in Ijesaland, South-Western Nigeria.**. *The Journal of Men’s Health and Gender* (2012.0) **9** 51-57
55. 55World Health Organization. World report on health and ageing. 2015. *World report on health and ageing* (2015.0)
|
---
title: Utility of silhouette showcards to assess adiposity in three countries across
the epidemiological transition
authors:
- Tyler O. Reese
- Pascal Bovet
- Candice Choo-Kang
- Kweku Bedu-Addo
- Terrence Forrester
- Jack A. Gilbert
- Julia H. Goedecke
- Estelle V. Lambert
- Brian T. Layden
- Lisa K. Micklesfield
- Jacob Plange-Rhule
- Dale Rae
- Bharathi Viswanathan
- Amy Luke
- Lara R. Dugas
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021870
doi: 10.1371/journal.pgph.0000127
license: CC BY 4.0
---
# Utility of silhouette showcards to assess adiposity in three countries across the epidemiological transition
## Abstract
The Pulvers’ silhouette showcards provide a non-invasive and easy-to-use way of assessing an individual’s body size perception using nine silhouette shapes. However, their utility across different populations has not been examined. This study aimed to assess: 1) the relationship between silhouette perception and measured anthropometrics, i.e., body mass index (BMI), waist circumference (WC), waist-height-ratio (WHtR), and 2) the ability to predict with silhouette showcards anthropometric adiposity measures, i.e., overweight and obesity (BMI ≥ 25 kg/m2), obesity alone (BMI ≥ 30 kg/m2), elevated WC (men ≥ 94 cm; women ≥ 80 cm), and WHtR (> 0.5) across the epidemiological transition. 751 African-origin participants, aged 20–68 years old, from the United States (US), Seychelles, and Ghana, completed anthropometrics and selected silhouettes corresponding to their perceived body size. Silhouette performance to anthropometrics was examined using a least-squares linear regression model. A receiver operator curve (ROC) was used to investigate the showcards ability to predict anthropometric adiposity measures. The relationship between silhouette ranking and BMI were similar between sexes of the same country but differed between countries: 3.65 [$95\%$ CI: 3.34–3.97] BMI units/silhouette unit in the US, 3.23 [2.93–3.74] in Seychelles, and 1.99 [1.72–2.26] in Ghana. Different silhouette cutoffs predicted obesity differently in the three countries. For example, a silhouette ≥ five had a sensitivity/specificity of $77.3\%$/$90.6\%$ to predict BMI ≥ 25 kg/m2 in the US, but $77.8\%$/$85.9\%$ in Seychelles and $84.9\%$/$71.4\%$ in Ghana. Ultimately, silhouettes predicted BMI, WC, and WHtR similarly within each country and sex but not across countries. Our data suggest that Pulvers’ silhouette showcards may be a helpful tool to predict anthropometric and adiposity measures in different populations when direct measurement cannot be performed. However, no universal silhouette cutoff can be used for detecting overweight or obesity status, and population-specific differences may stress the need to calibrate silhouette showcards when using them as a survey tool in different countries.
## Introduction
The prevalence of overweight and obesity is increasing in populations spanning the epidemiological transition and may be particularly high in individuals of African-origin [1–4]. In addition, elevated weight has been associated with the development of non-communicable diseases (NCDs) [5–8]. To assess for obesity, body mass index (BMI, kg/m2) is widely used because of its simplicity and ease of measurement. However, BMI does not discriminate well between adipose and lean mass. Therefore, waist circumference (WC) and waist-to-height ratio (WHtR) have been suggested to predict adiposity better and have been shown to correlate well with fat mass as assessed by accurate methods such as computed tomography (CT) [9–12].
Measures of adiposity that do not rely on actual measurements may be helpful in some situations. Examples include surveys and studies in public health, anthropology, economics, and marketing that must be performed without direct contact with a respondent (e.g., mail-order or internet-based) or situations to avoid the burden of asking participants to remove clothing. Furthermore, self-reported adiposity (e.g., height and weight) is prone to reporting bias, depends on access to anthropometric tools like scales, and can be influenced by cultural views on body size [13–19].
Initially developed by Stunkard and colleagues, sex-specific silhouette showcards (referred to as “silhouettes” hereafter) can be used to determine one’s perception of their body size. This tool relies on presenting a series of drawings of distinct body sizes in an increasing sequence. Respondents then select the silhouette they think best reflects their body size relative to objective measurements [20]. Silhouettes should be ethnically ambiguous enough to be used in different cultures but still detailed enough to be relatable. A variety of silhouette tools have been developed and validated for different populations [21–24]. For example, Pulvers and colleagues created culturally relevant silhouette showcards for African Americans (Fig 1) [25]. These silhouettes were validated in different populations of African-origin such as Seychelles, the Caribbean, and the US [25–28]. However, while many studies have shown a good association between the silhouettes and anthropometrics, including for the prediction of obesity, most studies have only assessed their validity in a single population at a time [21–32]. Also, no studies have directly compared the associations of silhouette ranking and anthropometrics between countries with different population mean BMI levels or stages of development. Thus, assessing the validity of silhouettes to predict adiposity in different populations may be challenging. As such, cross-cultural evaluation should rely on studies that use the same methodology in different countries [29, 33–36].
**Fig 1:** *Pulvers’ silhouettes designed for populations of African-origin.Source: Pulvers 2004, Obesity Res.*
Therefore, our study aims to assess: 1) the relationship between silhouette perception and measured anthropometrics, i.e., body mass index (BMI), waist circumference (WC), waist-height-ratio (WHtR), and 2) the ability to predict with silhouette showcards anthropometric adiposity measures, i.e., overweight and obesity (BMI ≥ 25 kg/m2), obesity alone (BMI ≥ 30 kg/m2), elevated WC (men ≥ 94 cm; women ≥ 80 cm), and WHtR (> 0.5) in three African-origin populations with differing population mean BMI levels, as well as, stages of social and economic development.
## Study populations and ethics approval
This study is a subset analysis from the METS-Microbiome study (R01-DK111848), for which the protocol has been published [37]. The first METS cohort consisted of 2,506 participants enrolled in December 2010 and January 2011, and 751 participants from this original cohort participated in this study. The METS-Microbiome study continues yearly measurements of participants initially recruited for the Modeling the Epidemiological Transition Study (METS; R01-DK080763). The METS-Microbiome study includes five African-origin populations spanning the epidemiologic transition varying by the United Nations Human Development Index (HDI) 2010. HDI is a statistical composite index of education, life expectancy, and per capita income indicators used to rank countries by human development [38, 39].
The current data presented was collected between 2018–2019 from participants in metropolitan Chicago, IL, US (HDI: 0.92), the mixed urban/rural Seychelles islands (0.80), and rural Ghana (0.59) [37, 39]. These three sites represent different social and economic development stages and have a largely different prevalence of obesity [39]. All participants were of African-origin except for Seychelles, where both African participants and participants of mixed ethnicity ancestry were included.
The METS-Microbiome protocol used in this study was approved by the Institutional Review Board of Loyola University Chicago, Chicago, IL, US (LU 209537), the National Research Ethics Committee of Seychelles, and the Committee of Human Research Publication and Ethics of Kwame Nkrumah University of Science and Technology, Kumasi, Ghana [37]. In addition, written informed consent was obtained from all participants.
## Survey and body size silhouette showcards
The survey component of the METS-Microbiome study consisted of a face-to-face interview performed by centrally trained personnel, capturing participants’ sociodemographic data, health-related behaviors, and medical history. Participants were also presented with sex and ethnicity-specific silhouette showcards created by Pulvers (Fig 1) [25, 27]. This nine-image tool displayed sex-specific body sizes in increasing order ranging from very thin to severely obese. Participant’s perceived body size was assessed by asking, “In the drawing, which figure best reflects how you think you look with regards to your body shape?”. Participants’ responses were recorded on a scale of 1 (representing the thinnest silhouette) to 9 (representing the most obese).
## Anthropometric and adiposity measurements
Anthropometric data, including measured height (m), weight (kg), and waist circumference (cm), was collected from each participant. Across all sites, standardized equipment and protocols were used, as previously published [37]. Body mass index (BMI, weight/height2) was calculated and classified as underweight (BMI < 18.5 kg/m2), normal weight (BMI 18.5–24.9 kg/m2), overweight (BMI 25.0–29.9 kg/m2) or obese (BMI ≥ 30 kg/m2) [40]. A dichotomous waist circumference (cm) variable was used to classify the presence of central obesity as defined by the International Diabetes Federation (≥ 94 cm in men, ≥ 80 cm in women) for European or African-origin individuals [11]. WHtR (waist in cm/ height in cm) was calculated and dichotomized using a widely used cut-off point for normal (WHtR ≤ 0.5) or increased central obesity (WHtR > 0.5) [41].
## Statistical analyses
Participant characteristics were summarized using means and $95\%$ confidence intervals (CI). Proportions were calculated and presented as a percent (%) and $95\%$ CI for categorical variables. Participant characteristics by sex were compared to the US using a two-sample t-test. In line with previous studies on this topic, Spearman’s rank correlation coefficients were used to describe the associations between the self-reported perceived silhouette ranking and BMI, WC, and WHtR.
Mean BMI and $95\%$ CI for each silhouette rank were determined by sex and by country. To assess how the change in one silhouette ranking by anthropometric measures, such as BMI units per silhouette unit, differed between countries and sex, we estimated the linear regression coefficient slopes by sex and country with accompanying $95\%$ CI. A robust regression analysis was also performed, which lessens the influence of outliers on the regression coefficient estimates. Estimates were almost identical to those in the least-squares linear regression.
The self-reported silhouette showcards were assessed for accuracy in predicting widely used dichotomized anthropometric adiposity measures, e.g., overweight and obesity (BMI ≥ 25 kg/m2) or obesity alone (BMI ≥ 30 kg/m2), elevated waist circumference (≥ 94 cm in men, ≥ 80 cm in women) and elevated waist-to-height ratio (WHtR > 0.5) using sex and country-specific receiver-operator curve (ROC) analysis [26]. In line with previous studies, we used the area under the curve (AUC, i.e., the c-statistic) and sensitivity and specificity associated with different cutoffs of the silhouettes to predict these dichotomous adiposity categories.
All statistical analyses were performed using STATA SE 12 (StataCorp, College Station, TX, US).
## Demographics
Table 1 shows the main characteristics of the 751 participants from the three countries. The study sample consisted of men and women aged 20–68 years old. Approximately $66\%$ of the whole sample identified as female. Mean age for men ranged from highest in the US (47.1 years) and significantly lower in Seychelles men (45.2) ($p \leq 0.05$). Women’s mean age was also highest in the US (45.3) and significantly lower in Ghanaian women (41.4).
**Table 1**
| Unnamed: 0 | United States | United States.1 | Seychelles | Seychelles.1 | Ghana | Ghana.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | Men (N = 88) | Women (N = 177) | Men (N = 100) | Women (N = 183) | Men (N = 67) | Women (N = 136) |
| Age (years) | 47.1 [45.9–48.3] | 45.3 [44.3–46.2] | 45.2 [44.2–46.2]* | 44.3 [43.4–45.2] | 45.6 [43.4–47.7] | 41.4 [40.0–42.8]* |
| Height (cm) | 174.8 [173.4–176.2] | 164.8 [163.7–165.8] | 173.6 [172.4–174.7] | 162.1 [161.1–163.2]* | 167.9 [166.3–169.6] * | 159.1 [158.2–160.0]* |
| Weight (kg) | 88.2 [83.6–92.9] | 97.3 [93.8–100.8] | 85.3 [82.0–88.5] | 80.3 [77.4–83.1]* | 67.3 [65.0–69.5]* | 71.9 [69.5–74.4]* |
| BMI (kg/m2) | 28.9 [27.4–30.4] | 35.8 [34.5–37.0] | 28.3 [27.3–29.3] | 30.5 [29.5–31.6]* | 23.9 [23.1–24.7]* | 28.5 [27.5–29.4]* |
| Waist circumference (cm) | 100.0 [96.1–103.6] | 109.3 [106.8–111.9] | 95.8 [93.6–98.1]* | 95.6 [93.6–97.5]* | 86.3 [83.9–88.7]* | 96.1 [94.0–98.1]* |
| Waist-to-height ratio | 0.57 [0.55–0.59] | 0.66 [0.65–0.68] | 0.55 [0.54–0.57] | 0.59 [0.58–0.60]* | 0.52 [0.50–0.53]* | 0.6 [0.59–0.62]* |
| Perceived silhouette (1–9) | 4.2 [3.9–4.6] | 6.2 [5.9–6.4] | 4.5 [4.3–4.8] | 5.4 [5.1–5.6]* | 4.1 [3.6–4.5] | 5.5 [5.1–5.8]* |
| Anthropometric adiposity measures, % | Anthropometric adiposity measures, % | Anthropometric adiposity measures, % | Anthropometric adiposity measures, % | Anthropometric adiposity measures, % | Anthropometric adiposity measures, % | Anthropometric adiposity measures, % |
| Underweight | 2.3 [0.0–5.4] | 0.6 [0.0–1.7] | 2 [0.0–4.8] | 0.5 [0.0–1.6] | 1.5 [0.0–4.4] | 1.5 [0.0–3.5] |
| (BMI < 18.5 kg/m2) | 2.3 [0.0–5.4] | 0.6 [0.0–1.7] | 2 [0.0–4.8] | 0.5 [0.0–1.6] | 1.5 [0.0–4.4] | 1.5 [0.0–3.5] |
| Normal weight | 26.1 [16.9–35.4] | 6.2 [2.6–9.8] | 26 [17.3–34.7] | 21.9 [15.8–27.9]* | 61.2 [49.4–73.0]* | 27.9 [20.4–35.5]* |
| (BMI 18.5–24.9 kg/m2) | 26.1 [16.9–35.4] | 6.2 [2.6–9.8] | 26 [17.3–34.7] | 21.9 [15.8–27.9]* | 61.2 [49.4–73.0]* | 27.9 [20.4–35.5]* |
| Overweight | 36.4 [26.2–46.5] | 19.2 [13.4–25.0] | 41 [31.3–50.7] | 29 [22.4–35.6]* | 34.3 [22.8–45.8] | 35.3 [27.2–43.4]* |
| (BMI 25–29.9 kg/m2) | 36.4 [26.2–46.5] | 19.2 [13.4–25.0] | 41 [31.3–50.7] | 29 [22.4–35.6]* | 34.3 [22.8–45.8] | 35.3 [27.2–43.4]* |
| Obese | 35.2 [25.1–45.3] | 74 [67.5–80.5] | 31 [21.8–40.2] | 48.6 [41.4–55.9]* | 3 [0.0–7.1]* | 36 [27.9–44.1]* |
| (BMI ≥ 30 kg/m2) | 35.2 [25.1–45.3] | 74 [67.5–80.5] | 31 [21.8–40.2] | 48.6 [41.4–55.9]* | 3 [0.0–7.1]* | 36 [27.9–44.1]* |
| Central obesity | 55.7 [45.1–66.3] | 96.6 [93.9–99.3] | 54 [44.1–63.9] | 88.5 [83.9–93.2]* | 29.9 [18.6–41.1]* | 92.6 [88.2–97.1] |
| (WC ≥ 94 cm M; ≥ 80 cm W) | 55.7 [45.1–66.3] | 96.6 [93.9–99.3] | 54 [44.1–63.9] | 88.5 [83.9–93.2]* | 29.9 [18.6–41.1]* | 92.6 [88.2–97.1] |
| Elevated waist-to-height ratio (> 0.5) | 73.9 [64.5–83.2] | 94.9 [91.6–98.2] | 79 [70.9–87.1] | 83.6 [78.2–89.0]* | 44.8 [32.6–57.0]* | 90.4 [85.4–95.4] |
## Anthropometric and adiposity measures
All measures of size, including height, weight, and adiposity, were significantly higher in the US, intermediate in Seychelles, and lowest in Ghana, and was higher in women than men in each country. For example, mean BMI (kg/m2) in men/women was $\frac{28.9}{35.8}$ in the US, $\frac{28.3}{30.5}$ in Seychelles, and $\frac{23.9}{28.5}$ in Ghana (Table 1), with similar trends for WC and WHtR. The mean values of anthropometric measures and the prevalence of the anthropometric adiposity measures were lowest in men from Ghana. The prevalence (%) of obesity in men/women (BMI ≥ 30) followed a similar decreasing trend: $\frac{35.2}{74.0}$ in the US, $\frac{31.0}{48.6}$ in Seychelles, and $\frac{3.0}{36.0}$ in Ghana (Table 1). The prevalence of elevated WC exceeded $88.5\%$ in women from all three countries. Men in the US and Seychelles had an intermediate prevalence of elevated WC ($55.7\%$ and $54.0\%$, respectively) versus Ghanaian men ($29.9\%$). Finally, the prevalence of elevated WHtR (> 0.5) exceeded $70\%$ in men and women from all countries, except in men from Ghana ($44.6\%$) (Table 1). The different adiposity markers used in this study inter-correlated quite strongly in men and women. Correlation coefficients were 0.91 for men/0.77 for women for the association between BMI and WC, $\frac{0.94}{0.80}$ for the association between BMI and WHtR, and $\frac{0.96}{0.95}$ for the association between WC and WHtR in the US; $\frac{0.89}{0.91}$, $\frac{0.91}{0.92}$, and $\frac{0.94}{0.95}$ in Seychelles; and $\frac{0.88}{0.91}$, $\frac{0.90}{0.90}$, and $\frac{0.92}{0.96}$ in Ghana.
## Correlations between silhouette showcards and continuous anthropometric measures
Table 2 shows Spearman’s correlation coefficients between the perceived self-reported silhouette rankings with BMI, WC, and WHtR, by country and sex. These coefficients ranged between 0.71 and 0.80 in men and women in all countries, except in men in Ghana (0.55-0-58) ($p \leq 0.001$ for all coefficients).
**Table 2**
| Unnamed: 0 | Men [95% CI] | Women [95% CI] |
| --- | --- | --- |
| United States, N | 88 | 177 |
| BMI (kg/m2) | 0.77 [0.66–0.87] | 0.79 [0.73–0.85] |
| WC (cm) | 0.72 [0.60–0.83] | 0.74 [0.67–0.82] |
| Waist-to-height ratio | 0.75 [0.64–0.86] | 0.75 [0.68–0.82] |
| Seychelles, N | 100 | 183 |
| BMI (kg/m2) | 0.78 [0.71–0.87] | 0.80 [0.74–0.85] |
| WC (cm) | 0.76 [0.66–0.86] | 0.77 [0.71–0.84] |
| Waist-to-height ratio | 0.79 [0.70–0.88] | 0.76 [0.70–0.83] |
| Ghana, N | 67 | 136 |
| BMI (kg/m2) | 0.56 [0.39–0.73] | 0.74 [0.67–0.82] |
| WC (cm) | 0.55 [0.37–0.73] | 0.73 [0.65–0.82] |
| Waist-to-height ratio | 0.58 [0.41–0.75] | 0.71 [0.63–0.80] |
## Relationship between silhouette ranking and measured BMI
Table 3 shows a graded increase in mean BMI according to silhouette ranking by sex and country. The table also depicts the least-squares linear regression coefficients by sex and country between participants’ measured BMI and the self-reported silhouettes. Regression coefficients (i.e., slopes of the regression lines) were higher in women than men in all three countries. However, regression coefficients were significantly lower in Ghana than in the other two countries for both men and women. For example, in the US and Seychelles, an increase in 1 silhouette unit was associated with an increase of 3.05–3.75 BMI units (kg/m2) but only 1.15–2.06 BMI units in Ghana. Nearly identical trends were observed for WC and WHtR (S1 and S2 Tables).
**Table 3**
| Unnamed: 0 | United States (N = 265) | United States (N = 265).1 | United States (N = 265).2 | United States (N = 265).3 | Seychelles (N = 283) | Seychelles (N = 283).1 | Seychelles (N = 283).2 | Seychelles (N = 283).3 | Ghana (N = 203) | Ghana (N = 203).1 | Ghana (N = 203).2 | Ghana (N = 203).3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | N (all) | Men | Women | All | N (all) | Men | Women | All | N (all) | Men | Women | All |
| | N (all) | Mean [95% CI] | Mean [95% CI] | Mean [95% CI] | N (all) | Mean [95% CI] | Mean [95% CI] | Mean [95% CI] | N (all) | Mean [95% CI] | Mean [95% CI] | Mean [95% CI] |
| Silhouette 1 | 4 | 20 [0–40.4] | 21 [15.3–27.4] | 20 [15.6–25.3] | 0 | * | * | * | 8 | 22 [19.9–23.3] | 21 [15.8–26.3] | 21 [19.4–23.2] |
| Silhouette 2 | 14 | 23 [20.6–25.4 | 25 [19.4–29.8] | 23 [21.2–25.4] | 20 | 21 [19.1–23.4] | 22 [20.6–22.3] | 21 [20.5–22.3] | 16 | 21 [19.6–21.8] | 22 [19.6–23.8] | 21 [20.1–22.4] |
| Silhouette 3 | 24 | 25 [23.1–26.2] | 25 [23.1–26.4] | 25 [23.4–25.9] | 21 | 24 [22.3–26.0] | 23 [20.7–25.2] | 24 [22.1–24.9] | 28 | 23 [21.2–23.9] | 24 [21.4–25.8] | 23 [21.9–24.2] |
| Silhouette 4 | 40 | 28 [26.3–29.1] | 28 [25.7–29.5] | 28 [26.5–28.8] | 67 | 26 [25.0–27.0] | 26 [24.8–26.9] | 26 [25.2–26.6] | 26 | 24 [21.8–25.9] | 25 [24.1–26.2] | 25 [23.6–25.6] |
| Silhouette 5 | 43 | 30 [27.4–32.0] | 31 [29.5–32.2] | 31 [29.3–31.7] | 63 | 30 [28.5–31.2] | 29 [27.5–30.5] | 29 [28.3–30.4] | 38 | 25 [23.0–26.5] | 27 [25.2–28.1] | 26 [24.7–26.9] |
| Silhouette 6 | 57 | 33 [29.7–35.6] | 35 [33.5–36.1] | 34 [33.2–35.6] | 55 | 33 [30.6–35.6] | 31 [30.1–32.7] | 32 [30.7–33.0] | 37 | 26 [23.1–29.1] | 29 [27.6–31.0] | 29 [27.1–30.1] |
| Silhouette 7 | 40 | 41 [35.1–47.1] | 38 [36.3–40.5] | 39 [36.8–40.8] | 38 | 36 [34.1–38.4] | 35 [33.4–36.8] | 35 [33.8–36.7] | 27 | 29 [24.5–32.7] | 32 [30.2–34.0] | 32 [29.9–33.3] |
| Silhouette 8 | 25 | 48 [0–126] | 41 [38.6–43.8] | 42 [39.0–44.5] | 17 | 37 [*] | 43 [38.7–47.2] | 43 [38.6–46.6] | 17 | 26 [*] | 33 [30.3–35.7] | 33 [29.9–35.3] |
| Silhouette 9 | 18 | 55 [*] | 50 [46.6–53.4] | 50 [46.7–53.7] | 2 | * | 45 [29.8–59.5] | 45 [29.8–59.5] | 6 | * | 38 [29.4–46.5] | 38 [29.4–46.5] |
| Reg coeff | 265 | 3.51 [2.98–4.04] | 3.75 [3.28–4.21] | 3.65 [3.34–3.97] | 238 | 3.05 [2.56–3.54] | 3.34 [2.94–3.74] | 3.23 [2.93–3.54] | 203 | 1.15 [0.73–1.56] | 2.06 [1.71–2.41] | 1.99 [1.72–2.26] |
## Self-reported silhouette as a detector of overweight or obesity
Based on the graded increase in mean BMI presented in Table 3, silhouettes 4, 5, 6, and 7 were used to attempt to detect men and women who were overweight and obese or only obese. Table 4 shows that the predictive value decreased with increasing silhouette ranking. For silhouette ≥ 4, the sensitivity to predict BMI ≥ 25 ranged between 91.4–$96.7\%$ and was 98.8–$100\%$ to predict BMI ≥ 30, while for silhouette ≥ 7, sensitivity ranged from 26.9–$41.2\%$ to predict BMI ≥ 25 and was 45–$68.6\%$ to predict BMI ≥ 30. For silhouette ≥ 4, the specificity of predicting BMI ≥ 25 ranged between 47.9–$68.8\%$ and was 24.5–$38.8\%$ to predict BMI ≥ 30. For silhouette ≥ 7, specificity ranged from 98.8–$100\%$ to predict BMI ≥ 25 and was 90.1–$98.2\%$ to predict BMI ≥ 30.
**Table 4**
| Unnamed: 0 | Unnamed: 1 | Overweight and Obese | Overweight and Obese.1 | Obesity | Obesity.1 |
| --- | --- | --- | --- | --- | --- |
| | | Sensitivity (%) | Specificity (%) | Sensitivity (%) | Specificity (%) |
| Silhouette ≥ 4 | US | 91.4 | 68.8 | 98.8 | 38.8 |
| Silhouette ≥ 4 | Seychelles | 96.7 | 47.9 | 99.2 | 24.5 |
| Silhouette ≥ 4 | Ghana | 93.3 | 52.4 | 100.0 | 34.2 |
| Silhouette ≥ 5 | US | 77.3 | 90.6 | 92.0 | 67.0 |
| Silhouette ≥ 5 | Seychelles | 77.8 | 85.9 | 95.0 | 62.6 |
| Silhouette ≥ 5 | Ghana | 84.9 | 71.4 | 100.0 | 51.3 |
| Silhouette ≥ 6 | US | 59.7 | 96.9 | 79.0 | 88.4 |
| Silhouette ≥ 6 | Seychelles | 52.4 | 98.6 | 75.0 | 86.5 |
| Silhouette ≥ 6 | Ghana | 66.4 | 90.5 | 92.2 | 73.7 |
| Silhouette ≥ 7 | US | 35.6 | 100.0 | 49.4 | 97.1 |
| Silhouette ≥ 7 | Seychelles | 26.9 | 100.0 | 45 | 98.2 |
| Silhouette ≥ 7 | Ghana | 41.2 | 98.8 | 68.6 | 90.1 |
Fig 2 depicts the proportion of participants categorized as normal weight, overweight, or obese for the middle four silhouettes (4–7). Silhouettes 4 and 5 captured the largest proportion of overweight participants in the US, and Seychelles, while silhouettes 5 and 6 captured the most overweight participants in Ghana. When assessing obesity status, silhouette 7 in the US, Seychelles, and Ghana captured most obese participants.
**Fig 2:** *Proportion with normal weight, overweight, and obese within each silhouette category in the US, Seychelles, and Ghana.N weight: normal weight (BMI 18.5–24.9 kg/m2); Overweight (BMI 25.0–29.9 kg/m2); Obese (BMI ≥ 30 kg/m2); Sey: Seychelles.*
## Performance between silhouette ranking to BMI, waist circumference, and waist-to-height ratio in predicting adiposity
Table 5 shows the sex and country-specific AUCs (i.e., c-statistic) of silhouette ranking to predict overweight and obesity status (BMI ≥ 25 kg/m2) or obesity alone (BMI ≥ 30 kg/m2). AUCs ranged between 0.79 and 0.92 in men and between 0.87 and 0.97 in women, with minor differences by sex or country. Similar AUC values were found for silhouette ranking to predict elevated WC and WHtR.
**Table 5**
| Unnamed: 0 | Unnamed: 1 | Overweight and Obese | Obesity | Elevated WC | Elevated WHtR |
| --- | --- | --- | --- | --- | --- |
| Country | Sex | AUC [95% CI] | AUC | AUC | AUC |
| US | M | 0.79 [0.70–0.89] | 0.88 [0.81–0.95] | 0.82 [0.74–0.90] | 0.84 [0.75–0.93] |
| US | W | 0.97 [0.95–0.99] | 0.88 [0.83–0.94] | 0.96 [0.93–1.00] | 0.94 [0.89–0.98] |
| US | All | 0.91 [0.86–0.95] | 0.91 [0.87–0.94] | 0.91 [0.88–0.95] | 0.91 [0.88–0.95] |
| Seychelles | M | 0.87 [0.80–0.93] | 0.86 [0.82–0.95] | 0.85 [0.77–0.92] | 0.88 [0.80–0.96] |
| Seychelles | W | 0.91 [0.87–0.95] | 0.89 [0.85–0.94] | 0.91 [0.84–0.97] | 0.89 [0.84–0.94] |
| Seychelles | All | 0.89 [0.86–0.93] | 0.89 [0.86–0.93] | 0.87 [0.83–0.92] | 0.88 [0.84–0.93] |
| Ghana | M | 0.85 [0.76–0.94] | 0.92 [0.83–1.00] | 0.77 [0.65–0.88] | 0.83 [0.73–0.92] |
| Ghana | W | 0.87 [0.80–0.93] | 0.87 [0.82–0.93] | 0.88 [0.79–0.96] | 0.86 [0.77–0.95] |
| Ghana | All | 0.87 [0.82–0.92] | 0.90 [0.86–0.94] | 0.81 [0.75–0.87] | 0.84 [0.79–0.90] |
## Discussion
This study continues the foundation established by Pulvers and colleagues in creating the silhouette showcards and subsequent validation in populations of African-origin [25–27]. However, our study is the first to use Pulvers silhouette showcards across different populations using the same methodology. We showed that the silhouette showcards have a strong relationship to measured anthropometrics, can detect overweight and obesity, and might be a helpful tool for predicting adiposity measures such as elevated BMI, WC, and WHtR in different populations of both adult men and women of mainly African-origin. However, the relationship between silhouettes and adiposity measures differed according to the country, and no universal silhouette cutoff can predict overweight or obesity across populations. Overall, our data suggest that silhouettes may be a useful tool to predict actual anthropometric and adiposity measures, conditional to adequate calibration for a specific population.
BMI and other anthropometrics correlated strongly with silhouette ranking in all populations. However, the magnitude of the linear regression coefficients between silhouette ranking and actual anthropometrics differed between the three countries in this study. For example, an increase of 1 silhouette unit was associated with an increase of 3–4 BMI units (kg/m2) in the US and Seychelles but only 1–2 BMI units in Ghana (Table 3). This difference suggests varying perceptions of one’s body shape, possibly according to mean population BMI. One may speculate that in the US and Seychelles, where mean population BMI is high, individuals with adiposity are more inclined to view a large body shape as normal compared to populations (e.g., Ghana) where mean population BMI is lower. Again, this altered view suggests that silhouette showcards need to be specific (i.e., calibrated) to different populations when used for predicting individuals’ actual anthropometrics. From a prevention perspective, the differences in perceptions of one’s body size across populations may suggest larger tolerance for larger body shapes in populations with high obesity prevalence. Overall, this underlies that silhouettes can have a role in assessing body size in populations when direct measurements cannot be made (i.e., for surveillance purposes, as evaluated in this study), but may also be used to assess perceptions and attitudes of people for weight control programs.
Our data shows that when self-reported silhouette ranking and continuous anthropometric measures correlate between men and women within a population, it is likely that the same predictive linear regression models can be used. Previous studies using different silhouette showcards have shown, Spearman correlation coefficients for the relationship between silhouettes and BMI (kg/m2 per silhouette unit) were, for example, 0.73 for men and 0.81 for women among white Americans (with a mean BMI of 25.5 kg/m2 in men and 24.1 kg/m2 in women) and 0.73 for men and 0.80 in Japanese women (with mean BMI of 23.3 kg/m2 in men and 21.5 kg/m2 in women) and 0.80 for men and 0.81 for women in Seychelles (with mean BMI of 26.4 kg/m2 in men and 29.3 kg/m2 in women) [21, 22, 26]. Our correlation coefficients for the US and Seychelles were like these previous studies (Table 2). Given that the regression coefficients were similar between men and women within their respective population, it can be proposed that a linear regression model can be used for both sexes to calibrate the association between silhouettes and BMI (or other adiposity markers) within the same population. This assumption can stand if the silhouette and anthropometric correlations between sexes are similar, like in the US and Seychelles. Inversely, as our data in Ghana suggest, different predictive regression models may need to be developed in men and women when Spearman correlation coefficients markedly differ between sexes (correlation coefficient 0.55–0.58 in men compared to 0.71–0.74 in women) in the same population (Table 2). Differences in the regression coefficients between silhouettes and BMI (and other adiposity markers) may also partly depend on different mean population BMI and sex-specific perceptions of body shape, and these questions necessitate further studies.
Different silhouette cutoffs detected obesity differently in the three countries. While the BMI categories for each silhouette rank showed a large dispersion, there were apparent differences in the country’s distribution pattern (Fig 2). In addition, each silhouette also had varying sensitivities and specificity between countries for detecting overweight or obesity status (Table 4). This variation suggests that no universal silhouette cutoff can be used for detecting overweight or obesity status.
The country and sex-specific associations between silhouettes and adiposity measures were similar for BMI, WC, and WHtR. This relationship is not unexpected as BMI, WC, and WHtR quite strongly and similarly inter-correlate with each other, e.g., correlation coefficients of 0.77 to 0.96 in our study, which is consistent with correlations found in other studies [42]. However, the associations between silhouettes and BMI, WC, and WHtR are still not extremely strong, implying that silhouettes would not be a reliable tool to predict adiposity at the individual level (sensitivity and specificity are not optimal). However, they can be helpful when assessing adiposity levels (e.g., the prevalence of obesity, mean BMI) at a population level, conditional on appropriate calibration in a specific population. *More* generally, our data suggest that a subjective two-dimensional pictorial body size assessment (silhouette drawings) can be a valuable tool for predicting a volumetric dimension (adiposity), at least at the population level.
This study’s main strength was using the identical methodology in the three countries, allowing us to directly compare three populations that differed largely according to mean adiposity levels and socioeconomic development stages. However, the study also has limitations. First, although the study was designed to include participants of African-origin in all sites, to control for ethnic differences, persons from mixed origins were also included in varying but small proportions, particularly in Seychelles. Second, the study included adults aged 20–68, and the findings may not necessarily extend to older or younger individuals. Third, Pulvers’ silhouette tool presents body size silhouettes from thinnest to heaviest, possibly leading to reporting bias. Future studies should examine if presenting the silhouettes in random order would gather different results. Fourth, survey administrators presented silhouettes to the participants; further studies should assess if results would differ if participants had assessed their silhouettes in the absence of assisting personnel. Finally, our analysis, according to sex, was limited because of the limited sample size.
## Conclusions
This study supports the utility of Pulvers’ silhouette showcards as a useful tool to predict anthropometric and adiposity measures in different populations and in settings where body size cannot be measured directly. However, no universal silhouette cutoff can be used to detect overweight or obesity status, and caution should be used to ensure adequate adjustment (i.e., calibration) for the associations between silhouette ranking and actual adiposity measures between sexes and countries. In addition, further assessment should be done to examine sex-specific differences in body perception and cultural ideals in body size across the epidemiological transition.
## References
1. Pi-Sunyer FX. **Medical hazards of obesity**. (1993.0) **119** 655-660. DOI: 10.7326/0003-4819-119-7_part_2-199310011-00006
2. Must A, Spadano J, Coakley EH, Field AE, Colditz G, Dietz WH. **The disease burden associated with overweight and obesity**. (1999.0) **282** 1523-1529. DOI: 10.1001/jama.282.16.1523
3. Jaacks LM, Vandevijvere S, Pan A. **The obesity transition: stages of the global epidemic**. (2019.0) **7** 231-240. DOI: 10.1016/S2213-8587(19)30026-9
4. Flegal KM, Carroll MD, Ogden CL, Johnson CL. **Prevalence and trends in obesity among US adults, 1999–2000**. (2002.0) **288** 1723-1727. DOI: 10.1001/jama.288.14.1723
5. Haslam DW, James WP. **Obesity**. (2005.0) **366** 1197-1209. DOI: 10.1016/S0140-6736(05)67483-1
6. 6Diet, nutrition and the prevention of chronic diseases. World Health Organ Tech Rep Ser. 2003;916:.
7. Young T, Peppard PE, Gottlieb DJ. **Epidemiology of obstructive sleep apnea: a population health perspective**. (2002.0) **165** 1217-1239. DOI: 10.1164/rccm.2109080
8. Stokes A.. **Using maximum weight to redefine body mass index categories in studies of the mortality risks of obesity**. *Popul Health Metr* (2014.0) **12** 6. DOI: 10.1186/1478-7954-12-6
9. Gurunathan U, Myles PS. **Limitations of body mass index as an obesity measure of perioperative risk**. (2016.0) **116** 319-321. DOI: 10.1093/bja/aev541
10. Kim YJ, Lee SH, Kim TY, Park JY, Choi SH, Kim KG. **Body fat assessment method using CT images with separation mask algorithm**. (2013.0) **26** 155-162. DOI: 10.1007/s10278-012-9488-0
11. 11IDF. Consensus Statements: Metabolic Syndrome. 2017; Consensus Statements: Metabolic Syndrome. https://www.idf.org/e-library/consensus-statements/60-idfconsensus-worldwide-definitionof-the-metabolic-syndrome.html.
12. Burkhauser RV, Cawley J. **Beyond BMI: the value of more accurate measures of fatness and obesity in social science research**. (2008.0) **27** 519-529. DOI: 10.1016/j.jhealeco.2007.05.005
13. **Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults—The Evidence Report. National Institutes of Health [published correction appears in Obes Res 1998 Nov;6(6):464]**. (1998.0) **6** 51S-209S. PMID: 9813653
14. 14WHO. STEPwise approach to surveillance (STEPS). 2014; STEPwise approach to surveillance (STEPS). http://www.who.int/chp/steps/en/.
15. Hattori A, Sturm R. **The obesity epidemic and changes in self-report biases in BMI**. *Obesity (Silver Spring)* (2013.0) **21** 856-60. DOI: 10.1002/oby.20313
16. Tsai EW, Perng W, Mora-Plazas M, Marín C, Baylin A, Villamor E. **Accuracy of self-reported weight and height in women from Bogotá, Colombia**. (2014.0) **41** 473-476. DOI: 10.3109/03014460.2013.856939
17. Sagna ML, Schopflocher D, Raine K, Nykiforuk C, Plotnikoff R. **Adjusting divergences between self-reported and measured height and weight in an adult Canadian population**. (2013.0) **37** 841-850. DOI: 10.5993/AJHB.37.6.13
18. Yoon K, Jang SN, Chun H, Cho SI. **Self-reported anthropometric information cannot vouch for the accurate assessment of obesity prevalence in populations of middle-aged and older Korean individuals**. (2014.0) **59** 584-592. DOI: 10.1016/j.archger.2014.08.008
19. Wen M, Kowaleski-Jones L. **Sex and ethnic differences in validity of self-reported adult height, weight and body mass index**. (2012.0) **22** 72-78. PMID: 22774312
20. Stunkard AJ, Sørensen T, Schulsinger F. **Use of the Danish Adoption Register for the study of obesity and thinness**. (1983.0) **60** 115-120. PMID: 6823524
21. Nagasaka K, Tamakoshi K, Matsushita K, Toyoshima H, Yatsuya H. **Development and validity of the Japanese version of body shape silhouette: relationship between self-rating silhouette and measured body mass index**. (2008.0) **70** 89-96. PMID: 18954027
22. Bulik CM, Wade TD, Heath AC, Martin NG, Stunkard AJ, Eaves LJ. **Relating body mass index to figural stimuli: population-based normative data for Caucasians**. (2001.0) **25** 1517-1524. DOI: 10.1038/sj.ijo.0801742
23. Conti MA, Ferreira ME, de Carvalho PH. **Stunkard Figure Rating Scale for Brazilian men**. (2013.0) **18** 317-322. DOI: 10.1007/s40519-013-0037-8
24. Mak KK, McManus AM, Lai CM. **Validity of self-estimated adiposity assessment against general and central adiposity in Hong Kong adolescents**. (2013.0) **40** 276-279. DOI: 10.3109/03014460.2013.766261
25. Pulvers KM, Lee RE, Kaur H. **development of a culturally relevant body image instrument among urban African Americans**. (2004.0) **12** 1641-1651. DOI: 10.1038/oby.2004.204
26. Yepes M, Viswanathan B, Bovet P, Maurer J. **Validity of silhouette showcards as a measure of body size and obesity in a population in the African region: A practical research tool for general-purpose surveys**. (2015.0) **13** 35. DOI: 10.1186/s12963-015-0069-6
27. Pulvers K, Bachand J, Nollen N, Guo H, Ahluwalia JS. **BMI-based norms for a culturally relevant body image scale among African Americans**. (2013.0) **14** 437-440. DOI: 10.1016/j.eatbeh.2013.07.005
28. 28Henry, Alice Victoria, “The Perception of Men’s Preferred Female Body Size and Weight Control Behaviors of Afro-Caribbean Women in the United States Virgin Islands” (2020). Walden Dissertations and Doctoral Studies. 7971.
29. Lynch EB, Kane J. **Body size perception among African American women**. (2014.0) **46** 412-417. DOI: 10.1016/j.jneb.2014.03.002
30. Kaufer-Horwitz M, Martínez J, Goti-Rodríguez LM, Avila-Rosas H. **Association between measured BMI and self-perceived body size in Mexican adults**. (2006.0) **33** 536-545. DOI: 10.1080/03014460600909281
31. Muñoz-Cachón MJ, Salces I, Arroyo M, Ansotegui L, Rocandio AM, Rebato E. **Overweight and obesity: prediction by silhouettes in young adults**. (2009.0) **17** 545-549. DOI: 10.1038/oby.2008.541
32. Acevedo P, López-Ejeda N, Alférez-García I. **Body mass index through self-reported data and body image perception in Spanish adults attending dietary consultation**. (2014.0) **30** 679-684. DOI: 10.1016/j.nut.2013.11.006
33. Peterson M, Ellenberg D, Crossan S. **Body-image perceptions: reliability of a BMI-based Silhouette Matching Test**. (2003.0) **27** 355-363. DOI: 10.5993/ajhb.27.4.7
34. Lønnebotn M, Svanes C, Igland J. **Body silhouettes as a tool to reflect obesity in the past**. (2018.0) **13** e0195697. DOI: 10.1371/journal.pone.0195697
35. Flynn KJ, Fitzgibbon M. **Body images and obesity risk among black females: a review of the literature**. *Ann Behav Med* (1998.0) **20** 13-24. DOI: 10.1007/BF02893804
36. Hosseini SA, Padhy RK. (2020.0)
37. Dugas LR, Lie L, Plange-Rhule J. **Gut microbiota, short chain fatty acids, and obesity across the epidemiologic transition: the METS-Microbiome study protocol**. (2018.0) **18** 978. DOI: 10.1186/s12889-018-5879-6
38. Luke A, Bovet P, Forrester TE. **Protocol for the modeling the epidemiologic transition study: a longitudinal observational study of energy balance and change in body weight, diabetes and cardiovascular disease risk**. (2011.0) **11** 927. DOI: 10.1186/1471-2458-11-927
39. 39Barro, R.J.; Lee, J.W. A New Data Set of Educational Attainment in the World, 1950–2010; The National Bureau of Economic Research: Cambridge, MA, US, 2011.
40. 40WHO. Obesity and overweight. Fact sheet N°311 2015; http://www.who.int/mediacentre/factsheets/fs311/en/.
41. Yoo EG. **Waist-to-height ratio as a screening tool for obesity and cardiometabolic risk**. (2016.0) **59** 425-431. DOI: 10.3345/kjp.2016.59.11.425
42. Bouchard C.. **BMI, fat mass, abdominal adiposity and visceral fat: where is the ‘beef’?**. (2007.0) **31** 1552-1553. DOI: 10.1038/sj.ijo.0803653
|
---
title: Prevalence and determinants of hypertension in underrepresented indigenous
populations of Nepal
authors:
- Tsedenia Workneh Denekew
- Yoshina Gautam
- Dinesh Bhandari
- Guru Prasad Gautam
- Jeevan Bahadur Sherchand
- Amod K. Pokhrel
- Aashish R. Jha
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021878
doi: 10.1371/journal.pgph.0000133
license: CC BY 4.0
---
# Prevalence and determinants of hypertension in underrepresented indigenous populations of Nepal
## Abstract
Indigenous populations residing in low- and middle-income countries (LMICs) are highly underrepresented in medicine and public health research. Specifically, data on non-communicable diseases (NCDs) from indigenous populations remains scarce. Despite the increasing burden of NCDs in the Himalayan region, their prevalence in many indigenous populations remains understudied. The nationally representative public health surveys often do not include the indigenous communities, especially those that reside in rural areas or exist in small numbers. This observational cross-sectional survey study aimed to assess the prevalence of three NCD risk factors namely obesity, hypertension, and tachycardia and identify dietary and lifestyle variables associated with them across underrepresented indigenous populations of Nepal. A total of 311 individuals ($53.3\%$ women, $46.7\%$ men) with mean age 43±15 years from 12 indigenous Nepali communities residing in rural ($47.9\%$) or semi-urban ($52.1\%$) areas volunteered to participate in this study. Univariate tests and multivariable logistic regressions were used to analyze the survey data. The mean systolic and diastolic blood pressures were 121.3±19.5 mmHg and 81.3±11.8 mmHg respectively. Overall, the prevalence of obesity and tachycardia was low ($0.64\%$ and $3.22\%$, respectively) but hypertension was prevalent at $23.8\%$. Hypertension was not significantly different across populations, but it was associated with age, BMI, and tobacco use, and collectively, these variables explained $13.9\%$ variation in hypertension prevalence. Although we were unable to detect direct associations between individual determinants of hypertension identified in non-indigenous Nepalis, such as education levels, alcohol consumption, and smoking in this study, having one or more determinants increased the odds of hypertension in the indigenous participants. Furthermore, ~$14\%$ of the hypertensive individuals had none of the universally identified hypertension risk factors. The lack of association between previously identified risk factors for hypertension in these individuals indicates that the additional determinants of hypertension remain to be identified in indigenous Nepali populations.
## Introduction
Non-communicable diseases (NCDs) are the leading cause of morbidity and mortality in the world [1]. The global non-communicable death burden currently stands at $71\%$, with 41 million out of the world’s 57 million deaths linked directly to NCDs [2]. Although communicable diseases still pose a major threat in low- and middle-income countries (LMICs), they are also witnessing a steady increase in the incidence of NCDs. The proportion of deaths due to NCDs has already surpassed communicable diseases in these nations [3–5]. The vast majority of NCD related deaths ($78\%$) occur in LMICs [1], indicating that the existing healthcare infrastructures in LMICs are currently poorly equipped to tackle the increasing burden of NCDs. Despite growing awareness of high NCD burden, research designed to understand their prevalence and determinants within these regions is minimal [3]. Indigenous populations are even more underrepresented in medical research from LMICs as nationally representative public health studies designed to reflect the overall public health of a nationwide population do not include all ethnic communities, especially those that reside in rural areas or exist in small numbers. Therefore, the prevalence of NCDs and their determinants in indigenous populations remains obscure [6–8].
The Himalaya consists of three ecological regions: high mountains with elevation over 1,500 meters above the sea levels (masl), hills with elevation 600–1,500 masl, and the Terai plains (<600 masl). Nepal is a Himalayan country with 29 million people comprising over 120 ethnolinguistic groups. Fifty-nine of these groups are originally recognized as “indigenous” by the Government of Nepal [9, 10] and several groups in Terai are collectively referred to as “Dalit”. A majority (>$90\%$) of the population reside in the hill and Terai regions and many ethnic groups have historically localized to certain geographical areas, except a handful of urban cities where recent migration has led to a mix of co-residing populations of diverse ethnicities. A recent nationally representative cross-sectional survey in Nepal showed that NCDs such as diabetes, obesity, and hypertension are highly prevalent in the general population [11–15]. An alarming $26\%$ of the general population in Nepal had hypertension (systolic blood pressure SBP ≥ 140mmHg or diastolic blood pressure DBP ≥ 90) and $21\%$ of the adult population were overweight or obese (BMI ≥ 25). Almost all of the 4,200 respondents ($99.6\%$) showed at least one previously established NCD associated major risk factors such as smoking, harmful use of alcohol, consumption of low fiber diet rich in salt and sugar, physical inactivity, overweightness/obesity, raised blood pressure, elevated blood glucose, and raised total cholesterol, underscoring a looming health crisis that the country’s health infrastructure is unable to handle [11]. Other nationally representative surveys from Nepal have also highlighted the link between economic status and hypertension and called for exploration of socio-economic status and disease burden [13]. Indigenous populations are among the most marginalized and economically challenged in Nepal [13]. However, due to their small population sizes and lack of accessibility, many of the *Nepali indigenous* populations have been underrepresented in nationally representative surveys. When included, individuals from several ethnically distinct *Nepali indigenous* groups with unique traditions and lifestyles have been grouped into a single category called “janajatis” [16]. Since they make up a third of the country’s population, understanding the prevalence of NCDs and their determinants in indigenous Nepali groups may provide new insights in tackling the increasing burden of NCDs in the nation as a whole [7].
This study aims to assess the prevalence of NCDs and their determinants in diverse underrepresented ethnic groups in Nepal. We found that NCDs such as hypertension is prevalent at appreciable frequencies in indigenous Nepali populations but its determinants in indigenous populations may differ from non-Indigenous Nepalis. Our results may help policymakers and public health officials develop preventative approaches to address the increasing burden of NCDs in these communities.
## Ethical considerations
This study is a part of a multifaceted study that was approved by the Stanford University Institutional Review Board (IRB) as well as the Ethical Review Board of Nepal Health Research Council (NHRC).
## Study design
This population-based cross-sectional study was conducted to evaluate the prevalence of non-communicable diseases and associated risk factors in diverse indigenous populations of Nepal. The study was conducted from February—May 2016. Individuals from 12 different ethnic groups residing in the hills and Terai plains of Nepal were recruited via household visits. Among these, 11 groups are officially recognized as “indigenous” by the Government of Nepal [9] and they include Bote, Chepang, Darai, Kusunda, Maajhi, Newar, Raji, Raute (Dadeldhura), Tamang, Thami, and Tharu. The Musahars are officially classified as “Dalit” and are known to be native inhabitants of the Terai plains [17]. Within each ethnic group, male and female adults over 18 years old whose parents and grandparents were reported to be from the same ethnic group (non-multiracial) were invited to participate via signing an informed consent form. Unrelated individuals, i.e. individuals who did not share a grandparent, were randomly selected from the community to participate in this study and only one individual per household was recruited. A total of 337 individuals aged 18–83 volunteered to participate in this study. Children, minors under the age of 18, and pregnant women were excluded from participating in this study.
## Survey data collection
Demographic, anthropogenic, environmental, and dietary data were obtained from the participants using a survey questionnaire by a trained enumerator (YG). The survey questionnaire was aligned with WHO STEPS Instrument with some modifications to reflect the traditional Nepali lifestyles. The survey questionnaire included demographic variables such as age, gender, education level, and marital status, diet and physical activity, medical histories, and behavioral practices such as smoking, use of non-smoking tobacco products (chewing tobacco), and alcohol consumption, along with several environmental variables such as area of residence, type of cooking fuel used in home, source of drinking water, etc. ( S1 Table). Participants’ responses to the relevant survey data questionnaires are included in (S2 Table).
## Phenotypic measurements
Trained enumerator measured systolic and diastolic blood pressures and heart rates using an automatic blood pressure monitor (Omron, BP791IT). Participants were seated comfortably, and four readings of blood pressure were recorded at 3-minute intervals. The last three readings were averaged for subsequent analysis. Participants with systolic blood pressure (SBP) ≥ 140 mmHg and/or diastolic blood pressure (DBP) ≥ 90 mmHg were considered hypertensive. Systolic and diastolic hypertension were defined separately as ≥ 140mmHg and ≥ 90mmHg, respectively. Weight was measured using a balanced scale (Seca 869). The scale was placed on a flat surface and legs were adjusted until the indicator level was at the center position. Height was measured while the participants were standing on the balanced weighing scale. Height and weight were measured three times and averaged. Body mass index (BMI) was calculated as averaged weight in kilograms divided by the square of averaged height in meters. Participants with BMI < 18.5 kg/m2 were considered underweight and those with BMI 18.5–24.9kg/m2 were considered normal. Participants with BMI ≥ 25 kg/m2 and ≥ 30 kg/m2 were considered overweight and obese, respectively.
## Statistical analysis
All statistical analyses were performed using STATA (version 16.0) and visualized using R (version 3.6.1). Fifty variables were initially obtained from the survey questionnaire, and 29 variables with >$95\%$ identical answers across the dataset were removed from subsequent analyses. Summary statistics were calculated and associations between categorical variables were first assessed using a chi-squared test. Individuals were classified as hypertensive (SBP ≥ 140 mmHg and/or diastolic blood DBP ≥ 90 mmHg) and non-hypertensive (SBP < 140 mmHg and diastolic blood DBP < 90 mmHg) using binary categorization (0 = non-hypertensive and 1 = hypertensive), which was used as dependent variable for calculating odds ratios. Associations between factors and hypertension were assessed by performing univariate and multivariable logistic regressions and crude and adjusted odds ratios were calculated. To perform logistic regression, the continuous variables—BMI, age, and household size–were classified into groups of equal proportions. These analyses were performed separately to identify risk factors associated with systolic hypertension (SBP ≥ 140 mmHg) and diastolic hypertension (DBP ≥ 90 mmHg). To assess associations between the previously established determinants and hypertension, we first created a new variable by counting the total number of the five previously established determinants in each participant such that each participant had a score between 0 and 5. Only 3 individuals had 4 risk factors and none had 5. Due to limited numbers of individuals with 4 and 5 risk factors, individuals with 3 or more risk factors were pooled together to create four categories (0, 1, 2, or ≥3). Next, we compared whether including this variable improved the fit compared to a null model (with no variables) using a Wald test. Finally, we used a logistic regression to assess the odds ratio for hypertension with increasing number of risk factors (0 vs 1, 2, or ≥3). Statistical tests with P-value < 0.05 were considered significant.
## Population characteristics
A total of 337 individuals were recruited, of which 26 individuals had incomplete information and were removed from subsequent analyses, resulting in 311 participants. These participants belonged to 12 populations of which 11 are officially recognized as “indigenous” by the Government of Nepal [9] and the Musahar, known to be natives of the Terai plains [17], are officially classified as Dalit. The populations included in this study are described in Table 1.
**Table 1**
| Populations | Census size (26,494,504) | Official Status | Economic Status | Sample size (311) | Sample location (District and Provinces) |
| --- | --- | --- | --- | --- | --- |
| Tharu | 1737470 | Indigenous | Marginalized | 84 | Sarlahi, Province 2, Dhangadhi, Sudurpaschim Province, and Chitwan, Bagmati Province |
| Tamang | 1539830 | Indigenous | Marginalized | 13 | Dolakha, Bagmati Province, |
| Newar | 1321933 | Indigenous | Advanced | 53 | Lalitpur and Makwanpur,Bagmati Province |
| Musahar | 234490 | Dalit | Not Available | 18 | Chitwan, Bagmati Province and Sarlahi, Province 2 |
| Maajhi | 83727 | Indigenous | Highly marginalized | 31 | Sindhupalchowk, Bagmati Province |
| Chepang | 68399 | Indigenous | Highly marginalized | 27 | Chitwan, Bagmati Province |
| Thami | 28671 | Indigenous | Highly marginalized | 13 | Dolakha, Bagmati Province |
| Darai | 16789 | Indigenous | Marginalized | 16 | Chitwan, Bagmati Province |
| Bote | 10397 | Indigenous | Highly marginalized | 6 | Chitwan, Bagmati Province |
| Raji | 4235 | Indigenous | Endangered | 21 | Bardiya, Lumbini Province |
| Raute | 618 | Indigenous | Endangered | 24 | Dadeldhura, Sudurpaschim Province |
| Kusunda | 273 | Indigenous | Endangered | 3 | Dang, Lumbini Province |
| Other | -- | -- | -- | 2 | -- |
The social and demographic characteristics of the participating populations are shown in Table 2. Of the 311 participants, $53.4\%$ were from the Hills, and $46.6\%$ were from the Terai. Participants were recruited from 12 indigenous groups residing in rural ($47.9\%$) or semi-urban ($52.1\%$) areas, and groups with less than 10 participants (Kusunda and Bote) were grouped as “other.” About half ($53.3\%$) of the participants were females. Participants were on average 43±15 years old with $43.7\%$, $40.2\%$, and $16\%$ between the ages of 18–39, 40–61, and 61–83 years. Half of the participants had no formal schooling, and another $21\%$ had only completed primary school. A vast majority of participants ($88\%$) did not access the internet. Most ($97\%$) of participants consumed fresh vegetables twice a day (S2 Table), which is consistent with previous observations in Nepal [19]. Consistent with their agrarian lifestyle, a majority of participants were physically active on a regular basis ($80.7\%$). Animal protein (e.g., fish, meat, and dairy) was rare in their diet and $67.5\%$ of the study participants ate home cooked meals on a daily basis, and a vast majority of participants consumed fermented foods on a regular basis ($91\%$). About half of the participants ($48.6\%$) responded yes when asked if they perceived scarcity of food within their respective households. Smoking and use of non-smoking (chewing) tobacco products were observed in $24\%$ and $22\%$ of participants, respectively, and $47\%$ of participants consumed some level of alcohol on a daily or weekly basis.
**Table 2**
| Variables | N (n = 311) | Percent Prevalence | Percent Prevalence by gender | Percent Prevalence by gender.1 |
| --- | --- | --- | --- | --- |
| Age | | Overall % | Male % (n = 145) | Female %(n = 166) |
| 18–39 | 136.0 | 43.73 | 33.10 | 53.01 |
| 40–61 | 125.0 | 40.19 | 46.20 | 34.94 |
| 62–83 | 50.0 | 16.08 | 20.68 | 12.05 |
| BMI | | | | |
| Underweight (<18.5) | 54.0 | 17.36 | 17.93 | 16.87 |
| Normal (18.5–25) | 208.0 | 66.88 | 61.37 | 71.68 |
| Overweight/Obese (> = 25) | 49.0 | 15.76 | 20.68 | 11.44 |
| Education | | | | |
| Never/Rarely formal education | 163.0 | 52.41 | 42.06 | 61.44 |
| Primary Education | 65.0 | 20.9 | 26.89 | 15.66 |
| Secondary Education | 58.0 | 18.65 | 21.38 | 16.26 |
| Post-Secondary | 25.0 | 8.04 | 9.60 | 6.60 |
| Internet Access | | | | |
| Never/Rarely | 274.0 | 88.1 | 86.20 | 89.76 |
| Weekly/Daily | 37.0 | 11.9 | 13.79 | 10.24 |
| Tobacco Use | | | | |
| Never/Rarely | 244.0 | 78.46 | 60.68 | 93.97 |
| Weekly/Daily | 67.0 | 21.54 | 39.31 | 6.02 |
| Smoking | | | | |
| Never/Rarely | 234.0 | 75.24 | 63.44 | 85.54 |
| Weekly/Daily | 77.0 | 24.76 | 36.55 | 14.46 |
| Alcohol Drinking | | | | |
| Never/Rarely | 164.0 | 52.73 | 40.68 | 63.25 |
| Weekly/Daily | 147.0 | 47.27 | 59.31 | 36.75 |
| Exercise | | | | |
| Never/Rarely | 60.0 | 19.29 | 22.07 | 16.68 |
| Weekly/Daily | 251.0 | 80.71 | 77.93 | 83.13 |
| Location | | | | |
| Semi urban | 162.0 | 52.09 | 53.79 | 50.60 |
| Rural | 149.0 | 47.91 | 46.21 | 49.39 |
| Geography | | | | |
| Hills | 166.0 | 53.38 | 60.68 | 46.98 |
| Terrai | 145.0 | 46.62 | 39.31 | 53.01 |
| Household size | | | | |
| (1–4) | 93.0 | 29.9 | 28.27 | 31.32 |
| (5–9) | 183.0 | 58.84 | 55.86 | 61.44 |
| (10+) | 35.0 | 11.25 | 15.86 | 7.23 |
| Altitude | | | | |
| <500 | 169.0 | 54.34 | 46.89 | 60.84 |
| [500–1500] | 105.0 | 33.76 | 38.62 | 29.52 |
| >1500 | 37.0 | 11.9 | 14.48 | 9.64 |
| Milk | | | | |
| Never/Rarely | 277.0 | 89.07 | 88.97 | 89.16 |
| Weekly/Daily | 34.0 | 10.93 | 11.03 | 10.84 |
| Yoghurt | | | | |
| Never/Rarely | 259.0 | 83.28 | 85.52 | 81.32 |
| Weekly/Daily | 52.0 | 16.72 | 14.48 | 18.67 |
| Fermented food | | | | |
| Never/Rarely | 27.0 | 8.68 | 12.41 | 5.40 |
| Weekly/Daily | 284.0 | 91.32 | 87.58 | 94.58 |
| Cooked food | | | | |
| Never/Rarely | 101.0 | 32.48 | 48.27 | 18.67 |
| Weekly/Daily | 210.0 | 67.52 | 51.72 | 81.32 |
| Scarcity of food | | | | |
| No | 160.0 | 51.45 | 55.17 | 48.19 |
| Yes | 151.0 | 48.55 | 44.80 | 51.80 |
| Fish | | | | |
| Never/Rarely | 277.0 | 89.07 | 85.52 | 92.16 |
| Weekly/Daily | 34.0 | 10.93 | 14.48 | 7.83 |
| Meat | | | | |
| Never/Rarely | 220.0 | 70.74 | 68.27 | 72.89 |
| Weekly/Daily | 91.0 | 29.26 | 31.72 | 27.11 |
## Prevalence of non-communicable diseases
We assessed blood pressure, body mass index (BMI), and heart rate in these participants to determine the prevalence of hypertension, obesity, and tachycardia (Fig 1). The mean systolic and diastolic pressures were 121.3±19.5 mmHg and 81.3±11.8 mmHg respectively. Overall, the prevalence of hypertension, defined as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg, was $23.8\%$. The prevalence of hypertension was $27.7\%$ and $19.3\%$ in the Hills and Terai respectively. Hypertension was more prevalent among men ($31.7\%$) than women ($16.8\%$, $$P \leq 0.003$$, chi-square test). On average, women had lower systolic and diastolic blood pressures ($$n = 166$$, mean SBP = 115.5±17.9 mmHg and mean DBP = 78.9±11.3 mmHg) compared to men ($$n = 145$$, mean SBP = 127.0±19 mmHg and mean DBP = 84.0±11.7 mmHg, $$P \leq 0.049$$ and 0.003 for SBP and DBP respectively, Student’s t test). The average BMI and heart rate in the overall dataset were 21.6±3.2 and 77.3±11.7 bpm (Fig 1). About $15\%$ ($$n = 47$$) of participants were overweight, while both obesity (BMI>30) and tachycardia (heart rate >100 bpm) were negligible ($$n = 2$$ and 10, respectively). Because of its higher observed prevalence in these participants, we focused on hypertension in subsequent analyses.
**Fig 1:** *Prevalence of Non-Communicable Disease (NCD) risk factors in indigenous peoples of Nepal.Density plots showing the distributions of the four NCD risk factors measured from 311 Nepali individuals. Clockwise from the top: systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate measured as beats per minute (BPM), body mass index (BMI). Dotted lines indicate threshold for disease status: SBP = 140 mmHG, DBP = 90 mmHG, BMI = 30, and heart rate = 100 BPM.*
## Hypertension associated risk factors
Several modifiable risk factors such as higher BMI, alcohol consumption, smoking, lack of physical activity, and a processed diet rich in sodium and carbohydrates are classic determinants of hypertension [20]. We sought to identify whether these and other factors are associated with hypertension in our participants. We evaluated the association between hypertension, which was defined as SBP ≥ 140 mmHg and/or diastolic blood DBP ≥ 90 mmHg, and 19 factors using a univariate logistic regression (Table 3). This analysis revealed age, BMI, sex, chewing tobacco use and milk consumption was associated with hypertension in these participants ($P \leq 0.05$, Table 3). Complex traits such as hypertension can be influenced by more than one variable and some of these variables may be correlated with one another. To account for interdependence between variables, we next performed a multivariable logistic regression and calculated the adjusted Odds Ratio. This analysis revealed that age, use of non-smoking tobacco products, and BMI were significantly associated with hypertension ($P \leq 0.05$, Table 3). Compared to participants aged 18–39, individuals within the age range 40–61 and 62–83 were more likely to develop hypertension (adjusted OR = 2.8, $95\%$ CI: 1.3–6.2 and 7.6, $95\%$ CI: 2.8–20.8 respectively). We did not find significant associations between smoking and hypertension but compared to those who do not use non-smoking (chewing) tobacco products ($$n = 244$$), chewing tobacco users ($$n = 67$$) were more likely to develop hypertension (adjusted OR = 2.5, $95\%$ CI:1.1–5.6). Furthermore, participants whose diet includes milk on a weekly or daily basis ($$n = 34$$) had higher chances of developing hypertension relative to individuals who rarely or never consumed dairy products (adjusted O.$R = 3.6$, $95\%$ CI: 1.3–10.1). Finally, we performed a linear multivariable regression analysis using hypertension as the dependent variable and the 19 variables as independent variables. This model explained ~$14\%$ of the variance in hypertension in these populations (df = 24, R2 = 0.21; adjusted R2 = 0.139), indicating that additional determinants of hypertension remain to be identified in these populations.
**Table 3**
| Unnamed: 0 | UNIVARIATE LOGISTIC REGRESSION | UNIVARIATE LOGISTIC REGRESSION.1 | UNIVARIATE LOGISTIC REGRESSION.2 | UNIVARIATE LOGISTIC REGRESSION.3 | UNIVARIATE LOGISTIC REGRESSION.4 | MULTIVARIABLE LOGISTIC REGRESSION | MULTIVARIABLE LOGISTIC REGRESSION.1 | MULTIVARIABLE LOGISTIC REGRESSION.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Hypertension | Crude Odds Ratio | [95%Conf.Interval] | [95%Conf.Interval] | P value | Adjusted Odds Ratio | [95%Conf.Interval] | [95%Conf.Interval] | P value |
| Age | | | | | | | | |
| 18–39 | 1 | | | | 1 | | | |
| 40–61 | 3.263 | 1.684 | 6.323 | 0.000 | 2.872 | 1.324 | 6.232 | 0.008 |
| 62–83 | 6.872 | 3.173 | 14.882 | 0.000 | 7.668 | 2.826 | 20.805 | 0.000 |
| BMI | | | | | | | | |
| Underweight (<18.5) | 1 | | | | 1 | | | |
| Normal (18.5–25) | 1.679 | 0.741 | 3.804 | 0.215 | 2.134 | 0.825 | 5.523 | 0.118 |
| Overweight/Obese (> = 25) | 3.642 | 1.415 | 9.374 | 0.007 | 4.048 | 1.234 | 13.272 | 0.021 |
| Sex | | | | | | | | |
| Male | 1 | | | | 1 | | | |
| Female | 0.437 | 0.255 | 0.746 | 0.002 | 0.765 | 0.354 | 1.651 | 0.495 |
| Education | | | | | | | | |
| Never/Rarely formal education | 1 | | | | 1 | | | |
| Primary Education | 1.474 | 0.791 | 2.746 | 0.222 | 1.571 | 0.703 | 3.51 | 0.271 |
| Secondary Education | 0.461 | 0.202 | 1.052 | 0.066 | 0.753 | 0.26 | 2.183 | 0.601 |
| Post-Secondary | 0.251 | 0.057 | 1.108 | 0.068 | 0.816 | 0.113 | 5.879 | 0.84 |
| Internet Access | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 0.353 | 0.121 | 1.033 | 0.057 | 0.361 | 0.081 | 1.616 | 0.183 |
| Tobacco Use | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 2.369 | 1.319 | 4.256 | 0.004 | 2.492 | 1.102 | 5.631 | 0.028 |
| Smoking | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 1.066 | 0.585 | 1.942 | 0.834 | 0.667 | 0.317 | 1.399 | 0.283 |
| Alcohol Drinking | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 2.054 | 1.205 | 3.499 | 0.008 | 1.588 | 0.812 | 3.104 | 0.176 |
| Exercise | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 0.606 | 0.326 | 1.126 | 0.113 | 0.546 | 0.265 | 1.195 | 0.135 |
| Location | | | | | | | | |
| Semi urban | 1 | | | | 1 | | | |
| Rural | 1.04 | 0.617 | 1.753 | 0.884 | 1.102 | 0.538 | 2.295 | 0.776 |
| Geography | | | | | | | | |
| Hills | 1 | | | | 1 | | | |
| Terrai | 0.624 | 0.366 | 1.065 | 0.084 | 0.645 | 0.351 | 1.887 | 0.632 |
| Household size | | | | | | | | |
| (1–4) | 1 | | | | 1 | | | |
| (5–9) | 1.053 | 0.581 | 1.908 | 0.865 | 1.131 | 0.578 | 2.331 | 0.675 |
| (10+) | 1.371 | 0.569 | 3.306 | 0.482 | 0.691 | 0.239 | 2.366 | 0.627 |
| Milk | | | | | | | | |
| Never/Rarely | 1 | | | | | | | |
| Weekly/Daily | 2.192 | 1.038 | 4.63 | 0.040 | 3.597 | 1.333 | 10.139 | 0.012 |
| Yoghurt | | | | | | | | |
| Never/Rarely | 1 | | | | | | | |
| Weekly/Daily | 0.835 | 0.405 | 1.721 | 0.624 | 0.461 | 0.179 | 1.239 | 0.127 |
| Fermented food | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 0.495 | 0.216 | 1.133 | 0.096 | 1.237 | 0.685 | 2.436 | 0.429 |
| Cooked food | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 0.855 | 0.493 | 1.482 | 0.576 | 0.697 | 0.349 | 1.53 | 0.405 |
| Scarcity of food | | | | | | | | |
| No | 1 | | | | 1 | | | |
| Yes | 1.541 | 0.911 | 2.608 | 0.107 | 1.406 | 0.742 | 2.82 | 0.278 |
| Fish | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 1.174 | 0.522 | 2.642 | 0.698 | 1.594 | 0.504 | 4.272 | 0.482 |
| Meat | | | | | | | | |
| Never/Rarely | 1 | | | | 1 | | | |
| Weekly/Daily | 1.553 | 0.893 | 2.701 | 0.119 | 1.719 | 0.834 | 3.521 | 0.143 |
| Constant | | | | | 0.07 | 0.008 | 0.353 | 0.002 |
We repeated this analysis separately for systolic and diastolic hypertension. Univariate logistic regression analyses revealed that systolic hypertension was associated with age, sex, milk consumption and diastolic hypertension was associated with age, BMI, sex, tobacco use, alcohol and meat consumptions (Tables 4 and 5). After accounting for multiple confounding variables using multivariable logistic regressions, we found that milk consumption was associated with systolic hypertension (adjusted OR of 4.3, $95\%$ CI: 1.3–15.1, Table 4) and meat consumption was associated with diastolic hypertension (adjusted OR of 2.2, $95\%$ CI: 1.1–4.6, Table 5). Individuals consuming milk and meat frequently had higher chances of developing systolic and diastolic hypertension respectively. However, these correlations should be interpreted with caution as further investigations with larger cohorts supplemented with molecular markers are needed to identify causal factors in these populations.
## Distribution of risk factors among the hypertensive individuals
Several modifiable risk factors such as high sodium intake, low potassium intake, alcohol consumption, smoking, overweightness/obesity, lack of physical activity, and unhealthy diet have been established as determinants of hypertension [20]. Therefore, to evaluate whether previously established risk factors contribute to hypertension in these indigenous Nepali populations, we assessed the distribution of five determinants in these individuals, namely alcohol consumption, diet with low vegetables, physical activity, overweightness/obesity, and smoking (Table 6). Overall, $30.9\%$ of participants had none of the determinants while $39.9\%$, $21.2\%$, $8.0\%$ and $1.0\%$ had 1, 2, 3, and 4 determinants respectively but none of the participants had all 5 risk factors. Among the hypertensive individuals, $43.2\%$ had 1 determinant, $32.4\%$ had 2 determinants, $10.8\%$ had 3 determinants, and none of them had more than 3 determinants. $13.5\%$ of the hypertensive individuals had none of these previously identified determinants (Table 6). Since most of these determinants were not individually associated with hypertension in our participants (Table 3), we evaluated whether they collectively contribute to hypertension in these populations by counting the total number of determinants in each participant. We found that including the determinant counts significantly improves the fit of the model compared to a null model with no variables ($$P \leq 0.0021$$, Wald test). Moreover, relative to the individuals with none of the determinants, those that had 1 or more previously established determinants were more likely to have hypertension (OR > 2, $P \leq 0.05$, Table 6). This model explained an additional ~$2.8\%$ of the variance in hypertension in these participants. Together, these results indicate that the previously established hypertension risk factors do contribute to hypertension in indigenous Nepali populations but their effect sizes are smaller compared to the general Nepali population.
**Table 6**
| No. of Risk Factors | All (n = 311) | Non-hypertensive (n = 237) | Hypertensive (n = 74) | Odds Ratio (OR) | [95% CI] | [95% CI].1 | P-values |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | 29.9% | 35.0% | 13.5% | 1.000 | | | |
| 1 | 39.9% | 38.8% | 43.2% | 2.887 | 1.337 | 6.232 | 0.007 |
| 2 | 21.2% | 17.7% | 32.4% | 4.743 | 2.077 | 10.831 | 0.000 |
| 3 | 8.0% | 7.2% | 10.8% | 3.32* | 1.162* | 9.488* | 0.025* |
| 4 | 1.0% | 1.3% | 0.0% | -- | -- | -- | -- |
| 5 | 0.0% | 0.0% | 0.0% | -- | -- | -- | -- |
| Constant | | | | 0.120 | 0.063 | 0.232 | 0.000 |
## Discussion
Previous studies examining the prevalence of NCDs using nationally representative datasets such as the Nepal Demographic and Health Survey 2016 (NDHS) have identified high burden of hypertension in Nepal [13, 21]. However, NHDS does not disclose the specific population groups and their sample sizes, which makes it difficult to determine the prevalence of NCDs in a particular population group within Nepal. By focusing on specific population groups, this study aimed to provide a finer understanding of prevalence of NCDs in specific communities within Nepal. The two major goals of this study were to assess the prevalence of NCDs and identify their determinants in the indigenous populations of Nepal that have been underrepresented in previous reports. Our results revealed an appreciable burden of NCDs in indigenous Nepali populations. We detected obesity and elevated heart rate at low abundance but about half of the participants ($52\%$) had elevated blood pressure (SBP ≥ 130mmHg or DBP ≥ 80mmHg) and a quarter of participants presented with hypertension (SBP ≥ 140mmHg and/or DBP ≥ 80mmHg). This finding is consistent with previous studies that have reported a high overall prevalence of hypertension in the general Nepali population [13] as well as in the neighboring regions [22, 23], indicating that hypertension is a major threat in South Asia that needs serious consideration at the national as well as the local levels. This is likely only the tip of the iceberg because additional chronic diseases such as diabetes, cardiovascular diseases, cancer, etc., likely exist in these populations but remain undetected.
Given the prevalence of hypertension in our dataset, we evaluated the link between hypertension and 19 variables, which included previously established determinants of hypertension such as sex, age, BMI, education, alcohol and tobacco use, physical inactivity, and processed diet that are rich in sodium or carbohydrates and low in fiber content. Many of our results are consistent with findings from previous studies. For example, in our univariate analysis, we found hypertension was more prevalent in indigenous men, which corroborates previous reports [23, 24]. The lower rates of hypertension as well as lowered systolic and diastolic blood pressures in women may be likely due to the vasorelaxation properties of estrogen [24–26]. Multivariable analysis revealed strong positive associations between age and BMI with hypertension, which is also consistent with previous studies [20, 27]. However, many of the factors such as education, internet access, geographical location (rural vs urban), smoking, alcohol consumption, and physical activity, that have been previously established as determinants of hypertension in the general Nepali [16, 24, 28] as well as worldwide populations [20] were not individually associated with hypertension in the indigenous participants in this study. The lack of association between previously established determinants of hypertension in this study could be due to insufficient statistical power resulting from the small sample size, although other studies with similar sample sizes have been able to link the established determinants with hypertension in non-indigenous Nepali populations [22, 23, 29, 30].
The weak associations of previously established hypertension determinants in this study population may reflect the lifestyle differences between the indigenous and the general Nepali population. For example, most indigenous populations reside in rural or peri-urban areas of Nepal, have little to no formal education, and lead a highly active agrarian life. All of the participants in this study resided in rural or semi urban areas, $73\%$ were not educated beyond the primary level, and $81\%$ were physically active. Therefore, geographical location, education, or physical activities are not major contributors to hypertension in these populations. Furthermore, the consumption of processed foods with high sodium or carbohydrate content is one of the major determinants of hypertension. Although our surveys did not include sodium or potassium intake of our participants, we found that processed food is not likely contributing to hypertension in these populations. A majority ($68\%$) of participants ate home-cooked meals low in animal protein, $97\%$ reported that diet consisted of fresh vegetables twice a day, and $49\%$ reported supplementing their diet with fermented foods daily or weekly. These findings are consistent with previous reports that reported high vegetable consumption in Nepal [19] and lower consumption of processed foods in rural Nepali populations due to financial constrains [13]. Similarly, smoking was not common in these populations, $75\%$ of participants did not smoke, and many of those who smoked were occasional smokers, likely because cigarettes are not always affordable. However, non-smoking tobacco use is common among the lowest socio-economic groups within Nepal [31]. Almost a quarter ($21.5\%$) of our participants consumed non-smoking tobacco products regularly, which was significantly associated with hypertension in this study. A vast majority of the non-smoking tobacco users in our dataset ($85\%$) were male, which could partly explain the elevated prevalence of hypertension in indigenous Nepali males.
To further assess whether the previously established determinants contributed to hypertension in our study participants, we assessed the link between number of risk factors at the individual level and hypertension risk. This analysis further revealed that the collective contributions of these determinants are detectable in these populations. Individuals with one or more of these factors were more likely to have hypertension compared to the individuals with none of these factors. This finding indicates that the globally relevant determinants also affect hypertension in the *Nepali indigenous* populations but their individual contributions are smaller in these indigenous peoples compared to the general Nepali populations. Finally, we found that $13.5\%$ of the hypertensive individuals in this study had none of the commonly known risk factors for hypertension. Furthermore, a multivariable regression model including all of the variables in our dataset was able to explain only ~$14\%$ variance in hypertension, indicating that additional determinants of hypertension in these populations remain unidentified. It is possible that the determinants of hypertension in these populations may include population specific cultural, dietary, lifestyle elements, which may not have been captured by our survey questionnaire despite our best effort to capture elements of traditional Nepali lifestyle (e.g., fermented food consumption). For instance, it was difficult to measure participants’ salt intake, which is a known determinant of hypertension. Identification of additional hypertension determinants may require future studies to incorporate population specific cultural elements in addition to the generalized questions included in the standardized survey questionnaires.
We focused on indigenous populations because they are underrepresented in nationally representative surveys and hospital-based public health research. Therefore, they are the highest risk groups of developing and dying from NCDs. Our study does not encompass all of the indigenous populations of Nepal, nor does it assess all of the NCDs. Despite these limitations, our study highlights that hypertension is prevalent in indigenous Nepali populations, and in addition to the previously established determinants that contribute to hypertension in the general Nepali population, novel hypertension associated risk factors likely exist in these populations but remain to be identified. Thus, intervention strategies developed for the general population may not be sufficient to address the growing burden of NCDs in indigenous Nepali peoples. Larger and more comprehensive future studies are needed to detect the array of NCDs and pinpoint their determinants in the indigenous peoples. As such, we may need to observe NCDs using a different lens when it comes to indigenous peoples.
## References
1. Heller O, Somerville C, Suggs LS, Lachat S, Piper J, Aya Pastrana N. **The process of prioritization of non-communicable diseases in the global health policy arena**. *Health policy and planning* (2019.0) **34** 370-83. DOI: 10.1093/heapol/czz043
2. Bigna JJ, Noubiap JJ. **The rising burden of non-communicable diseases in sub-Saharan Africa**. *The Lancet Global Health* (2019.0) **7** e1295-e6. DOI: 10.1016/S2214-109X(19)30370-5
3. Allen LN, Pullar J, Wickramasinghe KK, Williams J, Roberts N, Mikkelsen B. **Evaluation of research on interventions aligned to WHO ‘Best Buys’ for NCDs in low-income and lower-middle-income countries: a systematic review from 1990 to 2015**. *BMJ global health* (2018.0) **3**. DOI: 10.1136/bmjgh-2017-000535
4. Allotey P, Davey T, Reidpath DD. **NCDs in low and middle-income countries-assessing the capacity of health systems to respond to population needs**. *BMC public health* (2014.0) **14** 1-3. DOI: 10.1186/1471-2458-14-S2-S1
5. Kankeu HT, Saksena P, Xu K, Evans DB. **The financial burden from non-communicable diseases in low-and middle-income countries: a literature review**. *Health Research Policy and Systems* (2013.0) **11** 1-12. DOI: 10.1186/1478-4505-11-31
6. Flood D, Rohloff P. **Indigenous languages and global health**. *The Lancet Global Health* (2018.0) **6** e134-e5. DOI: 10.1016/S2214-109X(17)30493-X
7. Hernández A, Ruano AL, Marchal B, San Sebastián M, Flores W. **Engaging with complexity to improve the health of indigenous people: a call for the use of systems thinking to tackle health inequity**. *International Journal for Equity in Health* (2017.0) **16** 1-5. DOI: 10.1186/s12939-016-0499-1
8. Whitinui P, McIvor O, Robertson B, Morcom L, Cashman K, Arbon V. **The World Indigenous Research Alliance (WIRA): Mediating and Mobilizing Indigenous Peoples’ Educational Knowledge and Aspirations**. *education policy analysis archives* (2015.0) **23** n120
9. 9National Foundation for Development of Indigenous Nationalities (NFDIN): an introduction. Lalitapura: NFDIN, Kathmandu, Nepal, 2003.
10. Bhattachan YK, Sarah Webster OG. (2005.0)
11. Aryal KK, Mehata S, Neupane S, Vaidya A, Dhimal M, Dhakal P. **The burden and determinants of non communicable diseases risk factors in Nepal: findings from a nationwide STEPS survey**. *PloS one* (2015.0) **10** e0134834. DOI: 10.1371/journal.pone.0134834
12. Mehata S, Shrestha N, Mehta R, Vaidya A, Rawal LB, Bhattarai N. **Prevalence, awareness, treatment and control of hypertension in Nepal: data from nationally representative population-based cross-sectional study**. *Journal of hypertension* (2018.0) **36** 1680-8. DOI: 10.1097/HJH.0000000000001745
13. Mishra SR, Ghimire S, Shrestha N, Shrestha A, Virani SS. **Socio-economic inequalities in hypertension burden and cascade of services: nationwide cross-sectional study in Nepal.**. *Journal of human hypertension* (2019.0) **33** 613-25. DOI: 10.1038/s41371-019-0165-3
14. Shrestha N, Mishra SR, Ghimire S, Gyawali B, Pradhan PMS, Schwarz D. **Application of single-level and multi-level modeling approach to examine geographic and socioeconomic variation in underweight, overweight and obesity in Nepal: findings from NDHS 2016**. *Scientific reports* (2020.0) **10** 1-14. DOI: 10.1038/s41598-019-56847-4
15. Shrestha N, Mishra SR, Ghimire S, Gyawali B, Mehata S. **Burden of diabetes and prediabetes in Nepal: A systematic review and meta-analysis.**. *Diabetes Therapy* (2020.0) 1-12. DOI: 10.1007/s13300-020-00884-0
16. Bista B, Dhungana RR, Chalise B, Pandey AR. **Prevalence and determinants of non-communicable diseases risk factors among reproductive aged women of Nepal: Results from Nepal Demographic Health Survey 2016**. *PloS one* (2020.0) **15** e0218840. DOI: 10.1371/journal.pone.0218840
17. Bista DB. *People of Nepal.* (2004.0)
18. 18National population and housing census 2011. Kathmandu,Nepal: National Planning Commission Secretariat Central Bureau of Statistics, 2012.
19. Nepali S, Rijal A, Olsen MH, McLachlan CS, Kallestrup P, Neupane D. **Factors affecting the fruit and vegetable intake in Nepal and its association with history of self-reported major cardiovascular events**. *BMC Cardiovascular Disorders* (2020.0) **20** 1-10. DOI: 10.1186/s12872-019-01312-3
20. Mills KT, Stefanescu A, He J. **The global epidemiology of hypertension**. *Nature Reviews Nephrology* (2020.0) **16** 223-37. DOI: 10.1038/s41581-019-0244-2
21. Rauniyar SK, Rahman MM, Rahman MS, Abe SK, Nomura S, Shibuya K. **Inequalities and risk factors analysis in prevalence and management of hypertension in India and Nepal: a national and subnational study.**. *BMC public health* (2020.0) **20** 1-11. DOI: 10.1186/s12889-019-7969-5
22. Dhungana RR, Devkota S, Khanal MK, Gurung Y, Giri RK, Parajuli RK. **Prevalence of cardiovascular health risk behaviors in a remote rural community of Sindhuli district, Nepal**. *BMC cardiovascular disorders* (2014.0) **14** 1-8. PMID: 24400643
23. Neupane D, McLachlan CS, Sharma R, Gyawali B, Khanal V, Mishra SR. **Prevalence of hypertension in member countries of South Asian Association for Regional Cooperation (SAARC): systematic review and meta-analysis**. *Medicine* (2014.0) **93**. DOI: 10.1097/MD.0000000000000074
24. Neupane D, Shrestha A, Mishra SR, Bloch J, Christensen B, McLachlan CS. **Awareness, prevalence, treatment, and control of hypertension in western Nepal**. *American journal of hypertension* (2017.0) **30** 907-13. DOI: 10.1093/ajh/hpx074
25. Ghosh S, Mukhopadhyay S, Barik A. **Sex differences in the risk profile of hypertension: a cross-sectional study.**. *BMJ open* (2016.0) **6**. DOI: 10.1136/bmjopen-2015-010085
26. Orshal JM, Khalil RA. **Gender, sex hormones, and vascular tone**. *American Journal of Physiology-Regulatory, Integrative and Comparative Physiology* (2004.0) **286** R233-R49. DOI: 10.1152/ajpregu.00338.2003
27. Hasan M, Sutradhar I, Akter T, Das Gupta R, Joshi H, Haider MR. **Prevalence and determinants of hypertension among adult population in Nepal: Data from Nepal Demographic and Health Survey 2016**. *PloS one* (2018.0) **13** e0198028. DOI: 10.1371/journal.pone.0198028
28. Chataut J, Adhikari R, Sinha N. **Prevalence and risk factors for hypertension in adults living in central development region of Nepal**. *Kathmandu University Medical Journal* (2011.0) **9** 13-8. DOI: 10.3126/kumj.v9i1.6255
29. Dhungana RR, Pandey AR, Shrestha N. **Trends in the Prevalence, Awareness, Treatment, and Control of Hypertension in Nepal between 2000 and 2025: A Systematic Review and Meta-Analysis**. *International journal of hypertension* (2021.0) 2021
30. Khanal MK, Ahmed MM, Moniruzzaman M, Banik PC, Dhungana RR, Bhandari P. **Prevalence and clustering of cardiovascular disease risk factors in rural Nepalese population aged 40–80 years**. *BMC public health* (2018.0) **18** 1-13. DOI: 10.1186/s12889-018-5600-9
31. Shrestha N, Mehata S, Pradhan PMS, Joshi D, Mishra SR. **A nationally representative study on socio-demographic and geographic correlates, and trends in tobacco use in Nepal.**. *Scientific reports* (2019.0) **9** 1-11. DOI: 10.1038/s41598-018-37186-2
|
---
title: 'Non-communicable diseases attributed mortality and associated sociodemographic
factors in Papua New Guinea: Evidence from the Comprehensive Health and Epidemiological
Surveillance System'
authors:
- Bang Nguyen Pham
- Ronny Jorry
- Nora Abori
- Vinson D. Silas
- Anthony D. Okely
- William Pomat
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021879
doi: 10.1371/journal.pgph.0000118
license: CC BY 4.0
---
# Non-communicable diseases attributed mortality and associated sociodemographic factors in Papua New Guinea: Evidence from the Comprehensive Health and Epidemiological Surveillance System
## Abstract
### Background
Papua New Guinea (PNG) is undergoing an epidemiological transition with increased mortality from NCDs. This study examined NCDs-attributed mortality and associated sociodemographic factors in PNG.
### Method
Using WHO 2016 instrument, 926 verbal autopsy (VA) interviews were conducted in six major provinces from January 2018 to December 2020. InterVA-5 tool was used to assign causes of death (COD). Multivariable logistic regression analysis was performed to identify sociodemographic factors associated with mortalities from emerging and endemic NCDs.
### Finding
NCDs accounted for $47\%$ of the total deaths, including $20\%$ of deaths attributed to emerging NCDs and $27\%$ of deaths due to endemic NCDs. Leading CODs from emerging NCDs were identified including cardiac diseases, stroke, and diabetes. The risk of dying from emerging NCDs was significantly lower among populations under age 44y compared with population aged 75+y (OR: 0.14 [0.045–0.433]; p-value: 0.001). People living in urban areas were twice likely to die from emerging NCDs than those in rural areas (OR: 1.92 [1.116–3.31]; p-value: 0.018). People in Madang province were $70\%$ less likely to die from emerging NCDs compared to those from East New Britain province (OR: 0.314 [0.135–0.73]; p-value: 0.007). Leading CODs from endemic NCDs included digestive neoplasms, respiratory neoplasms, and other neoplasms. Only children aged 0-4y had significant lower risk of dying from endemic NCDs compared to the population aged 75+y (OR: 0.114 [$95\%$ CI: 0.014–0.896]; p-value: 0.039).
### Conclusion
Public health interventions are urgently needed, prioritizing urban population and those aged over 44y to reduce premature mortality from NCDs.
## Introduction
Non-communicable diseases (NCDs) are medical conditions or diseases that are, by definition, non-infectious and non-transmissible between people [1]. NCDs are the leading cause of death (COD) and kill 41 million people each year, equivalent to $71\%$ of all deaths globally. Sustainable Development Goal (SDG) 3.4 states: “By 2030, reduce by one third premature mortality from NCDs through prevention and treatment” [2]. The Global Burden of Disease Study showed that cardiovascular diseases, cancers, and chronic respiratory diseases were the main causes contributing to the global and regional premature mortality from NCDs [3].
There has been a change in the global distribution of the burden of NCDs. Compared to high-income countries, NCDs now disproportionately affect populations in low- and middle-income countries (LMICs) [4]. About $85\%$ of ‘premature deaths’ between the ages of 30 and 70 years and $77\%$ of all NCD deaths each year occur in LMICs [5]. The examination of NCDs in LMICs has only recently been a research priority. The highest risks of dying from NCDs were observed in LMICs, especially in sub-Saharan Africa, Central Asia and Eastern Europe [6]. This epidemiological shift provides a starting point for better understanding NCDs in LMICs [7]. Understanding the socioeconomic determinants of mortality and morbidity from NCDs among the working age population is important [8, 9]. Studies on socioeconomic factors of NCDs are often limited to analysis of the effect of income per capita on the level of mortality. The association between sociodemographic factors and NCDs has been well established in high-income countries, but these associations are less clear in LMICs [10]. Understanding sociodemographic factors associated with mortality attributed to NCDs is important for the development of public health policy and programming interventions to achieve SDG targets [11].
Papua New Guinea (PNG) has a very young population, with $38.2\%$ of the population under the age of 15 years [12]. The country has been reported undergoing an epidemiological transition, characterised by a shift from infectious diseases to NCDs [13, 14]. The increase in the prevalence of NCDs has placed additional burden on the healthcare system, particularly at the primary health level [14, 15]. The PNG Government has set a target to reduce premature NCD mortality by $25\%$ from 2010 to 2020 [16]. The increased prevalence of NCD associated risk factors in PNG is likely associated with recent changes in socioeconomic development [17, 18]. However, little evidence is available to support this argument. Previous studies have reported causes of death in PNG [19], but not on sociodemographic factors associated with NCDs mortality and patterns. How household socioeconomic and demographic factors have contributed to the emergence of NCDs, driving the current mortality transition in PNG is unclear. To what extent the mortality from NCDs is taking place across ages and sexes of the population, urban-rural sectors, provinces and social classes is also little known. Addressing this current knowledge gap would assist the PNG Government in restructuring the healthcare systems, and prioritizing healthcare services, potentially contributing to a reduction in premature NCDs mortality and further improvement in health and well-being among the population.
## Objective
This study aimed to identify sociodemographic factors associated with mortality from NCDs among the general population in PNG by analysing the association of sociodemographic characteristics of the deceased and their causes of deaths.
## Data source
Mortality surveillance data were extracted from the Comprehensive Health and Epidemiological Surveillance System (CHESS), which was designed as a population-based longitudinal study, and established by Papua New Guinea Institute of Medical Research (PNGIMR). CHESS covers both urban and rural populations living in eight sentinel surveillance sites in six major provinces of PNG: Central, Eastern Highlands, East Sepik, Madang, East New Britain (ENB) and Port Moresby (POM)—the National Capital District. These provinces represent the four geographical regions of PNG: Highlands, Southern, Momase and Islands. The selection of these sites was based on the previous integrated Health and Demographic Surveillance System (iHDSS), with adjustment for inclusion of new urban sites in consultation with national and local level stakeholders. The urban-rural sites were defined based on the National Statistics Office’s definition applied to the National Census 2010 [20]. CHESS covered a population of approximately 54,399 at the baseline, equivalent to $0.65\%$ of the total population in PNG in 2018 [21, 22]. The age and sex structures of the surveillance population are similar to those of the entire population in PNG [23]. The design and methods of CHESS have been previously published [24, 25].
## Data collection
The mortality surveillance data were collected using the WHO 2016 verbal autopsy (VA) instrument [26]. The instrument is programmed for conducting VA interviews by using portable electronic devices such as smart phones, iPADs and tablets. This tool does not require interviewers to have medical background to conduct VA interviews. Hence, the tool offers considerable improvement in the implementation [27].
The WHO 2016 VA tool has been adapted for suitable use in PNG. Aside from asking standard questions about the signs and symptoms the deceased experienced prior to death, an additional data module on the deceased identification information was included in the questionnaire, including household location (GPS data) and individual background information. This information allow linkages between mortality data and other existing data components available from the CHESS database such as morbidity data and household socioeconomic data, to enhance the scope of data for analysing sociodemographic factors associated with NCDs mortality in PNG [18]. VA interviews were conducted by the CHESS demographic team to collect mortality data from the communities from January 2018 to December 2020.
## Analysing cause of death
The InterVA-5 diagnostic tool was used for cause of death (COD) analysis. This analytical tool is a computer-based programme that can assign 64 specific CODs and categories aligned with the International Classification of Diseases version 10 [28]. The programme successfully assigned CODs for 926 deaths. The specific ascribed CODs were then grouped into four main categories: The rationale for grouping NCD-attributed CODs into emerging NCDs and endemic NDCs has been described elsewhere [14].
## Analysing sociodemographic factors
Sociodemographic analyses were conducted incorporating three steps. First, COD data were linked with household socioeconomic status (SES) data derived from CHESS database for the corresponding period of time. Data on CODs of 689 deceased were successfully linked with household SES data, using the unique household identification codes. No household SES data from East Sepik Province (ESP) were available as the site was only opened in 2019.
A new variable for household wealth index was then created for all COD records. Household wealth index was an overall marker of household SES. This variable was constructed based on household socioeconomic variables including housing characteristics, access to water and sanitation, and household assets, using the principal component analysis (PCA) method. Variables on marital status, education, relationship to the household head, employment status, and access to healthcare services were not statistically significant in the PCA models hence they were excluded from the statistical model. Household wealth indices were then divided into five quintiles, representing the poorest, poor, middle, richer and the richest quintile [23].
Finally, multinomial logistic regression (MLR) analyses were conducted to identify sociodemographic factors associated with mortalities from emerging and endemic NCDs. Since social determinants of mortalities from emerging NCDs and endemic NCDs could be different [13, 18], analyses were conducted separately for these two categories to predict the risks of dying from these COD categories across sub-populations. Two binary variables (Yes/No) on emerging NCDs and endemic NCDs attributed deaths were created in the dataset. These variables were included in MLR models as dependent outcome variables with deaths to other CODs being used as the reference category. Sociodemographic variables were used as independent factors.
Significant factors remained in the MLR model, including sex (male and female), age at death (grouped into 0–4, 5–14, 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, and 75+ years), urban-rural sector, provinces, and household wealth quintiles, assuming other predictors remained constant. The main effect was selected in these MLR models to produce maximum likelihood estimates of odds ratios (OR) for mortalities attributed to emerging NCDs and endemic NCDs. The statistical likelihood test was selected to calculate $95\%$ confidence intervals (CI) of OR estimates. A p-value of 0.05 was used to determine statistical significance. No correction for multiple testing was applied to the p-values in study. All the p-values, odds ratios and they were directly produced from the MLR models. Data analyses were performed using Statistical Package for Social Science (SPSS) (version 20) (see S1 Data).
## Ethics approval and informed consent
The CHESS was granted ethics approvals from Instructional Review Board of PNG Institute of Medical Research (IRB’s Approval no. 18.05) and the Medical Research Advisory Committee of Papua New Guinea (MRAC’s Approval no. 18.06). These approvals covered all the data components under the CHESS, including the mortality data which were used in this manuscript.
VA interviews were conducted with informed consent. Participant information and consent were integrated as part of the VA questionnaire of the WHO 2016 VA tool. They were informed about their right to withdraw from the study at any stage. Informed consent was sought from self-identified close relatives of the deceased. Informed consent was obtained verbally from all respondents preceding the VA interview. The consents were written in the tablets with name of participant, name of witness and date of consent. VA interviews were conducted at the respondent’s home unless otherwise requested, and in the witnesses of other family or community member. All VA interview respondents were matured adults with close family relationship to the deceased. Among the 1021 deaths identified by the data reporters in the communities through household visits, consents were obtained for conducing 1003 VA interviews, resulted in a participation rate of $98\%$.
## Results
Among the 1021 deaths in the communities that were identified by the CHESS’s data reporters based in the surveillance sites through household visits, consents were obtained for conducing 1003 VA interviews, resulted in a participation rate of $98\%$. The flow chart for death identification, selection for interview and data analysis is shown in Fig 1.
**Fig 1:** *Flow chart for identification of deaths, selection for interview, and analyses of sociodemographic factors of mortalities from emerging and endemic NCDs in the population in PNG, PNGIMR’s CHESS, 2020.*
## Distribution of CODs
NCDs accounted for the largest proportion ($47\%$) of the total deaths, including $20\%$ of deaths attributed to emerging NCDs and $27\%$ of deaths due to endemic NCDs. Infectious diseases accounted for the second largest proportion, $34\%$ of the total deaths, followed by injuries and other external CODs ($12\%$). Neonatal deaths and still births accounted for $3\%$ and maternal deaths was $1\%$. About $3\%$ of the deaths were unable to identify a specific COD.
## Leading causes of death from NCDs
Fig 2 shows the leading COD from emerging NCDs. Acute cardiac diseases (ACD) was identified as the first leading CODs, followed by unspecified cardiac diseases, stroke, and lastly diabetes. The numbers of male deaths to the emerging NCDs were slightly higher the female deaths, but the difference was not statistically significant (Chi-squared test: p-value at 0.086).
**Fig 2:** *Numbers of deaths from emerging non-communicable diseases, PNGIMR’s CHESS, 2020.*
Fig 3 showed the leading CODs from endemic NCDs, with digestive neoplasms as the first leading COD, followed by respiratory neoplasms, and other unspecified neoplasms. The number of male deaths to these cancers was significantly higher than female counterparts (Chi-squared test p-value at 0.012).
**Fig 3:** *Numbers of deaths from endemic non-communicable diseases, PNGIMR’s CHESS, 2020.*
## Socioeconomic demographic factors associated with mortality from NCDs
Table 1 reports the distribution of deaths attributed to emerging NCDs, endemic NCDs and other CODs by sociodemographic characteristics of the deceased. The proportions of male and female deaths from emerging NCDs and from endemic NDCs were similar, around $20\%$-$24\%$ of the total deaths. The proportions of deaths from emerging and endemic NCDs increased with age, but the increase in emerging NCD-attributed deaths was most obvious from ages 35–44 ($5\%$) to ages 45–54 years ($30\%$). The increased proportion of deaths from endemic NCDs occurred in a more linear fashion from ages 15–24 through to ages 45–64. Different patterns of mortalities from emerging and endemic NCDs were observed between urban and rural populations, with higher proportions of emerging NCDs related deaths in the former and higher proportions of endemic NCDs related deaths in the latter. Port Moresby had the highest proportions of deaths from both emerging and endemic NCDs, each accounting for $30\%$ of the total deaths. While Madang had a higher proportion of deaths from endemic NCDs than emerging NCDs, $18\%$ compared to $13\%$, ENB had higher proportion of emerging NCDs deaths than endemic NCDs deaths, $26\%$ versus $18\%$, respectively. In relation to household wealth quintile, people from the richest households (5th quintile) had the highest proportion of deaths from emerging NCDs ($28\%$) and people from rich households (4th quintile) had the highest proportion of deaths from endemic NCDs ($25\%$).
**Table 1**
| Unnamed: 0 | Unnamed: 1 | Emerging NCD attributed CODs | Endemic NCD attributed CODs | Other CODs | All CODs |
| --- | --- | --- | --- | --- | --- |
| Sex | Male | 124 (24.1%) | 116 (22.6%) | 274 (53.3%) | 514 (100.0%) |
| | Female | 83 (20.1%) | 97 (23.5%) | 232 (56.3%) | 412 (100.0%) |
| Total | | 207(22.4%) | 213 (23.0%) | 506 (54.6%) | 926 (100.0%) |
| Age group | 0–4 | 1(1.5%) | 2 (3.0%) | 64 (95.5%) | 67 (100.0%) |
| Age group | 5–14 | 2 (7.7%) | 2 (7.7%) | 22 (84.6%) | 26 (100.0%) |
| Age group | 15–24 | 8 (12.9%) | 7 (11.3%) | 47 (75.8%) | 62 (100.0%) |
| Age group | 25–34 | 10 (10.1%) | 18 (18.2%) | 71 (71.7%) | 99 (100.0%) |
| Age group | 35–44 | 5 (5.3%) | 22 (23.4%) | 67 (71.3%) | 94 (100.0%) |
| Age group | 45–54 | 44 (29.5%) | 44 (29.5%) | 61 (40.9%) | 149 (100.0%) |
| Age group | 55–64 | 56 (32.9%) | 42 (24.7%) | 72 (42.4%) | 170 (100.0%) |
| Age group | 65–74 | 47 (32.9%) | 48 (33.6%) | 48 (33.6%) | 143 (100.0%) |
| Age group | 75+ | 34 (30.1%) | 28 (24.8%) | 51 (45.1%) | 113 (100.0%) |
| Total | | 207 (22.4%) | 213 (23.1%) | 503 (54.5%) | 923 (100.0%) |
| Sector | Urban | 56 (24.6%) | 48 (21.1%) | 124 (54.4%) | 228 (100.0%) |
| Sector | Rural | 147 (21.7%) | 159 (23.5%) | 371 (54.8%) | 677 (100.0%) |
| Total | | 203 (22.4%) | 207 (22.9%) | 495 (54.7%) | 905 (100.0%) |
| Province | POM | 9 (30.0%) | 9 (30.0%) | 12 (40.0%) | 30 (100.0%) |
| Province | Central | 66 (22.9%) | 65 (22.6%) | 157 (54.5%) | 288 (100.0%) |
| Province | EHP | 64 (21.3%) | 74 (24.7%) | 162 (54.0%) | 300 (100.0%) |
| Province | Madang | 10 (13.2%) | 14 (18.4%) | 52 (68.4%) | 76 (100.0%) |
| Province | ESP | 28 (24.1%) | 30 (25.9%) | 58 (50.0%) | 116 (100.0%) |
| Province | ENB | 30 (25.9%) | 21 (18.1%) | 65 (56.0%) | 116 (100.0%) |
| Total | | 207 (22.4%) | 213 (23.0%) | 506 (54.6%) | 926 (100.0%) |
| Household wealth quintiles | Poorest | 31 (22.5%) | 29 (21.0%) | 78 (56.5%) | 138 (100.0%) |
| Household wealth quintiles | Poor | 33 (23.9%) | 30 (21.7%) | 75 (54.3%) | 138 (100.0%) |
| Household wealth quintiles | Middle | 30 (21.7%) | 32 (23.2%) | 76 (55.1%) | 138 (100.0%) |
| Household wealth quintiles | Rich | 27 (19.6%) | 35 (25.4%) | 76 (55.1%) | 138 (100.0%) |
| Household wealth quintiles | Richest | 38 (27.7%) | 30 (21.9%) | 69 (50.4%) | 137 (100.0%) |
| Total | | 159 (23.1%) | 156 (22.6%) | 374 (54.3%) | 689 (100.0%) |
Table 2 shows the sociodemographic factors associated with mortalities from emerging and endemic NCDs. The risk of dying from emerging NCDs was significantly lower among the populations of age groups 5–14, 15–24, 25–34, 35–44, compared with the oldest age group (75+) (p-values < 0.05), meaning the risk of dying from these diseases significantly increased from the ages of 45 years and above. Compared to the population in rural areas, urban populations were nearly twice more likely to die from emerging NCDs (OR: 1.9 [$95\%$ CI: 1.1–3.3]; p-value: 0.018). People in Madang were $70\%$ less likely to die from emerging NCDs than those who live in ENB (OR: 0.3 [$95\%$ CI: 0.13–0.73]; p-value: 0.007). However, the risk of dying from emerging NCDs was not significantly different between two sexes and household wealth quintiles (p-values >0.05).
**Table 2**
| Socioeconomic demographic factor | Unnamed: 1 | Unnamed: 2 | Unnamed: 3 | Emerging NCDs attributed mortalitya | Emerging NCDs attributed mortalitya.1 | Emerging NCDs attributed mortalitya.2 | Emerging NCDs attributed mortalitya.3 | Endemic NCDs attributed mortality | Endemic NCDs attributed mortality.1 | Endemic NCDs attributed mortality.2 | Endemic NCDs attributed mortality.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Socioeconomic demographic factor | Category | N | % | Odd ratio | Lower bound | Upper bound | P-value | Odds ratio | Lower bound | Upper bound | P-value |
| Sex | Male | 387 | 58.2% | 1.219 | 0.822 | 1.807 | 0.324 | 0.872 | 0.571 | 1.331 | 0.525 |
| Sex | Female | 278 | 41.8% | Ref. | | | | Ref. | | | |
| Age group | 0–4 | 39 | 5.9% | | | | | 0.114 | 0.014 | 0.896 | 0.039 |
| Age group | 5–14 | 20 | 3.0% | 0.137 | 0.017 | 1.092 | 0.060 | 0.229 | 0.028 | 1.847 | 0.166 |
| Age group | 15–24 | 45 | 6.8% | 0.366 | 0.142 | 0.946 | 0.038 | 0.393 | 0.122 | 1.266 | 0.117 |
| Age group | 25–34 | 64 | 9.6% | 0.212 | 0.080 | 0.563 | 0.002 | 0.867 | 0.374 | 2.013 | 0.741 |
| Age group | 35–44 | 62 | 9.3% | 0.140 | 0.045 | 0.433 | 0.001 | 0.824 | 0.351 | 1.934 | 0.656 |
| Age group | 45–54 | 115 | 17.3% | 0.834 | 0.447 | 1.556 | 0.567 | 0.890 | 0.431 | 1.838 | 0.753 |
| Age group | 55–64 | 131 | 19.7% | 0.960 | 0.523 | 1.763 | 0.896 | 0.857 | 0.420 | 1.747 | 0.671 |
| Age group | 65–74 | 102 | 15.3% | 1.202 | 0.640 | 2.260 | 0.567 | 1.134 | 0.552 | 2.330 | 0.732 |
| Age group | 75+ | 87 | 13.1% | Ref. | | | | Ref. | | | |
| Sector | Urban | 226 | 34.0% | 1.922 | 1.116 | 3.310 | 0.018 | 1.250 | 0.687 | 2.275 | 0.464 |
| Sector | Rural | 439 | 66.0% | Ref. | | | | Ref. | | | |
| Province | POM | 30 | 4.5% | 1.078 | 0.409 | 2.840 | 0.879 | 2.394 | 0.850 | 6.741 | 0.098 |
| Province | Central | 167 | 25.1% | 1.635 | 0.840 | 3.182 | 0.148 | 1.920 | 0.853 | 4.322 | 0.115 |
| Province | EHP | 277 | 41.7% | 0.900 | 0.508 | 1.594 | 0.718 | 1.814 | 0.897 | 3.665 | 0.097 |
| Province | Madang | 75 | 11.3% | 0.314 | 0.135 | 0.730 | 0.007 | 1.260 | 0.516 | 3.076 | 0.611 |
| Province | ENB | 116 | 17.4% | Ref. | | | | Ref. | | | |
| Household wealth quintiles | Poorest | 135 | 20.3% | 0.773 | 0.428 | 1.396 | 0.394 | 1.516 | 0.786 | 2.927 | 0.215 |
| Household wealth quintiles | Poor | 135 | 20.3% | 0.818 | 0.457 | 1.461 | 0.497 | 1.264 | 0.649 | 2.460 | 0.491 |
| Household wealth quintiles | Middle | 130 | 19.5% | 0.721 | 0.396 | 1.314 | 0.286 | 0.916 | 0.447 | 1.876 | 0.809 |
| Household wealth quintiles | Rich | 130 | 19.5% | 0.571 | 0.309 | 1.058 | 0.075 | 1.344 | 0.688 | 2.626 | 0.387 |
| Household wealth quintiles | Richest | 135 | 20.3% | Ref. | | | | Ref. | | | |
| Total valid | | 665 | 100.0% | | | | | | | | |
For endemic NCDs, only children aged from 0–4 years had a significantly lower risk of dying compared with the oldest age group (75+) (OR: 0.1 [$95\%$ CI: 0.01–0.9]; p-value: 0.04). The risks of dying from endemic NCDs were not significantly different between sexes, age groups, urban-rural sectors, provinces, and household wealth quintiles.
## Discussion
This study is the first to present an overview of the scope and distribution of sociodemographic factors of mortalities attributed to emerging and endemic NCDs at the population level in PNG. The data show that the diversity of NCD mortalities is occurring across sub-populations. Epidemiological shifts were progressing differently between emerging NCDs and endemic NCDs across ages and sexes, urban and rural areas, provinces, and social classes of the population. The data have indicated an overall trend of increased mortalities associated with emerging NCDs in the urban population, who were twice more likely to die from these diseases than those from rural areas. The risk of dying from emerging NCDs was particularly high among those who aged 45 years and above, and resided in Port Moresby, and urbane areas of Central, Eastern Highlands, East Sepik, and East New Britain provinces. By contrast, the association of the endemic NCDs mortality with sociodemographic characteristics of the deceased was unclear, except for children under five years of age, who were less likely to die from these diseases because these chronic conditions often commence in later years of life and last for years before the patients died from it.
About $85\%$ of the PNG population reside in rural areas, mostly involving in subsistence-based agriculture [29, 30]. The urban surveillance sites in the catchment of CHESS can be basically classified into three types based on their level of urbanisation: (i) Towns in low level of urbanisation are more associated with household economy with incomes mostly from the agricultural industries such as farming and fishery i.e. Goroka in EHP and Maprik in East Sepik provinces; (ii) Towns in middle level of urbanisation are more associated with the development of the private sector with waged employment mostly from mining and food processing industries i.e. Hiri in Central, Newtown in Madang and Kokopo in ENB provinces; and (iii) Towns with high level of urbanisation are more associated with the development of the public sector and services i.e. Hohola in POM [23].
Urbanisation appears a major socioeconomic determinant of the mortality transition in PNG, shifting the CODs in the population from infectious diseases towards NCDs. Unlike the urbanisation processes in other LMICs where the epidemiological transition from infectious diseases to NCDs progress slowly over a period of time [30], this study shows evidence suggesting that the mortality transition in PNG had taken place at an early stage of the urbanisation. Given the recent urbanisation progress in PNG, the epidemiological transition has already had an impact with an increase in mortality attributed to emerging NCDs among the urban population.
The physical and human contrasts, extensive swamps, impenetrable bush-tangled rocky terrain, and high mountains are still effective barriers to human mobility, settlement and communication in PNG [31]. However, PNG people are becoming more mobile, even in remote areas of the country, and migration destinations are most likely associated with job and employment opportunities. Along with the expansion of urban areas and the development of new towns, farmers are leaving their homes in rural areas to move into urban centres, resulting in increased population growth across cities and towns in PNG. The internal migration flow was slow and small at the beginning, but has increased more recently in areas, where large development projects are taking place, attracting internal migration flows into and out of the project sites. The crude gross migration rate was as high as $16\%$ across the surveillance sites in 2018 [21].
Urbanization interplays with local contextual factors providing possible explanations for the variation and differences in trends in mortality in PNG. The burden of NCDs is not equally distributed across the provinces. POM had the highest proportion of mortalities from emerging and endemic NCDs, accounting for $60\%$ of the total deaths, followed by Central province ($45\%$). Central Province has experienced a rapid urbanisation in the 2010s with the development of the PNG Natural Liquefied Project. Urban populations have been reportedly moving away from traditional foods, diets, and life-styles [32]. In new resettlement areas, health risks are reportedly associated with urban poverty, including poor housing conditions [24], unhealthy lifestyles and eating behaviours, together with social stresses associated with the working environment and social conflicts [33]. The increase of cash flow and the local economic growth are thought to contribute to increased access to and consumption of processed foods [32, 34]. Children in these sites were reported as consuming high levels of sugar, soft drinks and salty fast foods [35]. Half of the adult population and $25\%$ of adult patients in Central Province were identified as overweight or obese. Nearly half of male adults smoked tobacco. NCDs such as acute cardiac diseases, stroke and diabetes have emerged among other NCDs as a result of the high-risk local environment [14].
Compared to mortality transitions in other countries in the South Pacific Region, the mortality transition in PNG is likely slower. With about $45\%$ of deaths due to emerging and endemic NCDs, our finding is similar to those reported in the recent study in PNG ($48.8\%$), lower than Solomon ($70\%$) [19], and Fiji ($80\%$) [36]. Our data support that emerging NCDs such as cardiovascular diseases, stroke, diabetes, and endemic NCDs including cancers and chronic respiratory diseases are among the leading CODs in the population, with the highest impact on socioeconomic development and healthcare systems in LMICs [37, 38], where weak governance, poor administration, and fragile health systems with limited human and financial resources are common reasons for ineffective responses to planning and management of urban health [39].
## Limitations
Mortality data from VA interviews were used to assign possible CODs, but the information about the deaths of the deceased provided by the relatives could be biased, particularly for those interviews which were conducted two or more years after the date of death [40]. The deaths in the communities were identified by the village-based data reporters, but these data were not included unless the deaths occurred within the catchment areas of the surveillance system during the data collection period. Our field work experiences and mortality surveillance records suggest that about $20\%$ of deaths in the urbane sites in POM and Madang and about $15\%$ of deaths in the rural sites in Eastern Highlands, East New Britain, East Sepik, and Central provinces occurred in health facilities [13, 18].
Since CHESS was designed to have a primary health facility included within the catchment areas of each surveillance site, $95\%$ of the surveillance population reported having access to primary healthcare services [25]. About $90\%$ of the population of working age (15–64 years) reported having some kind of jobs in the past two weeks and the education attainment was low (net enrollment rate in primary education below $60\%$) [41]. This could be a reason for why those variables on women’s access to health services, employment and education were non-significant in the statistical modeling.
## Conclusion
The study has provided insights into the current mortality transition in PNG. Urbanisation and local contextual factors could be the key socioeconomic correlates of the mortality transition in PNG. The study found that different sociodemographic factors contributed to the increased mortalities from emerging and endemic NCDs. Different health policies and interventions are needed to target urban and rural populations to change unhealthy life styles to reduce premature mortality from NCDs. The increased risk of dying from emerging NCDs such as cardiovascular diseases, stroke and diabetes mellitus is evident in urban areas. Urban residents are twice more likely to die from these emerging NCDs, particularly among those who are 45 years and older, and from provinces undergoing rapid urbanisation such as Port Moresby and Central province. By contrast, mortality attributed to endemic NCDs such as neoplasms, chronic respiratory diseases and malnutrition are evenly distributed across ages, sexes, urban-rural sectors, provinces and social classes.
As urbanisation may continues in PNG in coming years, profound demographic changes including domestic migration will have greater impacts on the health and well-being of the population. Policy level changes are needed to reduce risk factors among the most at risk populations. This study has provided evidence that can be used by PNG Government agencies and the health sector. Health systems are much in need to reform and restructure to effectively respond to new challenges and to meet with the increased demand for healthcare services for NCDs among the population. Equitable access to effective and quality preventive measures and curative services are needed to protect the population from premature deaths from NCDs. NCDs prevention programmes and interventions need to focus on emerging NCDs and target populations in urban areas and in provinces, where urbanisation is occurring at a rapid pace.
Socioeconomic development programmes and health interventions are needed to alleviate poor health conditions, change unhealthy lifestyles and eating behaviors, and reduce the social stress of urban living. Social planners, public health experts and stakeholders need to make greater efforts to improve the health and well-being of the population in an equitable, effective and sustainable manner. This calls for more data and analysis of causes of death to assist the PNG Government in developing strategies to effectively address multiple complex public health issues in such critical epidemiological transitional period in PNG.
## References
1. 1WHO. Sustainable Developmetn Goals Health and Health related Targets 2015 Available from: http://www.who.int/gho/publications/world_health_statistics/2016/EN_WHS2016_Chapter6.pdf
2. 2The United Nations. United Nations’ Millennium Development Goals: United Nations; 2015 Available from: http://www.un.org/millenniumgoals/childhealth.shtml
3. Martinez R, Lloyd-Sherlock P, Soliz P, Ebrahim S, Vega E, Ordunez P. **Trends in premature avertable mortality from non-communicable diseases for 195 countries and territories, 1990–2017: a population-based study**. *The Lancet Global health* (2020.0) **8** e511-e23. DOI: 10.1016/S2214-109X(20)30035-8
4. Abera SF, Gebru AA, Biesalski HK, Ejeta G, Wienke A, Scherbaum V. **Social determinants of adult mortality from non-communicable diseases in northern Ethiopia, 2009–2015: Evidence from health and demographic surveillance site.**. *PLOS ONE.* (2017.0) **12** e0188968. DOI: 10.1371/journal.pone.0188968
5. 5WHO. Noncommunicable diseases 2021 Available from: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases
6. Bennett J, Stevens G, Mathers C, Bonita R, Rehm J, Kruk M. **NCD Countdown 2030: worldwide trends in non-communicable disease mortality and progress towards Sustainable Development Goal target 3.4**. *Lancet* (2018.0) **392** 1072-88. DOI: 10.1016/S0140-6736(18)31992-5
7. Reubi D, Herrick C, Brown T. **The politics of non-communicable diseases in the global South.**. *Health & Place.* (2016.0) **39** 179-87. DOI: 10.1016/j.healthplace.2015.09.001
8. Chen S, Kuhn M, Prettner K, Bloom DE. **The macroeconomic burden of noncommunicable diseases in the United States: Estimates and projections.**. *PloS one.* (2018.0) **13** e0206702. DOI: 10.1371/journal.pone.0206702
9. Chen S, Bloom DE. **The macroeconomic burden of noncommunicable diseases associated with air pollution in China.**. *PLoS One* (2019.0) **14** e0215663. DOI: 10.1371/journal.pone.0215663
10. Allen L, Williams J, Townsend N, Mikkelsen B, Roberts N, Foster C. **Socioeconomic status and non-communicable disease behavioural risk factors in low-income and lower-middle-income countries: a systematic review**. *The Lancet Global Health* (2017.0) **5** e277-e89. DOI: 10.1016/S2214-109X(17)30058-X
11. 11WHO. Sustainable Development Goal 3: WHO; 2021 Available from: https://www.who.int/health-topics/sustainable-development-goals#tab=tab_1
12. 12National Statistics Office. Papua New Guinea Demographic and Health Survey 2016–18: Key Indicators Report.
Port Moresby, Papua New Guinea: National Statistics Office; 2019.. *Papua New Guinea Demographic and Health Survey 2016–18: Key Indicators Report.* (2019.0)
13. Pham NB, Abori N, Aga T, Jorry R, Jaukae J, Silas V. *Comprehensive Health and Epidemiological Surveillance System: March 2020 Edition on Morbidity Surveillance at Primary Health Facilities in PNG.* (2020.0)
14. Gouda HN, Hazard RH, Maraga S, Flaxman AD, Stewart A, Joseph JC. **The epidemiological transition in Papua New Guinea: new evidence from verbal autopsy studies**. *International Journal of Epidemiology* (2019.0) 1-12. DOI: 10.1093/ije/dyy184
15. Boli R, Pham NB, Siba P. **Assessing the changing burden of diseases at the primary healthcare level in rural Papua New Guinea**. *PNG Medical Journal* (2017.0) **60**
16. 16PNG Government. National Health Plan: 2010–2020 (Volume 1: Policies and Strategies.
Port Moresby: National Department of Health; 2010.. *National Health Plan: 2010–2020 (Volume 1: Policies and Strategies.* (2010.0)
17. Rarau P, Vengiau G, Gouda HN, Phuanukoonnon S, Kevau I, Bullen C. **Prevalence of non-communicable disease risk factors in three sites across Papua New Guinea: a cross-sectional study**. *BMJ Glob Health* (2017.0) **2** e000221. DOI: 10.1136/bmjgh-2016-000221
18. Pham BN, Jorry R, Abori N, Silas V, Maraga S, Kue L. *Comprehensive Health and Epidemiological Surveillance System Technical Report: Mortality Surveilance in Communities* (2020.0)
19. Hart JD, Mahesh P, Kwa V, Reeve M, Chowdhury HR, Jilini G. **Diversity of epidemiological transition in the Pacific: Findings from the application of verbal autopsy in Papua New Guinea and the Solomon Islands**. *The Lancet Regional Health West Pacific* (2021.0) **11**. DOI: 10.1016/j.lanwpc.2021.100150
20. 20National Statistics Office. National Population and Housing census 2011. Port Moresby; 2013.
21. Pham NB, Maraga S, Boli R, Aga T, Jorry R, Kue L. *Partnership in Health Programme: March 2018 Edition on Population Census and Demographic Changes in the PNG IMR’s Health and Epidemiological Surveillance Sites in the period 2015–2017* (2018.0)
22. Government of PNG, Moresby Port. *Medium Term Development Plan III 2018–2022.* (2018.0)
23. Pham NB, Maraga S, Boli R, Kue L, Ainiu N, Aga T. *Comprehensive Health and Epidemiological Surveillance System: September 2018 Edition on Household Socioeconomic and Demographic Characteristics* (2018.0)
24. 24PNG Institute of Medical Research. Comprehensive Health and Epidemiological Surveillance System Technical Report: Household Socioeconomic and Demographic Characteristics (Reporting period: January-June 2018) Goroka: PNG Institute of Medical Research; 2018 Available from: https://www.researchgate.net/publication/329706050_Comprehensive_Health_and_Epidemiological_Surveillance_System_Technical_Report_Household_Socioeconomic_and_Demographic_Characteristics
25. Pham NB, Whittaker M, Pomat W, Siba P. **CHESS: a new generation of population health surveillance for sustainable development of Papua New Guinea**. *PNG Med J* (2017.0) **60** 154-72
26. 26WHO. Verbal autopsy standards: ascertaining and attributing causes of death 2016 Available from: https://www.who.int/healthinfo/statistics/verbalautopsystandards/en/
27. Nichols E, Byass P, Chandramohan D, Clark SJ, Flaxman AD, Jakob R. **The WHO 2016 verbal autopsy instrument: An international standard suitable for automated analysis by InterVA, InSilicoVA, and Tariff 2.0.**. *PLoS Med.* (2018.0) **15** e1002486. DOI: 10.1371/journal.pmed.1002486
28. 28Centers for Diseases Control and Prevention. International Classification of Diseases, Tenth Revision (ICD-10): National Center for Health Statistics; 2021 Available from: https://www.cdc.gov/nchs/icd/icd10.htm
29. Rarau P, Pulford J, Gouda H, Phuanukoonnon S, Bullen C, Scragg R. **Correction: Socio-economic status and behavioural and cardiovascular risk factors in Papua New Guinea: A cross-sectional survey.**. *PLOS ONE.* (2019.0) **14** e0212894. DOI: 10.1371/journal.pone.0212894
30. Utzinger J, Keiser J. **Urbanization and tropical health—then and now**. *Ann Trop Med Parasitol* (2006.0) **100** 517-33. DOI: 10.1179/136485906X97372
31. 31PNG Institute of Medical Research. Partnership on Health Project Report: Population Census and Demographic Changes (Reporting period: July-December 2017).
Goroka: PNG Institute of Medical Research; 2018.. *Partnership on Health Project Report: Population Census and Demographic Changes (Reporting period: July-December 2017).* (2018.0)
32. Pham NB, Silas VD, Okely AD, Pomat W. **Breastfeeding rate, food supplementary and dietary diversity among children aged 6–59 months, and associated risk factors in Papua New Guinea.**. *Frontiers in Nutrition* (2021.0) **8**
33. Pham NB, Silas V, Jorry R, Okely AD, Pomat W. **Marijuana-use related Homicide: A case study in Papua New Guinea.**. *Annal Case Report* (2020.0) **14** 312
34. Pham NB, Silas VD, Okely AD, Pomat W. **Measuring wasting and stunting prevalence among children under 5 years of age and associated risk factors in Papua New Guinea.**. *Frontiers in Nutrition* (2021.0) **8** 622660. DOI: 10.3389/fnut.2021.622660
35. 35Pham NB, Maraga S, Degemba B, Kue L, Ainui N, Aga T, et al. Comprehensive Health and Epidemiological Surveillance System: March 2019 Edition on Child Health Goroka 2019 Available from: https://www.researchgate.net/publication/333678257_PNG_IMR’s_CHESS_Technical_Report_March_2019_Edition_Child_Health
36. Kessaram T, McKenzie J, Girin N, Roth A, Vivili P, Williams G. **Noncommunicable diseases and risk factors in adult populations of several Pacific Islands: results from the WHO STEPwise approach to surveillance.**. *Aust N Z J Public Health* (2015.0) **39** 336-43. DOI: 10.1111/1753-6405.12398
37. Chand SS, Singh B, Kumar S. **The economic burden of non-communicable disease mortality in the South Pacific: Evidence from Fiji.**. *PloS one.* (2020.0) **15** e0236068. DOI: 10.1371/journal.pone.0236068
38. Ediriweera DS, Karunapema P, Pathmeswaran A, Arnold M. **Increase in premature mortality due to non-communicable diseases in Sri Lanka during the first decade of the twenty-first century.**. *BMC Public Health* (2018.0) **18** 584. DOI: 10.1186/s12889-018-5503-9
39. Tabibzadeh I, Liisberg E. **Response of health systems to urbanization in developing countries.**. *World health forum.* (1997.0) **18** 287-93. PMID: 9478144
40. Serina P, Riley I, Stewart A, James SL, Flaxman AD, Lozano R. **Improving performance of the Tariff Method for assigning causes of death to verbal autopsies.**. *BMC Med.* (2015.0) **13**. DOI: 10.1186/s12916-015-0527-9
41. Pham NB, Okely AD, Whittaker M, Siba P, Pomat W. **Millennium development goals in Papua New Guinea: towards universal education**. *Educational Research for Policy and Practice* (2020.0) **19** 181-209
|
---
title: 'Impacts of tuberculosis services strengthening and the COVID-19 pandemic on
case detection and treatment outcomes in Mimika District, Papua, Indonesia: 2014–2021'
authors:
- Trisasi Lestari
- Kamaludin
- Christopher Lowbridge
- Enny Kenangalem
- Jeanne Rini Poespoprodjo
- Stephen M. Graham
- Anna P. Ralph
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021881
doi: 10.1371/journal.pgph.0001114
license: CC BY 4.0
---
# Impacts of tuberculosis services strengthening and the COVID-19 pandemic on case detection and treatment outcomes in Mimika District, Papua, Indonesia: 2014–2021
## Abstract
Indonesia is a high-burden tuberculosis (TB) country with a wide case detection gap, exacerbated by the COVID-19 pandemic. We aimed to review the epidemiology of TB in a high-endemic setting of Indonesia before and during the implementation of health system strengthening activities for TB, including during the first two years of the COVID-19 pandemic. We analysed TB program data from Mimika District, Papua, Indonesia from 2014 to 2021. Health system strengthening activities to improve the programmatic management of TB were implemented from 2017 onwards. Activities included decentralization of TB services, training and mentoring of healthcare workers, improved screening for co-morbidities, and introduction and optimisation of Xpert testing in 2018. A total of 11,803 TB cases were notified to the Mimika District Health Office over eight years (2014–21). Between 2015 and 2019, there was a $67\%$ increase in annual case notifications, an $89\%$ increase in bacteriologically confirmed cases and the proportion of TB cases detected in primary care increased from $26\%$ to $46\%$. In 2020, coinciding with the COVID-19 pandemic, investigation of people with presumptive TB fell by $38\%$, but the proportion of those tested with Xpert increased. TB case notifications decreased by $19\%$ from 1,796 in 2019 to 1,461 in 2020, but then increased by $17\%$ to 1,716 in 2021. Routine screening for co-morbidities (HIV, diabetes) among TB patients improved over time and was not affected by the pandemic. Treatment success overall was $71\%$ and remained relatively unchanged. Loss to follow-up and death were $18\%$ and $3.7\%$ respectively. Improvements in TB case finding were observed over a period in which a range of health system strengthening activities were implemented. While COVID-19 had a negative impact on the TB program in Mimika District, there are encouraging signs of recovery. Further work is needed to improve TB treatment outcomes.
## Introduction
Tuberculosis (TB) is a leading infectious cause of death, although rivalled in recent years by SARS-CoV-2 (COVID-19). Indonesia has the second highest TB burden globally with an estimated 845,000 TB cases and 98,000 deaths each year [1]. There have been improvements in TB case detection in recent decades, especially since a major case detection gap was identified by the National TB Prevalence Survey in 2014–2015. In Indonesia, overall TB case notifications rose by $69\%$ from 331,703 in 2015 to 562,049 in 2019 but a wide case detection gap remains due to a combination of underdiagnosis and underreporting of people diagnosed with TB. Indonesia accounted for $10\%$ of the estimated 2.9 million TB cases not detected or notified globally in 2019 [1]. In 2021, only $47\%$ of the estimated TB cases in Indonesia were notified [2]. Within Indonesia, there are diverse determinants of TB epidemiology, and variation in performance of regional TB programs. Mimika district in Indonesia’s easternmost province of *Papua is* a multi-cultural mining district with an annual TB case notification rate of 818 per 100,000 population in 2019 [2]. Details of TB epidemiology and program performance in Mimika have been scant, although past reports have indicated a very high burden of disease [3, 4].
In 2017, in a collaboration between the Mimika District TB Program, University of Gadjah Mada, Timika Research Facility and Menzies School of Health Research (Australia), we commenced a TB health-systems strengthening project, the Stronger Health Systems for multidrug-resistant tuberculosis and malaria (STRATUM) initiative and a follow-up initiative called PRIME-TB (Papua New Guinea and Indonesia for the Micro Elimination of TB). The focus of this project was on prevention of TB through implementation and scale-up of screening household contacts of people with TB and use of TB preventive treatment (TPT) among young (<5 years) child contacts without TB. However, a broader set of health system strengthening activities were also implemented, including improvements to TB program information management systems, and use of a Continuous Quality Improvement (CQI) framework to identify gaps and opportunities for improving the clinical and public health management of TB [5] Mimika district has also received TB program support through United States Agency for International Development (USAID) funded initiatives, including promotion of active case finding and technical assistance to develop a district action plan for TB control [6].
Since early 2020, the COVID-19 pandemic has impacted the delivery of TB services. Staff and resources were diverted to the pandemic response and TB patients have avoided health care contact to limit their exposure to COVID-19. Utilization of primary and tertiary health care decreased dramatically [7] Indonesian national TB case notifications dropped to 393,323 in 2020, a $31\%$ reduction compared to 2019 and less than half the estimated caseload [1]. In response to escalating COVID-19 cases in Mimika from March 2020 the local government implemented large-scale social restrictions [8]. Unintended impacts on access of the population to healthcare, including TB care, quickly become evident.
We aimed to evaluate the impact of strengthening of TB services on TB case notifications, diagnostic practices and treatment outcomes in this high-incidence setting of Indonesia, as well as the impact of the first two years of the COVID-19 pandemic on any progress made.
## Study design, setting and population
We conducted a retrospective epidemiological review of routinely collected TB program data. The study setting was Mimika District, located in Papua Province of Indonesia. The population of Mimika was 311,969 in 2020. It is a fast-growing district with urban and rural areas built around a copper and gold mine. It is a culturally and linguistically diverse area, with approximately half the population comprised of Indigenous Papuan peoples. There are five hospitals and 23 primary health centers, 20 of which provide TB diagnosis and treatment services, 13 located in rural settings [9]. Presumptive TB cases are requested to provide two sputum samples for Xpert assay (if available) or for smear microscopy if Xpert not accessible. Chest x-ray is only available in hospital and not routinely done for presumptive TB cases from primary health centers. The study population was limited to persons diagnosed with TB and notified to the TB program of Mimika District, Papua Province, Indonesia between 2014 and 2021.
## Data sources and variables
The National TB Program (NTP) of Indonesia uses manual paper-based data recording using national standardized forms to report all presumptive and confirmed TB cases. These data have been entered into an electronic platform that was first introduced into Mimika district in 2013, namely the ‘Integrated Tuberculosis Information System’ or Sistem Informasi Tuberkulosis Terpadu [SITT] for 2014 until 2020. It was then upgraded to the ‘Tuberculosis Information System’ or Sistem Informasi Tuberkulosis [SITB], which included additional details and forms, and option for real-time updates. Limited internet coverage in Mimika means data entry is often delayed and batched until TB staff can visit a site with internet access. The district TB coordinator regularly reviews submitted data and checks for completeness and consistency.
We used data from individual TB patient records and the electronic registers (SITT and SITB) for the period 1 January 2014 to 31 December 2021, and data on all people suspected of and investigated for TB for the period 1 January 2019 to 31 December 2021. Cohort data of treatment outcomes were documented up to 31 December 2020. To allow for delayed entry, data were extracted until March 2022. We undertook comprehensive data validation for TB patients by cross-checking paper forms with the electronic databases to minimize missing data, errors, and duplication. Ethnicity is not documented; therefore, we manually assigned ethnicity as indigenous Papuan or non-Papuan Indonesian on the basis of surname [10]. Surname is considered a reliable indicator of ethnicity locally. History of TB treatment was often misreported and was corrected by manual cross-checking. The main data source was the standardized individual TB register form (‘TB03’) which provides a record of patients who received TB treatment and their treatment outcome. Data collected included demographic characteristics, name of health facility, date of registration, date of treatment commencement, laboratory results, TB disease classification, HIV and diabetes status, treatment for drug-susceptible or drug-resistant TB, and treatment outcome. Extra-pulmonary TB (EPTB) is defined as TB disease involving organs other than the lungs. A case with evidence of TB in both pulmonary and extra-pulmonary sites is notified to TB program as pulmonary TB. Outcome was categorized as ‘treatment success’ if treatment cure or completion was recorded.
Only sputum specimens were evaluated for Xpert or smear, so no EPTB cases were bacteriologically confirmed. Laboratory results comprised smear microscopy for acid fast bacilli and nucleic acid amplification using the Xpert MTB/RIF (Cepheid, USA) (Xpert). Mycobacterial culture was largely unavailable. TB diagnosis in children is mostly a clinical diagnosis made using the ‘Indonesian pediatric TB scoring system’ (S4 Table); although limitations of this are recognized [11, 12]. Since 2016, Xpert has been recommended as the preferred first-line diagnostic option for samples from children with presumptive TB [11].
## Interventions to improve TB case finding
To understand local contextual factors impacting TB program performance in addition to our own health system strengthening work [5], we sought information from the Mimika District Health Office about TB activities led by government and non-government organisations during the years of the study. These are shown in Table 1. In summary, in 2016 the Community Empowerment of People Against Tuberculosis (CEPAT) project was introduced to improve TB case finding, monitoring and evaluation, and TB surveillance [6]. This project recruited and trained local community health workers to do door-to-door TB symptom screening and to facilitate attendance of symptomatic individuals at a Primary Health Centre (PHC, known as Puskesmas) for evaluation and laboratory testing. In 2017, the TB Challenge project provided technical assistance to the district TB coordinator and staff in health facilities to optimize TB reporting, public-private mix in TB management, and develop a district action plan for TB control [6]. The CEPAT and TB Challenge projects ended in 2018 and 2019, respectively.
**Table 1**
| Year | Activity |
| --- | --- |
| 2014 | 1. Electronic data entry for TB (SITT) in use for a full year after installment in 20132. Routine TB monitoring and evaluation meetings were conducted biannually |
| 2015 | Treatment was initiated for the first drug-resistant TB case |
| 2016 | Introduction of TB CEPAT (‘Community Empowerment of People Against Tuberculosis’)–funded by USAID |
| 2017 | 1. TB CEPAT project continued2. TB Challenge project was introduced–funded by USAID3. TDRRCI project (‘Tropical Disease Research Regional Collaboration initiative’)–funded by the Australian Government (the Indo-Pacific Centre for Health Security of the Department of Foreign Affairs and Trade) to support following activities: • Establishment of household contact screening and management with TPT using 6H for young (<5 years) child contacts under-5 years in five facilities • Local TB training, focusing on childhood TB and TPT • Introduction of quarterly Continuous Quality Improvement (CQI) meetings for TB program |
| 2018 | 1. TB CEPAT project ended2. GeneXpert MTB/RIF machine (first in the district) installed in the District Hospital3. TDRRCI project activities: • TB training provided–TB treatment; child TB; TB in pregnancy • Scale-up of household contact screening and TPT program to 11 health facilities • TB program competition between health facilities with prizes for best-performing facilities–case finding, contact screening, HIV testing, TPT • Diabetes screening kit was distributed to health facilities • TB Monitoring and Evaluation meetings quarterly |
| 2019 | 1. TB Challenge project ended2. The electronic TB Data Entry was updated (SITB)3. STRATUM project (‘Stronger Health Systems for multidrug-resistant tuberculosis and malaria’)–funded as a follow-up to TDRRCI by the Indo-Pacific Centre for Health Security of the Australian Government Department of Foreign Affairs and Trade to support: • Comprehensive care introduced for drug-resistant TB care • Scale-up of household contact screening and TPT program to 16 health facilities • Introduced monthly meetings for TB monitoring and evaluation, led by the District Health Office • TB training provided: infection control; treatment of infection |
| 2020 | 1. Public health response to COVID–diversion of health services and human resources; isolation/curfew at home from 2 pm; reduced clinic time at facilities; the GeneXpert machine in the district hospital temporarily used for SARS-CoV-2 diagnosis; contact screening implemented for COVID instead of TB2. A new GeneXpert machine installed in a Primary Health Center for TB diagnosis3. STRATUM project and follow-up PRIME-TB project (’Papua New Guinea & Indonesia for the Micro Elimination of TB’), also funded by Australian Government (the Indo-Pacific Centre for Health Security of the Department of Foreign Affairs and Trade): • Introduced online TB monitoring and evaluation meeting • Introduced online TB training • Activities to strengthen detection and treatment of MDR TB |
| 2021 | 1. Indonesian NTP introduced short regimen for TPT using 3HP and 3RH2. Three additional GeneXpert machines installed (for a total of five in Mimika District): one to the district hospital for TB diagnosis; one to the district hospital for SARS-CoV-2 diagnosis during the national sport event; and one to a primary health center for TB diagnosis3. PRIME-TB activities: • Hybrid, online and onsite, TB monitoring and evaluation meeting • Hybrid, online and onsite, TB training, consultation, and mentoring |
From 2017, our research team implemented a multi-component intervention funded by the Australian government to initiate and strengthen household contact screening and management [5]. This project chiefly aimed to increase TB case finding among contacts and to provide TB preventive treatment (TPT) to young child contacts. Screening for co-morbidities was also supported with the distribution of blood glucose testing kits in 2018 to screen for diabetes and encouragement to offer HIV screening. Our project’s interventions have included group training, developing and providing educational materials, technical assistance and mentoring. Regular CQI meetings have been held to review progress and discuss achievements, barriers and challenges. All TB staff participate in the CQI meetings, raise their concerns and propose ideas to solve the identified problems.
A four-module GeneXpert (Cepheid, USA) machine was procured by the NTP for the district hospital in mid-2018 to detect *Mycobacterium tuberculosis* and rifampicin resistance (using ‘Xpert MTB/RIF’ cartridges) (Table 1). A second GeneXpert machine was installed in March 2020 at the Puskesmas with the highest TB caseload and three additional machines were installed in 2021, two at the district hospital and one at a PHC facility.
## The COVID-19 response in Mimika District and impact on TB activities
Two pandemic waves occurred during the study period. The first, in 2020, triggered strong social restriction policies [7]. The second, between May and September 2021 brought limited social restrictions [13]. By end of 2021, a total of 110,079 COVID-19 cases had been reported from Mimika district [14]. To support COVID-19 diagnosis, the Indonesian Ministry of Health recommended using Xpert Xpress SARS-CoV-2 cartridges [15, 16]. Therefore, the GeneXpert machine at the district hospital was temporarily diverted from TB to COVID-19 testing. Restrictions enforced during 2020 included reduced opening hours for healthcare facilities. COVID-19 stigma resulted in people avoiding contact with healthcare providers and rejection of outreach activities from TB staff. Most healthcare staff were diverted to the COVID-19 response, including the District TB Coordinator and TB staff in primary care and hospitals. In October 2021, Papua province hosted a national sports event with participants from around Indonesia. Healthcare workers, including TB staff in Mimika district, were required to do COVID-19 screening of participants and provide healthcare at the sports venues. This directly affected provision of routine care.
During the first pandemic year the STRATUM TB program maintained CQI meetings and TB training videoconferences and webinars. This allowed additional TB staff from rural areas with internet access to participate. Coordination and communication were also maintained using the WhatsApp group mobile messaging platform to keep TB staff informed and motivated. Since 2021, when social distancing was relaxed, ongoing training and CQI meetings were run using a hybrid in-person and online model.
## Analysis
Data were managed in Microsoft Excel 365 and analysis was performed using Stata v13 (StataCorp 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP). Among TB patients, we used Pearson’s Chi-square test to compare outcomes between subgroups with $p \leq 0.05$ considered significant. Treatment outcomes were assessed by calculating odds ratios in univariable and multivariable models including age, ethnicity, residential location, health facility type, location of TB, mode of diagnosis, history of TB treatment, HIV and diabetes status. Longitudinal comparisons were tested with Stata’s ptrend command (chi-square statistic for trend) excluding the year 2014 due to incomplete data. To examine COVID-19 impact, we created an annotated plot of monthly numbers of potential (suspected) TB cases as well as TB case notifications from 2019 to 2021.
## Ethical approval
We obtained ethical clearance from the institutional review boards at the Universitas Gadjah Mada (KE/$\frac{0715}{06}$/2018; KE/$\frac{1188}{10}$/2019; and KE/$\frac{0090}{02}$/2021), the Northern Territory Department of Health and Menzies School of Health Research (2017–2777) and Charles Darwin University (H20110). Permission to access TB data was obtained from the Mimika District Health Office.
## Investigation of people with presumptive TB
Data on screening for TB were available for 2019–2021. In 2019, 7745 people were screened for TB. This represented three quarters of the annual target of 10,314 set by the national TB program [17]. The impact of COVID-19 is shown as an annotated plot in Fig 1 and S1 Table; there was a $38\%$ decrease in numbers of people screened for TB to 4808 in 2020, attributable to lockdown, drops in presentations to clinics and a fall in contact screening activities to identify potential cases. This coincided with an increase in the proportion of Xpert assays that tested positive, from $14.8\%$ of community members with presumptive TB in 2019 to $19.5\%$ in 2020 and $20.5\%$ in 2021.
**Fig 1:** *Monthly number of TB suspects and newly diagnosed patients with TB before and during the COVID-19 pandemic.*
## TB notifications and characteristics of people with TB
During the eight-year study period, 11,803 TB cases were notified in Mimika District. Table 2 lists characteristics of TB cases by year. Overall, $69.1\%$ were of Papuan ethnicity and $82.1\%$ were from an urban setting, reflecting the general Mimika population [18]. The prevalence of co-morbidities among those tested were $10.5\%$ with HIV infection and $5.4\%$ with diabetes which compares to a prevalence of $2.4\%$ and $1.1\%$ in the general population respectively [18]. Overall, $22\%$ of TB cases were EPTB, higher than the national average of $9\%$ [1] and more common in children ($27.3\%$), people living with HIV ($26.2\%$) and Papuan people ($24.7\%$).
**Table 2**
| Year of reporting | 2014 | 2015 | 2016 | 2017 | 2018 | 2019 | 2020 | 2021 | Total |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Total cases | 920 | 1078 | 1506 | 1567 | 1759 | 1796 | 1461 | 1716 | 11803 |
| Characteristics | Number (%) | Number (%) | Number (%) | Number (%) | Number (%) | Number (%) | Number (%) | Number (%) | Number (%) |
| Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex | Sex |
| Male | 520 (56.5) | 616 (57.1) | 865 (57.4) | 913 (58.3) | 1011 (57.4) | 985 (54.8) | 822 (56.3) | 968 (56.4) | 6699 (56.8) |
| Female | 400 (43.5) | 462 (42.9) | 641 (42.6) | 654 (41.7) | 748 (42.6) | 811 (45.2) | 639 (43.7) | 748 (43.6) | 5104 (43.2) |
| Age group | Age group | Age group | Age group | Age group | Age group | Age group | Age group | Age group | Age group |
| 0–4 years | 58 | 53 | 137 | 201 | 228 | 185 | 153 | 234 | 1249 |
| 0–4 years | 6.3 | 4.9 | 9.1 | 12.8 | 13.0 | 10.3 | 10.5 | 13.6 | 10.6 |
| 5–14 years | 113 | 138 | 195 | 159 | 193 | 211 | 158 | 169 | 1336 |
| 5–14 years | 12.3 | 12.8 | 12.9 | 10.2 | 11.0 | 11.8 | 10.8 | 9.9 | 11.3 |
| 15–44 years | 624 | 680 | 873 | 934 | 1027 | 1066 | 882 | 981 | 7067 |
| 15–44 years | 67.8 | 63.1 | 58.0 | 59.6 | 58.4 | 59.4 | 60.4 | 57.2 | 59.9 |
| 45–64 year (n) | 111 | 183 | 264 | 240 | 280 | 298 | 239 | 308 | 1923 |
| 45–64 year (n) | 12.1 | 17.0 | 17.5 | 15.3 | 15.9 | 16.6 | 16.4 | 18.0 | 16.3 |
| > = 65 year (n) | 14 | 24 | 37 | 33 | 31 | 36 | 29 | 24 | 228 |
| > = 65 year (n) | 1.5 | 2.2 | 2.5 | 2.1 | 1.8 | 2.0 | 2.0 | 1.4 | 1.9 |
| Ethnicity | Ethnicity | Ethnicity | Ethnicity | Ethnicity | Ethnicity | Ethnicity | Ethnicity | Ethnicity | Ethnicity |
| Papuan | 632 | 799 | 1037 | 1043 | 1227 | 1266 | 1044 | 1151 | 8159 |
| Papuan | 68.7 | 74.1 | 68.9 | 66.6 | 69.8 | 70.5 | 68.7 | 67.1 | 69.1 |
| Non-Papuan | 274 | 250 | 391 | 450 | 521 | 525 | 451 | 565 | 3427 |
| Non-Papuan | 29.8 | 23.2 | 25.9 | 28.7 | 29.6 | 29.2 | 30.9 | 32.9 | 29.0 |
| Unknown | 14 | 29 | 78 | 74 | 11 | 5 | 6 | 0 | 217 |
| Unknown | 1.5 | 2.7 | 5.2 | 4.7 | 0.6 | 0.3 | 0.4 | 0 | 1.9 |
| Previous TB | Previous TB | Previous TB | Previous TB | Previous TB | Previous TB | Previous TB | Previous TB | Previous TB | Previous TB |
| History of TB treatment | 44 | 51 | 87 | 148 | 185 | 177 | 151 | 179 | 1022 |
| History of TB treatment | 4.8 | 4.7 | 5.8 | 9.4 | 10.5 | 9.9 | 10.3 | 10.4 | 8.7 |
| Presenting facility | Presenting facility | Presenting facility | Presenting facility | Presenting facility | Presenting facility | Presenting facility | Presenting facility | Presenting facility | Presenting facility |
| Primary care | 303 | 277 | 468 | 584 | 762 | 822 | 739 | 772 | 4727 |
| Primary care | 32.9 | 25.7 | 31.1 | 37.3 | 43.3 | 45.8 | 50.6 | 45.0 | 40.1 |
| Hospital | 617 | 801 | 1038 | 983 | 997 | 974 | 722 | 944 | 7076 |
| Hospital | 67.1 | 74.3 | 68.9 | 62.7 | 56.7 | 54.2 | 49.4 | 55.0 | 59.9 |
| Residence | Residence | Residence | Residence | Residence | Residence | Residence | Residence | Residence | Residence |
| Urban Mimika | 735 | 869 | 1213 | 1291 | 1467 | 1455 | 1199 | | 8249 |
| Urban Mimika | 84.2 | 80.8 | 80.7 | 82.4 | 83.4 | 81.0 | 82.2 | | 82.1 |
| Rural Mimika | 137 | 202 | 290 | 276 | 283 | 339 | 254 | | 1784 |
| Rural Mimika | 15.7 | 18.8 | 19.3 | 17.6 | 16.1 | 18.9 | 17.4 | | 17.7 |
| Other district | 1 | 4 | 0 | 0 | 9 | 2 | 5 | | 21 |
| Other district | 0.1 | 0.4 | 0.0 | 0.0 | 0.5 | 0.1 | 0.3 | | 0.2 |
| Site of TB | Site of TB | Site of TB | Site of TB | Site of TB | Site of TB | Site of TB | Site of TB | Site of TB | Site of TB |
| Pulmonary | 723 | 792 | 1107 | 1226 | 1378 | 1232 | 1247 | 1526 | 9231 |
| Pulmonary | 78.6 | 73.5 | 73.5 | 78.2 | 78.3 | 68.6 | 85.3 | 88.9 | 78.2 |
| Extra pulmonary | 197 | 286 | 399 | 341 | 381 | 564 | 214 | 190 | 2572 |
| Extra pulmonary | 21.4 | 26.5 | 26.5 | 21.8 | 21.7 | 31.4 | 14.7 | 11.1 | 21.8 |
| HIV status | HIV status | HIV status | HIV status | HIV status | HIV status | HIV status | HIV status | HIV status | HIV status |
| Negative | 337 | 661 | 763 | 954 | 1290 | 1214 | 1052 | 1103 | 7374 |
| Negative | 36.6 | 61.3 | 50.7 | 60.9 | 73.3 | 67.6 | 72.0 | 64.3 | 62.5 |
| Positive | 74 | 146 | 166 | 124 | 125 | 78 | 101 | 81 | 895 |
| Positive | 8.0 | 13.5 | 11.0 | 7.9 | 7.1 | 4.3 | 6.9 | 4.7 | 7.6 |
| Not known | 502 | 252 | 535 | 469 | 333 | 436 | 255 | 482 | 3534 |
| Not known | 55.3 | 25.1 | 38.3 | 31.2 | 19.6 | 28.1 | 21.1 | 31.0 | 29.9 |
| Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus | Diabetes mellitus |
| Negative | | | | 485 | 1420 | 1242 | 1088 | 1356 | 5591 |
| Negative | | | | 30.9 | 80.7 | 69.2 | 74.5 | 79.0 | 67.4 |
| Positive | | | | 43 | 75 | 71 | 61 | 60 | 310 |
| Positive | | | | 2.7 | 4.3 | 4.0 | 4.2 | 3.5 | 3.7 |
| Not tested | | | | 1039 | 264 | 483 | 312 | 300 | 2398 |
| Not tested | | | | 66.3 | 15.0 | 26.9 | 21.3 | 17.5 | 28.9 |
| Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome | Treatment outcome |
| Treatment success | 475 | 766 | 1078 | 1183 | 1306 | 1280 | 1090 | | 7178 |
| Treatment success | 51.6 | 71.1 | 71.6 | 75.5 | 74.3 | 71.3 | 74.6 | | 71.2 |
| Died | 21 | 32 | 28 | 32 | 79 | 97 | 88 | | 377 |
| Died | 2.3 | 3.0 | 1.9 | 2.0 | 4.5 | 5.4 | 6.0 | | 3.7 |
| Loss to follow-up | 182 | 210 | 261 | 285 | 308 | 344 | 229 | | 1819 |
| Loss to follow-up | 19.8 | 19.5 | 17.3 | 18.2 | 17.5 | 19.2 | 15.7 | | 18.0 |
| Failed | 9 | 3 | 10 | 7 | 18 | 17 | 9 | | 73 |
| Failed | 1.0 | 0.3 | 0.7 | 0.5 | 1.0 | 1.0 | 0.6 | | 0.7 |
| Transfer out | 25 | 62 | 105 | 60 | 39 | 52 | 29 | | 372 |
| Transfer out | 2.7 | 5.8 | 7.0 | 3.8 | 2.2 | 2.9 | 2.0 | | 3.7 |
| Not recorded | 208 | 5 | 24 | 0 | 9 | 6 | 16 | | 268 |
| Not recorded | 22.6 | 0.5 | 1.6 | 0.0 | 0.5 | 0.3 | 1.1 | | 2.7 |
TB case detection and reporting increased annually until the onset of the COVID-19 pandemic in 2020 (Fig 2). Between 2015 and 2019 there was a $67\%$ increase in notified TB cases (1078 to 1796 cases), achieving $94\%$ of the estimated annual target case detection rate of 1910 cases for 2019 as set by the health authority. This was followed in 2020, coinciding with COVID-19 spread and lockdown measures, by a $18.7\%$ decrease in case notification from 1796 in 2019 to 1461 in 2020, which then increased to 1716 in 2021.
**Fig 2:** *Number of newly diagnosed patients with tuberculosis in Mimika district from 2014 to 2021.*
## Diagnosis and quality of care among people diagnosed with TB
The proportion of cases diagnosed in primary care increased over time from around a third to half of all cases (Fig 2). Although the proportion with bacteriological confirmation remained relatively constant at $35.8\%$ overall, there was an $89\%$ increase in the number of individuals with bacteriologically confirmed TB from 361 in 2015 to 682 in 2019 (Tables 3 and S1). All samples tested were sputum samples. The mode of diagnosis shifted from smear microscopy as the sole diagnostic option to Xpert being used in more than half the cases during the study period (Tables 3 and S1 and Fig 3). Bacteriological confirmation was uncommon in children (<15 years of age) but there was an increase each year in the proportion of children with laboratory confirmation from $1.5\%$ in 2016 to $8.6\%$ in 2020 (Table 3). This occurred as the proportion of children tested with Xpert increased; with diagnostic yield in children of $32.1\%$ ($\frac{62}{193}$) positive on Xpert compared with $26.5\%$ ($\frac{56}{211}$) for smear microscopy. Bacteriological confirmation in adults was $55.0\%$ overall, again with Xpert having a higher diagnostic yield than smear microscopy.
**Fig 3:** *Smear microscopy and GeneXpert testing showing number of cases with testing performed (left y axis), and proportion of tests done which tested positive (right y axis).* TABLE_PLACEHOLDER:Table 3 *Clinical diagnosis* was made in $64.2\%$ of total cases. Of 7,573 with a clinical diagnosis, $43.7\%$ were adults with pulmonary TB, $23.7\%$ were adults with EPTB and $32.6\%$ were children. Of adults with clinically diagnosed pulmonary TB, only $4.7\%$ had no microbiological testing undertaken; the rest ($95.3\%$ had a negative laboratory test (Xpert or smear). Among adults with presumptive pulmonary TB who seeks care at the hospital, $75.6\%$ had chest x-ray. Clinically diagnosed TB cases were treated as per standard treatment regimens for drug-susceptible TB. The Indonesian pediatric scoring system was used to support clinical diagnosis in children without bacteriological confirmation with $53.3\%$ recording a score and the majority ($95.4\%$) of the scores recorded being 6 or more.
Xpert identified 107 cases of rifampicin-resistant (likely multidrug-resistant) TB (S1 Table). Prior to the introduction of Xpert into District facilities, only samples from TB cases in which there was a high likelihood of multidrug-resistant TB on clinical suspicion were sent for Xpert testing at the distant reference laboratory situated in the provincial capital. Once Xpert testing became available locally and indications for Xpert testing broadened, the prevalence of rifampicin resistance among TB cases detected by Xpert was 5–$7\%$ between 2018 and 2021. The proportion of TB patients tested for HIV and diabetes improved during the study period and mostly remained strong during the pandemic years (Fig 4).
**Fig 4:** *Proportion of people with TB in whom HIV and diabetes were tested.*
## Treatment outcomes
Documentation of treatment outcomes improved after 2014 and treatment outcomes up until the 2020 cohort are listed in Table 2. Treatment success did not significantly change over time, ranging from $71.1\%$ to $74.6\%$ between 2015 and 2020 (χ2 test for trend $$p \leq 0.21$$). A significant increase in the proportion of deaths was seen over time from $3\%$ in 2015 to $6\%$ in 2020 ($p \leq 0.001$) and a fall in the proportion reported as lost to follow-up in recent years (Table 2). A rise in deaths of older individuals (>64 years) is noted in 2020 coinciding with COVID (S2 Table and S2 and S3 Figs). Treatment success rates were significantly higher in children ($78.1\%$) compared with adults ($69.2\%$), and in people treated in primary care ($77.5\%$) compared with hospitals ($67.1\%$) ($p \leq 0.001$) (Table 4 and S1 Fig). One in five TB cases who initiated treatment in hospital were lost to follow-up, compared to one in ten whose treatment was initiated at a PHC (Table 2). People with TB/HIV co-infection had significantly lower treatment success ($65.0\%$) than people without HIV ($74.4\%$) ($p \leq 0.001$) but ART status of those with TB/HIV was not routinely recorded. People with past TB had lower treatment success than those without ($$p \leq 0.001$$) (Table 4). Ethnicity was not associated with treatment outcome.
**Table 4**
| Characteristic, Number (N) with known data | Sub-group | Treatment success***N, % | Unfavorable outcome****N, % | Unadjusted Odds ratio (95% CI) | P value* | Adjusted odds ratio# | P value |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Age | 0–14 years | 1705 | 477 | 0.72 (0.65–0.79) | 0.0 | 0.63 (0.55–0.71) | 0.0 |
| N = 10,087 | 0–14 years | 78.1% | 21.9% | 0.72 (0.65–0.79) | 0.0 | 0.63 (0.55–0.71) | 0.0 |
| | > = 15 years | 5473 | 2432 | 0.72 (0.65–0.79) | 0.0 | 0.63 (0.55–0.71) | 0.0 |
| | > = 15 years | 69.2% | 30.8% | 0.72 (0.65–0.79) | 0.0 | 0.63 (0.55–0.71) | 0.0 |
| Ethnicity | Papuan | 5001 | 2007 | 0.96 (0.89–1.03) | 0.3 | 0.96 (0.87–1.05) | 0.384 |
| N = 9,870 | Papuan | 71.4% | 28.7% | 0.96 (0.89–1.03) | 0.3 | 0.96 (0.87–1.05) | 0.384 |
| | Non-Papuan | 2014 | 848 | 0.96 (0.89–1.03) | 0.3 | 0.96 (0.87–1.05) | 0.384 |
| | Non-Papuan | 70.4% | 29.6% | 0.96 (0.89–1.03) | 0.3 | 0.96 (0.87–1.05) | 0.384 |
| Residential location | Urban | 5952 | 2277 | 1.16 (1.04–1.30) | 0.007 | 1.09 (0.97–1.22) | 1.114 |
| N = 10,010 | Urban | 72.3 | 27.7% | 1.16 (1.04–1.30) | 0.007 | 1.09 (0.97–1.22) | 1.114 |
| | Rural | 1214 | 567 | 1.16 (1.04–1.30) | 0.007 | 1.09 (0.97–1.22) | 1.114 |
| | Rural | 68.2% | 31.8% | 1.16 (1.04–1.30) | 0.007 | 1.09 (0.97–1.22) | 1.114 |
| Health facility type | Hospital | 4114 | 2018 | 0.76 (0.70–0.82) | 0.0 | 0.54 (0.49–0.60) | 0.0 |
| N = 10,087 | Hospital | 67.1% | 32.9% | 0.76 (0.70–0.82) | 0.0 | 0.54 (0.49–0.60) | 0.0 |
| | Primary Health Care | 3064 | 891 | 0.76 (0.70–0.82) | 0.0 | 0.54 (0.49–0.60) | 0.0 |
| | Primary Health Care | 77.5% | 22.5% | 0.76 (0.70–0.82) | 0.0 | 0.54 (0.49–0.60) | 0.0 |
| TB type | Pulmonary | 5518 | 2187 | 1.15 (1.05–1.26) | 0.003 | 0.94 (0.84–1.05) | 0.29 |
| N = 10,087 | Pulmonary | 71.6% | 28.4% | 1.15 (1.05–1.26) | 0.003 | 0.94 (0.84–1.05) | 0.29 |
| | Extra Pulmonary | 1660 | 722 | 1.15 (1.05–1.26) | 0.003 | 0.94 (0.84–1.05) | 0.29 |
| | Extra Pulmonary | 69.7% | 30.3% | 1.15 (1.05–1.26) | 0.003 | 0.94 (0.84–1.05) | 0.29 |
| Mode of diagnosis | Clinical diagnosis | 4657 | 1809 | 0.90 (0.83–0.97) | 0.008 | 0.83 (0.75–0.93) | 0.002 |
| N = 10,087 | Clinical diagnosis | 72.0% | 28.0% | 0.90 (0.83–0.97) | 0.008 | 0.83 (0.75–0.93) | 0.002 |
| | Bacteriologically confirmed** | 2521 | 1100 | 0.90 (0.83–0.97) | 0.008 | 0.83 (0.75–0.93) | 0.002 |
| | Bacteriologically confirmed** | 69.6% | 30.4% | 0.90 (0.83–0.97) | 0.008 | 0.83 (0.75–0.93) | 0.002 |
| Record of prior TB | No past TB | 6619 | 2625 | 0.76 (0.66–0.86) | 0.001 | 0.75 (0.65–0.89) | 0.0 |
| N = 10,087 | No past TB | 71.6% | 28.4% | 0.76 (0.66–0.86) | 0.001 | 0.75 (0.65–0.89) | 0.0 |
| | Past TB | 559 | 284 | 0.76 (0.66–0.86) | 0.001 | 0.75 (0.65–0.89) | 0.0 |
| | Past TB | 66.3% | 33.7% | 0.76 (0.66–0.86) | 0.001 | 0.75 (0.65–0.89) | 0.0 |
| HIV status | HIV negative | 4666 | 1605 | 0.89 (0.86–0.91) | 0.0 | 0.91 (0.88–0.94) | 0.0 |
| N = 7,085 | HIV negative | 74.4% | 25.6% | 0.89 (0.86–0.91) | 0.0 | 0.91 (0.88–0.94) | 0.0 |
| | HIV positive | 529 | 285 | 0.89 (0.86–0.91) | 0.0 | 0.91 (0.88–0.94) | 0.0 |
| | HIV positive | 65.0% | 35.0% | 0.89 (0.86–0.91) | 0.0 | 0.91 (0.88–0.94) | 0.0 |
| Diabetes | Diabetes absent | 3192 | 1064 | 1.10 (1.07–1.15) | 0.0 | 0.90 (0.86–0.94) | 0.0 |
| N = 4,519 | Diabetes absent | 75.0% | 25.0% | 1.10 (1.07–1.15) | 0.0 | 0.90 (0.86–0.94) | 0.0 |
| | Diabetes present | 203 | 60 | 1.10 (1.07–1.15) | 0.0 | 0.90 (0.86–0.94) | 0.0 |
| | Diabetes present | 77.2% | 22.8% | 1.10 (1.07–1.15) | 0.0 | 0.90 (0.86–0.94) | 0.0 |
## Discussion
This study provides important insights into the epidemiology of TB in Mimika district, Papua province, revealing major improvements in TB case detection over an eight-year period. The $67\%$ increase in TB case notifications exceeded population growth of approximately $35\%$ over the same period [9]. Substantial downturns in TB detection occurred during the COVID-19 pandemic, but the TB program showed greater resilience than reported nationally: in 2020 compared with 2019, case notifications in Mimika district decreased by $18.7\%$ which compared to a $30.9\%$ decline in TB case notifications in Indonesia nationally [19]. Gains made in TB program quality indicators prior to the pandemic, such as proportion of patients accessing bacteriological TB diagnostics and HIV testing, as well as the proportion achieving treatment success, were largely maintained during the first two years of the pandemic. Health system strengthening activities underway in Mimika district may have contributed to these positive performance indicators. ‘ TB CEPAT’, an active case finding initiative that conducted door-to-door symptom screening in 2016–2017, is likely to have contributed to increased TB notifications whilst our CQI projects (‘STRATUM’ and ‘PRIME-TB’) have provided a package of health system strengthening activities, with a strong focus on regularly engaging, training and motivating TB program staff. The Mimika district TB program has been recognised nationally for its successes, being awarded the Best TB Program in Papua Province in 2018 and 2019 and celebrated at a national World TB Day celebration in 2019 and 2021 [20–22].
Damaging impacts of the COVID-19 pandemic on TB program performance have been reported nationally and internationally [1, 23]. Indonesian national TB case notifications dropped to 393,323 in 2020, a $31\%$ reduction compared to 2019 and less than half the estimated caseload. While case detection rates in Mimika dropped during the first year of the pandemic, the Mimika TB program successfully returned to close to its performance before the pandemic. The case detection rate dropped from $94\%$ in 2019 to $76\%$ in 2020, but recovered quickly and reached above $90\%$ in 2021. During the same time, the case detection rate at province level only increased by $2.3\%$, and by $6.1\%$ at national level [17], highlighting the stronger recovery in Mimika compared to other parts of Indonesia. The findings we report from Mimika suggest that investment in program strengthening activities, including health care provider education and CQI may mitigate external impacts on the TB program. Detailed evaluation of these interventions is underway. However, further improvements are needed with regards to screening for case detection. In Mimika district, reallocation of diagnostic equipment (GeneXpert machine) and human resources to COVID-19 management, including to support a National Sports Event week in October 2021, had demonstrable negative consequences for TB screening (Fig 2). This illustrates the importance of developing strategies to build surge capacity to protect routine service delivery while responding to health emergencies.
Our findings highlight that treatment outcomes and bacteriological confirmation are key areas for further improvement. An overall treatment success of $71.2\%$ does not meet the target of $85\%$ set by the Indonesian NTP or WHO [1, 24]. Furthermore, the treatment success rate did not improve from 2015 to 2020. The number of people with TB who are reported as lost to follow-up is high, especially among those receiving treatment through a hospital. Hospitals are often far from patient’s homes and have no outreach services, compared with primary care facilities which are better equipped to provide follow-up care in the community. In recognition of the need to decentralise care for treatment support, we worked with local healthcare providers to encourage treatment in primary care by implementing a referral process from hospitals to clinics. While the numbers of cases detected and treated at PHC level increased (Fig 1), the proportion of those lost to follow-up remained relatively unchanged until 2019 (Fig 5 and Table 2). A higher caseload provides additional strain on the health services to provide treatment support. Digital innovation to support active case finding, patient education, referrals, data recording and reporting is an urgent priority in settings where human resources are limited [25]. Specific challenges in Mimika district include a large and highly mobile mine worker population, and high language diversity among different ethnic groups. A large proportion of the population resident in Mimika are from other provinces in Indonesia and the high population mobility may also contribute to loss to follow-up and transfer out. More work is needed to address retention in care, such as through ensuring culturally appropriate approaches and outreach services. It is also noted that the proportion of deaths increased in 2018 and 2019. Specific reasons for this are not known, but notably, there were increases in MDR TB cases and in the proportion of TB cases being diagnosed in people aged 45–64 years. Deaths among people with TB increased further in 2020, mirroring WHO reporting on global TB-related deaths increasing by $7\%$ during the pandemic [1]. Finally, poorer TB treatment outcomes were noted for people living with HIV including higher death rates but ART status of those with TB/HIV was not recorded highlighting the need for improved integrated care for those with comorbidities.
**Fig 5:** *Treatment outcome among TB patients commenced on treatment, Mimika district, 2014–2020.*
Bacteriological confirmation of TB was low at $35.8\%$ compared with the national average in 2020 of $41\%$ [17]. The only sample type reported in our study is sputum. No EPTB cases had bacteriological confirmation and hence any drug-resistant EPTB cases would have received incorrect treatment. Xpert is now increasingly used in preference to smear microscopy as per national guidelines and the use of Xpert increased during the study period. However, Xpert cartridges remain under-utilised while the analyser is required for COVID-19 testing and there was a lack of presumptive case finding. Only around 5000 of the annual allocation of more than 7000 cartridges distributed by the NTP to Mimika are being used annually. Xpert positivity among presumptive TB cases is high. Of all people tested with Xpert in 2020–2021, one in five tested positive which may suggest that not enough testing is being done (S3 Table) [26]. Programmatic planning to increase Xpert utilisation for TB testing is of particular importance while the competing need for COVID-19 testing with Xpert persists. Rifampicin resistance was $7.3\%$ among those tested with Xpert in 2021, a concerning problem for Mimika given the high reliance on clinical diagnoses and empirical therapy with a standard drug-susceptible TB regimen. Xpert for EPTB diagnosis was unavailable in Mimika during this study due to lack of equipment (centrifuge and tubes), training or biosafety cabinet access for level 3 sample processing, and lack of finance from the NTP or health insurance to cover these costs.
Xpert clearly has high appeal given higher yield and lesser workload than microscopy and is now recommended in Indonesia for broad use. However, as utilisation of smear microscopy decreases and Xpert increases, there is a need to maintain microscopy expertise since this is still recommended for monitoring treatment response. By 2021, $27.7\%$ of presumptive TB cases were diagnosed on microscopy and $63.5\%$ using Xpert.
TB case finding in children is a marker of the quality of TB program performance. The proportion of TB cases in Mimika aged less than 15 years is consistently higher than national data. In 2020, $21.3\%$ of TB cases in Mimika district were children aged 0–14 years, compared to only $9\%$ nationally [2]. Infant BCG vaccination coverage in *Mimika is* low only $60.7\%$ in 2020 and $64.5\%$ in 2021. The high proportion of cases reported among children is similar to that reported from neighbouring Papua New Guinea ($25\%$ of cases aged less than 15 years), where TB is also highly endemic and where children make up a large proportion of the population [27]. Census data indicate that children comprise approximately $35\%$ of the population in Mimika district compared to $23\%$ in Indonesia overall. Only $5.6\%$ of children were confirmed bacteriologically in this cohort highlighting the need for better TB diagnostics for children, especially young children with paucibacillary disease. Most ($58\%$) of the bacteriologically confirmed TB in was diagnosed by Xpert and the proportion of bacteriologically confirmed child TB has increased since the introduction of Xpert, but remains low. Sputum induction or gastric aspirates for testing with Xpert can be done at the district hospital but these procedures require hospitalisation and are not routine for children with mild TB symptoms in the outpatient clinic. Furthermore, the WHO has recommended avoiding sputum induction to prevent airborne transmission during the COVID-19 pandemic. Recently updated WHO child TB guidelines [2022] recommend alternative samples for Xpert testing such as nasopharyngeal aspirates or stool [28]. The diagnosis of TB in children is usually clinical. The Indonesian pediatric TB scoring system is widely used and $52.1\%$ of child TB patients in Mimika had a score ≥6, which indicates the need for TB treatment [17]. However, the scoring system includes tuberculin skin test results (largely unavailable at our study site) and chest X-ray also (limited in availability, only at hospitals requiring a referral from primary care) which limit its diagnostic value.
The main limitation of this study is incomplete data capture especially in the first year; however, we undertook stringent cross-checking to minimise missing data and validate entries, applied consistently across all years of the study. Ethnicity is not captured in the TB electronic records; some errors in assigning ethnicity could have occurred. We did not have access to directly comparable national TB datasets and therefore comparisons between program performance in Mimika district with other parts of Indonesia were only able to be descriptive. Compared to the other 28 districts in Papua, Mimika district ranked first in TB case finding in 2020 and 2021 [17]. While the population comprised $7.3\%$ of the total population of Papua, it is estimated that about $10\%$ of TB cases will be notified from Mimika district. However, in 2021, Mimika district was able to contribute to $18.4\%$ of total TB case finding in Papua [17]. Strengths of this study include the highly comprehensive dataset from an under-reported, high burden setting, uniquely able to track program performance over a long period incorporating the unexpected disruption caused by COVID-19.
In summary, the implementation of several dedicated TB case finding and health system strengthening activities substantially improved TB case detection in Mimika District. Treatment success rate was sustained despite the increased burden that TB case numbers posed to the health services, but improvements in treatment outcome are needed. The analysis shows the negative impact of the COVID-19 pandemic on TB surveillance but there are already encouraging signs of recovery in case detection suggesting resilience in the TB services. There are clear areas for ongoing investment, including greater availability and uptake of rapid and accurate diagnostics, which international collaborative work in this study setting continues to address.
## References
1. 1WHO. Global Tuberculosis Report 2021. Geneva; 2021.
2. 2NTP Indonesia. Dashboard TB Indonesia. 16 Apr 2021 [cited 25 Jul 2021]. Available: https://tbindonesia.or.id/pustaka-tbc/dashboard-tb/
3. Pontororing GJ, Kenangalem E, Lolong DB, Waramori G, Sandjaja E. **The burden and treatment of HIV in tuberculosis patients in Papua Province, Indonesia: a prospective observational study**. *BMC Infectious Diseases* (2010.0) **10** 362. DOI: 10.1186/1471-2334-10-362
4. Ralph AP, Ardian M, Wiguna A, Maguire GP, Becker NG, Drogumuller G. **A simple, valid, numerical score for grading chest x-ray severity in adult smear-positive pulmonary tuberculosis**. *Thorax* (2010.0) **65** 863-869. DOI: 10.1136/thx.2010.136242
5. Lestari T, Graham S, van den Boogard C, Triasih R, Poespoprodjo JR, Ubra RR. **Bridging the knowledge-practice gap in tuberculosis contact management in a high-burden setting: a mixed-methods protocol for a multicenter health system strengthening study**. *Implementation Science* (2019.0) 14. DOI: 10.1186/s13012-019-0870-x
6. 6USAID. Pengendalian Tuberkulosis di Indonesia. In: News and information [Internet]. 6 Jun 2017 [cited 28 Mar 2022]. Available: https://www.usaid.gov/id/indonesia/fact-sheets/reducing-multidrug-resistant-tuberculosis-indonesia
7. Setiawaty V, Kosasih H, Mardian Y, Ajis E, Prasetyowati EB. **The Identification of First COVID-19 Cluster in Indonesia**. *The American Journal of Tropical Medicine and Hygiene* (2020.0) **103** 2339-2342. DOI: 10.4269/ajtmh.20-0554
8. 8Kompas. UPDATE: Kasus Positif Covid-19 Pertama di Mimika, 2 Pasien Dirawat di RSUD Mimika. Kompas. 30 Mar 2020. Available: https://regional.kompas.com/read/2020/03/30/06421481/update-kasus-positif-covid-19-pertama-di-mimika-2-pasien-dirawat-di-rsud?page=all. Accessed 21 Mar 2022.
9. 9BPS. Mimika Regency in Figures. Timika; 2021.
10. 10Papua Province Population and Civil Registry Social Service. Population Development Report. Recapitulation based on clan and gender. Jayapura; 2017.
11. 11MoH. National guideline for medical services: tuberculosis management. Jakarta: Ministry of Health Republic of Indonesia; 2020.
12. Migliori GB, Wu SJ, Matteelli A, Zenner D, Goletti D, Ahmedov S. **Clinical standards for the diagnosis, treatment and prevention of TB infection**. *The International Journal of Tuberculosis and Lung Disease* (2022.0) **26** 190-205. DOI: 10.5588/ijtld.21.0753
13. 13Mimika DHO. Situasi COVID-19 Mimika: Update 18 Maret 2022. Mimika; 2022 Mar.
14. 14Mimika DHO. Situasi COVID-19 Mimika. Timika; 2022 Jul.
15. 15Widyawati. Mesin TCM- TB untuk Covid-19 Sudah Bisa Digunakan. In: Biro Komunikasi dan Pelayanan Masyarakat [Internet]. 4 May 2020 [cited 31 Jul 2021]. Available: https://sehatnegeriku.kemkes.go.id/baca/umum/20200504/5333823/mesin-tcm-tb-covid-19-sudah-digunakan/
16. Mardian Y, Kosasih H, Karyana M, Neal A, Lau C-Y. **Review of Current COVID-19 Diagnostics and Opportunities for Further Development**. *Frontiers in Medicine* (2021.0) 8. DOI: 10.3389/fmed.2021.615099
17. 17MoH. Sistem Informasi TB. Jakarta: Ministry of Health; 2020.
18. 18MoH. Profil Kesehatan Indonesia Tahun 2019. Jakarta; 2020.
19. 19WHO. Global Tuberculosis Report 2020. Geneva; 2020.
20. 20HotNorth. PhD student leading tuberculosis prevention in Indonesia. In: News [Internet]. Nov 2018 [cited 30 Apr 2022]. Available: http://www.hotnorth.org.au/hot-north-researcher-leading-tuberculosis-prevention-in-papua-indonesia/
21. 21Menzies. TB project named best in province. In: Healthy Tomorrow [Internet]. Dec 2019 [cited 30 Apr 2022]. Available: https://www.menzies.edu.au/page/News_and_Events/E-Newsletter/Healthy_Tomorrow/December_2019/TB_project_named_best_in_province
22. 22Wulan DR. Dirjen P2P Membuka Pertemuan Nasional Monitoring dan Evaluasi Program Tuberkulosis Tahun 2022. 8 Aug 2022 [cited 18 Aug 2022]. Available: https://tbindonesia.or.id/berita/dirjen-p2p-membuka-pertemuan-nasional-monitoring-dan-evaluasi-program-tuberkulosis-tahun-2022/
23. Caren GJ, Iskandar D, Pitaloka DA, Abdulah R, Suwantika AA. **COVID-19 Pandemic Disruption on the Management of Tuberculosis Treatment in Indonesia**. *Journal of Multidisciplinary Healthcare* (2022.0) **15** 175-183. DOI: 10.2147/JMDH.S341130
24. 24MoH. Strategi Nasional Penanggulangan Tuberkulosis di Indonesia: 2020–2024. 2020.
25. Pai M, Kasaeva T, Swaminathan S. **Covid-19’s Devastating Effect on Tuberculosis Care—A Path to Recovery**. *New England Journal of Medicine* (2022.0) **386** 1490-1493. DOI: 10.1056/NEJMp2118145
26. 26WHO. Xpert MTB/RIF implementation manual: technical and operational ‘how-to’; practical considerations. Geneva: World Health Organization; 2014.
27. Aia P, Wangchuk L, Morishita F, Kisomb J, Yasi R, Kal M. **Epidemiology of tuberculosis in Papua New Guinea: analysis of case notification and treatment-outcome data, 2008–2016**. *Western Pacific Surveillance and Response Journal* (2018.0) **9** 9-19. DOI: 10.5365/wpsar.2018.9.1.006
28. 28WHO. WHO consolidated guidelines on tuberculosis. Module 5: Management of tuberculosis in children and adolescents. Geneva: World Health Organization; 2022.
|
---
title: 'Understanding the treatment burden of people with chronic conditions in Kenya:
A cross-sectional analysis using the Patient Experience with Treatment and Self-Management
(PETS) questionnaire'
authors:
- Hillary Koros
- Ellen Nolte
- Jemima Kamano
- Richard Mugo
- Adrianna Murphy
- Violet Naanyu
- Ruth Willis
- Triantafyllos Pliakas
- David T. Eton
- Edwine Barasa
- Pablo Perel
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021888
doi: 10.1371/journal.pgph.0001407
license: CC BY 4.0
---
# Understanding the treatment burden of people with chronic conditions in Kenya: A cross-sectional analysis using the Patient Experience with Treatment and Self-Management (PETS) questionnaire
## Abstract
In Kenya, non-communicable diseases (NCDs) are an increasingly important cause of morbidity and mortality, requiring both better access to health care services and self-care support. Evidence suggests that treatment burdens can negatively affect adherence to treatment and quality of life. In this study, we explored the treatment and self-management burden among people with NCDs in in two counties in Western Kenya. We conducted a cross-sectional survey of people newly diagnosed with diabetes and/or hypertension, using the Patient Experience with Treatment and Self-Management (PETS) instrument. A total of 301 people with diabetes and/or hypertension completed the survey ($63\%$ female, mean age = 57 years). They reported the highest treatment burdens in the domains of medical and health care expenses, monitoring health, exhaustion related to self-management, diet and exercise/physical therapy. Treatment burden scores differed by county, age, gender, education, income and number of chronic conditions. Younger respondents (<60 years) reported higher burden for medication side effects ($p \leq 0.05$), diet ($p \leq 0.05$), and medical appointments ($$p \leq 0.075$$). Those with no formal education or low income also reported higher burden for diet and for medical expenses. People with health insurance cover reported lower (albeit still comparatively high) burden for medical expenses compared to those without it. Our findings provide important insights for Kenya and similar settings where governments are working to achieve universal health coverage by highlighting the importance of financial protection not only to prevent the economic burden of seeking health care for chronic conditions but also to reduce the associated treatment burden.
## Introduction
The prevalence of chronic conditions is rising globally, and many countries in sub-Saharan Africa are now facing the double burden of infectious and non-communicable diseases (NCDs) [1]. In Kenya, NCDs account for almost a third of all deaths and this proportion is projected to rise by over $50\%$ during the next decade [2]. Cardiovascular diseases and cancers are among the most common causes of morbidity, after infectious diseases, and in 2015, about a quarter of the population had hypertension and $5\%$ had diabetes or impaired fasting glycaemia [2]. Availability of screening, early detection and management of NCDs in primary care settings in *Kenya is* limited. Although strengthening primary health care is a priority [3], the health system has remained hospital-centric, with long waiting times and reduced quality of care [4], resulting in poor retention in care of people who screen positively for NCDs and low treatment adherence [5,6].
Availability of and access to services are important for the successful management of NCDs. However the social and economic contexts within which people with NCDs live also greatly impact their capacity to seek care and self-manage their conditions [7]. Evidence from Malawi and South Africa has pointed to the struggles people with diabetes and/or hypertension face on a daily basis [8–10]. For example, Matima et al. explored the experiences of people with HIV and diabetes comorbidity using the cumulative complexity model developed by Shippee and colleagues, which emphasises how clinical and social factors accumulate and interact to complicate patient care [9,11]. They identified two sets of workloads people have to deal with: ‘clinic-related’ workload around lack of service integration and perceived power imbalances between the patient and the health care provider, and ‘self-care’ related workloads around nutritional needs, medication burden and stigma. Available evidence suggests that a high (perceived) burden resulting from patient workloads may negatively affect adherence to treatment and quality of life, in particular among people with multiple and chronic conditions [12–15]. These challenges are further exacerbated by financial concerns, with households in Kenya facing a significant economic burden associated with NCD diagnosis and treatment costs [16].
Much of the work on patient work and treatment burden has been conducted in high-income countries [17,18] and similar work is only beginning to be undertaken in low- and middle income countries [19]. This study seeks to contribute to this emergent evidence by exploring the treatment and self-management burden among people with chronic conditions, in particular hypertension or diabetes, in Kenya.
## Methods
This study was set in the context of a wider implementation study that sought to understand the impact and scalability of a novel approach to integrate promotive, preventive, and curative care for diabetes, hypertension, cervical and breast cancer at the primary health care level (Primary Health Integrated Care Project for Chronic Conditions, PIC4C) within the Academic Model Providing Access to Healthcare (AMPATH) programme in Western Kenya [20]. PIC4C was launched in 2018 by the Kenyan Ministry of Health in partnership with AMPATH/Moi, Access Accelerated and the World Bank. The model was piloted in Busia and Trans Nzoia counties in Western Kenya, which formed the location for the present study.
One key component of our study was to understand the treatment and self-management burden among a group of people diagnosed with hypertension and/or diabetes since PIC4C implementation and how the burden is distributed across a range of socio-demographic characteristics. We used the Patient Experience with Treatment and Self-Management (PETS) instrument [21] as it assess treatment burden in patients with chronic health conditions requiring self-management. This comprehensive measure allows understanding what aspects of treatment and self-management prove to be most burdensome to people with chronic conditions in Kenya. The survey served as a basis from which to recruit a subsample of people with hypertension and/or diabetes to further explore treatment burden using in-depth interviews.
The PETS builds on a conceptual measurement framework developed from interviews and discussions with people with multiple chronic conditions in the USA [22] and has since been used in a range of populations with chronic conditions in the USA and Norway, including multiple conditions [14,21], cancer [23,24], heart failure [25], diabetes [26], and hypertension [27]. To our knowledge, this patient-reported measure of treatment burden has not yet been applied in a lower middle-income country.
## Adaptation of the PETS instrument
We used the PETS Questionnaire 60-item version (Vs. 2.0) [28]. It contains 12 multi-item and two single item scales: [1] medical information, [2] medications, [3] medical appointments, [4] monitoring health, [5] diet, [6] exercise or physical therapy, [7] medical equipment, [8] relationships with others, [9] medical and health care expenses, [10] difficulty with health care services, [11] role and social activity limitations, and [12] physical and mental exhaustion. The single-item scales assess respectively bother due to medication side effects and bother due to having to rely on medication. Items are assessed by 4 or 5-point ordinal rating scales (e.g. strongly agree, agree, disagree, strongly disagree or very easy, easy, neither easy nor difficult, difficult, very difficult).
We used a rigorous process to translate the PETS into Swahili, following the FACIT (Functional Assessment of Chronic Illness Therapy) Translation & Linguistic Validation Methodology developed by Eremenco et al. [ 29] and working with the FACITtrans team to implement the method [30]. This involved seven steps: [1] two (independent) forward translations of the questionnaire from English into Swahili by two native speakers; [2] reconciliation of the two forward translations by a third native speaker; [3] back-translation (blinded) of the reconciled version into English by a native speaker fluent in English; [4] review of back-translation by the FACITtrans team; [5] review and finalisation by a fourth independent native speaker of Swahili; [6] quality review of translations; and [7] formatting and proofreading of the test version by a native speaker (see also S1 Fig). Each step is documented in a separate Word document (available on request). All native speakers involved in the translation of the PETS into Swahili were members of the study team and the wider AMPATH research programme.
In a final step, the translated PETS was tested with six patients with diabetes or hypertension to assess comprehensibility and general relevance of the PETS questions in the Kenyan context, using cognitive interviews. Cognitive interview participants were randomly recruited from outpatients attending the Chronic Disease Management Clinic at Moi Teaching and Referral Hospital (MTRH) during one of the clinic days; they were approached at the point of exiting the clinical consultation or the outpatient department. Interviews took place at MTRH premises, following patient consent, and were conducted in Swahili language (S1 Fig). Interviews found that the phrasing of PETS items was generally understood, except for the domain medical equipment (‘Do you currently use any medical equipment or devices’). Interview participants queried whether the use of medical equipment or devices referred to the clinical setting or to their own homes. To enhance comprehensibility, we amended the question to clarify that this referred to use outside clinics or facilities by adding ‘in your own home’. The English language PETS and its final translated version can be made available upon written request to Dr Eton, the principal developer of the measure, at dteton99@gmail.com.
In addition to the PETS, the survey included questions about socio-demographic characteristics (age, ethnic group, marital status, education, work history, household size and income) and medical history, namely whether the respondent had ever been told to have a chronic condition as chosen from a list (e.g., high cholesterol, diabetes, hypertension, myocardial infarction, peripheral artery disease, heart failure) and any medications they are currently taking in relation to the reported condition, location of treatment when unwell, and health insurance (National Hospital Insurance Fund, NHIF) cover.
## Participant recruitment
The overall study was set in Busia and Trans Nzoia counties in Western Kenya, where the PIC4C model of care was implemented across a total of 73 facilities (public dispensaries, health centres and county referral hospitals) [20]. We sought to capture a wide range of people with hypertension and/or diabetes in terms of age, gender and broad socio-demographic status and who were registered with PIC4C facilities. Our sampling strategy was based on random sampling of PIC4C facilities in the two counties. In addition, random sampling was done at each strata, with stratification based on the location (rural/urban) or level (size) and then at the facility-level based on chronic condition, gender and age. We used the ratio for diabetic, hypertensive and people with both diabetes and hypertension of 1:6:2 reflecting recorded prevalence derived from the PIC4C database. We sought to arrive at a total sample of around 300 participants. This sample size was used by the developers of the PETS survey instrument to test its validity [21]. We judged this sample size to be appropriate to assess the feasibility of using a formal instrument (i.e. PETS) to assess the treatment and self-management burden in a group of people identified to have diabetes and/or hypertension, and allow for sub-group analysis by a range of socio-demographic characteristics such as gender, ethnicity, educational attainment and income. Anticipating non-response of 30–$45\%$, we oversampled to around 200 in each county (Busia: 202; Tranz Nzoia: 213) so as to randomly select 150 to be surveyed in each. The study population included all patients who had been screened and had hypertension and/or diabetes confirmed through PIC4C service efforts between 2018 and 2020. We used the permanent and the daily registers held at individual health facilities to identify eligible patients. If the randomly selected patient was not available or did not consent, we moved to the next patient on the list of randomly selected participants. This process was repeated until the required sample in each cluster was achieved.
## Data collection
The survey was interviewer-administered, using the principles of computer-assisted personal interviewing (CAPI) and REDCap [31], a secure web application for building and managing online surveys and databases. Trained research assistants acted as focal persons to reach out to selected survey participants by phone; those who could not be reached by phone call were invited in person. Upon agreement to take part, a date for survey completion was set, seeking to identify dates most convenient for participants. For practical reasons, the survey was conducted in-person at a health facility closest to the invited participants’ homes. At the time of the survey, participants were provided with an information sheet about the study (read out to participants unable to read). Patients were reminded that participation was entirely voluntary and non-participation had no influence on clinical care. Participant consent was sought by means of their signature on two paper copies of the consent form. Participants were reimbursed to cover the cost of travel. Survey data were collect by six trained research assistants between 1 December 2020 and 12 February 2021; the average duration was 60 minutes (allowing for breaks where requested) and the completion rate of PETS items was high (>$95\%$). The only exception was the domain ‘medical equipment’, which was only completed by $5\%$ ($$n = 14$$) survey respondents. We therefore did not consider this domain in the further analysis.
## Analysis
Survey data were analysed using descriptive statistics. Where appropriate, we used independent samples t-tests to compare means, chi square of independence for binary/categorical variables, and one-way ANOVA for variables with more than two groups.
Following Eton et al., we scored all PETS scales in such a way that a higher score indicates greater treatment burden. This means that positively worded items were reverse-coded before scoring. We then generated raw scale scores by summing the unweighted items within each domain. Aggregated subscale scores were prorated for missing data when at least $50\%$ of items were available. This was to account for items that may not be applicable to all respondents [21]. As response scales vary across PETS domains (i.e., there is no single response scale used for all domains) and the number of items is not the same across all of the PETS domain scales, we converted all raw scale scores to a standard 0 to 100 metric to facilitate interpretability, with 0 indicating ‘no burden’ and 100 indicating ‘highest burden.’ Higher scores on any PETS domain scale always indicate higher burden. We computed internal consistency reliability for all PETS scales using Cronbach’s alpha, with an alpha of ≥0.70 taken to indicate adequate reliability [32]. All analysis were conducted using SPSS Version 28.0.
## Ethical approval
The study received approvals from Moi University Institutional Research and Ethics Committee (FAN:0003586) and the London School of Hygiene & Tropical Medicine [17940] as well as a research permit from the National Commission for Science, Technology and Innovation (NACOSTI/P/$\frac{20}{4880}$).
## Results
Table 1 presents socio-demographic and health-related characteristics of our patient sample by county. Almost two-thirds of respondents ($63\%$) were female; the mean age was 57 years (range 20–90) and $73\%$ were married. The two county samples differed significantly on several variables, including ethnic group, educational attainment, and current working. For example, while a large proportion identified as Luhya in both counties ($57\%$ in Busia and $48\%$ in Trans Nzoia), the second largest group in the Busia sample identified as Teso ($33\%$) whereas in Trans Nzoia sample, about $22\%$ identified as Kalenjin and $19\%$ as Kikuyu. The Busia sample also had a higher proportion of respondents with no or only primary education ($78\%$ vs. $60\%$), and a smaller proportion that reported to be currently working ($78\%$ vs. $90\%$). Among those currently working, over half of the Trans Nzoia sample reported to be self-employed in agriculture ($55\%$) compared to $39\%$ in Busia. Average household income was low in both counties, with about half of respondents reporting a monthly income of less than KShs 3,000 (US$27; £20; €23 in January 2021).
**Table 1**
| Sample characteristic | Value or frequency | Value or frequency.1 |
| --- | --- | --- |
| Sample characteristic | Busia (n = 150) | Trans Nzoia (n = 151) |
| Other | 9% (13) | 9% (14) |
| Marital status, % (N) | | |
| Married | 72% (108) | 75% (113) |
| Single | 23% (34) | 17% (26) |
| Separated/widowed | 5% (8) | 8% (12) |
| Educational attainment, % (N)** | Educational attainment, % (N)** | |
| No formal education | 19% (28) | 11% (17) |
| Primary education | 59% (89) | 49% (74) |
| Secondary/tertiary education | 22% (33) | 40% (60) |
| Currently working, % (N)** | Currently working, % (N)** | |
| Yes | 78% (117) | 90% (136) |
| No | 22% (33) | 10% (15) |
| Currently working: Type of work, % (N)** | Currently working: Type of work, % (N)** | Currently working: Type of work, % (N)** |
| Agriculture | 39% (58) | 55% (83) |
| Self-employed | 25% (38) | 20% (30) |
| Unemployed | 7% (11) | 3% (5) |
| Other | 7% (10) | 12% (18) |
| Household size, number of HH members, % (N) | Household size, number of HH members, % (N) | Household size, number of HH members, % (N) |
| 1 | 5% (8) | 2% (3) |
| 2 | 8% (12) | 10% (15) |
| 3 | 10% (15) | 11% (17) |
| 4–6 | 43% (54) | 42% (64) |
| 7–9 | 28% (42) | 24% (36) |
| 10 or more | 6% (9) | 11% (16) |
| Monthly household income past year, KSh, % (N) | Monthly household income past year, KSh, % (N) | Monthly household income past year, KSh, % (N) |
| <1,000 | 25% (37) | 23% (34) |
| 1,000–2,999 | 29% (44) | 27% (41) |
| 3,000–4,999 | 15% (23) | 17% (25) |
| 5,000–7,999 | 11% (17) | 12% (18) |
| 8,000–10,000 | 9% (14) | 7% (10) |
| 10,000–15,000 | 9% (13) | 11% (17) |
| Self-reported chronic condition, % (N) | Self-reported chronic condition, % (N) | Self-reported chronic condition, % (N) |
| Hypertension | 85% (127) | 82% (124) |
| Diabetes | 44% (66) | 37% (56) |
| High cholesterol | 5% (7) | 3% (4) |
| HIV | 5% (7) | 2% (3) |
| Stroke | 6% (9) | - |
| Heart disease | 4% (6) | 2% (3) |
| Arthritis | 4% (6) | 1% (1) |
| Ulcer | 2% (3) | 2% (3) |
| Other | 7% (10) | 2% (3) |
| No. chronic conditions, % (N)*** | No. chronic conditions, % (N)*** | No. chronic conditions, % (N)*** |
| 1 | 52% (78) | 71% (107) |
| 2 | 37% (56) | 26% (40) |
| 3 or more | 11% (16) | 3% (4) |
| Mean number | 1.6 | 1.3 |
| Reporting hypertension and diabetes, % (N) | Reporting hypertension and diabetes, % (N) | Reporting hypertension and diabetes, % (N) |
| Yes | 27% (40) | 21% (32) |
| No | 73% (110) | 79% (119) |
| Taking any medication (excl. herbal), % (n) | Taking any medication (excl. herbal), % (n) | Taking any medication (excl. herbal), % (n) |
| Yes | 99% (149) | 100% (151) |
| No | 1% (1) | - |
| No. of medications taken (excl. herbal), % (N) | No. of medications taken (excl. herbal), % (N) | No. of medications taken (excl. herbal), % (N) |
| 1 | 71% (107) | 80% (120) |
| 2 | 27% (41) | 20% (30) |
| 3 | 1% (1) | 1% (1) |
| Taking herbal drugs, % (N) | Taking herbal drugs, % (N) | Taking herbal drugs, % (N) |
| Yes | 7% (10) | 13% (20) |
| No | 93% (139) | 87% (131) |
| Location of treatment when unwell, % (N)** | Location of treatment when unwell, % (N)** | Location of treatment when unwell, % (N)** |
| County hospital | 10% (15) | 11% (17) |
| Sub-county hosp | 25% (38) | 38% (57) |
| Health centre | 33% (49) | 29% (44) |
| Dispensary | 27% (40) | 18% (27) |
| Private provider | 5% (8) | 4% (6) |
| Has NHIF cover, % (N) | Has NHIF cover, % (N) | Has NHIF cover, % (N) |
| Yes | 24% (36) | 25% (38) |
| No | 76% (114) | 74% (112) |
| Don’t know | - | 1% (1) |
The majority of respondents reported having one chronic condition, with the Busia sample having a significantly higher proportion of respondents with two or more conditions ($48\%$ vs. $30\%$). Over $80\%$ in each county reported having hypertension and between $37\%$ (Trans Nzoia) and $44\%$ (Busia) had diabetes; the mean number of reported chronic conditions was 1.6 for the Busia sample and 1.3 for the Trans Nzoia sample. Only about a quarter reported having had NHIF cover at the time of the survey in either county.
Table 2 shows reliability of PETS domain scales for the sample overall and by county (S1–S5 Tables show the frequency of responses to individual PETS domain items). Internal consistency reliability was generally good for all multi-item scales, and Cronbach’s coefficients were well above the threshold for adequate reliability (α ≥ 0.70). The only exceptions were ‘monitoring health’ (α = 0.55), ‘medical and health care expenses’ (α = 0.69) and ‘diet’ (α = 0.67), although the latter two were close to the adequate reliability threshold. Monitoring health asks about the ease or difficulty of monitoring health behaviours (e.g. tracking exercise, foods eaten, or medicines taken) and health condition (e.g. weighing, checking blood pressure or blood sugar levels). Our cognitive interviews found that while respondents generally found the question itself easy to understand, some required further explanations of what was meant by ‘monitoring’. Furthermore, the ability to track, say, blood pressure or blood sugar levels at home requires respondents to have the relevant equipment at their disposal but, as noted earlier, only $5\%$ of survey respondents reported using medical equipment that would enable self-monitoring at home.
**Table 2**
| PETS scale | Total PIC4C sample | Busia | Trans Nzoia |
| --- | --- | --- | --- |
| Medical information (7 items) | 0.83 | 0.92 | 0.87 |
| Medications (7 items) | 0.86 | 0.85 | 0.85 |
| Medical appointments (6 items) | 0.85 | 0.84 | 0.83 |
| Monitoring health (2 items) | 0.55 | 0.64 | 0.39 |
| Interpersonal challenges (4 items) | 0.84 | 0.81 | 0.88 |
| Medical and health care expenses (5 items) | 0.69 | 0.76 | 0.55 |
| Difficulties with health care services (7 items) | 0.82 | 0.56 | 0.95 |
| Role/social activity limitations (6 items) | 0.92 | 0.93 | 0.91 |
| Physical/mental fatigue (5 items) | 0.87 | 0.87 | 0.88 |
| Diet (3 items) | 0.67 | 0.72 | 0.46 |
| Exercise and physical therapy (4 items) | 0.82 | 0.83 | 0.81 |
Lower internal consistency reliability for the domains medical and health care expenses and diet was largely driven by the Trans Nzoia sample (Table 2). The former domain includes one question about the ease or difficulty to understand what is and what is not covered by health insurance. Leaving this item out would increase Cronbach’s α for the domain to 0.918 in the Trans Nzoia sample (0.889 in the Busia sample). The item had a relatively large number of ‘not applicable’ responses ($20\%$ in Busia, $25\%$ in Trans Nzoia). As only $25\%$ of sample reported having NHIF cover it is conceivable that the item was not relevant to most respondents, although we have retained it in our subsequent analysis as removing it did not change findings in any discernible way. The domain diet includes the item ‘*It is* hard to find healthy foods’, which returned a negative Cronbach’s α when deleted for the Trans Nzoia sample, possibly reflecting the limited spread of answers in this sample across three categories (‘strongly agree’, ‘agree’, ‘disagree’) only. Finally, the domain difficulty with health care services showed good internal consistency reliability for the total sample but not for the Busia sample (α = 0.56). This might be explained by the relatively large proportion of ‘not applicable’ responses for three of the items in this domain (S1 Table).
Table 3 presents descriptive statistics of 11 PETS domains and two single item scores for the total sample and by county. The highest treatment burdens were reported in the domains of medical and health care expenses, monitoring health, and physical and mental exhaustion related to self-management as well as diet and exercise/physical therapy. Several items within other domains were also rated as especially burdensome, such as finding transport to get to (S1 Table) and long waits at medical appointments (S2 Table), feeling dependent on others for health care needs (S3 Table) or self-management interfering with work, family responsibilities or daily activities (S4 Table). Fig 1 disaggregates domains with the highest mean burden scores by response item and county.
**Fig 1:** *Response frequency for selected PETS domain items, by county.* TABLE_PLACEHOLDER:Table 3 Compared with the Busia sample, Trans Nzoia mean burden scores were significantly higher for medical information, medications, medical appointments and difficulty with health services. Conversely, Busia respondents reported a significantly higher burden for interpersonal challenges, medical expenses, role/social activity limitations and diet (Table 3). Scale scores were positively skewed toward a lower burden in most domains except for medical expenses, which was slightly negatively skewed toward a higher burden. Diet was also negatively skewed in the Busia sample. Floor effects were generally lower in the Trans Nzoia sample.
There were significant differences in mean burden scores by ethnic group in the domains medications, medical appointments and difficulties with health care services, with those identifying as Teso reporting a substantially lower burden. For example, in the domain difficulties with health care services, Teso respondents had a mean score of 21.3 (SD 20.0) compared with a score of around 40 for the other three main ethnic groups (S6 Table). At the same time, Teso respondents reported a significantly higher burden in the relationship with others and medical expenses domains (S6 Table).
There were generally fewer differences in mean burden scores by age, gender, education, income or number of chronic conditions. Younger respondents (<60 years) reported a significantly higher burden for bother with medication side effects and for diet ($p \leq .05$); they also reported a higher burden for medical appointments, which was borderline significant ($$p \leq 0.075$$). Respondents with no formal education or those on low income also reported a significantly higher burden for diet as well as for medical expenses (S6 Table). Perhaps not surprisingly, those reporting having health insurance cover reported a significantly lower (albeit still comparatively high) burden for medical expenses compared to those without health insurance cover (53.1 (SD 25.7) vs. 66.3 (SD 21.1).
Fig 2 disaggregates PETS scores for the total sample by number of chronic conditions. Perhaps somewhat counterintuitively, we found higher scores for people reporting one chronic condition compared to those with two or more conditions in several domains although differences were not statistically significant. The only domains where burden scores were significantly higher with more chronic conditions were role/social activity limitations and diet. There was a clearer relationship between the number of drugs taken and reported treatment burden in most domains although differences were significant for bother with medicine reliance, bother with medication side effects, role/social activity limitations and diet only (S6 Table).
**Fig 2:** *PETS scores by number of chronic conditions, total sample.Note. * role/social activity limitations: p = 0.005; diet: p = 0.030. See also S7 Table.*
Exploring treatment and self-management burden for type of chronic condition, we found that people with diabetes (with or without other conditions) reported a higher burden in all domains compared to people with any other NCD (including hypertension), and a significantly higher burden for medical appointments, role/social activity limitations, bother with medication side effects, and diet (S6 Table).
## Discussion
To our knowledge, this is one of the first studies that have used a patient-reported measure of treatment and self-management burden among people with chronic conditions in a lower middle-income country. We found that patients in Western Kenya who were recently diagnosed with hypertension or diabetes reported a considerable treatment burden in a range of areas, with particularly high burdens around difficulty paying for health care, monitoring health, and physical or mental exhaustion from self-management, alongside affording or following a healthy diet and engaging with exercise or physical therapy. Other areas perceived as especially difficult or bothersome included finding transport to get to and long waits at medical appointments, feeling dependent on others for health care needs and the impact that self-management had on work, family responsibilities or daily activities.
The reported treatment and self-management burden differed between patient populations in the two counties in most domains, and this appeared to be driven by a combination of factors, including ethnic group, educational level and burden of multiple chronic diseases. However, the group sizes were too small to allow for further robust analysis of underlying patterns.
Empirical application of the PETS has so far only been documented for populations in the USA and Norway, and the reported treatment and self-management burden is consistent with our findings in so far as the highest burden among people with diabetes or multimorbidity was in the domains of medical expenses, monitoring health and physical or mental exhaustion from self-management [21,33]. PETS scores in US patient populations tended to be lower than in the Kenyan samples queried in this study, although in Eton et al. ’s 2017 study, a subgroup of participants recruited from an urban safety-net hospital (Hennepin County Medical Center in Minneapolis, Minnesota), which provides care for low-income, uninsured, and vulnerable persons, had mean PETS scores that were much closer to those reported in our sample [21]. An additional study by Eton et al. [ 34] developed and administered a briefer short-form version of the PETS with the same patient population and found even higher mean scores in several burden domains. However, caution must be exercised when making direct comparisons given the different version of the PETS used in this prior study. Contrary to other measures of treatment burden, such as the Treatment Burden Questionnaire (TBQ) [35] or the Multimorbidity Treatment Burden Questionnaire (MTBQ) [13], psychometric testing of the PETS has not supported a global summary burden score, although more recent work has distinguished ‘workload’ and ‘impact’ summary scores which aggregate some of the PETS domain scales [14]. Determination of severity thresholds for PETS domain scores (i.e., low, medium, and high burden) are currently pending and may make comparisons with scores from other measures more feasible in the future.
Overall, our findings for treatment burden in a population of people with diabetes and/or hypertension in Western *Kenya* generally align well with other studies that used comparable measures. Work that has assessed treatment burden in different populations found that younger people [13,34,36], those with greater financial difficulties [13,21,35], and those with multiple conditions tended to report higher treatment burdens [14,21], although the international evidence is somewhat mixed on the latter [13,14,35]. We also showed treatment burden to be significantly associated with younger age although this was in selected domains around medical appointments, bother with reliance on medicines and diet only. Duncan et al. argued that a higher burden might reflect role differences, with younger people having to organise medical appointments around work commitments and looking after dependents while perhaps also having different expectations in terms of managing their own health [13]. Similar to studies in the USA, lower income was associated with a higher treatment burden in our sample, in particular around medical expenses [14,21]. Studies of populations in countries with universal health systems did not find such associations [13,35,36]. Moreover, we found that respondents who had health insurance cover reported a significantly lower (albeit still high) treatment burden as it related to medical expenses. Taken together, these observations highlight the importance of financial safety netting not only to protect people from financial risk related to managing chronic conditions but also to lower the associated treatment burden.
We did not find treatment burden to be higher among people with multiple chronic conditions except in the domains bother with medicine reliance, role/social activity limitations and diet. Possible explanations for an apparent lack of association between treatment burden and number of chronic conditions include that those with more than one condition might find it easier to call upon and navigate medical and social support because they are more experienced and ‘already in the system’ [37], and, possibly, because of the integrated provision of services within PIC4C, which would otherwise have required repeat visits to different clinics. We were unable, in this study, to assess disease severity; we did, however, find a dose-response relationship between the number of drugs taken and treatment burden, which could be indicative of greater perceived or experienced severity. Perhaps not surprisingly, we found that people with diabetes reported a higher treatment burden compared to those with hypertension. While it is difficult to compare directly, studies of people with (multiple) chronic conditions in Switzerland (using the TBQ) [38] and Victoria, Australia (MTBQ) [39] also found a positive association of treatment burden with diabetes. Herzig et al. [ 38] suggested that a perceived high treatment burden for diabetes might reflect the wider range of activities that patients have to engage with to effectively manage the condition, from regular drug intake to adapting diet and exercise, all impacting on perceived quality of life. Evidence from low resource settings specifically points to the key challenges of affording and accessing a healthy diet among people with diabetes [40,41].
## Strengths and limitations
A key strength of our study was the high completion rate of all items of the PETS instrument (>$95\%$). All items appeared to be relevant for all patients and the proportion of ‘does not apply’ responses was low for most. One exception were selected items in the ‘difficulty with health care services’ domain, where about $35\%$ of the total sample responded not applicable to the first two response items (‘different providers not communicating’; ‘seeing too many different specialists’). Reliability scores and the overall coherence of our findings in relation to what is known in the Kenyan context and internationally supports the transferability and applicability of the concept of treatment burden, and the use of the PETS in describing it, to Kenya.
Our sample was not representative of the patient population in Busia and Trans Nzoia counties and we cannot, therefore, generalise across the wider patient population in either county. However, this was not the aim of this survey. Indeed, the sample was meant to capture people using PIC4C services with targeted sampling of those with hypertension and/or diabetes specifically. Some two-thirds of our sample were women, which broadly reflects the general pattern of service use across PIC4C facilities, with women more likely to attend care for hypertension or diabetes as recorded in the PIC4C database. The Busia and Trans Nzoia samples differed significantly in terms of ethnic group composition; identification as belonging to a given ethnic group was self-reported although the proportions in our sample appear to reflect the ethnic composition of Busia and Trans Nzoia counties in broad terms. Indeed, differences in ethnic composition was a main reason for selecting the two counties as pilot regions for the wider PIC4C study.
About half of our respondents reported a monthly income of less than KShs 3,000, which is lower than the national poverty line of KShs 3,252 for rural populations as defined by the Kenyan National Bureau of Statistics [42], although similar to poverty levels reported for Busia and Trans Nzoia counties, at $61\%$ and $50\%$, respectively [2018] [20]. However, it is important to note that household income data were self-reported and not comprehensively captured by the survey; more than half of respondents were farmers ($56\%$), whose household income is difficult to estimate with considerable monthly fluctuation. Household expenditure is considered a preferable measure in settings characterised by mostly informal economic activities and income cannot easily be tracked or quantified [43]; however, this was not possible in the context of this study.
## Implications for practice and research
A key observation of our study is that people with diabetes and/or hypertension in Western Kenya reported a high treatment burden in a range of domains. While the substantial economic burden of chronic illness faced by individuals and households in *Kenya is* well known [44], the burden resulting from monitoring health and the physical/mental burden from self-management have as yet not been documented. The further development of integrated chronic care programmes such as PIC4C and similar programmes elsewhere should make provisions for supporting people to alleviate the added burdens in order to optimise NCD management and, ultimately, outcomes.
Future programmes should also consider targeting specific groups with higher burdens specifically. These include for example younger patients with dependents who have to balance work and caring commitments alongside managing their health condition/s. Effective management will require long-term engagement over the life time, which younger people may find especially challenging and they might benefit from targeted practical support.
Similarly, the main areas reported to be especially bothersome were finding transport to get to and long waits at medical appointments. These are areas where targeted approaches can potentially make a substantial difference to people, through for example, outreach services such as group medical visits as previously trialled in Western Kenya [45,46]. Recent efforts in the study region saw the piloting of tele-medicine services using community health workers and peer support as ‘clinician-extenders’ during the COVID-19 pandemic to maintain and improve access to NCD care [47]. Such approaches provide a useful starting point for the further development of NCD programmes in the region.
Finally, our observations provide important insights for Kenya as a whole as the government moves to roll out universal health coverage (UHC) [48]. In doing so, there is particular need for providing comprehensive coverage for NCDs that also involves enhanced support for monitoring and self-management to ensure reduced treatment burden.
The PETS has proved to be a useful tool for assessing the treatment and self-management burdens of people with NCDs in Western Kenya at one point in time. Further work should test the instrument in a wider range of populations in different settings and over time to understand its value as a measure of impact of interventions seeking to support people with chronic conditions.
## References
1. Gouda H, Charlson F, Sorsdahl K, Ahmadzada S, Ferrari A, Erskine H. **Burden of non-communicable diseases in sub-Saharan Africa, 1990–2017: results from the Global Burden of Disease Study 2017**. *Lancet Glob Health* (2019.0) **7** e1375-e1387. DOI: 10.1016/S2214-109X(19)30374-2
2. 2Ministry of Health. Kenya STEPwise survey for non communicable diseases risk factors 2015 report. Nairobi: Ministry of Health; 2015.. *Kenya STEPwise survey for non communicable diseases risk factors 2015 report* (2015.0)
3. 3Ministry of Health. Kenya primary health care strategic framework 2019–2024. Nairobi: Minstry of Health; 2019.. *Kenya primary health care strategic framework 2019–2024* (2019.0)
4. 4The World Bank. Laying the foundation for a robust health care system in Kenya. Kenya public expenditure review. Vol. II. Nairobi: The World Bank; 2014.. *Laying the foundation for a robust health care system in Kenya. Kenya public expenditure review* (2014.0) **II**
5. Naanyu V, Vedanthan R, Kamano J, Rotich J, Lagat K, Kiptoo P. **Barriers Influencing Linkage to Hypertension Care in Kenya: Qualitative Analysis from the LARK Hypertension Study**. *J Gen Intern Med* (2016.0) **31** 304-314. DOI: 10.1007/s11606-015-3566-1
6. Rachlis B, Naanyu V, Wachira J, Genberg B, Koech B, Kamene R. **Identifying common barriers and facilitators to linkage and retention in chronic disease care in western Kenya**. *BMC Public Health* (2016.0) **16** 741. DOI: 10.1186/s12889-016-3462-6
7. Nolte E, Anell A, Nolte E, Merkur S, Anell A. *Achieving person-centred health systems: evidence, strategies and challenges* (2020.0) 317-345
8. Angwenyi V, Aantjes C, Kajumi M, De Man J, Criel B, Bunders-Aelen J. **Patients experiences of self-management and strategies for dealing with chronic conditions in Malawi**. *PLoS One* (2018.0) **13** e0199977. PMID: 29965990
9. Matima R, Murphy K, Levitt N, BeLue R, Oni T. **A qualitative study on the experiences and perspectives of public sector patients in Cape Town in managing the workload of demands of HIV and type 2 diabetes multimorbidity**. *PLoS One* (2018.0) **13** e0194191. DOI: 10.1371/journal.pone.0194191
10. Murphy K, Chuma T, Methwes C, Steyn K, Levitt N. **A qualitative study of the experiences of care and motivation for effective self-management among diabetic and hypertensive patients attending public sector primary health care services in South Africa**. *BMC Health Serv Res* (2015.0) **15** 303. DOI: 10.1186/s12913-015-0969-y
11. Shippee N, Shah N, May C, Mair F, Montori V. **Cumulative complexity: a functional, patient-centered model of patient complexity can improve research and practice**. *J Clin Epidemiol* (2012.0) **65** 1041-1051. DOI: 10.1016/j.jclinepi.2012.05.005
12. Boyd C, Wolff J, Giovannetti E, Reider L, Weiss C, Xue Q. **Healthcare task difficulty among older adults with multimorbidity**. *Med Care* (2014.0) **52** S118-125. DOI: 10.1097/MLR.0b013e3182a977da
13. Duncan P, Murphy M, Man M, Chaplin K, Gaunt D, Salisbury C. **Development and validation of the Multimorbidity Treatment Burden Questionnaire (MTBQ)**. *BMJ Open* (2018.0) **8** e019413. DOI: 10.1136/bmjopen-2017-019413
14. Eton D, Lee M, St Sauver J, Anderson R. **Known-groups validity and responsiveness to change of the Patient Experience with Treatment and Self-management (PETS vs. 2.0): a patient-reported measure of treatment burden**. *Qual Life Res* (2020.0) **29** 3143-3154. DOI: 10.1007/s11136-020-02546-x
15. Schreiner N, DiGennaro S, Harwell C, Burant C, Daly B, Douglas S. **Treatment burden as a predictor of self-management adherence within the primary care population**. *Appl Nurs Res* (2020.0) **54** 151301. DOI: 10.1016/j.apnr.2020.151301
16. Subramanian S, Gakunga R, Kibachio J, Gathecha G, Edwards P, Ogola E. **Cost and affordability of non-communicable disease screening, diagnosis and treatment in Kenya: Patient payments in the private and public sectors**. *PLoS One* (2018.0) **13** e0190113. DOI: 10.1371/journal.pone.0190113
17. Sav A, Salehi A, Mair F, McMillan S. **Measuring the burden of treatment for chronic disease: implications of a scoping review of the literature**. *BMC Med Res Methodol* (2017.0) **12** 140. DOI: 10.1186/s12874-017-0411-8
18. Yin K, Jung J, Coiera E, Laranjo L, Blandford A, Khoja A. **Patient Work and Their Contexts: Scoping Review**. *J Med Internet Res* (2020.0) **22** e16656. DOI: 10.2196/16656
19. Hurst J, Agarwal G, van Boven J, Daivadanam M, Gould G, Wan-Chun Huang E. **Critical review of multimorbidity outcome measures suitable for low-income and middle-income country settings: perspectives from the Global Alliance for Chronic Diseases (GACD) researchers**. *BMJ Open* (2020.0) **10** e037079. DOI: 10.1136/bmjopen-2020-037079
20. Nolte E, Kamano J, Naanyu V, Etyang A, Gasparrini A, Hanson K. **Scaling up the Primary Health Integrated Care Project for Chronic Conditions in Kenya: Study Protocol for an implementation research project**. *BMJ Open* (2022.0) **12** e056261. DOI: 10.1136/bmjopen-2021-056261
21. Eton D, Yost K, Lai J, Ridgeway J, Egginton J, Rosedahl J. **Development and validation of the Patient Experience with Treatment and Self-management (PETS): a patient-reported measure of treatment burden**. *Qual Life Res* (2017.0) **26** 489-503. DOI: 10.1007/s11136-016-1397-0
22. Eton D, Ridgeway J, Egginton J, Tiedje K, Linzer M, Boehm D. **Finalizing a measurement framework for the burden of treatment in complex patients with chronic conditions**. *Patient Relat Outcome Meas* (2015.0) **6** 117-126. DOI: 10.2147/PROM.S78955
23. Anderson R, Eton D, Camacho F, Kennedy E, Brenin C, DeGuzman P. **Impact of comorbidities and treatment burden on general well-being among women’s cancer survivors**. *J Patient Rep Outcomes* (2021.0) **5** 2. DOI: 10.1186/s41687-020-00264-z
24. Husebø A, Dalen I, Richardson A, Bru E, Søreide J. **Factors influencing treatment burden in colorectal cancer patients undergoing curative surgery: A cross-sectional study**. *Eur J Cancer Care* (2021.0) **30** e13437. DOI: 10.1111/ecc.13437
25. Nordfonn O, Morken I, Bru L, Larsen A, Husebø A. **Burden of treatment in patients with chronic heart failure—A cross-sectional study**. *Heart Lung* (2021.0) **50** 369-374. DOI: 10.1016/j.hrtlng.2021.02.003
26. Rogers E, Yost K, Rosedahl J, Linzer M, Boehm D, Thakur A. **Validating the Patient Experience with Treatment and Self-Management (PETS), a patient-reported measure of treatment burden, in people with diabetes**. *Patient Relat Outcome Meas* (2017.0) **8** 143-156. DOI: 10.2147/PROM.S140851
27. Rogers E, Abi H, Linzer M, Eton D. **Treatment Burden in People with Hypertension is Correlated with Patient Experience with Self-Management**. *J Am Board Fam Med* (2021.0) **34** 1243-1245. DOI: 10.3122/jabfm.2021.06.210191
28. Eton D.. *Patient Experience with Treatment and Self-Management (PETS) Questionnaire 60-item version (Vs. 2.0)* (2016.0)
29. Eremenco S, Cella D, Arnold B. **A comprehensive method for the translation and cross-cultural validation of health status questionnaires**. *Eval Health Prof* (2005.0) **28** 212-232. DOI: 10.1177/0163278705275342
30. 30FACITtrans. FACITtrans History. [Cited 11 January 2021]. Available from: https://www.facit.org/facittrans-history.
31. 31REDCap. About. [Cited 7 February 2022]. Available from: https://projectredcap.org/about/.
32. Frost M, Reeve B, Liepa A. **What is sufficient evidence for reliability and validity of patient-reported outcome measures?**. *Value in Health* (2007.0) **10** S94-S105. PMID: 17995479
33. Spencer-Bonilla G, Serrano V, Gao C, Sanchez M, Carroll K, Gionfriddo M. **Patient Work and Treatment Burden in Type 2 Diabetes: A Mixed-Methods Study**. *Mayo Clin Proc Innov Qual Outcomes* (2021.0) **5** 359-367. DOI: 10.1016/j.mayocpiqo.2021.01.006
34. Eton D, Linzer M, Boehm D, Vanderboom C, Rogers E, Frost M. **Deriving and validating a brief measure of treatment burden to assess person-centered healthcare quality in primary care: a multi-method study**. *BMC Fam Pract* (2020.0) **21** 221. DOI: 10.1186/s12875-020-01291-x
35. Tran V, Harrington M, Montori V, Barnes C, Wicks P, Ravaud P. **Adaptation and validation of the Treatment Burden Questionnaire (TBQ) in English using an internet platform**. *BMC Med* (2014.0) **12** 109. DOI: 10.1186/1741-7015-12-109
36. Tran V, Montori V, Eton D, Baruch D, Falissard B, Ravaud P. **Development and description of measurement properties of an instrument to assess treatment burden among patients with multiple chronic conditions**. *BMC Med* (2012.0) **10** 68. DOI: 10.1186/1741-7015-10-68
37. Ridgeway J, Egginton J, Tiedje K. **Factors that lessen the burden of treatment in complex patients with chronic conditions: a qualitative study**. *Patient Prefer Adherence* (2014.0) **8** 339-351. DOI: 10.2147/PPA.S58014
38. Herzig L, Zeller A, Pasquier J, Streit S, Neuner-Jehle S, Excoffier S. **Factors associated with patients’ and GPs’ assessment of the burden of treatment in multimorbid patients: a cross-sectional study in primary care**. *BMC Fam Pract* (2019.0) **20** 88. DOI: 10.1186/s12875-019-0974-z
39. Hardman R, Begg S, Spelten E. **Exploring the ability of self-report measures to identify risk of high treatment burden in chronic disease patients: a cross-sectional study**. *BMC Public Health* (2022.0) **22** 163. DOI: 10.1186/s12889-022-12579-1
40. Murphy A, Biringanine M, Roberts B, Stringer B, Perel P, Jobanputra K. **Diabetes care in a complex humanitarian emergency setting: a qualitative evaluation**. *BMC Health Serv Res* (2017.0) **17** 431. DOI: 10.1186/s12913-017-2362-5
41. Zimmermann M, Bunn C, Namadingo H, Gray C, Lwanda J. **Experiences of type 2 diabetes in sub-Saharan Africa: a scoping review**. *Glob Health Res Policy* (2018.0) **3** 25. DOI: 10.1186/s41256-018-0082-y
42. 42World Bank Group. Poverty & Equity Brief. Sub-Saharan Africa. [Cited 7 February 2022]. Available from: https://databank.worldbank.org/data/download/poverty/33EF03BB-9722-4AE2-ABC7-AA2972D68AFE/Global_POVEQ_KEN.pdf.
43. Howe L, Galobardes B, Matijasevich A, Gordon D, Johnston D, Onwujekwe O. **Measuring socio-economic position for epidemiological studies in low- and middle-income countries: a methods of measurement in epidemiology paper**. *Int J Epidemiol* (2012.0) **41** 871-886. DOI: 10.1093/ije/dys037
44. Subramanian S, Gakunga R, Kibachio J, Gathecha G, Edwards P, Ogola E. **Cost and affordability of non-communicable disease screening, diagnosis and treatment in Kenya: Patient payments in the private and public sectors**. *PLoS One* (2018.0) **13** e0190113. DOI: 10.1371/journal.pone.0190113
45. Vedanthan R, Kamano J, Chrysanthopoulou S, Mugo R, Andama B, Bloomfield G. **Group Medical Visit and Microfinance Intervention for Patients With Diabetes or Hypertension in Kenya**. *J Am Coll Cardiol* (2021.0) **77** 2007-2018. DOI: 10.1016/j.jacc.2021.03.002
46. Dong R, Leung C, Naert M, Naanyu V, Kiptoo P, Matelong W. **Chronic disease stigma, skepticism of the health system, and socio-economic fragility: Qualitative assessment of factors impacting receptiveness to group medical visits and microfinance for non-communicable disease care in rural Kenya**. *PLoS One* (2021.0) **16** e0248496. DOI: 10.1371/journal.pone.0248496
47. Kamano J, Naanyu V, Ayah R, Limo O, Gathecha G, Saenyi E. **Maintaining care delivery for non-communicable diseases in the face of the COVID-19 pandemic in western Kenya**. *Pan Afr Med J* (2021.0) **39** 143. DOI: 10.11604/pamj.2021.39.143.29708
48. Kiragu Z, Rockers P, Onyango M, Mungai J, Mboya J, Laing R. **Household access to non-communicable disease medicines during universal health care roll-out in Kenya: A time series analysis**. *PLoS One* (2022.0) **17** e0266715. DOI: 10.1371/journal.pone.0266715
|
---
title: Using a health belief model to assess COVID-19 vaccine intention and hesitancy
in Jakarta, Indonesia
authors:
- Irma Hidayana
- Sulfikar Amir
- Dicky C. Pelupessy
- Zahira Rahvenia
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021901
doi: 10.1371/journal.pgph.0000934
license: CC BY 4.0
---
# Using a health belief model to assess COVID-19 vaccine intention and hesitancy in Jakarta, Indonesia
## Abstract
Since January 2021, Indonesia has administered a nationwide COVID-19 vaccination. This study examined vaccine intention and identified reasons for vaccine hesitancy in the capital city of Jakarta. This is a cross-sectional online survey using the Health Belief Model (HBM) to assess vaccine intent predictors and describe reasons for hesitancy among Jakarta residents. Among 11,611 respondents, $92.99\%$ (10.797) would like to get vaccinated. This study indicated that all HBM constructs predict vaccine intention ($P \leq 0.05$). Those with a high score of perceived susceptibility to the COVID-19 vaccine were significantly predicted vaccine hesitancy (OR = 0.18, $95\%$ CI: 0.16–0.21). Perceived higher benefits of COVID-19 vaccine (OR = 2.91, $95\%$ CI: 2.57–3.28), perceived severity of COVID-19 disease (OR: 1.41, $95\%$ CI: 1.24–1.60), and perceived susceptibility of the current pandemic (OR = 1.21, $95\%$ CI: 1.06–1.38) were significantly predicted vaccination intend. Needle fears, halal concerns, vaccine side effects, and the perception that vaccines could not protect against COVID-19 disease emerged as reasons why a small portion of the respondents ($$n = 814$$, $7.23\%$) are hesitant to get vaccinated. This study demonstrated a high COVID-19 vaccine intention and highlighted the reasons for vaccine refusal, including needle fears, susceptibility to vaccine efficacy, halal issues, and concern about vaccine side effects. The current findings on COVID-19 vaccination show that the government and policymakers should take all necessary steps to remove vaccine hesitancy by increasing awareness of vaccine efficacy and benefit interventions.
## Introduction
As of May 2021, Indonesia was one of the countries with the highest number of novel Coronavirus Diseases 2019 (COVID-19) cases with the lowest testing rate in Southeast Asia [1]. In December 2020, Indonesia received the first three million doses of the Sinovac-Coronavac vaccine [2]. Two days after the Indonesian authority of food and drug administration issued emergency use of Sinovac, the COVID-19 vaccination program was administered across the country with agglomerated urban areas as the main focus of vaccine distribution [3].
By the end of May 2021, Indonesia has reached 27 million shots of the COVID-19 vaccine. The Indonesian government has issued a regulation as a legal basis for vaccinating the Indonesian population prioritizing health care workers, older people, public servants, those with preexisting medical conditions, and those who live in areas with high transmission of COVID-19 [4]. Despite the efforts of the Indonesian government to vaccinate as many people as possible, vaccine hesitancy has existed among the population [5]. At about the same time, 15 different vaccines were granted emergency use authorization by the World Health Organization (WHO) [6]. However, the China Sinovac-Coronavac was not yet approved for emergency use by WHO [7]. Many of those approved for use have high efficacy [6]. In different demographics, vaccine efficacy is an essential driver of vaccine uptake [8, 9].
As the capital city of Indonesia, with a population of around 10.56 million, Jakarta has been the epicenter of COVID-19 in Indonesia since the beginning of March 2020 [10]. On 30 March 2021, the capital city recorded 381,090 cases with a total number of deaths of 6,327 [11]. In addition, only 1,178,243 persons ($39.3\%$) received the first dose out of 3,000,689 targets [11].
In the Southeast Asia region and worldwide, studies have been conducted to examine the intention of a vaccine against COVID-19. A previous study showed that COVID-19 vaccine uptake in Indonesia was influenced by the effectiveness of the vaccines [8]. Further, COVID-19 vaccine acceptance rates in ten lower-middle income countries in Asia, Africa, and South America were higher where vaccine safety and efficacy were high [12]. Similarly, vaccine hesitancy rates were low in Singapore [13] and New Zealand [14]. This paper assessed vaccine intention among the residents eligible to receive the shot and identified the specific reasons drawing hesitant attitudes towards COVID-19 vaccination during the first phase of the COVID-19 vaccination program in Jakarta.
## Health belief model as a theoretical framework
To pursue our goal, we applied the Health Belief Model (HBM) as the core framework in our study. As one of the most widely applied theories in health behaviors [15], the HBM consists of six domains that predict health behavior: perceived susceptibility, perceived severity, perceived benefits, perceived barriers, cues to action, and self-efficacy. As we studied one simple behavior, we excluded self-efficacy in this paper.
Perceived susceptibility refers to a belief about the possibility of getting a condition. This study addressed individuals’ beliefs about getting impacted by two conditions: the COVID-19 pandemics and the vaccine. Within this construct, we studied individuals’ perception of vaccine side effects, whether or not the vaccine could protect against infection, and halal concerns about the vaccine that may hinder individuals from getting vaccination against coronavirus infection. Perceived severity refers to feelings about the seriousness of having the COVID-19 disease. In a broader sense, we included severity related to social and financial consequences such as reduced income, loss of jobs, restricted family and social interactions. Moreover, as information and access to vaccination centers were found to be obstacles for some individuals [16], the construct of perceived barriers in this study is focused on technical aspects that individuals may have to access the vaccine. Perceived benefits refer to protection provided by COVID-19 vaccines. Finally, cues to action refers to a strategy or information source that promotes the adoption of a behavior [17].
## Study participants and survey design
This cross-sectional study was performed from 30 April to 15 May 2021. Quota sampling was used to analyze data collected from the proportion of gender represented across the five districts (West Jakarta, South Jakarta, East Jakarta, North Jakarta, and Central Jakarta) in the capital city, Jakarta. Data were collected through a web-based anonymous survey using a Qualtrics-based online questionnaire. The Jakarta Administration Bureau facilitated the distribution of questionnaires to the Jakarta population through JAKI, an application for administrative information for Jakarta residents. Inclusion criteria were that the respondents were Jakarta residents who were more than 18 years of age and with internet access, while those who work in Jakarta but live on the outskirts of the city were excluded from the study. The questionnaire was pilot tested and validated by local experts prior to the administration of the survey.
## Instruments
A 45-item structured questionnaire was developed to assess the study objectives. The survey consisted of questions that assessed demographic background (8 questions), health status and COVID-19 experience (3 questions), and HBM constructs (28 questions). A 5-point Likert scale (1 = strongly disagree to 5 = strongly agree) was used for the HBM portion of the questionnaire groups. Eight demographic variables were collected: gender, age, occupancy and whether the respondent works in the health area, their role in the local community, estimated monthly income, education level completed, and religious belief. Three questions assessed the comorbidities of the respondents and whether respondents and their families have existed or been diagnosed with COVID-19 (Yes/No). The survey was anonymous and contained no identifiable respondent information.
## Ethics statement
This study was approved by the Faculty of Psychology, Universitas Indonesia Research Ethics Committee in April 2021. Approval code: 039/FPsi. Komite Etik/PDP$\frac{.04.00}{2021}$/. The survey was conducted online. Informed consent was obtained before the respondent began participating in the study. Informed consent was documented on a digital platform. This study did not include minors.
## Statistical analysis
Descriptive statistics (mean, standard deviation, frequency) were obtained for all variables. The HBM-based statements were grouped according to their constructs (perceived susceptibility to the COVID-19 pandemic and the vaccine, perceived severity of COVID-19 disease, perceived barriers to vaccination, and perceived specific vaccine benefits, and cue to action). Cronbach’s alpha was calculated for the constructs; see supplemental materials for the detailed values. Spearman’s rho and Pearson Chi-Square correlation were used to assess the correlation between HBM construct and [1] demographic variables; [2] health status and COVID-19 experience variables. A logistic regression model was applied to examine HBM factors that significantly predicted COVID-19 vaccine intent and refusal. Additional regression test was done to study if COVID-19 health experience variables significantly predicted vaccination intention. All statistical analyses were performed using the IBM SPSS 26 software. A P-value of less than 0.05 (95 percent of confidence interval) was considered statistically significant in this study.
## Demographics characteristics
A total of 11,611 participants completed the survey. The study received proportional gender-based responses from all five districts within Jakarta province. As shown in Table 1, approximately an equal number of females ($49.67\%$) and males ($50.33\%$) respondents, who were majority had a high school degree as their highest education level ($52.70\%$, $$n = 6$$,119), were among those aged 40–50 old ($28.32\%$), and $53.68\%$ earned less than IDR 2.5 million (equal to USD 169) each month.
**Table 1**
| Variable | Category (N = 11611) | n | (%) |
| --- | --- | --- | --- |
| Sex | Male | 5844 | 50.33 |
| Sex | Female | 5767 | 49.67 |
| Age | 18–20 years old | 170 | 1.46 |
| Age | >20–30 years old | 1174 | 10.11 |
| Age | >30–40 years old | 2327 | 20.04 |
| Age | >40–50 years old | 3288 | 28.32 |
| Age | >50–60 years old | 2347 | 20.21 |
| Age | >60 years old | 2305 | 19.85 |
| Health-related jobs | Yes | 1378 | 11.87 |
| Health-related jobs | No | 10233 | 88.13 |
| Occupation | Student | 197 | 1.7 |
| Occupation | Housewife | 4017 | 34.6 |
| Occupation | Educational Staff | 275 | 2.37 |
| Occupation | Doctor/midwife/nurse/other health workers | 99 | 0.85 |
| Occupation | Day laborer (on-line driver, street trader, etc) | 1136 | 9.78 |
| Occupation | Military/Police | 59 | 0.51 |
| Occupation | Business owner | 671 | 5.78 |
| Occupation | State worker | 252 | 2.17 |
| Occupation | Private worker | 2134 | 18.38 |
| Occupation | NGO worker | 44 | 0.38 |
| Occupation | Artist | 39 | 0.34 |
| Occupation | Unemployed | 1256 | 10.82 |
| Occupation | Other | 1432 | 12.33 |
| Role in local community | Youth leader | 1411 | 12.15 |
| Role in local community | Woman leader | 1690 | 14.56 |
| Role in local community | Religious leader | 328 | 2.82 |
| Role in local community | Senior citizen | 3083 | 26.55 |
| Role in local community | | 5099 | 43.92 |
| Monthly income (Indonesian Rupiah) | < Rp. 2.500.000 | 6233 | 53.68 |
| Monthly income (Indonesian Rupiah) | IDR 2.500.001- IDR 5.000.000 | 3697 | 31.84 |
| Monthly income (Indonesian Rupiah) | IDR 5.000.001- IDR7.500.000 | 694 | 5.98 |
| Monthly income (Indonesian Rupiah) | IDR 7.500.001-IDR 10.000.000 | 324 | 2.79 |
| Monthly income (Indonesian Rupiah) | IDR 10.000.001-IDR 12.500.000 | 139 | 1.2 |
| Monthly income (Indonesian Rupiah) | IDR. 12.500.001-IDR 15.000.000 | 116 | 1.0 |
| Monthly income (Indonesian Rupiah) | > IDR 15.000.000 | 408 | 3.51 |
| Religion | Buddha | 214 | 1.84 |
| Religion | Hindu | 45 | 0.39 |
| Religion | Islam | 10168 | 87.57 |
| Religion | Catholic | 468 | 4.03 |
| Religion | Christian | 583 | 5.02 |
| Religion | Indigenous Beliefs | 14 | 0.12 |
| Religion | Other | 19 | 0.16 |
| Religion | Do not answer | 100 | 0.86 |
| Education | Not finished elementary/Middle/High school | 2704 | 23.29 |
| Education | High School | 6119 | 52.7 |
| Education | Diploma/College/Post Graduate | 2788 | 24.01 |
| Are currently being or had previously diagnosed with COVID-19 | Yes | 612 | 5.27 |
| Are currently being or had previously diagnosed with COVID-19 | No | 10999 | 94.73 |
| Have family members who are currently being or had previously diagnosed with COVID-19 | Yes | 834 | 7.182 |
| Have family members who are currently being or had previously diagnosed with COVID-19 | No | 10777 | 92.82 |
| Comorbidities | No | 8221 | 70.8 |
| Comorbidities | Yes | 3390 | 29.2 |
| Comorbidities | Cardiovascular Disease (CVD) | 270 | 2.33 |
| Comorbidities | Asthma | 290 | 2.5 |
| Comorbidities | Kidney Disease | 40 | 0.34 |
| Comorbidities | Diabetes Mellitus | 532 | 4.58 |
| Comorbidities | Hypertension | 1355 | 11.67 |
| Comorbidities | Autoimun | 31 | 0.27 |
| Comorbidities | Other | 473 | 4.07 |
| Comorbidities | Do not know | 833 | 7.17 |
More than half of the respondents (62,$45\%$) received their first dose of the COVID-19 vaccine. Only a small portion ($29.2\%$) reported having chronic diseases. The majority of respondents ($94.73\%$) and their families ($52.79\%$) were not being and had not previously been diagnosed with COVID-19.
The survey revealed that only a small portion of the respondents was unwilling to get vaccinated ($$n = 814$$, $7.01\%$) and identified five factors describing such hesitancy. Almost two percent ($1.73\%$) or 201 respondents showed a strong agreement of being afraid of needle injection, $2.5\%$ ($$n = 290$$) strongly agreed that the available COVID-19 vaccine is not halal, $3.49\%$ ($$n = 405$$) strongly agreed that the available vaccine does not provide protection from COVID-19 infection, and $3.62\%$ ($$n = 420$$) were concerned about the vaccine side effects. In addition, 279 respondents ($2.4\%$) expressed their concern that they were not included in the targeted vaccination population. See Fig 1.
**Fig 1:** *Reasons for vaccine hesitancy.*
## Health beliefs and vaccine intention
Correlation coefficient analyses were used to examine the relationship between demographic variables and the HBM constructs and COVID-19 experience variables (Table 2). All demographic variables except age were significantly correlated with the respondents’ perceived susceptibility to the COVID-19 pandemic and the vaccine, perceived severity of the COVID-19 disease, perceived barriers to vaccination, and perceived specific vaccine benefits ($P \leq 0.05$). Age was not associated with the respondents’ perceived severity of the COVID-19 disease. Table 2 details the relationship between whether respondents and their families were being or had previously been diagnosed with COVID-19, respondents’ comorbidities, and the HBM construct. In addition, more than half of the respondents had received vaccination during the survey ($$n = 7$$,251, $62.45\%$). Table 2 details the relationship between whether respondents and their families were being or had previously been diagnosed with COVID-19, respondents’ comorbidities, and the HBM construct.
**Table 2**
| Demographic variables | Demographic variables.1 | Demographic variables.2 | Perceived susceptibility of COVID-19 pandemics | Perceived susceptibility of COVID19 vaccine | Perceived severity of COVID-19 disease | Perceived barriers to COVID-19 vaccine | Perceived benefits of COVID-19 Vaccine |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Spearman’s rho | Age | Correlation Coefficient | 0.022* | -0.133* | 0.004 | -0.039* | 0.046* |
| Spearman’s rho | Age | Sig. (2-tailed) | 0.016 | 0.000 | 0.638 | 0.000 | 0.000 |
| Spearman’s rho | Monthly income (Indonesian Rupiah) | Correlation Coefficient | 0.028* | -0.120* | 0.045* | -0.137* | 0.052* |
| Spearman’s rho | Monthly income (Indonesian Rupiah) | Sig. (2-tailed) | 0.002 | 0.000 | 0.000 | 0.000 | 0.000 |
| Spearman’s rho | Highest Education level | Correlation Coefficient | 0.053* | -0.102* | 0.094* | -0.205* | 0.03 |
| Spearman’s rho | Highest Education level | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.007 |
| Pearson Chi-Square | Sex | Contingency Coefficient | 0.068* | 0.106* | 0.084* | 0.047* | 0.057* |
| Pearson Chi-Square | Sex | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Pearson Chi-Square | Health-related jobs | Contingency Coefficient | 0.069* | 0.044* | 0.070* | 0.107* | 0.05* |
| Pearson Chi-Square | Health-related jobs | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Pearson Chi-Square | Occupation | Contingency Coefficient | 0.096* | 0.146* | 0.114* | 0.139* | 0.101* |
| Pearson Chi-Square | Occupation | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Pearson Chi-Square | Role in Local community | Contingency Coefficient | 0.067* | 0.118* | 0.085* | 0.075* | 0.104* |
| Pearson Chi-Square | Role in Local community | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Pearson Chi-Square | Religion | Contingency Coefficient | 0.125* | 0.173* | 0.119* | 0.107* | 0.103* |
| Pearson Chi-Square | Religion | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| Pearson Chi-Square | Are currently being or had previously diagnosed with COVID-19 | Contingency Coefficient | 0.033* | 0.021 | 0.017 | 0.120* | 0.012 |
| Pearson Chi-Square | Are currently being or had previously diagnosed with COVID-19 | Sig. (2-tailed) | 0.012 | 0.253 | 0.519 | 0.025 | 0.806 |
| Pearson Chi-Square | Family members are currently being or had previously diagnosed with COVID-19 | Contingency Coefficient | 0.022 | 0.021 | 0.042* | 0.022 | 0.021 |
| Pearson Chi-Square | Family members are currently being or had previously diagnosed with COVID-19 | Sig. (2-tailed) | 0.214 | 0.297 | 0.000 | 0.250 | 0.288 |
| Pearson Chi-Square | Comorbidities | Contingency Coefficient | 0.051* | 0.163* | 0.048* | 0.023 | 0.06* |
| Pearson Chi-Square | Comorbidities | Sig. (2-tailed) | 0.000 | 0.024 | 0.000 | 0.186 | 0.000 |
| Pearson Chi-Square | COVID-19 Vaccination status | Contingency Coefficient | 0.037* | 0.272* | 0.073* | 0.128* | 0.139* |
| Pearson Chi-Square | COVID-19 Vaccination status | Sig. (2-tailed) | 0.003 | 0.000 | 0.000 | 0.000 | 0.000 |
| Pearson Chi-Square | Family consent to get vaccinated | Contingency Coefficient | 0.065* | 0.364* | 0.075* | 0.130* | 0.282* |
| | Family consent to get vaccinated | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| | Vaccination willingness | Contingency Coefficient | 0.060* | 0.374* | 0.073* | 0.118* | 0.282* |
| | Vaccination willingness | Sig. (2-tailed) | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
This study indicated that all HBM construct predicts vaccine intention ($P \leq 0.05$) as described in Table 3. Those with a high score of perceived susceptibility or concern to COVID-19 vaccine (OR = 0.18, $95\%$ CI: 0.16–0.21, $P \leq 0.05$) and perceived technical barriers (OR = 0.85, $95\%$ CI: 0.77–0.96, $P \leq 0.05$) were less likely to get vaccinated than those with less scores. Perceived higher benefits of COVID-19 vaccine (OR = 2.91, $95\%$ CI: 2.57–3.28, $P \leq 0.05$), perceived severity of the COVID-19 disease (OR: 1.41, $95\%$ CI: 1.24–1.60, $P \leq 0.05$), and perceived susceptibility of the current pandemic (OR = 1.21, $95\%$ CI: 1.06–1.38, $P \leq 0.05$) were significantly more likely to intend to get vaccinated. Two variables in respondents and family health conditions are found to be determinant factors for vaccine intent and hesitance as described in Table 4. Respondents with family members who were being or had previously been diagnosed with COVID-19 were one and a half times more likely to get vaccination than those who were not being or previously had not been diagnosed with COVID-19 (OR = 1.50, $95\%$ CI: 1.09–2.05, $P \leq 0.05$). In addition, those who have comorbidities were less likely to get vaccinated compared to those who have no comorbidities (OR = 0.50, $95\%$ CI: 0.44–0.58, $P \leq 0.05$).
## Discussion
The present study examined vaccine intention and described reasons for vaccine hesitancy vis-a-vis vaccine acceptance among Jakarta residents. This was conducted during the first phase of the COVID-19 vaccination rollout in Jakarta in which COVID-19 cases and deaths were the highest in the nation. By 30 April 2021, at the start of the present study, Jakarta recorded 408,620 cases, of which 6,733 died from COVID-19 [10]. Hence, the local health authority delivered the vaccination program massively and rapidly.
At the time of the study, vaccination priority was given to health care workers, older citizens, and those who work in public service areas [4]. Jakarta has targeted 3,000,689 people to receive COVID-19 vaccination during the first phase of vaccine rollout. As of 30 April 2021, 1,906,096 or $63.5\%$ of the target population have received the first dose, while 1,214,494 or $40.5\%$ have completed the second dose [10]. While the sample in this study comprises more of the non-priority vaccine population, when we looked into the respondents’ occupations, we found nearly $12\%$ ($$n = 1$$,378) had healthcare-related jobs and about forty percent of the respondents ($$n = 4$$,652, $40.06\%$) were elderly. The findings are in-line with the vaccination statistic data, for which $62.45\%$ or 7,251 respondents have received the first dose.
Consistent with the previous studies [8, 18–20], the respondents in our study demonstrated positive intention toward COVID-19 vaccination ($$n = 10$$,797, $92.77\%$). This substantial proportion of positive intention toward COVID-19 vaccination exceeds the finding that 96 countries achieved lower than the WHO target of $40\%$ of vaccination coverage by the end of 2021 and that lower-middle-income countries could achieve between $28\%$ to $80\%$ vaccination coverage [21].
However, although only a small portion of the respondents ($$n = 814$$, $7.01\%$) was unwilling to uptake the COVID-19 vaccine, scrutinizing the reasons for vaccine hesitancy helps better understand the barriers and formulate recommendations, especially communication to address the obstacles. Addressing vaccination barriers in Jakarta, the COVID-19 epicenter of Indonesia, is critical to ensure most of its population is protected by vaccines. This present study excavated five reasons as such barriers to vaccine hesitancy. One of the barriers this study revealed is needle fears or being afraid of injection among 201 respondents ($1.73\%$). This is not surprising because the previous study demonstrated that some Indonesian adults are afraid to inject a needle into the body [22]. In fact, fear of needle injection has been recognized in healthcare areas. Despite needle fears being common among children, a study in the USA estimated that 11.5 to 66 million U.S. adults might encounter this condition [23]. Consequently, this group often avoids seeking medical care which may lead to vaccination refusals [24].
Moreover, consistent with Baraniuk [25] and Singh and Upshur [26], 290 respondents ($2.5\%$) believe that the available COVID-19 vaccine is not halal, which led them to refuse to get the vaccination. It should be noted that China Sinovax’s Coronavax was the only vaccine available when the survey was conducted. As the most populous Muslim country, religious consideration, including a halal certification of a vaccine, is critical for vaccine acceptance [27]. The halal issue on vaccination has existed in the nation and is attributable to vaccine refusal [28, 29]. For example, a previous study has demonstrated a sharp decline in the measles and rubella vaccination when the population doubts whether the vaccine qualifies as halal [30]. Therefore, as a Muslim majority population, it is critical to issue halal certification as soon as the Emergency Use Authorization of a vaccine is announced to reduce refusal.
In addition to the halal issue, we found 405 or $3.49\%$ of respondents perceived that the available vaccine could not protect against COVID-19 disease. This finding has been consistent with the most recent study that assesses vaccine acceptance. Harapan et al. [ 8] indicated that the likelihood of people to uptake vaccination is if the COVID-19 vaccine had $95\%$ efficacy. Thus, this present study underlined the importance of higher effectiveness perceived as efficacy could impact vaccine uptake. Next is the concern about vaccine side effects ($$n = 420$$, $3.65\%$), which is consistent with findings from numerous studies [9, 19, 31, 32]. Wong et al. [ 32] indicated that fears about vaccine adverse effects are indicated as among the strongest barriers to vaccination, which was described by having heard of adverse effects after vaccination and having heard of death cases after vaccination. The last reason for vaccine hesitancy we found in this study is the concern of not being included in the vaccination program ($$n = 279$$, $2.4\%$). As described previously, the local authority only inoculated COVID-19 vaccination for the most vulnerable target population to protect against COVID-19 [4]. Thus, such concern is reasonable amidst limited available vaccine stock, and vaccination for a wider public is yet to be available.
Furthermore, although our findings suggest that people’s beliefs or perceptions about the susceptibility and severity of current COVID-19 pandemics and vaccines, including perceived benefits and technical barriers to access vaccines, were determinants of vaccine intent or refusal, greater attention should be emphasized to the perceived vaccine susceptibility (B = -1.72, $P \leq 0.05$) and the benefit of the vaccine to protect against COVID-19 ($B = 1.023$, $P \leq 0.05$). In this study, these two variables significantly provided major contributions to predicting vaccine intention and refusal compared to the other HBM variables. Again, these findings highlighted the pivotal role of removing barriers to halal issues and fears of needle injection. Moreover, the effectiveness of vaccines was one of the essential drivers for vaccine uptake [8, 9]. Therefore, as this study suggested, ensuring to provide vaccines with a higher efficacy level is more likely to reduce vaccine hesitancy.
Lastly, this study indicated that self and family health conditions significantly predicted vaccine intention. Those who have comorbidities were less likely to get vaccinated compared to those who have no comorbidities (OR = 0.50, $95\%$ CI: 0.44–0.58, $P \leq 0.05$). These findings are in line with studies conducted elsewhere: in Northern Italy [33] and Brazil [34]. The results of this study are consistent with others showing that COVID-19 vaccine hesitancy is more common among people with comorbidities.
## Conclusion
This study demonstrated a high COVID-19 vaccine intention ($$n = 10$$,797, $92.77\%$). Four major factors have been identified as predictors of such high uptake, i.e., perceived COVID-19 disease susceptibility (OR = 1.34, $$P \leq 0.00$$), the technical barrier to access vaccination (OR = 0.58, $$P \leq 0.00$$), family members who were currently being or previously had diagnosed with COVID-19 (OR = 1.42, $$P \leq 0.03$$), and self-comorbidities (OR = 1.89, $$P \leq 0.00$$). Additionally, this study underscored the importance of identifying reasons for vaccine refusal. Needle fears, susceptibility to vaccine efficacy, halal issues, concern about vaccine side effects and comorbidities, and not being included in the vaccination targeted group were indicated as barriers to vaccine uptake. Although only accounted for by a small number of respondents, it is plausible to address these specific barriers, given that Jakarta always had the highest COVID-19 cases and deaths. This study suggests that education on vaccine efficacy and benefit interventions, which encompasses removing vaccine hesitancy, is critically needed to promote vaccine uptake. Lastly, there is a need for further similar studies in the same population that might provide a comprehensive picture of vaccination intentions and barriers.
## Limitations
This study has two limitations. First, we used a simple stratification of the sample based on the sample’s gender proportion. However, the quota sampling employed could lead to sampling bias because the sample has not been chosen using random selection. *The* generalizability of the survey results may be impacted by how we distributed the online questionnaire. Second, the Jakarta administration team helped us to disseminate the questionnaire using an application for Jakarta residents. As a result, it might not reach people with no internet and no smartphone access, thus affecting data representation.
## References
1. Puno GR, Puno RCC, Maghuyop IV. **COVID-19 case fatality rates across Southeast Asian countries (SEA): a preliminary estimate using a simple linear regression model**. *Journal of Health Research* (2021.0) **35** 286-94
2. 2COVID-19 WRP. 11 Tahap kedatangan vaksin COVID-19 di Indonesia—masyarakat umum [Internet]. BNPB:2021 Available from: https://covid19.go.id/id/edukasi/masyarakat-umum/11-tahap-kedatangan-vaksin-covid-19-di-indonesia
3. Kementerian Sekretariat Negara RI. **Laksanakan program vaksinasi Covid-19 nasional, Indonesia serius tangani perlindungan kesehatan dan pemulihan ekonomi nasional |**. *Sekretariat Negara [Internet]. Setneg* (2021.0)
4. 4MOH. Permenkes No. 10 Tahun 2021 tentang pelaksanaan vaksinasi dalam rangka penanggulangan pandemi Corona Virus Disease 2019 (COVID-19) [JDIH BPK RI] [Internet]. BPK: 2021. Available from: https://peraturan.bpk.go.id/Home/Details/169665/permenkes-no-10-tahun-2021
5. MoH ITAGI, UNICEF WHO. **Survei penerimaan vaksin COVID-19 di Indonesia [Internet].**. *MoH* (2020.0)
6. 6Vaccines–COVID19 Vaccine Tracker [Internet]. Covid19 Vaccine Tracker. [cited 2021 Dec 5]. Available from: https://covid19.trackvaccines.org/vaccines/
7. 7WHO. WHO validates Sinovac COVID-19 vaccine for emergency use and issues interim policy recommendations [Internet]. WHO. 2021 [cited 2021 Dec 5]. Available from: https://www.who.int/news/item/01-06-2021-who-validates-sinovac-covid-19-vaccine-for-emergency-use-and-issues-interim-policy-recommendations
8. Harapan H, Wagner AL, Yufika A, Winardi W, Anwar S, Gan AK. **Acceptance of a COVID-19 vaccine in Southeast Asia: a cross-sectional study in Indonesia**. *Frontiers in Public Health* (2020.0) 8. PMID: 32117848
9. Kerekes S, Ji M, Shih SF, Chang HY, Harapan H, Rajamoorthy Y. **Differential effect of vaccine effectiveness and safety on COVID-19 vaccine acceptance across Socioeconomic groups in an international sample**. *Vaccines* (2021.0) **9** 1010. DOI: 10.3390/vaccines9091010
10. Jakarta DKI. **DKI Jakarta COVID-19 Statistics [Internet]**. *Corona Jakarta* (2020.0)
11. Communication DKI. **Informatics & Statistics Bureau. Perkembangan data kasus dan vaksinasi COVID-19 di Jakarta Per 30 Maret 2021 [Internet]**. *PPID.Jakarta* (2021.0)
12. Rosiello DF, Anwar S, Yufika A, Adam RY, Ismaeil MI, Ismail AY. **Acceptance of COVID-19 vaccination at different hypothetical efficacy and safety levels in ten countries in Asia, Africa, and South America**. *Narra J.* (2021.0) **1**
13. Griva K, Tan KYK, Chan FHF, Periakaruppan R, Ong BWL, Soh ASE. **Evaluating rates and determinants of COVID-19 vaccine hesitancy for adults and children in the Singapore population: strengthening our community’s resilience against threats from emerging infections (SOCRATEs) Cohort**. *Vaccines* (2021.0) **9** 1415. DOI: 10.3390/vaccines9121415
14. Prickett KC, Habibi H, Carr PA. **COVID-19 vaccine hesitancy and acceptance in a cohort of diverse New Zealanders**. *The Lancet Regional Health–Western Pacific* (2021.0) 14
15. Glanz K, Bishop DB. **The role of behavioral science theory in development and implementation of public health interventions**. *Annual Review of Public Health* (2010.0) **31** 399-418. DOI: 10.1146/annurev.publhealth.012809.103604
16. 16LaporCovid-19. Somasi atas permenkes no. 19 tahun 2021 [Internet]. Somasi Terbuka. 2021 [cited 2021 Oct 5]. Available from: https://laporcovid19.org/post/somasi-atas-permenkes-no-19-tahun-2021
17. Glanz K, Rimer BK, Viswanath K. *Theory, research, and practice* (2008.0)
18. Detoc M, Bruel S, Frappe P, Tardy B, Botelho-Nevers E, Gagneux-Brunon A. **Intention to participate in a COVID-19 vaccine clinical trial and to get vaccinated against COVID-19 in France during the pandemic**. *Vaccine* (2020.0) **38** 7002-6. DOI: 10.1016/j.vaccine.2020.09.041
19. Elnaem MH, Mohd Taufek NH, Ab Rahman NS, Mohd Nazar NI, Zin CS, Nuffer W. **COVID-19 vaccination attitudes, perceptions, and side effect experiences in Malaysia: do age, gender, and vaccine type matter?**. *Vaccines* (2021.0) **9** 1156. DOI: 10.3390/vaccines9101156
20. Lucia VC, Kelekar A, Afonso NM. **COVID-19 vaccine hesitancy among medical students**. *Journal of Public Health* (2021.0) **43** 445-9. DOI: 10.1093/pubmed/fdaa230
21. Watson OJ, Barnsley G, Toor J, Hogan AB, Winskill P, Ghani AC. **Global impact of the first year of COVID-19 vaccination: a mathematical modelling study**. *The Lancet Infectious Diseases* (2022.0) **22** 1293-302. DOI: 10.1016/S1473-3099(22)00320-6
22. Ligita T, Wicking K, Francis K, Harvey N, Nurjannah I. **How people living with diabetes in Indonesia learn about their disease: a grounded theory study**. *PLOS ONE* (2019.0) **14** e0212019. DOI: 10.1371/journal.pone.0212019
23. Love AS, Love RJ. **Considering needle phobia among adult patients during mass COVID-19 vaccinations**. *J Prim Care Community Health* **202** 21501327211007390. DOI: 10.1177/21501327211007393
24. Cook LS. **Needle phobia**. *Journal of Infusion Nursing* (2016.0) **39** 273-9. DOI: 10.1097/NAN.0000000000000184
25. Baraniuk C.. **How to vaccinate the world against covid-19**. *BMJ* **372** n211. DOI: 10.1136/bmj.n211
26. Singh JA, Upshur REG. **The granting of emergency use designation to COVID-19 candidate vaccines: implications for COVID-19 vaccine trials**. *The Lancet Infectious Diseases* (2021.0) **21** e103-9. DOI: 10.1016/S1473-3099(20)30923-3
27. Padmawati RS, Heywood A, Sitaresmi MN, Atthobari J, MacIntyre CR, Soenarto Y. **Religious and community leaders’ acceptance of rotavirus vaccine introduction in Yogyakarta, Indonesia: a qualitative study**. *BMC Public Health.* (2019.0) **19** 368. DOI: 10.1186/s12889-019-6706-4
28. DeRoeck D, Deen J, Clemens JD. **Policymakers’ views on dengue fever/dengue haemorrhagic fever and the need for dengue vaccines in four southeast Asian countries**. *Vaccine* (2003.0) **22** 121-9. DOI: 10.1016/s0264-410x(03)00533-4
29. Seale H, Sitaresmi MN, Atthobari J, Heywood AE, Kaur R, MacIntyre RC. **Knowledge and attitudes towards rotavirus diarrhea and the vaccine amongst healthcare providers in Yogyakarta Indonesia**. *BMC Health Services Research* (2015.0) **15** 528. DOI: 10.1186/s12913-015-1187-3
30. Rochmyaningsih D.. **Indonesian fatwa causes immunization rates to drop**. *Science* (2018.0) **362** 628-9. DOI: 10.1126/science.362.6415.628
31. Rachman FF, Pramana S. **Analysis of Indonesian people’s sentiments about the side effects of the COVID-19 vaccine on twitter**. *Journal of Data Science and Its Applications* (2021.0) **4** 1-10
32. Wong MCS, Wong ELY, Cheung AWL, Huang J, Lai CKC, Yeoh EK. **COVID-19 vaccine hesitancy in a city with free choice and sufficient doses**. *Vaccines* (2021.0) **9** 1250. DOI: 10.3390/vaccines9111250
33. Reno C, Maietti E, Fantini MP, Savoia E, Manzoli L, Montalti M. **Enhancing COVID-19 vaccines acceptance: results from a survey on vaccine hesitancy in northern Italy**. *Vaccines* (2021.0) **9** 378. DOI: 10.3390/vaccines9040378
34. Vieira Rezende RP, Braz AS, Guimarães MFB, Ribeiro SLE, Abreu Vieira RMR, Bica BE. **Characteristics associated with COVID-19 vaccine hesitancy: a nationwide survey of 1000 patients with immune-mediated inflammatory diseases**. *Vaccine* (2021.0) **39** 6454-9. DOI: 10.1016/j.vaccine.2021.09.057
|
---
title: Prevalence and factors associated with diagnosed diabetes mellitus among Asian
Indian adults in the United States
authors:
- Ranjita Misra
- Suresh S. Madhavan
- Trupti Dhumal
- Usha Sambamoorthi
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10021922
doi: 10.1371/journal.pgph.0001551
license: CC BY 4.0
---
# Prevalence and factors associated with diagnosed diabetes mellitus among Asian Indian adults in the United States
## Abstract
Higher prevalence of diabetes mellitus (DM) has been documented among South Asians living in the United States. However, combining the south Asian subgroups into one category masks the heterogeneity in the diagnosed DM, after controlling for known protective and risk factors. We assessed the association of Asian Indian ethnicity to diagnosed DM using a nationally representative sample of 1,986 Asian Indian adults in the US compared to 109,072 Non-Hispanic Whites (NHWs) using disaggregated data from the National Health Interview Survey (2012–2016) (NHIS). 2010 US census figures were used for age-sex standardization. Age-sex adjusted prevalence of DM was $8.3\%$ in Asian Indians as compared to $5.8\%$ in NHW. In adjusted multivariable logistic regression models, Asian Indians had higher odds ratios of reporting diagnosed DM compared to NHWs (AOR = 1.39, $95\%$ CI: 1.12, 1.71). This association remained strong and significant even after controlling for other risk factors in the model (AOR = 1.47, $95\%$ CI: 1.16, 1.85). Results suggest a favorable socio-economic profile of Asian Indians was not protective on diagnosed DM. In addition, they were more likely to have diagnosed DM due to higher prevalence of obesity despite healthier behaviors of smoking and exercise.
## Introduction
The world-wide prevalence of diabetes mellitus (DM), a disabling chronic condition, is growing and is projected to increase to 700 million by 2045 [1]. South Asian countries (India, Bangladesh, Pakistan, Sri Lanka, Bhutan, Nepal and Maldives), and China account for approximately $60\%$ of the world’s population with diabetes [2, 3]. DM prevalence in these countries has increased more than 2.5 fold during the last decade and is expected to increase exponentially with time [3]. Migrant Asian Indians in the U.S. have high rates of insulin resistance due to an inherent genetic predisposition, and an increased disease incidence at lower age and body mass index (BMI) [4, 5] even with a favorable socio-economic profile (i.e. high income and high levels of education). Hence, migrant Asian Indians in the U.S. are reported to have the highest ethnic-specific DM rates and higher risk of DM as compared to other ethnic groups [6–11]. Findings from a 20-year longitudinal study from the United Kingdom suggest that DM prevalence is three times higher among Asian Indians compared to the European controls [12]. Similar high prevalence of DM among immigrant *Indians is* reported in Singapore as compared to the native population [13].
Asian Indians (or Indian Americans) encompass 4.4 million people in the United States [14]. They are the second largest and fastest growing Asian subgroup in the US with a growth rate of $70\%$ between 2000–2010 and expected to almost triple by 2050 [15]. Published studies have focused on DM in South Asians due to of small sample sizes and lack of disaggregated national data, potentially masking the disease burden within the subgroups [16]. Using National Health and Nutrition Examination Survey (NHANES) data from 2011 to 2016, Cheng et. al. reported that the prevalence of age-sex adjusted diagnosed DM among south Asians was higher ($16.0\%$ vs $8.2\%$) than prevalence among NHWs [17]. The heterogeneity of South Asians living in the US in terms of ethnic, religious, linguistic, and socioeconomic characteristics is a factor is masking disease burden within the subgroups. Hence, disaggregating the data has clear advantages for examining epidemiological trends over time, reduce health disparities, and understanding the DM prevalence in various racial/ ethnic groups. DM is increasing faster among Asian Americans subgroups than non-Hispanic Whites, non-Hispanic Blacks, and Hispanic Americans [18–20].
Despite a surge in investigation of chronic diseases in Asian Americans and its subgroups, population-based studies on DM prevalence among migrant Asian Indians in the United States are limited due to small sample size, use of purposive sampling and/or specific subgroups in various regions that limits the generalizability of the results, increase measurement errors in risk factors, or lack a comparison group [21–24]. For example, a 2004 community-based survey that included only Asian Indians living in the metro area of Georgia, Atlanta, found that $18.3\%$ had DM [25]. Based on National Health Interview Survey (NHIS) data from 1997 through 2005, Asian Indians were three times as likely as NHWs to report DM even after adjusting for variables like age, sex, and obesity [26]. Using cross-sectional data from the NHIS data from 1997 through 2000, Mohanty et al., concluded that despite lower-rates of obesity, Asian Indians were more likely to have diagnosed diabetes compared to NHWs after controlling for age and obesity [21]. However, this study did not use the appropriate BMI for Asian populations as recommended by the World Health Organization (WHO) [27]. The Diabetes among Indian Americans (DIA) study estimated the prevalence of diabetes among U.S. Asian Indians residing in 7 US cities at $17.4\%$ [22].
These studies have suggested a growing burden of DM in Asian Indians that represent a public health challenge. An examination of the prevalence of diagnosed DM and associated protective and risk factors among Asian Indians can assist in targeted prevention and treatment efforts in this ethnic group. As Asian Indian population is expected to increase, almost triple its size by 2050 [28], examination of diabetes prevalence in this group is important. Therefore, the primary objective of this study is to evaluate the association of Asian Indian ethnicity to diagnosed diabetes, after controlling for known protective and risk factors, as compared to non-Hispanic whites (NHWs) in the United States. We will use disaggregated data for Asian Indians from the NHIS, a nationally representative sample of non-institutionalized civilian population of households.
## Methods
The Institutional Review Board of West Virginia University and the University of North Texas Health Sciences Center determined that the study was exempt from IRB approval because the study used publicly available data.
## Study design
A cross-sectional study design was adopted using pooled secondary data from multiple years (2012–2016) of the NHIS. Data was pooled to ensure adequate cell size and to increase reliability of results. NHIS investigators recommend pooling multiple years of data to achieve adequate sample size and minimize the relative standard error (less than $30\%$) [29]. We included only Asian Indians and NHWs as our primary interest was to compare these two groups.
## Data source
NHIS is an ongoing, continuous, nationwide, cross-sectional annual survey of household civilian noninstitutionalized population in the US [30]. The nationally representative data is obtained by utilizing a multistage sampling technique, wherein the target universe is divided into numerous nested levels of strata and clusters [30]. This study used data from family, person, sample adult core, and imputed income files. The person file captures attributes such as sociodemographic characteristics, health status, and health insurance, whereas, information on the poverty status of the household was provided by the family files. Information about chronic physical conditions, psychological distress, access to care, and utilization of healthcare services were utilized from the sample adult core files [31].
## Study sample
The study sample consisted of NHIS participants who were 18 years or older, who were either Asian Indian or NHWs, and participated in the sample adult core. As our primary interest is comparing Asian Indians and NHWs, other racial/ethnic groups were excluded from the analysis. The final sample consisted of 111,058 adults (1,986 Asian Indians and 109,072 NHWs).
## Dependent variable: Self-reported diagnosed diabetes—Yes/No
To ascertain an individual’s diabetes status, a positive response to the following question was used: “Have you ever been told by a doctor or health professional (other than during pregnancy, if female) that you have diabetes?”. Those who answered border-line DM were considered as not having diabetes. Also, pregnant women with gestational diabetes were excluded.
## Key independent variable: Asian Indians versus non-Hispanic Whites
Data on Asian American subgroups have been collected by the NHIS since 1992 but desegregated data for Asian Indians was only available since 2011 [32]. NHIS asks a number of queries to ascertain an individual’s race or ethnicity. We used responses from a question that queried whether the respondent is 1) Hispanic, Latino/a, or Spanish origin, and 2) what his/her race is. Individuals responding, “not of Hispanic, Latino and/or Spanish origin” and “White” were categorized as NHWs. Asian subgroups were designated to seven subcategories: Asian Indian, Chinese, Filipino, Japanese, Korean, Vietnamese, and other Asian. For the purpose of this study, we only included individuals who self-identified themselves as Asian Indians.
## Demographic and lifestyle characteristics
We included variables that are known to be associated with DM risk among adults. These comprised of biological variables such as age (18–44 years, 45–64, and 65 or older) and sex (women/men). Socio-economic factors included education (less than high school, high school/GED, some college, and college), federal poverty level (FPL) (< $100\%$, 100 to < $200\%$, 200 to < $400\%$ and > = $400\%$) and employment (employed vs not employed). Access to healthcare (health insurance) defined as having health insure and no health insurance. General health was self-reported on a five-point Likert scale, ranging from excellent to poor and diagnosis of chronic conditions such as diabetes, hypertension, high cholesterol, COPD, and heart disease by a doctor or health professional as yes/no. Lifestyle health practices included smoking status (nonsmoker, former smoker, and current smoker), alcohol use (lifetime abstainer, former drinker, and current drinker), physical activity (daily, weekly, monthly/yearly, and unable), and body mass index (BMI). Obesity was assessed by BMI cutoffs using both standard criteria and the World Health Organization Western Pacific Region (WHO-WPR) (World Health Organization Western Pacific Region, International Association for the Study of Obesity, International Obesity Task Force, 2000). The Centers for Disease Control’s BMI criteria was used to classify NHWs as follows: 1) Underweight/normal (0–25.0 kg/m2); 2) Overweight (25.0–30.0 kg/m2); and 3) Obese (≥ 30 kg/m2). For Asian Indians, we used both the standard and WHO recommended revised cut points for Asians that defined overweight as a BMI of 23.0–24.9 and obesity as a BMI≥25 [33–35]. For all variables with missing values, we included indicator variables representing missing group in adjusted analyses. Self-reported general health (excellent/very good, good, fair/poor) and high cholesterol (people who answered “yes” to the question “have you ever been told by a doctor or other health professional that you had high cholesterol?). Health insurance coverage (defined as insured and uninsured) and have usual source of care.
## Statistical analysis
To account for the complex sampling design, clustering, stratification, and weight variables were used, and all statistical analyses were conducted using SAS 9.4 (SAS institute INC, Cary, North Carolina) survey procedures. Unadjusted group differences in sample characteristics between Asian Indians and NHWs were analyzed using Rao-Scott chi-square tests. To derive age-sex adjusted prevalence of DM, we used 2010 US census distribution of age and sex [36]. Multivariable logistic regressions were used to analyze the associations of Asian Indians ethnicity to diagnosed DM, adjusted for known factors associated with DM—age, sex, education, employment, poverty status, health insurance, marital status, hypertension, high cholesterol, physical activity and smoking. Statistical inferences were based on a significance level of P (two-sided) ≤ 0.05.
## Results
Study sample consisted of 109,072 NHWs ($98.2\%$) and 1,986 Asian Indians ($1.7\%$) (Table 1). A little over half ($51.1\%$) of the participants were female ($49.5\%$ AIs and $51.5\%$ NHWs) and younger than 65 years of age ($78\%$ total, $92\%$ AIs and $77.7\%$ NHWs). A quarter ($27.8\%$) were obese ($9.8\%$ AIs and $27.4\%$ NHWs) according to the standard CDC criteria (BMI ≥30.0). In addition, the World Health Organization obesity classification for South Asians was used to calculate obesity for Asian Indians (≥25.0). Using the WHO criteria, obesity rate was significantly higher i.e., $46.1\%$ for AIs. Current smoking status was reported by only $17.9\%$ of the participants. *In* general, over a quarter of the participants ($29.9\%$) and $\frac{2}{3}$rd ($68.3\%$) reported a diagnosis of hyperlipidemia and hypertension, respectively.
**Table 1**
| ALL | Total | Total.1 | Asian Indians | Asian Indians.1 | Non-Hispanic Whites | Non-Hispanic Whites.1 | Non-Hispanic Whites.2 | Non-Hispanic Whites.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| ALL | N | Wt % | N | Wt % | N | Wt % | Chi-sq | Prob b |
| ALL | 111058 | 100.0 | 1986 | 100.0 | 109072 | 100.0 | | |
| Diabetes Mellitus | | | | | | | 2.8 | 0.095 |
| Yes | 10671 | 8.8 | 137 | 7.5 | 10534 | 8.8 | | |
| No | 100387 | 91.2 | 1849 | 92.5 | 98538 | 91.2 | | |
| Sex | | | | | | | 1.8 | 0.186 |
| Women | 60020 | 51.5 | 914 | 49.5 | 59106 | 51.5 | | |
| Men | 51038 | 48.5 | 1072 | 50.5 | 49966 | 48.5 | | |
| Age Groups | | | | | | | 265.2 | < 0.001 |
| 18–39 | 33775 | 33.9 | 1186 | 53.3 | 32589 | 33.5 | | |
| 40–49 | 16342 | 16.4 | 364 | 21.3 | 15978 | 16.3 | | |
| 50–64 | 30501 | 27.7 | 277 | 17.3 | 30224 | 28.0 | | |
| >= 65 | 30440 | 22.0 | 159 | 8.0 | 30281 | 22.3 | | |
| Marital status | | | | | | | 182.0 | <0.001 |
| Married | 58718 | 52.8 | 1358 | 68.3 | 57360 | 52.5 | | |
| Wid/Sep/Div | 30496 | 27.4 | 142 | 7.15 | 30354 | 27.8 | | |
| Never married | 21615 | 19.4 | 483 | 24.3 | 21.132 | 19.3 | | |
| Education | | | | | | | 771.6 | < 0.001 |
| LT HS | 9855 | 8.5 | 88 | 5.2 | 9767 | 8.5 | | |
| HS | 27661 | 24.9 | 175 | 10.1 | 27486 | 25.2 | | |
| Some College | 35645 | 31.8 | 207 | 11.4 | 35438 | 32.2 | | |
| College | 37587 | 34.5 | 1508 | 72.9 | 36079 | 33.8 | | |
| Poverty Status | | | | | | | 60.4 | < 0.001 |
| < 100% FPL | 12239 | 8.5 | 219 | 8.3 | 12020 | 8.5 | | |
| 100-<200% | 17590 | 13.8 | 217 | 10.3 | 17373 | 13.8 | | |
| 200-<400% | 30177 | 26.9 | 379 | 20.6 | 29798 | 27.0 | | |
| >= 400% | 41616 | 42.1 | 1003 | 52.6 | 40613 | 41.9 | | |
| Health Insurance | | | | | | | 0.3 | 0.845 |
| Yes | 100731 | 90.7 | 1804 | 90.7 | 98927 | 90.7 | | |
| No | 10013 | 8.9 | 175 | 9.0 | 9838 | 8.9 | | |
| Hypertension | | | | | | | 115.5 | < 0.001 |
| Hypertension | 38098 | 31.6 | 320 | 18.0 | 37778 | 31.8 | | |
| No hypertension | 72861 | 68.3 | 1665 | 82.0 | 71196 | 68.1 | | |
| High Cholesterol | | | | | | | 51.7 | <0.001 |
| Yes | 33291 | 29.9 | 380 | 19.1 | 32911 | 30.1 | | |
| No | 77347 | 69.6 | 1602 | 80.7 | 75745 | 69.4 | | |
| Body Mass Index | | | | | | | 198.9 | < 0.001 |
| Und/normal | 39525 | 35.6 | 650 | 30.5 | 38875 | 35.7 | | |
| Overweight | 36710 | 33.0 | 415 | 21.2 | 36295 | 33.3 | | |
| Obese a | 30940 | 27.8 | 882 | 46.1 | 30058 | 27.4 | | |
| Smoking Status | | | | | | | 698.9 | < 0.001 |
| Never Smoked | 60439 | 56.1 | 1719 | 87.9 | 58720 | 55.5 | | |
| Former Smoker | 29756 | 25.5 | 140 | 6.8 | 29616 | 25.9 | | |
| Current Smoker | 20330 | 17.9 | 119 | 5.1 | 20211 | 18.1 | | |
| Alcohol Use | | | | | | | 965.8 | < 0.001 |
| Abstainer | 17158 | 15.7 | 959 | 50.3 | 16199 | 15.1 | | |
| Former Drinker | 17574 | 14.2 | 97 | 5.4 | 17477 | 14.4 | | |
| Current Use | 74817 | 68.6 | 903 | 42.6 | 73914 | 69.1 | | |
| Physical Activity | | | | | | | 28.4 | < 0.001 |
| Daily | 7570 | 6.9 | 131 | 6.6 | 7439 | 6.9 | | |
| Weekly | 39527 | 37.3 | 838 | 40.6 | 38689 | 37.2 | | |
| Monthly/Year | 60143 | 52.8 | 994 | 51.4 | 59149 | 52.9 | | |
| Region | | | | | | | 47.4 | < 0.001 |
| Northeast | 20271 | 19.1 | 450 | 24.4 | 19821 | 19.0 | | |
| Midwest | 28973 | 27.4 | 356 | 17.6 | 28617 | 27.6 | | |
| South | 35173 | 33.9 | 638 | 31.1 | 34535 | 34.0 | | |
| West | 26641 | 19.7 | 542 | 26.8 | 26099 | 19.5 | | |
We observed statistically significant subgroup differences among Asian Indians and NHWs except for sex and health insurance (Table 1). Asian Indians were more educated (highest percentage of college education) ($72.9\%$) as compared to NHWs ($34.4\%$). Moreover, Asian Indians were less likely to be current smokers ($5.1\%$ vs. $18.1\%$), more likely to engage in weekly physical activity ($40.6\%$ vs $37.2\%$), less likely to report hypertension ($18\%$ vs $31.8\%$), and more likely to be obese ($46.1\%$ vs $27.4\%$) compared to NHWs. World Health Organization obesity classification for South Asians was used to calculate obesity for Asian Indians (≥25.0) and standard criteria was used for NHWs (≥30.0).
Unadjusted prevalence of diagnosed DM did not differ between Asian Indians and NHWs ($7.5\%$ vs $8.8\%$; $$P \leq .09$$). However, the age-sex adjusted prevalence of diagnosed DM was significantly higher among Asian Indians ($8.3\%$ vs $5.8\%$ for NHWs).
## Association of Asian Indian ethnicity to diagnosed diabetes
Unadjusted odds ratios (UOR) and adjusted odds ratios (AOR) and their associated $95\%$ confidence intervals (CI) from unadjusted and adjusted multivariable logistic regression analyses are summarized in Table 2. Without adjustments for age and sex, Asian Indians were as likely as NHWs to report diagnosed DM (UOR = 0.84, $95\%$ CI: 0.68, 1.03; $$P \leq 0.0958$$). However, age-sex adjusted multivariable logistic regression indicated higher odds of reporting diagnosed DM among Asian Indians compared to NHWs (AOR = 1.39, $95\%$ CI: 1.12, 1.71, $P \leq 0.01$). This association remained strong and significant after controlling for variables such as age, sex, education, employment, poverty status, and other health insurance (AOR = 1.68, $95\%$ CI: 1.36, 2.09, $P \leq 0.01$). In the fully adjusted model, Asian Indians were 1.5 times as likely as NHWs to report DM (AOR = 1.47, $95\%$ CI: 1.16, 1.85, $P \leq 0.05$).
**Table 2**
| Unnamed: 0 | UOR | 95% CI | Prob | Unnamed: 4 |
| --- | --- | --- | --- | --- |
| Model 1—Unadjusted | Model 1—Unadjusted | Model 1—Unadjusted | Model 1—Unadjusted | Model 1—Unadjusted |
| Race/Ethnicity | | | | |
| Asian Indians | 0.84 | [0.68, 1.03] | 0.0958 | |
| NHW (Reference Group) | | | | |
| | AOR | 95% CI | Prob | |
| Model 2—Adjusted for age and sex | Model 2—Adjusted for age and sex | Model 2—Adjusted for age and sex | Model 2—Adjusted for age and sex | Model 2—Adjusted for age and sex |
| Race/Ethnicity | | | | |
| Asian Indians | 1.39 | [1.12, 1.71] | 0.0024 | ** |
| NHW (Reference Group) | | | | |
| Model 3—Adjusted for age, sex, education, employment, poverty status, and health insurance | Model 3—Adjusted for age, sex, education, employment, poverty status, and health insurance | Model 3—Adjusted for age, sex, education, employment, poverty status, and health insurance | Model 3—Adjusted for age, sex, education, employment, poverty status, and health insurance | Model 3—Adjusted for age, sex, education, employment, poverty status, and health insurance |
| Race/Ethnicity | | | | |
| Asian Indians | 1.68 | [1.36, 2.09] | < 0.001 | *** |
| NHW (Reference Group) | | | | |
| Model 4—Adjusted for age, sex, education, employment, poverty status, health insurance, marital status, hypertension, high cholesterol, smoking, physical activity and smoking | Model 4—Adjusted for age, sex, education, employment, poverty status, health insurance, marital status, hypertension, high cholesterol, smoking, physical activity and smoking | Model 4—Adjusted for age, sex, education, employment, poverty status, health insurance, marital status, hypertension, high cholesterol, smoking, physical activity and smoking | Model 4—Adjusted for age, sex, education, employment, poverty status, health insurance, marital status, hypertension, high cholesterol, smoking, physical activity and smoking | Model 4—Adjusted for age, sex, education, employment, poverty status, health insurance, marital status, hypertension, high cholesterol, smoking, physical activity and smoking |
| Race/Ethnicity | | | | |
| Asian Indians | 1.47 | [1.16, 1.85] | 0.0013 | ** |
| NHW (Reference Group) | | | | |
A noteworthy finding was the protective effect of college education and high income on diagnosed DM. Those with less than college education were more likely to be diagnosed with diabetes compared to those with college education (less than high school AOR = 1.29; $95\%$ CI: 1.16, 1.42; high school education AOR = 1.22; $95\%$ CI: 1.22, 1.32; some college education AOR = 1.18; $95\%$ CI: 1.10, 1.28. In addition, adults with income $100\%$ below the federal poverty level were more likely to have diagnosed DM (AOR = 1.51; $95\%$ CI: 1.34, 1.70) compared to those with $400\%$ above federal poverty level.
## Association of race/ethnicity and diagnosed diabetes mellitus
We conducted a secondary analysis of stratified logistic regressions by race/ethnicity on diagnosed DM status. These logistic regressions revealed similar factors (sex, age, hypertension, heart disease, cholesterol, and marital status) among both AIs and NHWs. While higher income and education were significantly associated with decreased odds of diagnosed DM among NHWs, they were not a protective factor for Asian Indians. Further, obese Asian Indians were 2 times as likely to report diagnosed DM as those with normal BMI (AOR = 1.85, $95\%$ CI: 1.06, 3.24).
## Discussion
The overall aim of this study was to estimated age-sex adjusted prevalence of diagnosed DM, using disaggregated NHIS race/ethnicity data, among Asian Indians living in the US. Results indicated diagnosed DM was $8.3\%$ and higher than earlier NHIS years of 1997–2000 ($6.7\%$) [21] that did not adjust for age and sex. However, the observed DM prevalence among Asian *Indians is* much lower than $14\%$ as reported in the DIA study [2010] [22] and $38\%$ age-sex adjusted rate of Asian Indian participants living in Chicago and San Francisco aged 40–84 years of the MASALA (Mediators of Atherosclerosis in South Asians Living in America) study [37]. The significantly higher rates in the DIA study can be explained by the diagnosed and undiagnosed DM used for the prevalence rate; the MASALA study used purposive sampling, older age and participants limited to a few counties in the greater Chicago and San Francisco area. The estimated diagnosed DM rate in this study is also lower than the $12.6\%$ prevalence reported by the National Diabetes Statistics Report (2017–2018) that used multiple data sources. Variation in prevalence rates may also be explained for differing time-period of the data sources (2017–2018 vs 2012–2016) for the National Diabetes Statistics report and the current study as well as the use of non-institutionalized civilian households for estimating diagnosed diabetes [38].
Studies have reported that Asian Indians have higher insulin resistance due to an inherent genetic predisposition, and an increased disease incidence at a younger age and lower body mass index (BMI) [4, 5] Although an estimated 77 million adults in India lived with diagnosed DM in 2019 [39], findings from a cross-sectional multi-level analysis and other research show Asian Indians with a favorable socio-economic profile (i.e. high income and high levels of education) are more likely to report DM compared to other ethnic groups [6–9]. In other words, obesity and sedentary lifestyle predisposes native Indians in India with higher socio-economic status for diabetes [6]. This is contrary to lower risk of DM documented among higher socio-economic status strata in the United States [40].
Asian Indians in the current study had higher socio economic status in the United States, and is consistent with a prior NHIS study [21]. Higher level of socioeconomic status was associated with lower diagnosed DM among Asian Indians. Therefore, life-style factors such as westernized diet, physical inactivity and obesity may play a significant greater role in increasing the risk of DM among Asian Indians in the United States. Raising awareness among the population regarding the deleterious effects of a high-fat, high-carbohydrate, high-calorie diet and encouraging them to continue the more healthy traditional foods could help individuals make healthy dietary choices, helping to reduce the risk of not only type 2 diabetes but related comorbid chronic conditions as well.
Asian Indians had lower rates of smoking and higher levels of exercise as compared to NHWs, which concurs with published studies [21]. Using data estimates from the National Survey on Drug Use and Health, it has been found that, compared to 14 racial/ethnic groups, Asian Indians had one of the lowest rates of tobacco use [41]. Furthermore, obesity increased the odds of a diagnosed DM in both groups. However, estimated obesity rates among Asian Indians were lower than NHWs when the standard criteria were used and elevated with the WHO Asian criteria. Similar observations were also made in the DIA study that reported an obesity rate of 11 and $49.8\%$ using the standard and WHO Asian criteria, respectively [22]. Other studies have also reported similar trends [21, 23].
Studies have documented that Asian Indians may be susceptible to central obesity and higher body fat percentage for the same BMI, in comparison to other groups [42, 43] *There is* also evidence that South Asians often perceive themselves to be normal weight and underestimate their risk for chronic diseases even when they have higher BMI [44]. Acculturation and adaptation of westernized culture has shown to result in changes to ethnic dietary habits, e.g., an increased consumption of refined grains, high energy dense food and sugary drinks and beverages leading to high glycemic load that increase the risk for diabetes [45]. In addition, health outcomes of immigrant tend to decline with longer duration of stay in the US and Asian Indians who do not prioritize preventive health/wellness checks, health diet and exercise or physical activity may be at higher risk for diabetes. Since access to healthcare and English language proficiency is not a barrier for health promotion/ lifestyle interventions, tailored messages to improve awareness of disease risk, health beliefs or perceptions and support for behavior modification can be recommended by healthcare providers for both primary and secondary prevention, especially in 1st generation Asian Indians due to an emphasis on careers and family priorities. These findings suggest that culturally appropriate diabetes prevention programs, shown to be efficacious in preventing diabetes among Asian Indians should be widely disseminated [46, 47].
## Strengths and limitations
The strength of the current study includes is the utilization of disaggregated national data for Asian Indians for multiple years. Use of WHO *Asian criteria* for BMI cut point for Asian Indian and adjustment of a comprehensive list of risk factors in the multivariate model are additional strengths. However, the limitations include its cross-sectional design, a small sample size despite the aggregation of multiple years that did not allow assessment of heterogeneity among the Asian Indian subgroup. Also, our definition of diabetes included only self-reporting of diagnosed diabetes and may underestimate the true prevalence of diabetes in both groups. For example, some persons may have been classified as not having diabetes when in fact they had undiagnosed diabetes; as much as one-quarter to more than one-third of all diabetes may be undiagnosed. Unfortunately, the NHIS data did not allow us to capture undiagnosed diabetes. However, additional analysis on diagnosed diabetes, pre-or borderline diabetes, and no diabetes between the two groups showed no systematic differences in the rates ($$P \leq .224$$). In addition, there was a lack of cross-cultural validity of the physical activity measurement. More specifically, recall bias of specific vigorous physical activities, the description of exercises would vary according to different ethnic groups. Lastly, important confounders such as clinical information, dietary habits, or family history could potentially affect the interpretation and association of Asian Indian ethnicity with diagnosed DM. Family history of DM is a significant risk factor that could not be adjusted as the data was available only in 2016.
## Conclusion
In conclusion, the current study confirmed the higher prevalence of diagnosed DM among Asian Indians as compared to NHWs, despite favorable socio-economic profiles and low smoking rates. The study’s findings also highlighted that among Asian Indians obesity rates were high and linked with diabetes. Findings demonstrate the need for health education and culturally tailored diabetes prevention programs that are critical in preventing DM among Asian Indians living in the US.
## Integrity statement
The primary author, Ranjita *Misra is* the guarantor of this work and takes responsibility for the integrity of the data and the accuracy of the data analysis. The article in the current form is approved by all authors and confirms to the criteria by the International Committee for Medical Journal Editors (ICMJE).
## Prior presentation information
The study was presented at the American Public Health Association Annual Conference.
## References
1. Ogez D, Bourque C-J, Péloquin K, Ribeiro R, Bertout L, Curnier D. **Definition and improvement of the concept and tools of a psychosocial intervention program for parents in pediatric oncology: a mixed-methods feasibility study conducted with parents and healthcare professionals**. *Pilot Feasibility Stud* (2019.0) **5** 20. DOI: 10.1186/s40814-019-0407-8
2. Nanditha A, Ma RC, Ramachandran A, Snehalatha C, Chan JC, Chia KS. **Diabetes in Asia and the Pacific: implications for the global epidemic**. *Diabetes care* (2016.0) **39** 472-85. DOI: 10.2337/dc15-1536
3. Ramachandran A, Snehalatha C, Shetty AS, Nanditha A. **Trends in prevalence of diabetes in Asian countries**. *World J Diabetes* (2012.0) **3** 110-7. DOI: 10.4239/wjd.v3.i6.110
4. Nakagami T, Qiao Q, Carstensen B, Nhr-Hansen C, Hu G, Tuomilehto J. **Age, body mass index and Type 2 diabetes-associations modified by ethnicity**. *Diabetologia* (2003.0) **46** 1063-70. DOI: 10.1007/s00125-003-1158-9
5. Abate N, Chandalia M. **Ethnicity and type 2 diabetes: focus on Asian Indians**. *J Diabetes Complications* (2001.0) **15** 320-7. DOI: 10.1016/s1056-8727(01)00161-1
6. Corsi DJ, Subramanian SV. **Association between socioeconomic status and self-reported diabetes in India: a cross-sectional multilevel analysis**. *BMJ Open* (2012.0) **2** e000895. DOI: 10.1136/bmjopen-2012-000895
7. Ramachandran A, Snehalatha C, Vijay V, King H. **Impact of poverty on the prevalence of diabetes and its complications in urban southern India**. *Diabet Med* (2002.0) **19** 130-5. DOI: 10.1046/j.1464-5491.2002.00656.x
8. Kinra S, Bowen LJ, Lyngdoh T, Prabhakaran D, Reddy KS, Ramakrishnan L. **Sociodemographic patterning of non-communicable disease risk factors in rural India: a cross sectional study**. *BMJ* (2010.0) **341** c4974. DOI: 10.1136/bmj.c4974
9. Mohan V, Mathur P, Deepa R, Deepa M, Shukla DK, Menon GR. **Urban rural differences in prevalence of self-reported diabetes in India—the WHO-ICMR Indian NCD risk factor surveillance**. *Diabetes Res Clin Pract* (2008.0) **80** 159-68. DOI: 10.1016/j.diabres.2007.11.018
10. Montesi L, Caletti MT, Marchesini G. **Diabetes in migrants and ethnic minorities in a changing world**. *World J Diabetes* (2016.0) **7** 34. DOI: 10.4239/wjd.v7.i3.34
11. Sattar N, Gill JM. **Type 2 diabetes in migrant south Asians: mechanisms, mitigation, and management**. *The lancet Diabetes & endocrinology* (2015.0) **3** 1004-16. DOI: 10.1016/S2213-8587(15)00326-5
12. Tillin T, Hughes AD, Godsland IF, Whincup P, Forouhi NG, Welsh P. **Insulin resistance and truncal obesity as important determinants of the greater incidence of diabetes in Indian Asians and African Caribbeans compared with Europeans: the Southall And Brent REvisited (SABRE) cohort**. *Diabetes Care* (2013.0) **36** 383-93. DOI: 10.2337/dc12-0544
13. Zheng Y, Lamoureux EL, Ikram MK, Mitchell P, Wang JJ, Younan C. **Impact of migration and acculturation on prevalence of type 2 diabetes and related eye complications in Indians living in a newly urbanised society**. *PLoS One* (2012.0) **7** e34829-e. DOI: 10.1371/journal.pone.0034829
14. 14Asamnews. https://asamnews.com/2019/05/16/south-asians-grew-40-in-the-u-s-in-just-7-years/2019
15. 15Rangaswamy P. Indian Americans: New York: Chelsea House; 2007 2007. 158 p.
16. 16Indian Immigrants in the United States: Migration Policy Institute; [https://www.migrationpolicy.org/article/indian-immigrants-united-states].
17. Cheng YJ, Kanaya AM, Araneta MRG, Saydah SH, Kahn HS, Gregg EW. **Prevalence of Diabetes by Race and Ethnicity in the United States, 2011–2016**. *Jama* (2019.0) **322** 2389-98. DOI: 10.1001/jama.2019.19365
18. 18Diamant AL, Babey SH, Hastert TA, Brown ER. Diabetes: the growing epidemic. Policy Brief UCLA Cent Health Policy Res. 2007(PB2007-9):1–12.
19. Lee JW, Brancati FL, Yeh HC. **Trends in the prevalence of type 2 diabetes in Asians versus whites: results from the United States National Health Interview Survey, 1997–2008**. *Diabetes Care* (2011.0) **34** 353-7. DOI: 10.2337/dc10-0746
20. McBean AM, Li S, Gilbertson DT, Collins AJ. **Differences in diabetes prevalence, incidence, and mortality among the elderly of four racial/ethnic groups: whites, blacks, hispanics, and asians**. *Diabetes Care* (2004.0) **27** 2317-24. DOI: 10.2337/diacare.27.10.2317
21. Mohanty SA, Woolhandler S, Himmelstein DU, Bor DH. **Diabetes and cardiovascular disease among Asian Indians in the United States**. *J Gen Intern Med* (2005.0) **20** 474-8. DOI: 10.1111/j.1525-1497.2005.40294.x
22. Misra R, Patel T, Kotha P, Raji A, Ganda O, Banerji M. **Prevalence of diabetes, metabolic syndrome, and cardiovascular risk factors in US Asian Indians: results from a national study**. *J Diabetes Complications* (2010.0) **24** 145-53. DOI: 10.1016/j.jdiacomp.2009.01.003
23. Thomas A, Ashcraft A. **Type 2 Diabetes Risk among Asian Indians in the US: A Pilot Study**. *Nurs Res Pract* (2013.0) **2013** 492893. DOI: 10.1155/2013/492893
24. Chandalia M, Lin P, Seenivasan T, Livingston EH, Snell PG, Grundy SM. **Insulin resistance and body fat distribution in South Asian men compared to Caucasian men**. *PLoS One* (2007.0) **2**. DOI: 10.1371/journal.pone.0000812
25. Venkataraman R, Nanda NC, Baweja G, Parikh N, Bhatia V. **Prevalence of diabetes mellitus and related conditions in Asian Indians living in the United States**. *Am J Cardiol* (2004.0) **94** 977-80. DOI: 10.1016/j.amjcard.2004.06.048
26. Oza-Frank R, Ali MK, Vaccarino V, Narayan KMV. **Asian Americans: diabetes prevalence across U.S. and World Health Organization weight classifications**. *Diabetes care* (2009.0) **32** 1644-6. DOI: 10.2337/dc09-0573
27. Who EC. **Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies**. *Lancet (London, England)* (2004.0) **363** 157. DOI: 10.1016/S0140-6736(03)15268-3
28. 28Center PR. The rise of Asian Americans. Chapter 1: Portrait of Asian Americans [https://www.pewsocialtrends.org/2012/06/19/chapter-1-portrait-of-asian-americans/].
29. 29Lewis T, editor Estimation Strategies Involving Pooled Survey Data. Proceedings of the SAS Global Forum 2017 Conference; 2017.
30. Guerriere D, Husain A, Zagorski B, Marshall D, Seow H, Brazil K. **Predictors of caregiver burden across the home-based palliative care trajectory in Ontario, Canada**. *Health & Social Care In The Community* (2016.0) **24** 428-38. DOI: 10.1111/hsc.12219
31. Alhussain K, Meraya AM, Sambamoorthi U. **Serious Psychological Distress and Emergency Room Use among Adults with Multimorbidity in the United States**. *Psychiatry J* (2017.0) **2017** 8565186. PMID: 29085831
32. 32Centers for Disease Control and Prevention. NHIS—Race and Hispanic Origin Information. [https://www.cdc.gov/nchs/nhis/rhoi/rhoi_history.htm].
33. Aziz N, Kallur SD, Nirmalan PK. **Implications of the revised consensus body mass indices for asian indians on clinical obstetric practice**. *J Clin Diagn Res* (2014.0) **8** Oc01-3. DOI: 10.7860/JCDR/2014/8062.4212
34. Applebaum AJ, Panjwani AA, Buda K, O’Toole MS, Hoyt MA, Garcia A. **Emotion regulation therapy for cancer caregivers-an open trial of a mechanism-targeted approach to addressing caregiver distress**. *Transl Behav Med* (2018.0). DOI: 10.1093/tbm/iby104
35. 35Organization WH. Regional office for the Western Pacific. The Asia-Pacific perspective: redefining obesity and its treatment Sydney: Health Communications Australia. 2000:11–2.
36. 36Bureau U. Age and sex composition in the United States. 2010. The United States Census Bureau, (https://www.census.gov/data/tables/2018/demo/age-and-sex/2018-age-sex-composition.html) [Google Scholar]. 2018.
37. Gujral UP, Narayan KV, Pradeepa RG, Deepa M, Ali MK, Anjana RM. **Comparing type 2 diabetes, prediabetes, and their associated risk factors in Asian Indians in India and in the US: the CARRS and MASALA studies**. *Diabetes care* (2015.0) **38** 1312-8. PMID: 25877810
38. McKenzie H, White K, Hayes L, Fitzpatrick S, Cox K, River J. **’Shadowing’ as a management strategy for chemotherapy outpatient primary support persons**. *Scandinavian Journal of Caring Sciences* (2017.0) **31** 887-94. DOI: 10.1111/scs.12410
39. 39International diabetes federation atlas 2019 [https://www.diabetesatlas.org/en/].
40. Saydah S, Lochner K. **Socioeconomic status and risk of diabetes-related mortality in the U.S**. *Public Health Rep* (2010.0) **125** 377-88. DOI: 10.1177/003335491012500306
41. **Prevalence of cigarette use among 14 racial/ethnic populations—United States, 1999–2001**. *MMWR Morb Mortal Wkly Rep* (2004.0) **53** 49-52. PMID: 14749612
42. Bajaj HS, Pereira MA, Anjana RM, Deepa R, Mohan V, Mueller NT. **Comparison of relative waist circumference between Asian Indian and US adults**. *J Obes* (2014.0) **2014**. DOI: 10.1155/2014/461956
43. Rush E, Plank L, Chandu V, Laulu M, Simmons D, Swinburn B. **Body size, body composition, and fat distribution: a comparison of young New Zealand men of European, Pacific Island, and Asian Indian ethnicities**. *N Z Med J* (2004.0) **117** U1203. PMID: 15608799
44. Tang JW, Mason M, Kushner RF, Tirodkar MA, Khurana N, Kandula NR. **Peer reviewed: South Asian American perspectives on overweight, obesity, and the relationship between weight and health**. *Prev Chronic Dis* (2012.0) **9**
45. Shobana S, Ramya MB, Sudha V, Unnikrishnan R, Pradeepa R, Anjana R. **Nutrition and its link with diabetes in asian indians: Challenges and solutions**. *Proceedings of the Indian National Science Academy* (2018.0) **84** 955-63
46. Patel RM, Misra R, Raj S, Balasubramanyam A. **Effectiveness of a Group-Based Culturally Tailored Lifestyle Intervention Program on Changes in Risk Factors for Type 2 Diabetes among Asian Indians in the United States**. *J Diabetes Res* (2017.0) **2017** 2751980. DOI: 10.1155/2017/2751980
47. Islam NS, Zanowiak JM, Wyatt LC, Kavathe R, Singh H, Kwon SC. **Diabetes prevention in the New York City Sikh Asian Indian community: a pilot study**. *Int J Environ Res Public Health* (2014.0) **11** 5462-86. DOI: 10.3390/ijerph110505462
|
---
title: 'Prevalence of, and risk factors for, diabetes and prediabetes in Bangladesh:
Evidence from the national survey using a multilevel Poisson regression model with
a robust variance'
authors:
- Mohammad Bellal Hossain
- Md. Nuruzzaman Khan
- John C. Oldroyd
- Juwel Rana
- Dianna J. Magliago
- Enayet K. Chowdhury
- Md Nazmul Karim
- Rakibul M. Islam
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021925
doi: 10.1371/journal.pgph.0000461
license: CC BY 4.0
---
# Prevalence of, and risk factors for, diabetes and prediabetes in Bangladesh: Evidence from the national survey using a multilevel Poisson regression model with a robust variance
## Abstract
To estimate the age-standardized prevalence of diabetes and prediabetes and identify factors associated with these conditions at individual, household, and community levels. Data from 11952 Bangladeshi adults aged 18–95 years available from the most recent Bangladesh Demographic and Health Survey 2017–18 were used. Anthropometric measurements and fasting blood glucose samples were taken as part of the survey. Prevalence estimates of diabetes and prediabetes were age-standardized with direct standardization, and risk factors were identified using multilevel mix-effects Poisson regression models with robust variance. The overall age-standardised prevalence of diabetes was $9.2\%$ ($95\%$CI 8.7–9.7) (men: $8.8\%$, women: $9.6\%$), and prediabetes was $13.3\%$ ($95\%$CI 12.7–13.9) (men: $13.0\%$, women: $13.6\%$). Among people with diabetes, $61.5\%$ were unaware that they had the condition. $35.2\%$ took treatment regularly, and only $30.4\%$ of them had controlled diabetes. Factors associated with an increased prevalence of having diabetes were increasing age, male, overweight/obesity, hypertension, being in the highest wealth quintile, and living in the Dhaka division. People currently employed and living in the Rangpur division were less likely to have diabetes than those currently not employed and living in the Barishal division. Diabetes and prediabetes affect a substantial proportion (over one-quarter) of the Bangladeshi adult population. Continuing surveillance and effective prevention and control measures, focusing on obesity reduction and hypertension management, are urgently needed.
## Introduction
Diabetes mellitus remains a significant contributor to the global burden of disease [1]. People with diabetes have an increased risk of developing several serious life-threatening micro-and macro-vascular complications resulting in higher medical care costs, reduced quality of life, and increased mortality [2]. The International Diabetes Federation (IDF) has estimated that 463 million adults live with diabetes worldwide in 2019, with a projected increase to 700 million by 2045 [3]. Seventy-nine percent of those with diabetes live in low- and middle-income countries (LMICs) [4]. It is projected that diabetes cases will increase by $74\%$ in Southeast Asian countries in the next two decades, from 88 million in 2019 to 153 million by 2045 [4].
In Bangladesh, 8.4 million adults lived with diabetes in 2019 and projected to be almost double (15.0 million) by 2045 [4]. Studies, including a systematic review and meta-analysis, and national survey reports, showed that the prevalence of diabetes among adults had increased substantially in Bangladesh, from ~$5\%$ in 2001 to ~$14\%$ in 2017 [5–8]. Several studies were also conducted using BDHS 2011 data, showing that people with older age, overweight/obesity, hypertension, and higher socioeconomic status (e.g., education level and wealth status) were associated with increasing likelihood of diabetes [9–12]. However, these studies are relatively old and overlooked the prevalence of, and risk factors for, prediabetes. In 2019, it was estimated that 3.8 million people had prediabetes in Bangladesh, which is a major challenge for the health system in Bangladesh when to with existing cases of diabetes [4].
Thus, it is necessary to study the age-standardized prevalence of and risk factors for diabetes and prediabetes in Bangladeshi adults using the latest Bangladesh Demographic and Health Survey (BDHS) 2017–18. To our knowledge, Kibria [9] has published an article using BDHS 2017–18 with significant flaws, including an inconsistent sample size compared to the BDHS 2017–2018 published report and inappropriate age-standardization. Besides, Kibria [9] did not estimate prediabetes and estimated the odds ratio for diabetes using logistic regression, which is not robust when the outcome is common [13, 14]. Based on these limitations, we aimed to estimate the age-standardized prevalence of diabetes and prediabetes in Bangladeshi adults aged 18 years and older using the latest BDHS. We also investigated factors associated with diabetes and prediabetes in Bangladeshi adults using a multilevel Poisson regression model with robust variance. Results are examined in detail according to the individual, household, and community-level characteristics.
## Study population and data collection
Bangladesh, a Southeast Asian country, currently has a 111 million population aged 18 years and older [15]. In 2017–18, the National Institute of Population Research and Training (NIPORT), the Ministry of Health and Family Welfare, Bangladesh, conducted the second BDHS survey of its kind that collected data on blood pressure, fasting blood glucose (FBG) biomarker measurements, and relevant information in addition to socio-demographic characteristics [6].
The BDHS is a nationally representative survey conducted using a two-stage stratified sample of households, including strata for rural and urban areas. Detailed survey sampling and the data collection procedure have been published in the BDHS survey report [6]. The primary sampling units (PSUs), each containing 120 households on average, were taken from the most recent 2011 *Bangladesh census* enumeration areas. In BDHS 2017–18, a total of 675 PSUs was selected with probability proportional to PSU size; however, 672 PSUs were included (three PSUs were not sampled due to flooding), of which 192 and 480 were from urban and rural areas, respectively. In the second stage, 30 households per PSU were selected using systematic random sampling to provide statistically reliable estimates of health outcomes for the country as a whole for each of the eight divisions and urban and rural areas separately. Of the 20,160 selected households, interviews were completed in 19,457 households with an overall $96.5\%$ household response rate [6]. Of these, one-fourth of the households [4864] were selected for the collection of biomarkers. A total of 14,704 (8013 women, 6691 men) respondents aged 18+ were available in the 4864 selected households for blood glucose measurement. However, 12,100 (6919 women, 5181 men) respondents aged 18 years and older had their blood glucose tested ($82.3\%$ response rate) (Fig 1).
**Fig 1:** *Schematic representation of the sampling procedure of the Bangladesh Demographic and Health Survey, 2017–18.*
## Analytic sample
Of the respondents who had their blood glucose tested, we excluded those for whom body mass index data were missing ($$n = 143$$) and those who were pregnant at the time of blood glucose measurement ($$n = 7$$). After exclusion, we had 11,952 respondents who had their blood glucose tested which was our analytical sample for analysing the prevalence and risk factors of diabetes. However, in the subsequent analyses of pre-diabetes prevalence and risk factors, respondents having diabetes ($$n = 1$$,174) were excluded, and the data of the remaining 10,779 respondents were analysed.
## Outcome: Diabetes and prediabetes
The primary outcomes in this study were diabetes and prediabetes calculated based on the fasting plasma glucose (FPG) level [6]. The HemoCue Glucose 201 DM system with plasma conversion was used to test a drop of capillary blood obtained from consenting eligible respondents from the middle or ring finger. The system automatically converted the fasting whole blood glucose measurements taken in the survey to FPG equivalent values. The respondents were asked not to eat or drink anything other than plain water for at least 8 hours before testing. The details of blood sample collection have been described in the BDHS survey report [6]. The World Health Organization (WHO) criteria for diabetes and prediabetes classification were used [16]. The variable diabetes included respondents with diabetes defined as FPG level greater than or equal to 7.0 mmol/L or those who reported using medication for diabetes. Respondents without diabetes were those with FPG levels less than 7.0 mmol/L and not taking any diabetes controlling medication. The variable prediabetes included respondents with prediabetes defined as FPG levels between 6.1 mmol/L and 6.9 mmol/L and not taking any diabetes controlling medication. Respondents without prediabetes were those with FPG levels less than 6.1 mmol/L and not taking diabetes controlling medication [16].
## Explanatory variables
Three different levels of explanatory variables of diabetes and prediabetes were identified through a comprehensive review of the literature [5, 9, 10, 17]. Individual-level factors included were participants’ age, sex, BMI, educational level, working status, and hypertension. The BMI was categorized based on Asian cut-off as suggested by the WHO expert consultation due to the high risk of type 2 diabetes and cardiovascular disease in Asian people at lower BMIs than the existing WHO cut-off [18]. The presence of hypertension was defined as a systolic blood pressure ≥ 140 mmHg and/or a diastolic blood pressure ≥ 90 mmHg, and/or currently on treatment with antihypertensive medication [19]. The household wealth quintile (lowest to highest) was the household-level factor. It was derived from the household wealth index reported in the BDHS, which was constructed using principal component analysis of household’s durable and non-durable assets (e.g., televisions, bicycles, sources of drinking water, sanitation facilities, and construction materials) [20]. Community-level factors included were the place of residence and administrative divisions of the country.
The literature review also identified family history of diabetes, mass media exposure, cigarette smoking, alcohol consumption, sleep duration, diet, and physical activity as significant behavioral and lifestyle-related individual-level risk factors for diabetes and pre-diabetes. However, this study could not include these variables in the analyses as they were not included in the original survey.
## Statistical analysis
The crude prevalence of diabetes and prediabetes were estimated, allowing for the complex survey design and survey sampling weights. To account for different age distributions between groups and over time, we age-standardized estimates to the 2011 Census population of Bangladesh using the direct method, with age categories of 18–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64 and ≥65 years. Differences between continuous and categorical variables were tested using the Mann-Whitney and chi-square tests, respectively.
We used a multilevel Poisson regression model with a robust variance to identify factors associated with diabetes and prediabetes, and the results were presented as a prevalence ratio (PR) with a $95\%$ confidence interval (CI). We used this model since the odds ratio estimated using logistic regression from a cross-sectional study may significantly overestimate relative risk when the outcome is common [13, 14]. Secondly, in the case of convergence failure with the log-binomial model, Poisson regression with a robust variance performs better in estimating the prevalence ratio from a cross-sectional study [21]. Furthermore, in the BDHS, individuals were nested within the household; households were nested within the PSU/cluster. Hence, our multilevel mixed-effects Poisson regression model accounts for these multiple hierarchies and dependency in data and the problem of overestimation [22].
With progressive model-building techniques, four models were run for diabetes, and separately four models for pre-diabetes, each introducing different confounding factors at the individual, household, and community levels. Model 1 was run without confounding factors to determine the cluster level variation of diabetes and pre-diabetes in Bangladesh. Model 2 and 3 were adjusted for individual, and individual plus household level factors, respectively. Model 4 was the final model that included individual, household, and community-level factors simultaneously. The Intra-Class Correlation (ICC), Akaike Information Criteria (AIC), and Bayesian Information Criteria (BIC) were used to assess model performance. All statistical tests were two-sided, and a p-value < 0·05 was considered statistically significant. The study was designed and reported following strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [23]. All analyses were performed using statistical software packages Stata (version 15·10; Stata Corp LP, College Station, Texas).
## Ethical approval
This study is a secondary analysis of publicly available household survey data. Thus, we did not require any ethical approval for this study. However, institutional review boards (IRBs) at ICF and the Bangladesh Medical Research Council (BMRC) approved the survey methodology, biomarker measurements, and other survey instruments. In addition, the BDHS 2017–18 datasets are publicly available (https://dhsprogram.com/methodology/survey/survey-display-536.cfm), and we received authorization from the DHS to use the datasets.
## Results
Of 12,100 participants who provided FBG, 11,952 were included in the analysis (Fig 1). The median (IQR) age of the participants was 36 [24] years (Table 1). Of the study participants, $57.1\%$ [6,826] were female, $26.6\%$ [3,179] lived in urban areas, $40.1\%$ [4,794] were overweight/obese and $27.4\%$ [3,274] had hypertension. People with diabetes were significantly older than those without ($p \leq 0.001$), they were significantly more likely to be overweight or obese ($p \leq 0.001$), and more likely to be hypertensive ($p \leq 0.001$). Furthermore, people with diabetes were more likely to live in urban areas and come from a household with higher wealth quintiles ($p \leq 0.001$) (Table 1).
**Table 1**
| Characteristics | N (%)a | N (%)a.1 | N (%)a.2 | P-valueb |
| --- | --- | --- | --- | --- |
| Characteristics | All (N = 11,952) | No. without diabetes (N = 10,778) | No. with diabetes (N = 1,174) | P-valueb |
| Individual level | Individual level | Individual level | Individual level | Individual level |
| Age in years, Median (IQR) | 39 (24) | 35 (23) | 46 (23) | <0.001 |
| 18–34 | 5,391 (45.1) | 5,114 (47.4) | 277 (23.6) | |
| 35–39 | 1,371 (11.5) | 1,233 (11.4) | 138 (11.8) | |
| 40–44 | 1,047 (8.7) | 916 (8.5) | 131 (11.2) | |
| 45–49 | 994 (8.3) | 861 (8.0) | 133 (11.3) | |
| 50–54 | 672 (5.6) | 556 (5.2) | 115 (9.8) | |
| 55–59 | 676 (5.7) | 577 (5.4) | 100 (8.5) | |
| 60–64 | 675 (5.7) | 563 (5.2) | 112 (9.5) | |
| ≥65 | 1,126 (9.42) | 958 (8.9) | 168 (14.3) | |
| Sex | | | | |
| Men | 5,126 (42.9) | 4,589 (42.6) | 537 (45.8) | 0.550 |
| Women | 6,826 (57.1) | 6,189 (57.4) | 637 (54.2) | 0.550 |
| Body Mass Index (kg/m 2 ) | | | | |
| Underweight (<18.5) | 2,067 (17.3) | 1,938 (18.0) | 129 (11.0) | <0.001 |
| Normal weight (18.5–23.0) | 5091 (42.6) | 4696 (43.6) | 395 (33.7) | <0.001 |
| Overweight (23.0–27.5) | 3532 (29.6) | 3107 (28.8) | 425 (36.2) | <0.001 |
| Obese (≥27.5) | 1262 (10.5) | 1037 (9.6) | 225 (19.1) | <0.001 |
| Level of education | | | | |
| No education/preschool | 3,033 (25.4) | 2,736 (25.4) | 296 (25.2) | 0.937 |
| Primary education | 3,590 (30.0) | 3,225 (29.9) | 365 (31.1) | 0.937 |
| Secondary education | 3,539 (29.6) | 3,203 (29.7) | 337 (28.7) | 0.937 |
| Higher education | 1,790 (15.0) | 1,614 (15.0) | 176 (15.0) | 0.937 |
| Currently working | | | | |
| Yes | 4,621 (38.7) | 6,695 (62.1) | 636 (54.1) | <0.001 |
| No | 7,331 (61.3) | 4,083 (37.9) | 538 (45.9) | <0.001 |
| Hypertension | | | | |
| Yes | 3,416 (28.6) | 2,865 (26.6) | 551 (46.9) | <0.001 |
| No | 8,536 (71.4) | 7,913 (73.4) | 623 (53.1) | <0.001 |
| Household level | Household level | Household level | Household level | Household level |
| Wealth quintile | | | | |
| Lowest | 2,311 (19.3) | 2,183 (20.3) | 128 (10.9) | <0.001 |
| Second | 2,354 (19.7) | 2,213 (20.5) | 141 (12.0) | <0.001 |
| Middle | 2,466 (20.6) | 2,272 (21.1) | 194 (16.5) | <0.001 |
| Fourth | 2,377 (19.9) | 2,111 (19.6) | 266 (22.7) | <0.001 |
| Highest | 2,444 (20.5) | 1,999 (18.5) | 445 (37.9) | <0.001 |
| Community level | Community level | Community level | Community level | Community level |
| Place of residence | | | | |
| Urban | 3,179 (26.6) | 2,764 (25.6) | 416 (35.4) | <0.001 |
| Rural | 8,773 (73.4) | 8,014 (74.4) | 758 (64.6) | <0.001 |
| Administrative division | | | | |
| Barishal | 660 (5.5) | 597 (5.5) | 62 (5.3) | <0.001 |
| Chattogram | 2,053 (17.2) | 1,828 (17.0) | 224 (19.1) | <0.001 |
| Dhaka | 2,767 (23.2) | 2,373 (22.0) | 394 (33.6) | <0.001 |
| Khulna | 1,489 (12.5) | 1,368 (12.7) | 121 (10.3) | <0.001 |
| Mymensingh | 973 (8.1) | 897 (8.3) | 77 (6.5) | <0.001 |
| Rajshahi | 1,727 (14.4) | 1,590 (14.8) | 138 (11.7) | <0.001 |
| Rangpur | 1,503 (12.6) | 1,419 (13.2) | 84 (7.2) | <0.001 |
| Sylhet | 780 (6.5) | 706 (6.5) | 74 (6.3) | <0.001 |
The crude and age-standardized prevalence of diabetes and prediabetes by individual, household, and community-level characteristics are presented in Table 2. The overall age-standardized prevalence of diabetes was $9.2\%$ ($95\%$CI, 8.7–$9.7\%$) with comparable estimates for men: $8.8\%$, $95\%$CI 8.1–9.6, and women: $9.6\%$, $95\%$CI 8.9–10.3. The age-standardized diabetes prevalence was higher in urban ($11.8\%$, $95\%$CI 10.9–12.7) than in rural residents ($7.9\%$, $95\%$CI 7.3–8.5). Prevalence of diabetes was highest in people who were obese ($18.4\%$, $95\%$CI 16.3–20.5), hypertensive ($13.7\%$, $95\%$CI 12.3–15.0), in the highest wealth quintile ($16.5\%$, $95\%$CI 15.9–17.9), and living in the Dhaka division ($15.0\%$, $95\%$CI 13.3–16.7) compared to respective reference categories.
**Table 2**
| Unnamed: 0 | Diabetes, % (95% CI) | Diabetes, % (95% CI).1 | Pre-diabetes, % (95% CI) | Pre-diabetes, % (95% CI).1 |
| --- | --- | --- | --- | --- |
| | Crude prevalence | Standardized prevalence | Crude prevalence | Standardized prevalence |
| Overall | 9.8 (9.1–10.6) | 9.2 (8.7–9.7) | 14.2 (13.3–15.1) | 13.3 (12.7–13.9) |
| Individual level | Individual level | Individual level | Individual level | Individual level |
| Age (years) | | | | |
| 18–34 | 5.1 (4.4–5.9) | 5.1 (4.53–5.70) | 12.7 (11.5–14.0) | 11.5 (10.6–12.4) |
| 35–39 | 10.0 (8.4–12.0) | 10.0 (8.4–11.5) | 15.6 (13.5–17.9) | 15.2 (13.3–17.1) |
| 40–44 | 12.5 (10.4–15.1) | 11.8 (9.8–13.7) | 15.6 (13.1–18.5) | 14.9 (12.7–17.1) |
| 45–49 | 13.4 (11.2–15.9) | 12.9 (10.8–14.9) | 15.7 (13.4–18.3) | 15.4 (13.2–17.6) |
| 50–54 | 17.2 (14.2–20.7) | 16.4 (13.6–19.2) | 14.3 (11.6–17.6) | 14.7 (12.1–17.4) |
| 55–59 | 14.8 (12.0–18.0) | 15.9 (13.2–18.6) | 15.7 (12.9–18.9) | 15.3 (12.6–18.0) |
| 60–64 | 16.5 (13.7–19.8) | 15.3 (12.6–18.0) | 15.7 (12.8–19.2) | 15.4 (12.7–18.1) |
| ≥65 | 14.9 (12.6–17.5) | 14.8 (12.8–16.9) | 14.9 (12.5–17.7) | 14.7 (12.7–16.8) |
| Sex | | | | |
| Men | 10.5 (9.5–11.5) | 8.8 (8.1–9.6) | 14.0 (12.8–15.2) | 13.0 (12.1–13.9) |
| Women | 9.3 (8.5–10.2) | 9.6 (8.9–10.3) | 14.4 (13.3–15.5) | 13.6 (12.8–14.5) |
| Body Mass Index (kg/m 2 ) | | | | |
| Underweight (<18.5) | 6.2 (5.0–7.6) | 5.3 (4.3–6.3) | 12.5 (10.8–14.4) | 11.9 (10.4–13.3) |
| Normal weight (18.5–23.0) | 7.8 (6.9–8.7) | 7.3 (6.6–8.0) | 12.9 (11.7–14.1) | 12.2 (11.3–13.1) |
| Overweight (23.0–27.5) | 12.0 (10.8–13.4) | 11.6 (10.5–12.6) | 15.3 (13.8–16.8) | 14.1 (12.9–15.2) |
| Obese (≥27.5) | 17.8 (15.6–20.3) | 18.4 (16.3–20.5) | 19.4 (16.9–22.0) | 18.5 (16.3–20.7) |
| Level of education | | | | |
| No education, preschool | 9.7 (8.6–11.1) | 6.9 (8.4–10.5) | 14.8 (13.2–16.6) | 13.0 (11.3–14.8) |
| Primary | 10.2 (9.0–11.5) | 9.3 (8.8–10.7) | 13.4 (12.1–14.8) | 12.7 (11.6–13.8) |
| Secondary | 9.5 (8.4–10.7) | 11.7 (8.6–10.5) | 14.3 (12.8–15.8) | 13.6 (12.4–14.8) |
| Higher education | 9.8 (8.3–11.6) | 13.1 (8.6–11.2) | 14.4 (12.5–16.5) | 15.7 (13.9–17.5) |
| Currently working | | | | |
| Yes | 8.7 (7.9–9.5) | 8.1 (7.5–8.7) | 13.5 (12.4–14.5) | 12.5 (11.8–13.3) |
| No | 11.6 (10.5–12.9) | 11.8 (10.9–12.8) | 15.3 (14.0–16.8) | 14.6 (13.6–15.7) |
| Hypertension | | | | |
| Yes | 16.4 (15.0–18.0) | 13.7 (12.3–15.0) | 14.9 (13.5–16.5) | 13.4 (12.0–14.8) |
| No | 7.3 (6.6–8.1) | 7.4 (6.8–8.0) | 13.9 (12.9–15.0) | 13.2 (12.5–13.9) |
| Household level | Household level | Household level | Household level | Household level |
| Wealth quintile | | | | |
| Lowest | 5.5 (4.4–6.9) | 5.3 (4.4–6.2) | 11.8 (10.1–13.8) | 11.5 (10.2–12.8) |
| Second | 6.0 (5.0–7.2) | 5.6 (4.7–6.6) | 10.3 (8.8–11.9) | 10.0 (8.8–11.2) |
| Middle | 7.8 (6.7–9.2) | 7.6 (6.5–8.6) | 11.7 (10.2–13.6) | 11.2 (9.8–12.3) |
| Fourth | 11.2 (9.7–12.9) | 10.4 (9.1–11.6) | 16.2 (14.4–18.2) | 14.6 (13.2–16.1) |
| Highest | 18.2 (16.5–20.1) | 16.5 (15.9–17.9) | 20.6 (18.7–22.7) | 18.6 (17.1–20.1) |
| Community level | Community level | Community level | Community level | Community level |
| Place of residence | | | | |
| Urban | 13.1 (11.8–14.5) | 11.8 (10.9–12.7) | 17.7 (16.0–19.6) | 15.1 (14.0–16.2) |
| Rural | 8.6 (7.8–9.6) | 7.9 (7.3–8.5) | 12.9 (11.8–14.0) | 12.3 (11.6–13.0) |
| Administrative division | | | | |
| Barishal | 9.4 (7.5–11.8) | 9.2 (7.7–10.8) | 16.3 (13.6–19.5) | 16.6 (14.5–18.7) |
| Chattogram | 10.9 (9.0–13.2) | 10.9 (9.4–12.3) | 14.6 (12.7–16.8) | 14.4 (12.7–16.1) |
| Dhaka | 14.2 (12.3–16.5) | 15.0 (13.3–16.7) | 22.0 (19.3–25.1) | 22.8 (20.7–24.8) |
| Khulna | 8.2 (6.8–9.7) | 8.0 (6.8–9.2) | 10.6 (8.8–12.9) | 10.0 (8.6–11.5) |
| Mymensingh | 7.9 (6.3–9.9) | 7.8 (6.3–9.2) | 12.2 (10.5–14.2) | 12.3 (10.5–14.0) |
| Rajshahi | 8.0 (6.4–9.8) | 8.2 (6.9–9.5) | 9.7 (7.8–11.9) | 9.3 (7.9–10.7) |
| Rangpur | 5.6 (4.4–7.0) | 5.6 (4.5–6.7) | 9.0 (7.3–11.1) | 9.1 (7.7–10.6) |
| Sylhet | 9.4 (7.3–12.2) | 9.6 (8.2–11.1) | 12.6 (10.5–15.0) | 12.6 (10.8–14.3) |
The overall age-standardized prevalence of prediabetes was $13.3\%$ ($95\%$CI, 12.7–13.9), with a similar prevalence in women ($13.6\%$, $95\%$CI 12.8–14.5) and in men ($13.0\%$, $95\%$CI 12.1–13.9). The age-standardized prevalence estimates of prediabetes were higher in people who were obese ($18.5\%$, $95\%$CI 16.3–20.7), in the highest wealth quintile ($18.6\%$, $95\%$CI 17.1–20.1), and were living in the Dhaka division ($22.8\%$, $95\%$CI 20.7–24.8) (Table 2). Three out of five ($61.5\%$, $95\%$CI 57.9–64.9) people living with diabetes were unaware of their condition. One-third ($35.2\%$, $95\%$CI 32.0–38.5) received appropriate treatment, and only $30.4\%$ ($95\%$CI 26.0–35.2) of them had controlled diabetes.
Each of the four mixed-effects multilevel Poisson models was run to identify factors associated with diabetes and prediabetes. We compared intra-class correlation (ICC), Akaike’s information criterion (AIC), and Bayesian information criterion (BIC) to select the best fitting model: the preferred model having the smallest ICC, AIC and BIC. According to these indicators, Model 4 (including individual, household, and community-level factors) had the best fitting model. Model 1 (crude) produced an ICC of $30.52\%$ and $42.54\%$ for diabetes and prediabetes, respectively (Table 3). This result indicates the degree of the variance seen across clusters without taking other factors into account. However, the ICC was reduced to $12.42\%$ for diabetes and $14.89\%$ for prediabetes once individual, household, and community-level factors were included in the final model.
**Table 3**
| Random effects (measure of variation for diabetes)a | Diabetes± | Diabetes±.1 | Diabetes±.2 | Diabetes±.3 | Prediabetes± | Prediabetes±.1 | Prediabetes±.2 | Prediabetes±.3 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Random effects (measure of variation for diabetes)a | Model 1 | Model 2 | Model 3 | Model 4 | Model 1 | Model 2 | Model 3 | Model 4 |
| Cluster-level variance (SE)b | 0.31 (0.04) | 0.23 (0.04) | 0.16 (0.04) | 0.13 (0.04) | 0.28 (0.03) | 0.26 (0.03) | 0.21 (0.03) | 0.15 (0.03) |
| Intra-class correlation (ICC, %) | 30.52% | 22.57% | 15.53% | 12.47% | 42.54% | 25.27% | 20.54% | 14.89% |
| Explained variance (PCV)(%) | Reference | 25.81% | 48.38% | 61.29% | Reference | 7.14% | 25.00% | 46.42% |
| Model summary | Model summary | Model summary | Model summary | Model summary | Model summary | Model summary | Model summary | Model summary |
| AIC | 7695.98 | 7264.89 | 7176.57 | 7156.65 | 9857.34 | 9849.54 | 9800.16 | 9733.91 |
| BIC | 7710.79 | 7398.19 | 7339.49 | 7378.80 | 9872.15 | 99.82.83 | 9963.07 | 9956.06 |
The results for model 4 are shown in Table 4, while the results for all other models are shown in S1 Table. The final model showed that diabetes was associated with age, sex, BMI, employment status, hypertension, wealth quintile, and administrative division of the country but not with the place of residence (urban /rural) or level of education. Compared with individuals aged 18 to 34 years, individuals aged 40 to 49 years were over two times more likely to have diabetes, while individuals aged ≥50 years were about three times more likely to have diabetes (Table 4). Men were more likely to have diabetes (PR 1.17, $95\%$CI 1.01–1.36) than women. Diabetes was significantly positively associated with being overweight (PR 1.23, $95\%$CI 1.06–1.43) or obese (PR 1.45, $95\%$CI 1.21–1.75) compared with normal weight, being hypertensive (PR 1.47, $95\%$CI 1.30–1.68) compared with normotensive, belonging to either of the fourth (PR 1.60, $95\%$CI 1.23–2.09); or the highest wealth quintile (PR 2.21, $95\%$CI 1.70–2.86), compared with the lowest quintile, and living in the Dhaka division (PR 1.32, $95\%$CI 1.02–1.71) compared with the Barishal division. Individuals currently employed (PR 0.81, $95\%$CI 0.69–0.94) and living in the Rangpur division (PR 0.67, $95\%$CI 0.50–0.91) were less likely to have diabetes than being employed and living in the Barishal division.
**Table 4**
| Characteristics | Diabetes, PR (95% CI)± | Pre-diabetes, PR (95% CI)± |
| --- | --- | --- |
| Individual level | Individual level | Individual level |
| Age in years, (ref: 18–34) | | |
| 35–39 | 1.80 (1.45–2.25) | 1.19 (1.01–1.41) |
| 40–44 | 2.37 (1.87–3.00) | 1.22 (1.01–1.48) |
| 45–49 | 2.39 (1.89–3.01) | 1.22 (1.02–1.45) |
| 50–54 | 3.16 (2.45–4.06) | 1.12 (0.90–1.41) |
| 55–59 | 2.70 (2.13–3.43) | 1.26 (1.01–1.58) |
| 60–64 | 3.11 (2.46–3.92) | 1.26 (1.00–1.59) |
| ≥65 | 2.77 (2.16–3.55) | 1.21 (0.99–1.49) |
| Sex, (ref: women) | 1.17 (1.01–1.36) | 0.98 (0.87–1.10) |
| Body Mass Index (kg/m 2 ), (ref: normal weight) | | |
| Underweight (<18.5) | 0.83 (0.67–1.03) | 0.97 (0.83–1.13) |
| Overweight (23.0–27.5) | 1.23 (1.06–1.43) | 1.07 (0.95–1.21) |
| Obese (≥27.5) | 1.45 (1.21–1.75) | 1.23 (1.05–1.44) |
| Level of education, (ref: higher education) | | |
| No education, preschool | 0.93 (0.74–1.18) | 1.16 (0.95–1.42) |
| Primary | 1.16 (0.95–1.42) | 1.04 (0.88–1.24) |
| Secondary | 1.08 (0.90–1.30) | 1.05 (0.89–1.23) |
| Currently working, (ref: no) | 0.81 (0.69–0.94) | 1.01 (0.90–1.14) |
| Hypertension, (ref: no) | 1.47 (1.30–1.68) | 0.98 (0.87–1.10) |
| Household level | Household level | Household level |
| Wealth quintile, (ref: lowest) | | |
| Second | 1.04 (0.79–1.35) | 0.82 (0.66–1.03) |
| Middle | 1.24 (0.96–1.60) | 0.94 (0.77–1.15) |
| Fourth | 1.60 (1.23–2.09) | 1.16 (0.95–1.42) |
| Highest | 2.21 (1.70–2.86) | 1.36 (1.10–1.68) |
| Community level | Community level | Community level |
| Place of residence, (ref: rural) | 1.02 (0.89–1.18) | 1.00 (0.88–1.14) |
| Administrative division, (ref: Barishal) | | |
| Chattogram | 0.98 (0.74–1.29) | 0.82 (0.66–1.03) |
| Dhaka | 1.32 (1.02–1.71) | 1.19 (0.94–1.49) |
| Khulna | 0.78 (0.60–1.01) | 0.59 (0.46–0.77) |
| Mymensingh | 0.94 (0.70–1.27) | 0.74 (0.59–0.93) |
| Rajshahi | 0.90 (0.68–1.20) | 0.57 (0.43–0.75) |
| Rangpur | 0.67 (0.50–0.91) | 0.54 (0.42–0.71) |
| Sylhet | 1.00 (0.72–1.38) | 0.73 (0.56–0.94) |
The fully adjusted model showed that compared with the younger age group, individuals aged between 35 to 49 years were $19\%$ to $23\%$ more likely to have prediabetes, and individuals aged 55 to 64 years were $26\%$ more likely to have prediabetes. Being obese (PR 1.23, $95\%$CI 1.05–1.44) and belonging to the highest wealth quantile (PR 1.36, $95\%$CI 1.10–1.68) were associated with prediabetes compared with being normal weight and being in the lowest wealth quintile. The respondents living in Khulna, Mymensingh, Rajshahi, Rangpur, and Sylhet divisions were about $25\%$ to $45\%$ less likely to have prediabetes than the Barishal division. Sex, level of education, working status, hypertension, and place of residence were not associated with prediabetes.
## Discussion
Diabetes and prediabetes affect a substantial proportion of the Bangladeshi population. Based on data from the latest BDHS 2017–18, over one-quarter of individuals aged 18 years and older had diabetes or prediabetes in Bangladesh, representing more than 19 million individuals in 2020. Factors associated with diabetes were age, sex, BMI, wealth quintile, employment status, hypertension, and administrative division of the country but not the place of residence (urban /rural) or education level. These findings confirm a continuing high burden of diabetes and prediabetes in Bangladesh.
We reported that the prevalence of diabetes is lower than the overall age-adjusted diabetes prevalence of $8.7\%$ in the Southeast Asian region, estimated by the IDF in 2021 [4]. The IDF has identified that the countries with the largest numbers of adults with diabetes aged 20–79 years in 2019 in the region are China (116 million cases) and India (77 million cases) [4]. In 2021, the IDF ranked Bangladesh 8th of countries with the highest number of adults (20–79 years) with diabetes (13.1 million cases), and it is expected to be ranked 7th in 2045 [4], consistent with our estimates. In our analysis, about 1 in 10 adults (18+) had diabetes, representing an estimated over 7.9 million individuals in Bangladesh in 2020. Note that our data included the younger population compared with the IDF estimates; as such, the total number of cases is deflated due to a very low prevalence of diabetes in the younger population. Nevertheless, this large number of diabetes cases in Bangladesh indicates that it is one of the leading countries for diabetes burden in the Southeast Asian region and highlights the urgent need for policies supporting the rollout of diabetes prevention in this country.
Our study undertook an identical methodological approach (e.g., anthropometric measurements and fasting blood samples) to the 2011 study and demonstrated that the prevalence of diabetes had increased markedly in Bangladesh over seven years [5]. Similar increasing trends of diabetes have been observed in other Southeast Asian countries [24]. However, the extent to which changes in traditional diabetes risk factors can explain the increasing trends in the prevalence of diabetes in this setting needs further investigation. A higher diabetes prevalence suggests that despite greater global awareness of diabetes and interventions for improved non-communicable disease management in primary health care [25], diabetes in *Bangladesh is* increasing. Furthermore, it suggests that health promotion may be failing in the face of dietary and lifestyle patterns. Thus, more resources are needed to be invested in primary health care to address the prevention of diabetes in Bangladesh.
Our estimates suggest that the prevalence of prediabetes has decreased in Bangladesh in the seven years between 2011 and $\frac{2017}{18}$ [5]. Prediabetes is important as during this stage, micro-vascular complications occur, often without people knowing they are glucose intolerant. The literature shows that up to $40.5\%$ of individuals with prediabetes convert to diabetes during follow-up [26]. A high conversion rate of prediabetes to diabetes is indicative of the potential for an uncontrolled increase in the prevalence of diabetes. The observation that diabetes prevalence has increased but prediabetes has decreased in seven years may indicate higher than expected conversion rates due to rapidly changing environmental conditions [25]. There is a well-established relationship between increasing age and the risk of diabetes which is consistent with our study [4, 27]. The risks are even higher among the respondents’ aged ≥50 years. The most important factors leading to such increasing risks are (i) deficiency of insulin secretion developing with age and (ii) growing insulin resistance caused by a change in body composition and sarcopaenia [28, 29]. The implication is that with increasing life expectancy in Bangladesh (current life expectancy at birth 72.3 years) [30], the increasing numbers of older people will result in even more cases and a higher burden of diabetes.
Consistent with our study, the literature shows that obesity is a leading risk factor for type 2 diabetes [31]. However, the association between obesity and diabetes is complicated as obesity is also related to socioeconomic status. We observed a significantly higher prevalence of diabetes in the highest wealth quintile than the lowest wealth quintile. A possible explanation is that those in the highest wealth quintile in LMICs use disposable income to purchase western, high-energy food (‘nutrition transition’) and avoid physically demanding tasks as symbols of status [32]. This results in obesity which in turn is associated with diabetes. Increasing obesity in Bangladesh [33] may also be due to reduced physical activity associated with changing traditional agricultural/domestic works replaced by technology, watching television, and using the internet. Irrespective of the explanation, increasing obesity in Bangladesh suggests that diabetes will increase further with its strengthening economy. A further policy implication is that interventions for diabetes prevention in Bangladesh need to focus on obesity, particularly by reducing the consumption of unhealthy diets and increasing physical activity. This should be given priority as there is evidence that even a modest weight reduction of 5–$7\%$ in high-risk individuals result in a decline in the incidence of diabetes [34] as recent studies demonstrated that the incidence of diagnosed diabetes is stabilizing or declining in many high-income countries since 2010 [35, 36].
Our data also show that diabetes is higher in people with hypertension, which agrees with other studies [37, 38]. In addition, hypertension is exacerbated by other risk factors such as obesity, advanced age and significantly contributes to micro and macrovascular complications resulting in renal failure and cardiovascular disease [37, 39]. Pathways through which these complications may occur include insulin resistance, inflammation, and obesity [40]. The implication of the strong association between hypertension and diabetes is that efforts are needed in Bangladesh to delay or prevent comorbid hypertension in diabetes through frequent follow-up and aggressive management.
Our study found that awareness, treatment, and control of diabetes are low in Bangladesh. Renewed efforts are needed to increase awareness, treatment, and control to improve diabetes outcomes and reduce/delay complications. Receiving appropriate treatment may be partly influenced by out-of-pocket health costs, large in Bangladesh [41]. This places a financial burden on households and has the effect of preventing people from accessing care (health care is viewed as a ‘luxury’ not a ‘necessity’) or seeking alternative providers who are cheaper but untrained and cause adverse effects [42]. Such barriers need to be addressed as they may undermine diabetes prevention efforts.
The likelihood of diabetes was high among respondents of the Dhaka division and low among respondents of the Rangpur and Barishal divisions. Many factors might contribute to such differences in risks of diabetes across divisions. For instance, being the capital of Bangladesh, the Dhaka division is highly urbanised and inhabited by people with high levels of education. Consequently, sedentary lifestyles and dependency on western foods are highly prevalent among residents of Dhaka as compared to the residents of other divisions in Bangladesh. Dhaka is also commonly characterised by very poor environmental conditions, including extreme air pollution, and few parks and open spaces. Together these factors contribute to an increased risk of diabetes among residents of the Dhaka division. In contrast, the Rangpur division is mostly rural, and people are largely employed in agricultural activity. These characteristics could be responsible for a lower risk of diabetes among residents of the Rangpur division. These regional variations in diabetes and prediabetes prevalence in Bangladesh are consistent with studies conducted in other South Asian countries, including India and Nepal [43, 44].
The strengths of the study are that it used a large, nationally representative dataset suggesting the findings have external validity. A further strength is that clinical variables, including FBG, blood pressure, body weight, and height, were measured using high-quality techniques. The WHO criteria for the classification of diabetes and prediabetes were used, while hypertension was defined using the seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of high blood pressure criteria. Our multilevel mixed-effects Poisson regression corrects the overestimation of the effects size produced by conventional logistic regression employed in cross-sectional studies and increases the precision of the findings. However, this was a cross-sectional study, which limits our ability to infer causal relationships. The oral glucose tolerance test (OGTT) or HbA1c tests are the gold standards for diagnosing type 2 diabetes. However, using these gold standards to diagnose diabetes was not possible for population-level data in the context of a resource-poor country. Behavioral factors (e.g., alcohol consumption, smoking status, sleep duration), dietary factors (e.g., type and amount of food taken), physical activity, and genetic factors (e.g., family history of diabetes) are crucial risk factors for diabetes. However, these data were not collected and could not be controlled for in the analyses.
## Conclusions
This study implies that efforts to control diabetes and prediabetes in Bangladesh need to be strengthened and optimized, investing further resources. Given that diabetes and prediabetes are preventable diseases by modifying diet and physical activities, Bangladesh needs to intensify its efforts to implement diabetes prevention. This may require the health care system changes in which non-communicable disease prevention is prioritized and household medical care payments reviewed to reduce out-of-pocket expenses. These measures will be worth the investment as they will maximize access to high-quality public health programs. Our analysis further implies that diabetes prevention should focus on reducing obesity and managing hypertension, suggesting that their management will bring the greatest benefits. Without effective preventive measures, diabetes will continue to increase in Bangladesh.
Diabetes and prediabetes affect a substantial proportion (over one-quarter) of the Bangladeshi adult population. Despite worldwide recognition of the increasing burden of diabetes in LMICs and widespread awareness of the need for prevention through lifestyle interventions, these conditions remain a significant public health burden in Bangladesh. Along with obesity and hypertension management, newer approaches to prevention are needed, which address obesogenic environments. These will include creating walkable neighborhoods, encouraging healthy food choices in schools, and workplaces, motivating physical activity and supporting active transport. These should be part of policies to address non-communicable diseases, including diabetes, in Bangladesh.
## References
1. Lin X, Xu Y, X P, Jingya Xu J, Ding Y, Sun X. **Global, regional, and national burden and trend of diabetes in 195 countries and territories: an analysis from 1990 to 2025**. *Nature* (2020.0) **10** 14790
2. Baena-Díez JM, Peñafiel J, Subirana I, Ramos R, Elosua R, Marín-Ibañez A. **Risk of cause-specific death in individuals with diabetes: a competing risks analysis**. *Diabetes Care* (2016.0) **39** 1987-95. DOI: 10.2337/dc16-0614
3. Saeedi P, Petersohn I, Salpea P, Malanda B, Karuranga S, Unwin N. **Global and regional diabetes prevalence estimates for 2019 and projections for 2030 and 2045: Results from the International Diabetes Federation Diabetes Atlas**. *Diabetes research and clinical practice* (2019.0) **157** 107843. DOI: 10.1016/j.diabres.2019.107843
4. 4International Diabetes Federation. IDF Diabetes Atlas, 9th edn. Brussels. Belgium: 2019. https://www.diabetesatlas.org, accessed 18 November 2020: 2019.
5. Akter S, Rahman MM, Abeb SK, Sultanac P. **Prevalence of diabetes and prediabetes and their risk factors among Bangladeshi adults: a nationwide survey**. *Bull World Health Organ* (2014.0) **92** 204-13A. DOI: 10.2471/BLT.13.128371
6. 6National Institute of Population Research and Training (NIPORT) Ma, and ICF International,. Bangladesh Demographic andHealth Survey 2011. Dhaka, Bangladesh and Calverton, Maryland, USA:NIPORT, Mitra and Associates, and ICF International: 2013.
7. 7Research NIoP, Training—NIPORT, Health Mo, Family Welfare, ICF. Bangladesh Demographic and Health Survey 2017–18. Dhaka, Bangladesh: NIPORT/ICF, 2020.. *Bangladesh Demographic and Health Survey 2017–18* (2020.0)
8. Saquib N, Saquib J, Ahmed T, Khanam MA, Cullen MR. **Cardiovascular diseases and type 2 diabetes in Bangladesh: a systematic review and meta-analysis of studies between 1995 and 2010**. *BMC public health* (2012.0) **12** 434. DOI: 10.1186/1471-2458-12-434
9. Al Kibria GM. **Prevalence and Factors Associated with Diabetes among Bangladeshi Adults: An Analysis of Demographic and Health Survey 2017–18**. *Diabetes Epidemiology and Management* (2021.0) **2** 100012. DOI: 10.1186/s40885-021-00174-2
10. Chowdhury MAB, Uddin MJ, Khan HM, Haque MR. **Type 2 diabetes and its correlates among adults in Bangladesh: a population based study**. *BMC Public Health* (2015.0) **15** 1-11. DOI: 10.1186/s12889-015-2413-y
11. Harshfield E, Chowdhury R, Harhay MN, Bergquist H, Harhay MO. **Association of hypertension and hyperglycaemia with socioeconomic contexts in resource-poor settings: the Bangladesh Demographic and Health Survey**. *International journal of epidemiology* (2015.0) **44** 1625-36. DOI: 10.1093/ije/dyv087
12. 12Swasey K, Al Kibria GM, Stafford K, editors. A conservative approach to estimating the socioeconomic factors associated with diabetes in Bangladesh. APHA’s 2018 Annual Meeting & Expo (Nov 10-Nov 14); 2018: American Public Health Association.
13. Barros AJ, Hirakata VN. **Alternatives for logistic regression in cross-sectional studies: an empirical comparison of models that directly estimate the prevalence ratio**. *BMC medical research methodology* (2003.0) **3** 21. DOI: 10.1186/1471-2288-3-21
14. Tamhane AR, Westfall AO, Burkholder GA, Cutter GR. **Prevalence odds ratio versus prevalence ratio: choice comes with consequences**. *Statistics in medicine* (2016.0) **35** 5730-5. DOI: 10.1002/sim.7059
15. 15United Nations, Department of Economic and Social Affairs, Population Division. World Population Prospects 2019: Highlights (ST/ESA/SER.A/423), https://population.un.org/wpp/Publications/Files/WPP2019_Highlights.pdf accessed on 24 Jan 2021. 2019.
16. 16World Health Organization. Definition and diagnosis of diabetes mellitus and intermediate hyperglycaemia: report of a WHO/IDF consultation. 2006.
17. Talukder A, Hossain M. **Prevalence of diabetes mellitus and its associated factors in Bangladesh: application of two-level logistic regression model**. *Scientific Reports* (2020.0) **10** 1-7. PMID: 31913322
18. Consultation WE. **Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies**. *Lancet (London, England)* (2004.0) **363** 157-63. DOI: 10.1016/S0140-6736(03)15268-3
19. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL. **The seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure: the JNC 7 report**. *Jama* (2003.0) **289** 2560-71. DOI: 10.1001/jama.289.19.2560
20. 20Rutsein SO, Johnson K. The DHS Wealth Index. DHS Comparative Reports No. 6. Calverton, Maryland, USA: ORC Macro. http://dhsprogram.com/publications/publication-cr6-comparative-reports.cfm. Accessed November 07, 2015.: 2004.
21. Yelland LN, Salter AB, Ryan P. **Performance of the modified Poisson regression approach for estimating relative risks from clustered prospective data**. *American journal of epidemiology* (2011.0) **174** 984-92. DOI: 10.1093/aje/kwr183
22. O’Connell AA, McCoach DB. *Multilevel modeling of educational data: IAP* (2008.0)
23. Ye L, Cao W, Yao J, Peng G, Zhou R. **Systematic review of the effects of birth spacing after cesarean delivery on maternal and perinatal outcomes**. *International Journal of Gynecology & Obstetrics* (2019.0) **147** 19-28. DOI: 10.1002/ijgo.12895
24. Collaboration NRF. **Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4.4 million participants**. *Lancet* (2016.0) **387**. DOI: 10.1016/S0140-6736(16)00618-8
25. Zaman M, Ullah A, Bhuiyan M, Karim M, Moniruzzaman SMA. *Noncommunicable Disease Prevention and Control Situation in a Primary Health Care Setting of Bangladesh: Design and Baseline Findings of an Intervention Chronic Dis Int* (2016.0) **3** 1021
26. Mohan V, Deepa M, Anjana RM, Lanthorn H, Deepa R. **Incidence of diabetes and pre-diabetes in a selected urban south Indian population (CUPS-19)**. *J Assoc Physicians India* (2008.0) **56** 152-7. PMID: 18697630
27. Chen Y, Zhang X-P, Yuan J, Cai B, Wang X-L, Wu X-L. **Association of body mass index and age with incident diabetes in Chinese adults: a population-based cohort study**. *BMJ open* (2018.0) **8** e021768. DOI: 10.1136/bmjopen-2018-021768
28. Shou J, Chen P-J, Xiao W-H. **Mechanism of increased risk of insulin resistance in aging skeletal muscle**. *Diabetology & Metabolic Syndrome* (2020.0) **12** 1-10. DOI: 10.1186/s13098-020-0523-x
29. Krentz A, Viljoen A, Sinclair A. **Insulin resistance: a risk marker for disease and disability in the older person**. *Diabetic medicine* (2013.0) **30** 535-48. DOI: 10.1111/dme.12063
30. 30World Bank. https://data.worldbank.org/indicator/SP.POP.TOTL?locations=BD. 2021.
31. Barnes AS, Coulter SA. *The Epidemic of Obesity and Diabetes Trends and Treatments Tex Heart Inst J* (2011.0) **38** 142-4. PMID: 21494521
32. Pampel FC, Denney JT, Krueger PM. **Obesity, SES, and Economic Development: A Test of the Reversal Hypothesis**. *Social Science and Medicine* (2012.0) **74** 1073-81. DOI: 10.1016/j.socscimed.2011.12.028
33. Banik S, Rahman M. **Prevalence of overweight and obesity in Bangladesh: a systematic review of the literature**. *Current obesity reports* (2018.0) **7** 247-53. DOI: 10.1007/s13679-018-0323-x
34. Knowler WC, Barrett-Connor E, Fowler SE, Hamman RF, Lachin JM, Walker EA. **Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin**. *The New England journal of medicine* (2002.0) **346** 393-403. DOI: 10.1056/NEJMoa012512
35. Magliano DJ, Chen L, Islam RM. **Multi-country analysis of trends in the incidence of diagnosed diabetes from 22 million diagnoses in higher- income settings**. *The Lancet Diabetes & Endocrinology* (2021.0)
36. Magliano DJ, Islam RM, Barr EL, Gregg EW, Pavkov ME, Harding JL. **Trends in incidence of total or type 2 diabetes: systematic review**. *bmj* (2019.0) **366** l5003. DOI: 10.1136/bmj.l5003
37. Lastra G, Syed S, Kurukulasuriya LR, Manrique C, Sowers JR. **Type 2 diabetes mellitus and hypertension: an update**. *Endocrinology and Metabolism Clinics* (2014.0) **43** 103-22. DOI: 10.1016/j.ecl.2013.09.005
38. Suh D-C, Kim C-M, Choi I-S, Plauschinat CA, Barone JA. **Trends in blood pressure control and treatment among type 2 diabetes with comorbid hypertension in the United States: 1988–2004**. *Journal of hypertension* (2009.0) **27** 1908-16. DOI: 10.1097/HJH.0b013e32832d4aee
39. Simonson DC. **Etiology and prevalence of hypertension in diabetic patients**. *Diabetes care* (1988.0) **11** 821-7. DOI: 10.2337/diacare.11.10.821
40. Cheung BM, Li C. **Diabetes and hypertension: is there a common metabolic pathway?**. *Current atherosclerosis reports* (2012.0) **14** 160-6. DOI: 10.1007/s11883-012-0227-2
41. Afroz A, Habib SH, Chowdhury HA, Paul D, Shahjahan Md, Hafez A. **Healthcare cost of type 2 diabetes mellitus in Bangladesh: a hospital-based study**. *Int J Diabetes Dev Ctries* (2016.0) **36** 235-41
42. Khan JA, Ahmed S, Sultana M, Sarker AR, Chakrovorty S, Rahman MH. **The effect of a community-based health insurance on the out-of-pocket payments for utilizing medically trained providers in Bangladesh**. *International health* (2020.0) **12** 287-98. DOI: 10.1093/inthealth/ihz083
43. Shrestha N, Mishra SR, Ghimire S, Gyawali B, Mehata S. **Burden of diabetes and prediabetes in Nepal: a systematic review and meta-analysis**. *Diabetes Therapy* (2020.0) **11** 1935-46. DOI: 10.1007/s13300-020-00884-0
44. Tandon N, Anjana RM, Mohan V, Kaur T, Afshin A, Ong K. **The increasing burden of diabetes and variations among the states of India: the Global Burden of Disease Study 1990–2016**. *The Lancet Global Health* (2018.0) **6** e1352-e62. DOI: 10.1016/S2214-109X(18)30387-5
|
---
title: 'The organisation of primary health care service delivery for non-communicable
diseases in Nigeria: A case-study analysis'
authors:
- Whenayon Simeon Ajisegiri
- Seye Abimbola
- Azeb Gebresilassie Tesema
- Olumuyiwa O. Odusanya
- David Peiris
- Rohina Joshi
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021956
doi: 10.1371/journal.pgph.0000566
license: CC BY 4.0
---
# The organisation of primary health care service delivery for non-communicable diseases in Nigeria: A case-study analysis
## Abstract
As chronic diseases, non-communicable diseases (NCDs) require sustained person-centred and community-based care. Given its direct link to communities and households, Primary Health Care (PHC) is well positioned to achieve such care. In Nigeria, the national government has prioritized PHC system strengthening as a means of achieving national NCD targets. However, strengthening PHC systems for NCDs require re-organization of PHC service delivery, based on contextual understanding of existing facilitators and barriers to PHC service delivery for NCDs. We conducted a mixed method case study to explore NCD service delivery with 13 PHC facilities serving as the cases of interest. The study was conducted in two northern and two southern states in Nigeria–and included qualitative interviews with 25 participants, 13 focus group discussion among 107 participants and direct observation at the 13 PHCs. We found that interprofessional role conflict among healthcare workers, perverse incentives to sustain the functioning of PHC facilities in the face of government under-investment, and the perception of PHC as an inferior health system were major barriers to improved organisation of NCD management. Conversely, the presence of physicians at PHC facilities and involvement of civil society organizations in aiding community linkage were key enablers. These marked differences in performance and capacity between PHC facilities in northern compared to southern states, with those in the south better organised to deliver NCD services. PHC reforms that are tailored to the socio-political and economic variations across Nigeria are needed to improve capacity to address NCDs.
## Introduction
In Nigeria, NCDs account for $29\%$ of all deaths, out of which cardiovascular diseases accounts for $11\%$ [1]. The prevalence of hypertension and diabetes is estimated to be $28.9\%$ and $4.1\%$ respectively [2]. In line with the Global Action Plans [3] and Sustainable Development Goal 3.4 [4] (to strengthen responses for prevention and control of NCDs), Nigeria has also set national NCD targets. These include relative reduction of raised blood pressure and diabetes mellitus by $25\%$ in year 2025 [1]. However, as recommended in World Health Organization’s (WHO) “best-buys”, effective and feasible implementation of NCD prevention and control strategies in low- and middle-income countries (LMICs), require strengthening and orientating the health system through people-centred primary health care (PHC) [5].
Nigeria’s PHC system, the bedrock of the health system [6], was adopted into the 1988 national health policy in an attempt to improve access and utilization of basic health services [7]. Between 1986 to 1990, the PHC experienced several reforms including its expansion to all LGAs and the creation of colleges of health technology for the training of community health workers [8]. In 2014, the National Health Act placed the PHCs under the local government authorities [9]. This however created a bottleneck for PHC funding as federal laws are not binding on the states in a decentralized health system [8].
After several evolutions, the current strategic drive for the PHC is the ward health system (WHS). It was developed by the National Primary Health Care Development Agency (NPHCDA) in order to improve access to healthcare [10]. The WHS comprises of several interventions known as the Ward Health Minimum Package. These interventions are expected to address communicable, non-communicable diseases, and maternal and child health services. [ 11]. The direct linkage of the PHC to communities and households positions it as the pedestal to reach the ‘last mile’ population who are mostly located in rural areas [12]. Despite its pivotal role in addressing access barriers, PHC attracts the least investment in the national health system [6]. This is in part due to the devolution of PHC to the local government, the level of governance with the weakest technical and financial capacity.
Due to the chronic nature of NCDs, people living with these diseases and their risk factors require sustained person-centred and community-based care [13]. This would ideally be achieved through the PHC health system. However, NCD care appears to be the most neglected aspect of the PHC sector as comprehensive NCD care is omitted from the Minimum Package of Health Services that will be funded by the Basic Health Care Provision Fund [1]. This may partly be due to the fact that the WHS was established to align with Millennium Development Goals which largely omitted NCDs as it was perceived to contribute towards a proportionately smaller burden of disease at that time [7]. It may also be because most government and development partners’ interventions to strengthen PHC are focused mainly on maternal, child and reproductive health as well as infectious diseases [11], still aligning with the primary reason for establishing PHC systems especially in the 1980s [6].
Access to quality and essential NCD interventions is further compounded by inadequate and maldistribution of skilled health workers, particularly physicians and nurses [13]. Consequently, NCD care at the PHC level is mainly handled by community health workers (CHWs), whose training and skills are generally considered insufficient for NCD management and prevention [14]. Indeed, WHO’s country profile on NCDs reported that no PHC facility in Nigeria offers cardiovascular disease (CVD) risk stratification or utilizes CVD guidelines; only $30\%$ of facilities reported essential NCD medicines as “generally available”; and similarly only $33\%$ reported essential NCD technologies as “generally available” [15]. Given the national government’s commitment to prioritize PHC system strengthening as a means of achieving universal health coverage [6] and national NCD targets [1], a corresponding organizational restructuring of service delivery is critical. Such re-organisation requires contextual understanding of existing organisation and the facilitators and barriers to PHC service delivery for NCDs across different settings in Nigeria.
Most studies on NCDs in Nigeria centre on disease burden [2, 16–19] while some others have assessed health care workers’ knowledge on these diseases [13, 20]. Only few studies have examined enablers and barriers to NCD service delivery in PHC facilities. One study revealed that NCDs have the lowest service-readiness scores with major gaps in staff capacity to treat NCDs [8]. Another study among PHC health workers and insurance managers, showed that health insurance was perceived as an important facilitator of implementing high-quality hypertension care; while high staff workload; administrative challenges; and difficulty in adapting some guideline recommendations were key inhibitors of high-quality care [21]. In another study, availability of drugs at subsidized rates, trained workforce and regular training opportunities were identified as factors promoting quality; while cultural barriers and patients’ socioeconomic factors were identified as major barriers to receiving high quality care [22].
While identifying enablers and barriers for quality PHC service delivery for NCDs is important, national and sub-national decision-making requires a nuanced understanding of the context and structure within which these services are delivered. In this study, we sought to characterise the organisation of PHC service delivery for NCDs and identify what factors promote or hinder NCD service delivery, with special focus on hypertension and diabetes mellitus. We define such factors (i.e., enablers or barriers) as the people, processes, structures, skills and strategies that may facilitate or constrain high-quality NCD service delivery at the PHC level.
## Materials and methods
We conducted a mixed-method case study to explore NCD service delivery with each PHC facility serving as the case.
## Study setting
Three PHC facility types exist based on the Ward Health System [23]: Our study was conducted mainly at the highest level of the PHC system, the Primary Health Care Centres, where the most comprehensive form of service delivery with well-equipped and adequately skilled staff is expected.
## Study design
This case study used mixed methods (survey and interviews) to identify enablers and barriers to NCD (hypertension and diabetes mellitus) diagnosis and management at PHC level. A case study is a pragmatic design that seeks to explore contemporary phenomena in real-world settings [24]. It has potential to provide in-depth answers to the “how” and “why” of NCD service delivery at the facility level.
## Study area
A total of 13 PHC centres in four Nigerian states participated in the study. Two states were purposively selected in each of the northern and southern regions to reflect the regional socio-political, economic, and religious differences that are known to influence healthcare demands, household health seeking behaviours, and availability of medication and equipment in health facilities [25, 26]. An additional criterion for selecting two states (one in northern and one in southern Nigeria) was based on recent implementation of a new health intervention programme (performance-based financing) in those states, designed to re-organise PHC service delivery.
## Quantitative data collection
The WHO Service Availability and Readiness Assessment (SARA) tool was used to collect data on services available for NCDs (hypertension and diabetes) at each PHC facility [27]. The tool was adapted for the hypertension programme in Nigeria, with input from government agencies and relevant stakeholders. The adaptation specifically focused on diagnosis and management of hypertension and diabetes mellitus. It was subsequently pilot tested and data collected in 60 PHC facilities [28]. Data collection focussed on: (i.) service availability (ii) patient access, (iii) staffing capacity, (iv) infrastructure, (v) basic client amenities, (vi) infection control, (vii) healthcare waste management, (viii) clinical mentoring, (ix) basic equipment, (x) available services for non-communicable diseases and diagnostics, (xi) supply chain, (xii) medicines and vaccines, and (xiii) commodities [28]. This corresponded to the 13 sections of the SARA tool. In addition, data collected with the SARA survey instrument from PHC respondents were corroborated by direct observation by research staff of medications, equipment, and supplies in the PHC facilities.
## Qualitative data collection
In each PHC, in-depth, semi-structured interviews (IDI) with key PHC staff (nurses, community health workers or doctors) and focus group discussions (FGD) with about 6–10 participants were conducted among health workers who have worked for a minimum of three months at the facility. This is to ensure participants had worked for a sufficient duration to have detailed understanding of how their facilities operate. The focus of the interviews was to understand the factors that constitute barriers and enablers (defined below) to NCD management. All participants were interviewed face-to-face at their respective PHC facilities. Interview durations ranged from 35–60 minutes while FGDs last for 45–70 minutes. Both interviews and FGDs were audio-recorded, transcribed and field notes were also taken.
Data were collected from August 2019 to September 2019 guided by the consolidated criteria for reporting qualitative research guidelines for qualitative research [29].
## Data analysis
The unit of analysis in the study was a case as represented by each PHC facility. The quantitative data for each case was analyzed first to provide a general overview of each case. Findings from the quantitative data guided the initial analysis of the qualitative data.
## Quantitative data
Statistical analysis was done in Microsoft Excel and continuous, non-parametric measures were summarized by median and interquartile range. The facility-based service availability for hypertension and diabetes and other domains of interest such as equipment and supplies, personnel and medications were tabulated as frequencies.
## Qualitative data
Verbatim transcription of recorded FGDs and interviews was done. These transcripts were imported into NVivo 12 for data coding, which used the pre-determined coding from SARA results. The codes were categorized based on how related they were. Several meetings were subsequently held by the research team members to analyse and interpret the data from each case. These meetings helped to iteratively identify and refine emerging themes, and inferences, and to deal with apparently contradictory information across the cases. Phrases or full sentences that most accurately expressed or illustrated the categories under each theme were then identified and presented as quotes in the results section.
## Ethical considerations
Ethical approval was granted by the National Health Research Ethics Committee of Nigeria (Approval no: NHREC/$\frac{01}{01}$/2007) and the University of New South Wales Human Research Ethics Committee (HC: 190051). Informed written consent was obtained from all participants before conducting the interview. Anonymity and confidentiality of all respondents were maintained throughout the process. Participants names and the names of the states and PHC facilities included in the study were replaced with codes during data analysis (Table 1).
**Table 1**
| Unnamed: 0 | Region | KII1 | KII2 | KII3 | KII 4 | FGD participants |
| --- | --- | --- | --- | --- | --- | --- |
| Case 1 | North | M, 54, CHEW | F, 42, Nurse | – | – | 4 M, 6 F |
| Case 2 | North | F, 40, CHO | F, 52, Nurse | – | – | 5 F |
| Case 3 | North | F, 45/, CHO | F, 52, Nurse | – | – | 3 M, 6 F |
| Case 4 | North | F, 45, CHEW | – | – | – | 2 M, 6 F |
| Case 5 | North | F, 49, CHO | – | – | – | 1 M, 5 F |
| Case 6 | North | F, 55, CHO | – | – | – | 4 M, 6 F |
| Case 7 | North | F, 48, CHO | – | – | – | 2 M, 4 F |
| Case 8 | South | M, 55, CHO | F, 34, Nurse | F, 40, Physician | M, 47, Medical Officer of Health | 4 F, 6 F |
| Case 9 | South | M, 55, CHO | F, 47, Nurse | F, 37, Physician | – | 3 F, 6 F |
| Case 10 | South | F, 40, CHO | F, 37, Physician | – | – | 10 F |
| Case 11 | South | F, 39, Nurse | M, 40, Physician | – | – | 10 F |
| Case 12 | South | F, 38, CHO | F, 45, Nurse | – | – | 8 F |
| Case 13 | South | F, 41, CHO | F, 48, Nurse | – | – | 6 F |
## Results
The case study analysis described available services and explored the potential barriers and enablers to service delivery. A more descriptive information on each of the cases is provided in S1 Annex. Based on findings from our study, Fig 1 and Table 2 respectively provides information on the combined and disaggregated SARA findings of NCD-related services at the PHC facilities.
**Fig 1:** *Summary of NCD-related SARA findings from all 13 PHCs facilities.* TABLE_PLACEHOLDER:Table 2
## PHC team composition and capacity
Physicians are present in five facilities, while 3 facilities had neither physician nor nurse. Five facilities had staff who had been trained in screening or management of NCDs within the previous 2 years (Table 2). The composition of skilled health workers in PHC facilities varied across states. CHWs constituted most health workers in all the PHC facilities, and were the only health workers in Cases 5, 6 and 7, all of which were northern states. Patients could access care at PHC facilities at any time with or without appointment (Fig 1). PHC facilities, including those with only CHWs provided various services for diabetes and hypertension, ranging from screening, diagnosis, and counselling to drug prescription, referral, and follow-up. However, as a participant in Case 3 said: “health talk is the number one primary thing that we do at the primary health level, we give health talk to patients and we screen them” (Case 3, FGD, CHEW).
Knowledge of patient management for hypertension and diabetes varied across PHC facilities and appeared to be related to team capability at each PHC facility. In cases 4, 5, 6 and 7 where the teams comprised mainly CHWs, they were able to obtain clinical history from patients, refer them for laboratory investigations and to a higher-level facility if required: In Cases 8, 9 and 10; all facilities with full-time physicians, prescription, treatment, and follow-up appointments were provided to those diagnosed with hypertension and diabetes–and those with complications were referred to secondary health facilities for further management.
Insufficient staffing was identified as a major issue. To make up for staffing shortfalls, PHC facilities engaged CHWs as volunteers or contract staff who did not earn salary from the government or receive government support for further training or career progression. This situation was particularly prevalent in the facilities in the northern states and was identified as contributing to lack of motivation: “the volunteers work to support the hospital, they are trying [their best], but sometimes they will feel “I’m just doing this, after all, at the end of the day what will I get?”, you understand…it makes one weak [discouraged]” [Case 3, KII, 1]. The issue of insufficient staff was raised at most PHC facilities, as it led to increased workload and poor service delivery according to a participant in Case 5: In Case 6, insufficient staffing also negatively impacted service delivery as CHWs did not have the time to effectively raise awareness for disease prevention and health promotion activities, including that of NCDs: By contrast, in the southern states—cases 8, 9 and 10 in particular—physicians operated three shifts daily to provide 24-hour services to patients. Physician attrition and migration abroad was, however, a challenge, as a Medical Officer of Health described in Case 8: Insufficient staffing was linked to insufficient training to deliver NCD services. Apart from Cases 8, 9, 10 11 and 13 (all in southern states), no PHC staff had received in-service training on the management of NCDs within the last two years. Most CHWs relied on the knowledge acquired during their training in the College of Health Technology, which they considered inadequate for NCD management: “…for hypertension and diabetes…the knowledge we are using is the one we got in school [and]…we are not updated…. There are so many (new) drugs that are in use now, we are using the old knowledge.” [ Case 5, FGD]. Those who had received some in-service training seemed to have been provided only with skills in NCD screening. In Case 1, a CHW said: “there was a training we went to, but basically [it] was not for management, but for screening” [Case 1, KII 2, CHW]. Another constraint was that permanent CHWs were usually prioritized and volunteer or contract CHWs were not provided with training opportunities. In Case 6, a CHO said: “if it is in-service, the government will support you with your salary…. but those that are contract staff have no one to support them, that is one of the problems that is preventing (them) from going further” (Case 7, KII, CHO).
While lack of in-service training for CHWs in NCD management appeared to be a major issue, the presence of a physician in Cases 8, 9, 10, 11 and 12 (southern states) possibly mitigated this issue. This was evident with the use of a more systematic approach in the management of patients with hypertension and diabetes compared to facilities that lack physicians. For instance, in Case 10, classification, stages and systemic features that are pointers to complication for patients with NCDs were mentioned as criteria that guide service delivery:
## NCD facility supplies and treatment guidelines
All facilities screen for hypertension and diabetics but only 7 provide follow-up and long-term treatment. Blood pressure lowering medication are available in 12 facilities while blood-sugar lowering medications are available in 9 facilities (Table 2). None of the 13 PHC facilities had NCD guideline or information, communication, and education materials on NCD (Table 2). While all the PHC facilities claimed to have basic equipment for screening, diagnosing, and monitoring diabetes and hypertension, their availability did not guarantee functionality. In Case 3, a nurse said “… when you enter my office there, there are 2 BP apparatuses… you use it [for] a day or two, it will then develop problems. And that is how it continues. ( Case 3, KII 1, Nurse-in-Charge). In Case 6, a CHO said “We don’t have much equipment… If you don’t have equipment on ground, you will not be able to do your work.” ( Case 6, KII 1, CHO).
In all but one facility (Case 2) BP-lowering and blood-sugar lowering medicines were stocked, prescribed, and dispensed. However, the facilities had varying supply and delivery chain structures which helped to ensure that supplies were in place. For example, in a PHC facility included in the Performance Based Financing [PBF] initiative, a CHO said “…this facility is a PBF facility, … so, they have pharmaceuticals companies that are registered, and you can you buy from any of [those] registered pharmacies.” ( Case 5, KII1, CHO). In another example, the supply chain for Cases 8, 9, 10 and 11 used an established state government process with built-in accountability mechanisms. The supply structure, subsidised the costs of medicines and tracked the revenue generated: Other PHC facilities adopted an informal practice in their supply system. For instance, in Case 3, a senior nurse was the officer-in-charge of the facility’s administrative activities, while a pharmacist was in charge of medicine supply. The arrangement meant that transactions were transparently done through the government’s account, with restricted access to potential unauthorized use of internally generated revenue by any facility official. However, the matron-in-charge operated a parallel unofficial supply system with individual pharmaceutical vendors. So, rather than prescribing medicines to be dispensed through the PHC facility’s pharmacy, she prescribed and dispensed medicines directly to patients during consultation from her private unofficial supply in exchange of money paid to her but without a receipt.
The supply chain influenced facility operations and team dynamics. For instance, in Cases 1, 4, 5, 6 and 7, a CHW was the officer-in charge–i.e. oversaw the PHC facility’s administration which included medication and equipment supply. The CHW generated internal revenue to run the facility by hiring other CHWs on contract and sharing local profits. In some PHCs facilities that have nurses as staff members but a CHW as officers-in-charge of the facility, such as Cases 1 and 4, nurses were mainly in charge of maternity services while the CHWs attended to patients with NCDs and minor ailments. The nurses were protective of their “maternity territory” for two reasons. First, they did not consider the CHWs to be sufficiently competent: Second, because maternity services are revenue-generating services, those who provided the services had greater access to the revenue generated from ante-natal care and delivery services. Hence, this was another reason for role protection, as was highlighted in Case 6, where a CHW said: “we don’t have problem with the doctors, but nurses, at times they see this facility as community health workers are taking all the patients and leaving them without any, that’s the only problem (Case 6, FGD participant—CHW). The role conflict between nurses and CHWs was enabled by not implementing the defined scope of practice within the PHC facilities.
The Standing *Order is* the legal document that defines the scope of CHWs’ practice. Despite the restriction it places on the management of NCDs by CHWs, many prescribe and dispense NCD medication, monitor NCD patients and refer those they considered complicated condition to secondary health facilities which is beyond their scope of practice. These services are provided without any management guideline or dedicated in-service training. CHWs in all the northern state facilities stated that there were no NCD guidelines or treatment protocols available: “…there is no guideline, the only guide that I will talk of is our standing order… and [there] is no place where our standing order says you should treat hypertension, most of this our standing order, they will say refer” Case 1, FGD, CHW). The non-availability of treatment guideline was also raised as a concern among physician as expressed by one of the participants: “No, [no guideline] … we should know the new management guideline for managing high blood pressure…. It is good for every doctor to be on the same page. … if we have a guideline, we’ll be on the same page.” ( Case 8, KII, physician).
## PHC referral linkages and feedback relations
All PHC facilities had linkage with at least one secondary health facility for patient referral. Participants believed that this helped to ensure continuity of care. In Case 5, a CHW said: “If the person has diabetes, we refer to general hospital, because we work based on the standing order.” ( Case 5, KII 1). And in Case 3, a nurse said: “normally, we send them to the general hospital, so if they go and they’re attended to, they come back here to be checking their BP” (Case 3, FGD). While PHC facilities in southern states had access to transportation services for referrals to secondary health facilities, the northern state PHC facilities (Cases 1–6) did not: “…we don’t have any means of referral so if a patient is having a relation, they will go and look for a vehicle [to]… come and take the person to the facility” (Case 3, KII). In one southern state PHC facility (Case 9) patient transportation was available with a functional ambulance for transporting referred patients diagnosed with a serious illness, and they are usually accompanied by a healthcare worker “So, any patient that we refer, …We call our ambulance driver to take the patient to general hospital… Yes, one of our staff will follow them to the general hospital” (Case 9, FGD).
Despite the use of a two-way referral form, feedback on referred patients from secondary health facilities was generally limited. This was partly driven by competing workload at the referral centres: “…[for] referral, we [should] get feedback but it’s not so… The [doctor], maybe his table is full, he cannot even attend to you. I went there third time this week. I had to stay till today to get it” (Case 4, KII1, CHEW). However, in Case 11, the challenge of feedback appeared to be related to secondary health facilities ‘looking down’ on the PHC facilities as less important and not deserving of feedback. “… they look down on PHC workers as quacks, and so, they feel reluctant [to write feedback] …they say, “there’s no need of writing back to them, they don’t know what they’re doing” (Case 11, KII1, Physician). In Case 6, participants (CHWs) said, not only do physicians and nurses in secondary facilities see PHC as a second rate, but they also deride their services in front of patients and their relatives: Despite the challenges, staff from some PHC facilities (Cases 4, 5, 6 and 7) managed to obtain feedback for referred patients by visiting the secondary health facilities. The motivation to do so was driven by the Performance-Based Financing (PBF) program that gives financial incentives to PHC facilities for every referred patient with documentary evidence of feedback. According to a participant in Case 4, referral would take place even if there was no physician available: “…that [the availability of a physician] will not stop us from referring [to secondary facilities] …. because the referral service is in our [PBF] checklist, and it is also a source of money for us.” ( Case 4, KII participants, CHEW). The utility of the obtained feedback in the continuum of care for the patients was difficult to ascertain. However, obtaining feedback was easier in Cases 8, 9 and10 where the patient was referred between physicians. For instance, one site (Case 10) had an established social media platform for communication between physicians at PHC and secondary health facilities without connotations of inferiority: Administrative feedback from the Local Government Area office occurred in response to data submitted from PHC facilities. Patient information were mainly held on paper records and entered into a daily register. Selected NCD cases, along with some other infectious diseases, were entered into the Monthly Health Facility Summary Form and Integrated Diseases Surveillance for onward transmission to higher level of the health system. According to a participant in Case 5, feedback given to the PHC facility mainly focused on data quality with little attention on quality of services public health actions that need to be taken: There were also community referral linkages. The PHC system operates within the Ward Health System structure, which provides direct linkage to the community, helps to focus the PHC facility on their needs, raise awareness about government programmes, and mobilise community members to participate. Community outreaches (part of routine PHC services) were therefore an avenue to diagnose undetected diabetes and hypertension:
These community outreaches sometimes received ad-hoc support from non-governmental and commercial organizations which sometimes partner with PHC facilities to conduct free screening for hypertension and diabetes in the community and subsequently linked newly diagnosed patients to the PHC facilities for further management: Fig 2 shows patient flow for NCD service delivery at the PHC level, with enablers and barriers along the pathway which we describe below according to the identified themes.
**Fig 2:** *Patient flow for NCD service delivery at the PHC level, with enablers and barriers along the pathway.*
## Discussion
This mixed methods case study of 13 PHC facilities in Nigeria examined the organisation of PHC service delivery for NCDs in Nigeria. It provides insight into a range of factors that serves as potential enablers and barriers to delivery of care for NCDs. Key findings include: [1] role conflict among non-physician health workers; [2] Inadequate PHC workforce, and perception of PHC as inferior; [3] the role of the physician as catalyst for NCD service delivery [4] the use of perverse incentives to sustain the functioning of PHC facilities, and [5] the variation in PHC service delivery by geographical region. We discuss each of these in detail below.
## Interprofessional role conflicts
The delivery of quality health care is dependent on the contributions from the various cadres of healthcare workers that constitute the team. This involves a complex process, particularly at the PHC level where care needs of patients and service delivery environments are diverse [30]. Interprofessional role conflicts among the healthcare team member arise from this complexity with potential to decrease quality care for patients and reduce team effectiveness [31]. Possible sources of such conflicts could be substantive (particularly in relation to scope of practice and financial renumeration) and emotive (particular when driven by individual personalities and entrenched power differentials) [32]. In our study, non-availability of guidelines, lack of clear roles, control of internally generated incomes and the display of professional cadre superiority were some issues identified as potential sources of role conflicts among the non-physician PHC team members.
Previous studies, from high income countries, have discussed interprofessional role conflicts among PHC teams as a barrier to quality care–with discrepancies in guidelines [33], scope of practice, role boundary issues and accountability among the issues identified as sources of conflicts [30, 34]. Similar study from a LMIC setting revealed interprofessional role conflict among non-physician health workers providing care for patients with cardiovascular diseases [35]. PHC facilities in Nigeria are characteristically staffed with CHWs and nurses [36], with the latter often in the majority [14], it is important that conflict resolution strategies that support individual and the team as a whole are put in place. These include the development of an explicit, contextualized guideline with clear role description, standardised facility revenue management processes, task-shifting and task-sharing programmes [37, 38], and effective leadership [39].
## Inadequate PHC workforce, and the perception of PHC as an inferior component of the health system
An adequate health workforce in terms of supply and skill mix is essential for effective NCD service delivery. A major barrier observed in this study is that most PHCs facilities do not meet the minimum staffing requirements stipulated by the Ward Health System within which each PHC operates [10]. This persistent workforce shortage combined with a large pool of volunteer community health workers likely reflects limited funding allocations for the PHC sector. National and subnational governments need to accelerate and scale up staff recruitment combined with appropriate skills training [40], and proportionate distribution to areas of relative low workforce density [41]. This requires development of evidence-informed workforce policies, effective deployment mechanisms [41] and increased funding allocation.
Being the closest entry point of the community into the formal health system, the PHC sector is best placed to reach the ‘last mile population’ who predominantly reside in rural areas [12]. The existence of established referral linkages between the PHC and secondary facilities is considered an enabler for the management of patient with NCDs. Referrals are aimed at ensuring patients receive the appropriate quality and continuity of care within the health system [42]. However, referral can only be effective if there exists a close and congruent relationship among the various levels of care within the health system [43]. The labelling and perception of the PHC level, and its staff, as inferior rather than as a partner-in-care within the health system by some secondary health facilities is a major challenge. A reason for this attitude identified in our study was the cadre of health care workers that make up the PHC staff. We found this led to delayed or absolute lack of feedback from secondary health facilities and this has been observed in other studies [44, 45]. The impact of a one-way referral system without feedback to the PHC is disruption to the continuum of care for a patient [46].
Previous studies also identified weak referral linkage for NCD management between primary and secondary health facilities as a barrier to high quality service delivery [47]. It is important that governments at every level address this issue. Enacting and implementing appropriate referral policies, training of health workers, and activities to promote inter-professional collaboration accordingly could address health worker attitudes and reorientate secondary health care workers on the importance of feedback [46]. Appropriate use of referral guidelines could also help to clarify areas of disagreement between different levels of health system, halt or reverse personnel’s view of PHC as second-rate care and respectfully acknowledge PHC workers as members of the healthcare team [48].
Conversely, a promising enabler of referrals and linkages to the community identified in our study was the interest of civil society organizations (CSOs) in supporting NCD services, at the PHC level–and especially in supporting community-based outreaches that promote diagnosis and referrals. This finding is in line with previous studies in Nigeria that show the role and potential of community efforts (via community health committees) in supporting the day-to-day functioning of PHC services, linking community members to PHC facilities, and building community trust in PHC services [49, 50]. Such examples of social collaboration promote community participation, with potential to empower PHC workers in prioritizing the needs of community members within this space of PHC-CSO/community co-responsibility [51].
## Physician at PHC facilities as catalyst for NCD service delivery
Aside the private sector, physicians are not typically seen at PHC settings in sub-Saharan Africa [52]. Where they are found, they tend to enable high quality of care and support management of complex NCD cases, providing continuous quality care with resultant cost reductions for both the patient and health system as well as increasing community trust in the PHC system [53]. The presence of physicians in PHC facilities, as observed in this study, can therefore be considered an enabler to quality NCD service delivery. In addition to more advanced care for patients, PHC physician referral and feedback may also be enhanced among colleagues at the secondary health facilities. However, it is important that physician-centric biomedical care delivery models are avoided. This requires physician’s roles to be expanded beyond direct clinical care for patients [54]. Such roles may include managerial and administrative roles, formal and on-the-job training roles, and supportive supervision for the health workforce team.
Despite their important role, employment of physicians at PHC facilities is not realistic in most states of Nigeria due to critical shortage, maldistribution (between rural and urban facilities, and between northern and southern states) and migration (out of the country) [36]. It is therefore important that, in PHC facilities with physicians, effective strategies and structures be put in place to limit the risk of physician attrition. Where employment of physicians is not feasible, there is the need for effective implementation of task-shifting and task-sharing with nurses and CHWs. The current situation in which tasks are shifted to nurses and CHWs will need to be transformed to one in which task-shifting and task-sharing is deliberate and supported with widely disseminated and regularly updated decision-support job aids. This has potential to improve NCD management as was done successfully with maternal health, HIV [55] and contraception service provision [37, 38].
## “Perverse incentives” for sustaining the functioning of PHC facilities
There is a general perception that health system funding is inadequate, and PHC facilities are the worst hit by this [56]. Multiple government agencies are involved in the financing of PHC service delivery but local governments are primarily responsible for the funding the day-to-day functioning of the health facility [57, 58]. Due to inconsistent, insufficient, or absolute non-release of funds, most local governments are unable to support PHC facilities beyond payment of salaries [14]. This may be one of the reasons why the management of various PHC facilities have devised unofficial or informal means as a workaround strategy to generate revenue internally for sustaining the daily function of the PHC facilities.
Such informal practices have previously been documented as often essential for the day-to-day functioning of PHC facilities in Nigeria [49, 50]. As shown in our study, for instance, funds generated from unauthorized sales of drugs have been used to engage and retain unemployed health workers on a contract basis, so as to make up for staff shortfalls. While this can superficially be judged as corruption, those directly affected may argue that it is a rational adaptation to the existing PHC governance environment where funding is heavily constrained to meet community need. If funding and human resources at these facilities remain the same, efforts to control or police perverse incentives (which help to sustain the functioning of otherwise sub-optimally supported PHC facilities) are likely to prove impractical.
## Intra-regional variation
A key finding from our study was the distinct delineation of functions and structures between PHC facilities in northern as compared southern states in Nigeria. This was evident in terms of staffing capacity, cadres of health workers in the PHC teams, training, medication availability and supply chain structures. This disparity of PHC facilities’ functionality across geographical zones, is also reflected in wide disparities in service provision–for example, $0.5\%$ of PHC facilities in the northeast Nigeria provides immunisation services compared to $90\%$ in southwest Nigeria [14]. Much of the inequalities have been linked to historical, political, ethno-religious and socioeconomic reasons [25]. It is estimated that over $70\%$ of Northern population lives below poverty line compared to less than $35\%$ in the South [59].
The uneven distribution of doctors, about 160 per million population in the North and 443 per million populations in the South have also been attributed to financial, conflict and social reasons [60]. Another possible explanation for these disparities is the greater capacity for public financing of health in the southern states as a result of disproportionate economic development when compared to the northern states [61]. This disparity reflects the situation of the health workforce in northern states where CHWs are more likely to be in charge of PHC facilities than nurses or physicians. Although we observed relative differences between north and south states, the absolute lack of physician-run PHC facilities is a national problem, as is the use of a large pool of volunteer CHW staff, a phenomenon that threatens service delivery quality, continuity, and sustainability not only for NCDs but for other PHC services.
An important policy direction for PHC in *Nigeria is* the development of new state-level PHC agencies to take over PHC governance from both local and national governments [62]. On the one hand, the state-level PHC agencies will centralise PHC management within a state by taking over responsibilities from local government councils–given their weak financial and technical capacity in most states. On the other hand, state-level PHC agencies may also take over direct policy guidance from the national PHC agency–thereby better tailoring PHC policies to local needs and circumstances [63].
Our study shows different models of PHC (CHEW vs Physician led) which vary significantly by context: northern states vs southern states; and rural communities vs urban communities. Our findings can inform the ongoing efforts of state PHC agencies to re-organise NCD service delivery in a way that not only reflects their greater capacity relative to local governments, but also reflects greater contextualisation to local needs and capacity.
## Limitations
While the findings of this study are not generalisable to all PHC facilities in Nigeria, the facilities included represent some of the geographical variation across the country. The findings provide a textured and contextualized understanding of the organisation of PHC services for NCDs in Nigeria drawing on a broad range of perspectives and observations. More so, our study focused primarily on the operational structure within which services are being delivered for NCDs. However, a limitation is that this study did not address patient care directly which would have been desirable to understand the perception of care recipients, and what may constitute barriers and enables of NCD services at the PHC level from their perspective. We recommend that future studies should explore PHC service organisation with a focus on the socioeconomics, political and governance structures that apply to NCD service delivery.
## Conclusion
Our study highlighted essential considerations in efforts to strengthen the PHC system for NCD service delivery in Nigeria. Priority considerations include: [1] Adequate funding and staffing of the PHC system to ensure optimal health workforce strength considering the regional, socio-political and economic variations. [ 2] Continuous capacity building of PHC health workers with focus on NCD prevention and management. [ 3] Implementing task-sharing and task-shifting policies for NCDs among non-physician health workers, with clear role delineation and promotion of inter-professional networks and collaboration [4] Development of NCD treatment guidelines and protocols, and making them available and accessible at PHC facilities, adapted to the cadre and mix of the workforce available at each PHC facility [5] Financial and technical investment into basic NCD equipment, essential NCD medicines (with the essential medicine list at the PHC level revised to reflect this) and medicine supply chain structures. [ 6] Strengthening referral linkages between PHC and higher-level facility, and between communities and PHC facilities while also effectively integrating NCDs services into existing PHC structures. We acknowledge that all these need to take place within the appropriate political and technical leadership that govern the PHC system.
## References
1. 1Federal Ministry of Health, Nigeria. National Multi-Sectoral Action Plan for the Prevention and Control of Non-communicable Diseases (2019–2025) [Internet]. Abuja, Nigeria: Federal Ministry of Health Nigeria; 2019 [cited 2020 Jun 17]. https://www.health.gov.ng/doc/NCDs_Multisectoral_Action_Plan.pdf. *National Multi-Sectoral Action Plan for the Prevention and Control of Non-communicable Diseases (2019–2025)* (2019.0)
2. Adeloye D, Basquill C, Aderemi A, Thompson J, Obi F. **An estimate of the prevalence of hypertension in Nigeria: a systematic review and meta-analysis**. *Journal of Hypertension* (2015.0) **33** 230-42. DOI: 10.1097/HJH.0000000000000413
3. 3World Health Organization. Global Action Plan for the Prevention and Control of Noncommunicable Diseases 2013–2020 [Internet]. Geneva: World Health Organization; 2013 [cited 2020 Jun 9]. https://apps.who.int/iris/bitstream/handle/10665/94384/9789241506236_eng.pdf?sequence=1. *Global Action Plan for the Prevention and Control of Noncommunicable Diseases 2013–2020* (2013.0)
4. Nations United. *Transforming our world: the 2030 Agenda for Sustainable Development* (2015.0)
5. 5World Health Organization. Tackling NCDs: “best buys” and other recommended interventions for the prevention and control of noncommunicable diseases. 2017 [cited 2021 Jul 14]; https://apps.who.int/iris/handle/10665/259232
6. 6Federal Ministry of Health, Nigeria. Second National Strategic Health Development Plan 2018–2022 [Internet]. Abuja, Nigeria: Federal Ministry of Health Nigeria; 2018 [cited 2020 Jun 25]. 150 p. https://health.gov.ng/doc/NSHDP_II_ME_Plan.pdf. *Second National Strategic Health Development Plan 2018–2022* (2018.0) 150
7. Aigbiremolen A, Alenoghena I, Eboreime E, Abejegah C. **Primary Health Care in Nigeria: From Conceptualization to Implementation**. *Journal of Medical and Applied Biosciences* (2014.0) **6**
8. Ekenna A, Itanyi I, Nwokoro U, Hirschhorn L, Uzochukwu B. **How ready is the system to deliver primary healthcare? Results of a primary health facility assessment in Enugu State, Nigeria**. *Health Policy and Planning* (2020.0) **35** 97-106. PMID: 33165588
9. 9Federal Government of Nigeria. National Health Act, 2014. 2014.
10. 10National Primary Health Care Development Agency. Ward minimum Health Package 2007–2012. 2007.
11. 11Federal Ministry of Health. National Health Facility Survey (NHFS) 2016 [Internet]. Abuja, Nigeria; 2016 [cited 2021 Jul 15] p. 174. http://somlpforr.org.ng/wp-content/uploads/2017/05/NHFS-Final-Report-for-Printing_VI.pdf
12. Shiroya V, Shawa N, Matanje B, Haloka J, Safary E, Nkhweliwa C. **Reorienting Primary Health Care Services for Non-Communicable Diseases: A Comparative Preparedness Assessment of Two Healthcare Networks in Malawi and Zambia**. *Int J Environ Res Public Health* (2021.0) **18** 5044. DOI: 10.3390/ijerph18095044
13. Akinwumi A, Esimai O, Fajobi O, Idowu A, Esan O, Ojo T. **Knowledge of primary healthcare workers regarding the prevention and control of non-communicable diseases in Osun State, Nigeria: A rural-urban comparison**. *Afr J Prim Health Care Fam Med* (2021.0) **13** 2873. PMID: 34212741
14. 14World Health Organization. Primary Health Care Systems (PRIMASYS): Case study from Nigeria. [Internet]. Geneva: World Health Organization; 2017 [cited 2020 Jul 15] p. 44. https://www.who.int/alliance-hpsr/projects/alliancehpsr_nigeriaprimasys.pdf?ua=1. *Primary Health Care Systems (PRIMASYS): Case study from Nigeria* (2017.0) 44
15. 15World Health Organization. Non-communicable Diseases Country Profiles 2018 [Internet]. Geneva: World Health Organization; 2018 [cited 2020 Jun 9] p. 224. https://www.who.int/nmh/publications/ncd-profiles-2018/en/. *Non-communicable Diseases Country Profiles 2018* (2018.0) 224
16. Maiyaki M, Garbati M. **The burden of non-communicable diseases in Nigeria; in the context of globalization**. *Ann Afr Med* (2014.0) **13** 1-10. DOI: 10.4103/1596-3519.126933
17. Adedoyin R, Mbada C, Balogun M, Martins T, Adebayo R, Akintomide A. **Prevalence and pattern of hypertension in a semiurban community in Nigeria**. *Eur J Cardiovasc Prev Rehabil* (2008.0) **15** 683-7. DOI: 10.1097/HJR.0b013e32830edc32
18. Erhabor G, Agbroko S, Bamigboye P, Awopeju O. **Prevalence of asthma symptoms among university students 15 to 35 years of age in Obafemi Awolowo University, Ile-Ife, Osun State**. *J Asthma* (2006.0) **43** 161-4. DOI: 10.1080/02770900500499046
19. Asekun-Olarinmoye E, Akinwusi P, Adebimpe W, Isawumi M, Hassan M, Olowe O. **Prevalence of hypertension in the rural adult population of Osun State, southwestern Nigeria**. *Int J Gen Med* (2013.0) **6** 317-22. DOI: 10.2147/IJGM.S42905
20. **Knowledge of diabetes and hypertension care among health care workers in southwest Nigeria**. *Postgrad Med* (2009.0) **121** 173-7. DOI: 10.3810/pgm.2009.01.1965
21. Odusola AO, Stronks K, Hendriks ME, Schultsz C, Akande T, Osibogun A. **Enablers and barriers for implementing high-quality hypertension care in a rural primary care setting in Nigeria: perspectives of primary care staff and health insurance managers**. *Global Health Action* (2016.0) **9** 29041. DOI: 10.3402/gha.v9.29041
22. Ajike S, Obot M. **Quality hypertension care: Barriers and drivers for implementation among primary health care staff**. *World Journal of Advanced Research and Reviews* (2021.0) **9** 018-26
23. 23National Primary Health Care Development Agency. Minimum Standards for Primary Health Care in Nigeria [Internet]. [cited 2021 Jul 21]. http://dc.sourceafrica.net/documents/120589-Minimum-Standards-for-Primary-Health-Care-in.html
24. Yin R.. *Case Study Research Design and Methods* (2014.0) 282
25. Eboreime E, Abimbola S, Bozzani F. **Access to Routine Immunization: A Comparative Analysis of Supply-Side Disparities between Northern and Southern Nigeria**. *PLOS ONE* (2015.0) **10** e0144876. PMID: 26692215
26. Oyekale A.. **Assessment of primary health care facilities’ service readiness in Nigeria**. *BMC Health Services Research* (2017.0) **17** 172. DOI: 10.1186/s12913-017-2112-8
27. 27Baldridge A, Huffman M, Guo M, Hirschhorn L, Kandula N. Adapted Service Availability and Readiness Assessment for the HTN Program. DigitalHub. Galter Health Sciences Library & Learning Center;
28. Orji I, Baldridge A, Omitiran K, Guo M, Ajisegiri W, Ojo T. **Capacity and site readiness for hypertension control program implementation in the Federal Capital Territory of Nigeria: a cross-sectional study**. *BMC Health Services Research* (2021.0) **21** 322. DOI: 10.1186/s12913-021-06320-8
29. Tong A, Sainsbury P, Craig J. **Consolidated criteria for reporting qualitative research (COREQ): a 32-item checklist for interviews and focus groups**. *Int J Qual Health Care* (2007.0) **19** 349-57. DOI: 10.1093/intqhc/mzm042
30. Brown J, Lewis L, Ellis K, Stewart M, Freeman T, Kasperski M. **Conflict on interprofessional primary health care teams—can it be resolved?**. *J Interprof Care* (2011.0) **25** 4-10. DOI: 10.3109/13561820.2010.497750
31. Grumbach K, Bodenheimer T. **Can health care teams improve primary care practice?**. *JAMA* (2004.0) **291** 1246-51. DOI: 10.1001/jama.291.10.1246
32. Payne M.. *Teamwork in Multiprofessional Care* (2000.0)
33. Alyahya S, Al-Mansour K, Alkohaiz M, Almalki M. **Association between role conflict and ambiguity and stress among nurses in primary health care centers in Saudi Arabia during the coronavirus disease 2019 pandemic: A cross-sectional study**. *Medicine* (2021.0) **100** e27294. PMID: 34664892
34. Bailey P, Jones L, Way D. **Family physician/nurse practitioner: stories of collaboration**. *J Adv Nurs* (2006.0) **53** 381-91. DOI: 10.1111/j.1365-2648.2006.03734.x
35. Praveen D, Patel A, Raghu A, Clifford G, Maulik P, Abdul A. **SMARTHealth India: Development and Field Evaluation of a Mobile Clinical Decision Support System for Cardiovascular Diseases in Rural India**. *JMIR mHealth and uHealth* (2014.0) **2** e3568. PMID: 25487047
36. 36Federal Ministry Of Health, Nigeria. National Human Resources for Health Strategic Plan 2008–2012 [Internet]. Federal Ministry of Health Nigeria; 2007 [cited 2021 Nov 12]. https://www.who.int/workforcealliance/countries/Nigeria_HRHStrategicPlan_2008_2012.pdf
37. Charyeva Z, Oguntunde O, Orobaton N, Otolorin E, Inuwa F, Alalade O. **Task Shifting Provision of Contraceptive Implants to Community Health Extension Workers: Results of Operations Research in Northern Nigeria**. *Global Health: Science and Practice* (2015.0) **3** 382-94. PMID: 26374800
38. Akinyemi O, Harris B, Kawonga M. **Health system readiness for innovation scale-up: the experience of community-based distribution of injectable contraceptives in Nigeria**. *BMC Health Services Research* (2019.0) **19** 938. DOI: 10.1186/s12913-019-4786-6
39. Porter-O’Grady T.. **Embracing Conflict: Building a Healthy Community**. *Health Care Management Review* (2004.0) **29** 181-7. PMID: 15357228
40. Ali M, Rabadán-Diehl C, Flanigan J, Blanchard C, Narayan K, Engelgau M. **Systems and capacity to address noncommunicable diseases in low- and middle-income countries**. *Sci Transl Med* (2013.0) **5** 181cm4. DOI: 10.1126/scitranslmed.3005121
41. Mahipala P, Dorji G, Tisocki K, Rani M. **A critical review of addressing cardiovascular and other non-communicable diseases through a primary health care approach in the South-East Asia Region**. *Cardiovascular Diagnosis and Therapy* (2019.0) **9** 15057-157. PMID: 31143636
42. Dunmade A, Afolabi O, Eletta A. **Challenges of Otolaryngologic Referral in a Nigerian Tertiary Hospital**. *East and Central African Journal of Surgery* (2010.0) **15** 87-92
43. Afolaranmi T, Hassan Z, Filibus D, Al-Mansur U, Lagi L, Kumbak F. **Referral System: An Assessment of Primary Health Care Centres in Plateau State, North Central Nigeria**. *WJRR* (2018.0) **6** 262704
44. Peck R, Mghamba J, Vanobberghen F, Kavishe B, Rugarabamu V, Smeeth L. **Preparedness of Tanzanian health facilities for outpatient primary care of hypertension and diabetes: a cross-sectional survey**. *Lancet Glob Health* (2014.0) **2** e285-292. DOI: 10.1016/S2214-109X(14)70033-6
45. Katende D, Mutungi G, Baisley K, Biraro S, Ikoona E, Peck R. **Readiness of Ugandan health services for the management of outpatients with chronic diseases**. *Trop Med Int Health* (2015.0) **20** 1385-95. DOI: 10.1111/tmi.12560
46. Li M, Zhang Y, Lu Y, Yu W, Nong X, Zhang L. **Factors influencing two-way referral between hospitals and the community in China: A system dynamics simulation model. SIMULATION**. *Internet]* (2018.0) **94** 765-82. DOI: 10.1177/0037549717741349
47. Okpetu E, Abimbola S, Koot J, Kane S. **Implementing prevention interventions for non-communicable diseases within the Primary Health Care system in the Federal Capital Territory, Nigeria**. *Journal of Community Medicine and Primary Health Care* (2018.0) **30** 1-18
48. Kamau K, Onyango-Osuga B, Njuguna S. **Challenges Facing Implementation Of Referral System For Quality Health Care Services In Kiambu County, Kenya**. *Health Systems and Policy Research* (2017.0) **4**
49. Abimbola S, Molemodile S, Okonkwo O, Negin J, Jan S, Martiniuk AL. **‘The government cannot do it all alone’: realist analysis of the minutes of community health committee meetings in Nigeria**. *Health Policy and Planning* (2016.0) **31** 332-45. DOI: 10.1093/heapol/czv066
50. Abimbola S, Ogunsina K, Charles-Okoli A, Negin J, Martiniuk A, Jan S. **Information, regulation and coordination: realist analysis of the efforts of community health committees to limit informal health care providers in Nigeria**. *Health Economics Review* (2016.0) **6** 51. DOI: 10.1186/s13561-016-0131-5
51. Jimenez-Carrillo M, León-García M, Vidal N, Bermúdez K, De Vos P. **Comprehensive primary health care and non-communicable diseases management: a case study of El Salvador**. *International Journal for Equity in Health* (2020.0) **19** 50. DOI: 10.1186/s12939-020-1140-x
52. Mash R, Howe A, Olayemi O, Makwero M, Ray S, Zerihun M. **Goodyear-Smith F. Reflections on family medicine and primary healthcare in sub-Saharan Africa**. *BMJ Global Health* (2018.0) **3** e000662. PMID: 29765778
53. Flinkenflögel M, Sethlare V, Cubaka V, Makasa M, Guyse A, De Maeseneer J. **A scoping review on family medicine in sub-Saharan Africa: practice, positioning and impact in African health care systems**. *Hum Resour Health* (2020.0) **18** 27. DOI: 10.1186/s12960-020-0455-4
54. Nkosi P, Horwood C, Vermaak K, Cosser C. **The role of doctors in provision of support for primary health care clinics in KwaZulu-Natal, South Africa**. *South African Family Practice* (2009.0) **51**
55. Tesema A, Ajisegiri W, Abimbola S, Balane C, Kengne A, Shiferaw F. **How well are non-communicable disease services being integrated into primary health care in Africa: A review of progress against World Health Organization’s African regional targets**. *PLOS ONE* (2020.0) **15** e0240984. PMID: 33091037
56. Ajisegiri W, Abimbola S, Tesema A, Odusanya O, Ojji D, Peiris D. **Aligning policymaking in decentralized health systems: Evaluation of strategies to prevent and control non-communicable diseases in Nigeria**. *PLOS Global Public Health* (2021.0) **1** e0000050
57. Gupta D, Gauri V, Khemani S. *Decentralized Delivery of Primary Health Services in Nigeria: Survey Evidence from the States of Lagos and Kogi* (2004.0) 86
58. 58Audu M. Primary Health Care In Nigeria: 42 Years After Alma Ata Declaration [Internet]. Sydani Initiative. 2020 [cited 2022 Jan 11]. https://sydani.org/primary-health-care-in-nigeria-42-years-after-alma-ata-declaration/
59. Ngbea G, Achunike H. **Poverty in Northern Nigeria**. *Asian Journal of Humanities and Social Studies* (2014.0) **02**
60. Adeyanju O, Tubeuf S, Ensor T. **Socio-economic inequalities in access to maternal and child healthcare in Nigeria: changes over time and decomposition analysis**. *Health Policy and Planning* (2017.0) **32** 1111-8. DOI: 10.1093/heapol/czx049
61. Sambo M, Ejembi C, Adamu Y, Aliyu A. **Out-of-pocket health expenditure for under-five illnesses in a semi-urban community in Northern Nigeria**. *Journal of Community Medicine and Primary Health Care* (2004.0) **16** 29-32
62. Eboreime E, Abimbola S, Obi F, Ebirim O, Olubajo O, Eyles J. **Evaluating the sub-national fidelity of national Initiatives in decentralized health systems: Integrated Primary Health Care Governance in Nigeria**. *BMC Health Serv Res* (2017.0) **17**. PMID: 28327123
63. Abimbola S, Baatiema L, Bigdeli M. **The impacts of decentralization on health system equity, efficiency and resilience: a realist synthesis of the evidence**. *Health Policy Plan* (2019.0) **34** 605-17. DOI: 10.1093/heapol/czz055
|
---
title: Estimation of seroprevalence of melioidosis among adult high risk groups in
Southeastern India by indirect Hemagglutination assay
authors:
- Sruthi Raj
- Sujatha Sistla
- Deepthy Melepurakkal Sadanandan
- Tamilarasu Kadhiravan
- Palanivel Chinnakali
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021966
doi: 10.1371/journal.pgph.0000431
license: CC BY 4.0
---
# Estimation of seroprevalence of melioidosis among adult high risk groups in Southeastern India by indirect Hemagglutination assay
## Abstract
Burkholderia pseudomallei is an environmental saprophyte known to cause melioidosis, a disease endemic in northern Australia and Southeast Asia. With the increasing number of melioidosis cases, there is a lack of data on seroprevalence rates and extent of exposure in high risk population of melioidosis from different endemic regions in India. The present cross sectional study was undertaken to estimate the seroprevalence of melioidosis in high risk populations in and around Puducherry, a coastal town in Southeastern India. Blood samples were collected from 275 diabetic individuals attending a tertiary care centre in Southern India and 275 farmers residing under the rural field practice area of our hospital. The antibody levels were estimated using an Indirect Hemagglutination Assay. The overall seropositivity was found to be $19.8\%$ with a titer ≥1:20. Farmers were 2.8 times more likely to be seropositive than non-farmers. Rates of seroprevalence among diabetic subjects were less compared to the non-diabetic individuals. The seropositivity rates in non-diabetic farmers were higher ($\frac{56}{203}$, $27.6\%$) compared to diabetic farmers ($\frac{34}{164}$, $20.7\%$). The lowest seropositivity was seen among diabetic non-farmers at $10.4\%$. Multivariable logistic regression analysis revealed domicile (adjusted odds ratio—aOR: 2.32, $95\%$ Confidence interval—CI: 1.05, 5.13) and contact with animals (aOR: 1.89, $95\%$ CI:1.04, 3.44) as significant predictors of seropositivity. None of the other socio-demographic factors including gender and age were significantly associated with seropositivity. This study demonstrates widespread exposure to B. pseudomallei among adults residing in and around Puducherry, including those engaged in non-farming occupations.
## Introduction
Burkholderia pseudomallei, the causative agent of melioidosis, is known to exist in soil and water in endemic regions of Southeast Asia and northern Australia [1]. Clinically, melioidosis presents with a wide spectrum of disease manifestations ranging from acute systemic infections to chronic localised forms, that mimic other infections and leads to misdiagnosis and treatment failure [1]. Melioidosis is acquired through inoculation, inhalation or aspiration and ingestion. Diabetes mellitus and agricultural activities are the two important risk factors for acquiring melioidosis [2]. A higher risk of exposure to B. pseudomallei is seen in rice farmers as they regularly come in contact with contaminated soil [3].
In India, an increasing number of melioidosis cases have been reported from different regions with the maximum number of cases being recognised from Karnataka and Tamil Nadu [4]. This need not necessarily exhibit the true picture of melioidosis in India due to lack of awareness among microbiologists and clinicians. Moreover, limited research and lack of access to well-equipped diagnostic laboratories add to the misdiagnosis of melioidosis cases [4]. India has the dubious distinction of being the diabetes capital of the world with a majority of the population residing in rural areas engaged in agricultural activities [1,4].
In the absence of disease, individuals exposed to B. pseudomallei present in soil and water become seropositive and develop antibodies to B. pseudomallei. However, these antibodies are non-protective [5,6]. Seroprevalence studies can be used to estimate the exposure to B. pseudomallei in a given geographic region. Several methods have been employed for this purpose including the indirect hemagglutination assay (IHA) using crude antigen derived from B. pseudomallei strains. From India, except for one study from Karnataka [3], no other seroprevalence studies have been reported. There is a lack of data in relation to seropositivity and high-risk population including diabetics and farmers. The seroprevalence of melioidosis is important especially in a coastal town like Puducherry since the number of melioidosis cases has been increasing over the years. As reported elsewhere, 34 melioidosis cases were found between January 2014 and December 2018, from a single institution in Puducherry [7]. During the same period, 31 cases were identified from another center in the same region [8]. Therefore, we carried out an indirect hemagglutination test, to estimate the seroprevalence of melioidosis among the two high-risk populations in and around Puducherry.
## Study population and source of sera
A cross-sectional study was conducted from January 2020 to January 2021. The sample size was estimated with an anticipated prevalence of $29\%$ (Manipal study) [3], an alpha error of $5\%$ ($95\%$ confidence level) and absolute precision of $6\%$, the required sample size was estimated to be 220. Considering non-response to blood sample collection (approx$.20\%$), the minimum sample size was estimated to be 275.
A total of 550 adults between 18 and 90 years of age, were enrolled in the study. The 275 adults who had attended the diabetic clinic, in a tertiary care center in Southern India were selected using consecutive sampling. For the remaining 275 adults involved in agricultural work and residing under the rural field practice area of our hospital (JIRHC, Ramanathapuram), one of the four villages under Ramanathapuram was randomly selected Convenient sampling was performed using a house-to-house survey and farmers were recruited based on their willingness to participate in the study. Two ml of blood sample was collected using venipuncture. Details of the study subjects such as their age, gender, occupation, area of residence, personal habits such as alcohol consumption and other socio-economic details were recorded in a structured proforma by a trained personal. Residence was characterised as urban if they were located within a metropolitan district or as a main city and as rural if they resided outside these areas [6]. Serum samples of the subjects were stored at -80°C until tested. IHA was performed as described previously [9]. Crude antigens were prepared in our laboratory from two clinical isolates of B. pseudomallei. Positive control was prepared using pooled serum from three culture-proven melioidosis cases with the titer of ≥1:10240. In the present study, chick erythrocytes were used instead of sheep red blood cells, as they are heavier (being nucleated) and tend to settle down faster during the agglutination reaction. Fresh chick erythrocytes were used in the assay without Double-aldehyde stabilisation (DAS). The optimal antigen dilution was determined against the known positive serum. Each sample was tested in triplicate to validate the test and an average was taken. A titer of ≥1:20 with agglutination of erythrocytes was considered positive [3].
## Ethics statement
This study was approved by the Institutional Ethics Committee for Human Studies, JIPMER, Puducherry (JIP/IEC/$\frac{2018}{0230}$) and written informed consent for all procedures was obtained from the participants.
## Statistical analysis
All statistical analyses were performed using SPSS software version 19.0 (IBM; Armonk, NY, USA). All the categorical variables were summarised as frequency and percentages. The age which does not follow normality was summarised as median with interquartile range. The normality assumption was checked using the Kolmogrov-smirnov test. Mann-Whitney U test was used to compare the median age across the seropositivity status. An Independent student t test was carried out to compare the log transformed titer values across gender. Chi-square or Fisher’s exact tests were performed to find the association of seropositivity with strata and with possible exposure to soil among the diabetic non-farmers. With the use of Univariate logistic regression, association between seropositivity status and known risk factors were determined. The unadjusted odds ratios and the $95\%$ confidence intervals (CIs) were estimated. The variables which were found to have a p-value <0.20 in univariate analyses were included in the multivariable logistic regression. The multivariable logistic regression analysis was performed and adjusted odds ratios along with their $95\%$ CIs were reported. All the statistical analyses were carried out at $5\%$ level of significance and a p-value less than 0.05 was considered to be statistically significant.
## Results
The overall seroprevalence using antibody titers ≥ 1:20 was found to be $19.8\%$ ($95\%$ CI: 13.72, 25.92). The picture of the IHA plate is depicted in (Fig 1). Out of 550 subjects enrolled in the study, there were 257 males and 293 females aged from 18 years to 90 years. The distribution of antibody titers among high risk population is shown in (Fig 2). The high risk population included 347 diabetic subjects and 367 farmers, as there were 72 diabetic subjects among 275 farmers and 92 farmers among 275 diabetic subjects. A total of 164 study subjects were diabetic farmers, 183 subjects were diabetic non-farmers and non-diabetic farmers were 203.
**Fig 1:** *IHA plate with antibody titers.PC- positive control, NC- negative control, T1 -T6 are the test samples. Antibody titers are provided on the right.* **Fig 2:** *Distribution of IHA titers among high risk population.*
The median ages of seropositive and seronegative participants were found to be 55(IQR 44,65 years) and 56(IQR 47,63.5 years) respectively. Among the different age groups, seropositivity of $5.5\%$ was found in the younger age group (18–30) compared to $56\%$ in the elderly, aged ≥ 51 years. Seropositivity among the three different strata; diabetic farmers ($$n = 164$$), diabetic non-farmers ($$n = 183$$) and non-diabetic farmers ($$n = 203$$) were found to be $20.7\%$ ($95\%$ CI:14.53–26.94), $10.4\%$ ($95\%$ CI: 5.71,15.05) and $27.6\%$ ($95\%$ CI: 20.75, 34.43) respectively. This result was found to be statistically significant ($p \leq 0.001$). Among the diabetic non-farmers ($$n = 183$$), the majority of individuals $89.1\%$ ($\frac{163}{183}$) were professionals (Government and private employees, bank employees, engineers, teachers, etc). Among the diabetic non-farmers, $10.0\%$ ($\frac{2}{20}$) of subjects with possible exposure to soil were seropositive compared to $10.4\%$ ($\frac{17}{163}$) of subjects who were unlikely to have occupational exposed to soil/water (p-value = 1.00).
An antibody titer of ≥1:10 was found in 32 additional subjects, which would increase the overall seropositivity to $25.6\%$ ($95\%$ CI: 21.99, 29.29). Table 1 displays the results of the univariate logistic regression model for each risk factor, their associated seropositivity status, unadjusted odds ratios and its $95\%$ confidence intervals. Seropositivity was not significantly associated with various factors such as gender, different age groups, smoking and alcohol intake, activities near water bodies, gardening and use of footwear. There was no significant difference between the geometric mean titre values among males 33.02 ($95\%$ CI:25.89, 42.12) and females 42.78 ($95\%$ CI:34.03, 53.77), ($$p \leq 0.14$$). Odds of seropositivity was 2.8 folds higher in farmers than non-farmers (OR:2.80,$95\%$ CI: 1.65–4.77, $p \leq 0.001$). The odds of seropositivity was 2.85 folds higher in subjects who had contact with animals than who did not have animal contact (OR:2.85,$95\%$ CI: 1.77, 4.60, $p \leq 0.001$) and in rural dwellers odds of seropositivity was 3.69 folds higher than those who lived in urban districts (OR:3.69,$95\%$ CI:2.00, 6.81, $p \leq 0.001$). However, the odds of having seropositivity was $53\%$ lower in the diabetic group when compared to non-diabetic group (OR:0.47,$95\%$ CI: 0.31, 0.72, $$p \leq 0.001$$). Contact with animals and domicile were identified to be significant independent predictors through multivariable logistic regression analysis (Table 1). Rural dwellers had 2.32 folds higher odds of having seropositivity than those who lived in urban districts (aOR: 2.32, $95\%$ CI:1.05, 5.13, $$p \leq 0.04$$) when adjusted for other variables in the model. The odds of having seropositivity was 1.89 folds higher in people who had contact with animals when compared to those who had no contact with animals when adjusted for other variables in the model. ( aOR: 1.89, $95\%$ CI: 1.04, 3.44, $$p \leq 0.04$$).
**Table 1**
| Risk factorsn = 550 | Seropositivity# (≥20)n(%) = 109 (19.8)* | Seronegativity (<20)n(%) = 441(80.2)* | Unadjusted OR(95% CI) | Adjusted OR$(95% CI) |
| --- | --- | --- | --- | --- |
| GenderMale (257)Female (293) | 47(43.1)62(56.9) | 210(47.6)231(52.4) | 1.001.20 (0.79–1.83) | - |
| Age categories18–30 (23)31–40 (55)41–50 (124)≥51 (348) | 6 (5.5)13 (11.9)29 (26.6)61 (56.0) | 17 (3.9)42 (9.5)95 (21.5)287 (65.1) | 1.000.88 (0.29, 2.69)0.87 (0.31, 2.40)0.60 (0.23, 1.59) | - |
| DiabeticsYes (347)No (203) | 53 (48.6)56 (51.4) | 294 (66.7)147 (33.3) | 0.47 (0.31, 0.72)1 | 0.82 (0.49, 1.37)1 |
| OccupationFarmers (367)Non-farmers (183) | 90 (82.6)19 (17.4) | 277 (62.8)164 (37.2) | 2.80 (1.65, 4.77)1 | 1.89 (0.78, 4.58)1 |
| SmokingYes (86)No (464) | 17 (15.6)92 (84.4) | 69 (15.6)372 (84.4) | 1.00 (0.56, 1.78)1 | - |
| Alcohol intakeYes (123)No (427) | 26 (23.9)83 (76.1) | 97 (22.0)344 (78.0) | 1.11 (0.68, 1.82)1 | - |
| Activities near water bodiesYes (349)No (201) | 76 (69.7)33 (30.3) | 273 (61.9)168 (38.1) | 1.42 (0.90, 2.23)1 | 0.81 (0.48, 1.36)1 |
| GardeningYes (382)No (168) | 84 (77.1)25 (22.9) | 298 (67.6)143 (32.4) | 1.61 (0.99, 2.63)1 | 0.55 (0.27, 1.11)1 |
| Wear footwearNo(318)Yes (232) | 72 (66.1)37 (33.9) | 246 (55.8)195 (44.2) | 1.54 (1.00, 2.39)1 | 0.94 (0.55, 1.60)1 |
| Contact with animalsYes (316)No (234) | 83 (76.1)26 (23.9) | 233 (52.8)208 (47.2) | 2.85 (1.77, 4.60)1 | 1.89 (1.04, 3.44)1 |
| DomicileRural (390)Urban (160) | 96 (88.1)13 (11.9) | 294 (66.7)147 (33.3) | 3.69 (2.00, 6.81)1 | 2.32 (1.05, 5.13)1 |
## Discussion
India is an endemic country for melioidosis with the successful isolation of B. pseudomallei from soil and water [1,10–12]. The present study provides new data involving a high risk population of diabetics and farmers, to determine the extent of exposure to B. pseudomallei. Overall seropositivity of $19.8\%$ was noted in the study population. The IHA is a common serological test employed for seroepidemiological studies since it was first described in 1965, as it is simple, rapid and effective for the detection of antibodies [13,14]. However, IHA is also known for poor specificity and sensitivity due to high background seropositivity in melioidosis endemic countries [15]. In the present study a cut off titer of ≥ 1:20 was taken, to enhance the specificity of the test as described in a study from Southern India [3]. According to these authors, increasing the cut-off point may improve the specificity slightly, but it may affect the test sensitivity [16]. On the other hand, lowering the cut-off increases sensitivity as evidenced in the present study (seropositivity increased to $25.6\%$ when a titer of ≥1:10 is taken as cut-off). The chick erythrocytes were not stabilised during the assay since fresh chick blood was used throughout the test, which prevented the lysis of RBCs. However, this did not affect the test as the positive well displayed loose button formations with ragged or folded edges (Fig 1).
The basic principle of IHA is the agglutination of erythrocytes, in the presence of serum antibodies to a variety of polysaccharide and lipopolysaccharide antigens, prepared using clinical isolates of B. pseudomallei [17]. Due to a wide degree of genetic diversity and antigenic variation, the use of pooled antigens from different clinical strains is preferred over a single strain [6]. The crude antigen used is a polysaccharide component, from the slime layer of the pathogen [18]. The sensitivity and specificity of IHA rely on concentrations and combinations of the crude antigens used in the assay [5]. There are speculations regarding cross-reacting antibodies to B. thailandensis, commonly found in soil along with B. pseudomallei.
The overall seropositivity was found to be $19.8\%$ in the current study, which is comparable to the previous study reported from Southern India ($29\%$) [3], and from endemic countries such as Thailand ($38\%$) [6], Australia ($3\%$) [19], Vietnam (6.4–$31.8\%$) [20], Bangladesh ($28.9\%$) [21]. These variations in seropositivity rates may be associated with the use of more sensitive serological techniques and the nature of crude antigens used in the IHA test [22]. Indirect enzyme-linked immunosorbent assay (ELISA) and complement fixation test are other serological tests used for serosurveillance studies [22,23]. ELISA was found to be the more sensitive method. However, the presence of non-specific cross-reactive antibodies that reacted with the crude antigen in indirect ELISA, resulted in a higher seropositivity rate in a study from Bangladesh [23]. Cross-reacting antibodies were also reported with the use of complement fixation test [22].
India has the world’s largest population diagnosed with diabetes and $25\%$ of this population lives in rural settings with a predicted increase in diabetes from 51 million in 2010 to 87 million in 2030. [ 1,3]. Although diabetes is a major risk factor for acquiring melioidosis, no remarkable difference in seropositivity was found among diabetic individuals $48.6\%$ ($\frac{53}{109}$) compared to non-diabetic individuals $51.4\%$ ($\frac{56}{109}$). Similar results were obtained in a study from Karnataka [3]. Individuals with diabetes are known to be immunocompromised [6,17]. Hence, even after exposure the specific antibodies are produced at low levels and may not be detected in the IHA test.
A strong association is found between seropositivity and exposure to soil and water [24]. Farming is a major occupation in Southern India. Farmers are at a higher risk of acquiring melioidosis as their exposure to organisms present in the environment occurs repeatedly over a long period [2]. In the current study, we found farmers were significantly associated with seropositivity compared to non-farmers. Similar reports were found in farmers in Bangladesh [21]. In Thailand, farmers were 4.6 times more likely to be seropositive compared to non-farmers [6]. In contrast, no difference in seropositivity among farmers and non-farmers was reported from Karnataka [3] and Bangladesh [23]. According to a study from Thailand, diabetic farmers had a six to nine times higher risk of acquiring melioidosis compared to non-diabetic non-farmers, as impaired immunity in diabetes and high exposure in farmers act synergistically as risk factors to develop melioidosis [2]. In the present study, seropositivity in non-diabetic farmers was higher $27.6\%$ ($\frac{56}{203}$) compared to diabetic farmers $20.7\%$ ($\frac{34}{164}$). However, seropositivity was least among diabetic non-farmers $10.4\%$ ($\frac{19}{183}$). This evidence further supports that farmers are highly exposed and are at higher risk of acquiring melioidosis. Even though diabetic farmers are equally exposed to contaminated soil and water, impaired immunity in diabetes lowers the detectable amount of antibodies to IHA and hence reduces their seropositivity rates. However, majority ($89.1\%$) of individuals among the diabetic non-farmers were professionals where exposure to soil/water was unlikely, which explains the low seropositivity among them. However, no significant difference was observed in the seropositivity between the groups with possible and unlikely occupational exposure to soil/water. This lack of difference may be explained in two ways. One was the vast difference in the numbers (20 with occupational exposure vs 163 without occupational exposure) making comparison difficult. The second reason could be undocumented non occupational exposure to soil and water in the latter group.
Among other demographic factors studied, individuals living in rural areas and those having contact with animals were found to be significantly associated with seropositivity. Most of the residents in rural settings are farmers and they come in contact with animals at home and during work. They frequently get exposed to contaminated soil and water during agricultural activities and while helping animals graze the grasslands. This is further supported by a study from Thailand, where rural dwellers were three-fold more likely to be seropositive compared to the residents of urban districts [6]. Subjects from the urban districts lack exposure, moreover, a defective immune response is likely possible [25]. Also, over representation of rural dwellers in the study could lead to a lower comparative rate. None of the other personal habits such as smoking, alcohol intake, activities near water bodies, gardening and use of footwear were significantly associated with seropositivity in the present study. However, Vandana et al., in the Manipal study, reported gardening and activities near the river were positively associated with exposure [3]. Another study from Taiwan found that walking barefoot and flooding caused high risk of exposure [26].
The relationship between age and seropositivity is unclear. We found higher seropositivity rates in the elderly, aged ≥51 years compared to younger age groups which could reflect repeated exposure. A similar finding was reported from Bangladesh [23] where seropositivity of $30.4\%$ was detected in patients over 50 years of age. In Thailand, seropositivity rates increased with advancing age in children [27–29]. On the other hand, an inverse relationship between seroprevalence rate and advancing age was found in Australia [30]. Another study from Southern India reported no significant differences in seropositivity rates in adults among different age groups [3]. However, IHA is considered to be more specific in young children compared to older age groups due to low background seroprevalence in neonates and young infants [25].
The study found no significant association of seropositivity with gender. There are conflicting reports of gender wise seropositivity rates with Australia [5] and Bangladesh [23] showing similar rates in both sexes while a study from southern India [3] reported a significantly higher seropositivity rate in females. Equal or higher seropositivity rates in females are contradictory to the prevalence of melioidosis which is far higher in males. According to Vandana et al., the possible reason for this could be fewer outdoor activities and exposure to low inoculum among females which might cause a seroconversion in the absence of disease, compared to males, who are likely to be exposed to a larger inoculum resulting in overt disease [3].
Globally, constant serosurveillance among high risk and rural populations is essential to monitor the pathogens that are of public health concern. As per the current study, seropositivity rates were higher among rice farmers. In a country like India where agriculture plays a vital role in the livelihood of most of the rural population, the farmers working in the fields could be recognised and monitored for the prevalence and incidence of melioidosis in their work location. Additionally, the current serosurveillance study provides useful information to plan and improvise targeted surveillance and to initiate the environmental control measures which could benefit the policymakers in the decision-making process. The public health professionals could focus on methods to create awareness among farmers about the risk of acquiring melioidosis and educate the population at risk.
The study had a few limitations. The human population surveyed in our study consisted of both diabetics and farmers who were at a higher risk of exposure to B. pseudomallei. The healthy adult population and children were excluded. Hence, seropositivity in non-diabetic non-farmers and for the younger age group could not be determined. Additionally, IHA used in the present study is known to have variable sensitivity and specificity with respect to the crude antigens used and the cross-reacting antibodies.
Further studies are required to determine seroprevalence using more sensitive and specific techniques and identify high risk populations for melioidosis from different regions in India. Along with this, environmental sampling for the presence of B. pseudomallei will help in estimating the true burden of melioidosis in India.
## Conclusions
This study has unequivocally demonstrated the evidence of exposure to B. pseudomallei in adults in and around Puducherry and strongly indicates that these individuals have the potential to develop melioidosis during their lifetime. There is a need for strengthening the laboratory facilities in rural areas. Besides, clinicians must consider melioidosis as a differential diagnosis in these exposed high risk populations to initiate early treatment for successful management of these cases. The observed seropositivity rates in high risk groups ascertain the usefulness of IHA as a promising diagnostic tool for seroepidemiological surveys.
## References
1. Gassiep I, Armstrong M, Norton R. **Human melioidosis.**. *Clin Microbiol Rev* (2020.0) **33** e00006-19. DOI: 10.1128/CMR.00006-19
2. Suputtamongkol Y, Chaowagul W, Chetchotisakd P, Lertpatanasuwun N, Intaranongpai S, Ruchutrakool T. **Risk Factors for Melioidosis and Bacteremic Melioidosis**. *Clin Infect Dis* (1999.0) **29** 408-413. DOI: 10.1086/520223
3. Vandana KE, Mukhopadhyay C, Tellapragada C, Kamath A, Tipre M, Bhat V. **Seroprevalence of**. *PLoS Negl Trop Dis* (2016.0) **10** e0004610. DOI: 10.1371/journal.pntd.0004610
4. Mukhopadhyay C, Shaw T, Varghese G, Dance D. **Melioidosis in South Asia (India, Nepal, Pakistan, Bhutan and Afghanistan).**. *Trop Med Infect Dis* (2018.0) **3** 51. DOI: 10.3390/tropicalmed3020051
5. Lazzaroni SM, Barnes JL, Williams NL, Govan BL, Norton RE, LaBrooy JT. **Seropositivity to**. *Trans R Soc Trop Med Hyg* (2008.0) **102** S66-S70. DOI: 10.1016/S0035–9203(08)70018-X
6. Chaichana P, Jenjaroen K, Amornchai P, Chumseng S, Langla S, Rongkard P. **Antibodies in melioidosis: The Role of the Indirect Hemagglutination Assay in Evaluating Patients and Exposed Populations**. *Am J Trop Med Hyg* (2018.0) **99** 1378-1385. DOI: 10.4269/ajtmh.17-0998
7. Sharma G, Viswanathan S. **Melioidosis: A 5-year Review from a Single Institution in Pondicherry.**. *J Assoc Physicians India.* (2021.0) **69** 11-12. PMID: 34585891
8. Basheer A, Iqbal N, C S, Kanungo R, Kandasamy R. **Melioidosis: distinctive clinico-epidemiological characteristics in southern India.**. *Trop Doct* (2021.0) **51** 174-177. DOI: 10.1177/0049475520943698
9. 9Sharon Peacock, V. Wuthiekanun. Standard operating procedure (SOP) of indirect haemagglutination assay (IHA) for melioidosis. Mahidol-Oxford Tropical Medicine Research Unit (MORU). Faculty of Tropical Medicine, Mahidol University. 2011. Available:http://www.melioidosis.info/download/MICRO_SOP_IHA_ENG_v1%203_8Dec11_SDB.pdf.
10. Prakash A, Thavaselvam D, Kumar A, Kumar A, Arora S, Tiwari S. **Isolation, identification and characterization of**. *SpringerPlus.* (2014.0) **3** 438. DOI: 10.1186/2193-1801-3-438
11. Peddayelachagiri BV, Paul S, Nagaraj S, Gogoi M, Sripathy MH, Batra HV. **Prevalence and Identification of**. *PLoS Negl Trop Dis.* (2016.0) **10** e0004956. DOI: 10.1371/journal.pntd.0004956
12. Chandrasekar S, Dias M. **Soil Sampling of**. *J Bacteriol Mycol* (2017.0) **4** 1046
13. Harris PNA, Williams NL, Morris JL, Ketheesan N, Norton R E. **Evidence of**. *Clin. Vaccine Immunol* (2011.0) **18** 1288-1291. DOI: 10.1128/CVI.00077-11
14. Cheng AC, O’Brien M, Freeman K, Lum G, Currie BJ. **Indirect Hemagglutination Assay In Patients With Melioidosis In Northern Australia**. *Am J Trop Med Hyg* (2006.0) **74** 330-334. DOI: 10.4269/ajtmh.2006.74.330
15. Tiyawisutsri R, Peacock SJ, Langa S, Limmathurotsakul D, Cheng AC, Chierakul W. **Antibodies from Patients with Melioidosis Recognize**. *J Clin Microbiol* (2005.0) **43** 4872-4874. DOI: 10.1128/JCM.43.9.4872-4874.2005
16. Norazah A, Rohani MY, Chang PT, Kamel AGM. **Indirect Hemagglutination antibodies against**. *Malasiya. Southeast Asian J Trop Med Public Health1996* **27** 263-266. PMID: 9279987
17. Harris PNA, Ketheesan N, Owens L, Norton RE. **Clinical Features That Affects Indirect- Hemagglutination-Assay Responses to**. *Clin Vaccine Immunol* (2009.0) **16** 924-930. DOI: 10.1128/CVI.00026-09
18. Gilmore G, Barnes J, Ketheesan N, Norton R. **Use of Antigens Derived from**. *Clin Vaccine Immunol* (2007.0) **14** 1529-1531. DOI: 10.1128/CVI.00197-07
19. James GL, Delaney B, Ward L, Freeman K, Mayo M, Currie BJ. **Surprisingly Low Seroprevalence of**. *Clin Vaccine Immunol* (2013.0) **20** 759-760. DOI: 10.1128/CVI.00021-13
20. Van Phung L, Quynh HT, Yabuuchi E, Dance DA. **Pilot study of exposure to**. *Trans R Soc Trop Med Hyg* (1993.0) **87** 416. DOI: 10.1016/0035-9203(93)90017-k
21. Maude RR, Maude RJ, Ghose A, Amin MR, Islam MB, Ali M. **Seroepidemiological surveillance of Burkholderia pseudomallei in Bangladesh**. *Trans R Soc Trop Med Hyg* (2012.0) **106** 576-578. DOI: 10.1016/j.trstmh.2012.06.003
22. Embi N, Suhaimi A, Mohamed R, Ismail G. **Prevalence of Antibodies to**. *Microbiol Immunol* (1992.0) **36** 899904. DOI: 10.1111/j.1348-0421.1992.tb02092.x
23. Jilani MS, Robayet JA, Mohiuddin M, Hasan MR, Ahsan CR, Haq JA. *PLoS Negl Trop Dis* (2016.0) **10** e0004301. DOI: 10.1371/journal.pntd.0004301
24. Cheng AC, Wuthiekanun V, Limmathurotsakul D, Chierakul W, Peacock SJ. **Intensity of exposure and incidence of melioidosis in Thai children**. *Trans R Soc Trop Med Hyg* (2008.0) **102** S37-S39. DOI: 10.1016/S0035-9203(08)70010-5
25. Kanaphun P, Thirawattanasuk N, Suputtamongkol Y, Naigowit P, Dance DA, Smith MD. **Serology and Carriage of**. *J Infect Dis* (1993.0) **167** 230-233. DOI: 10.1093/infdis/167.1.230
26. Su HP, Yang HW, Chen YL, Ferng TL, Chou YL, Chung TC. **Prevalence of Melioidosis in the Er-Ren River Basin, Taiwan: Implications for Transmission**. *J Clin Microbiol* (2007.0) **45** 2599-2603. DOI: 10.1128/JCM.00228-07
27. Wuthiekanun V, Chierakul W, Langa S, Chaowagul W, Panpitpat C, Saipan P. **Development of Antibodies to**. *Am J Trop Med Hyg* (2006.0) **74** 1074-1075. PMID: 16760522
28. Charoenwong P, Lumbiganon P, Puapermpoonsiri S. **The Prevalence of the Indirect Hemagglutination Test for Melioidosis in Children in an Endemic Area.**. *Southeast Asian J Trop Med Public Health* (1992.0) **23** 698-701. PMID: 1284319
29. Wuthiekanun V, Pheaktra N, Putchhat H, Sin L, Sen B, Kumar V. *Emerg Infect Dis* (2008.0) **14** 301-303. DOI: 10.3201/eid1402.070811
30. Armstrong PK, Anstey NM, Kelly PM, Currie BJ, Martins N, Dasari P. **Seroprevalence of**. *Southeast Asian J Trop Med Public Health* (2005.0) **36** 1496-1502. PMID: 16610652
|
---
title: 'Grpel2 maintains cardiomyocyte survival in diabetic cardiomyopathy through
DLST-mediated mitochondrial dysfunction: a proof-of-concept study'
authors:
- Rongjin Yang
- Xiaomeng Zhang
- Yunyun Zhang
- Yingfan Wang
- Man Li
- Yuancui Meng
- Jianbang Wang
- Xue Wen
- Jun Yu
- Pan Chang
journal: Journal of Translational Medicine
year: 2023
pmcid: PMC10021968
doi: 10.1186/s12967-023-04049-y
license: CC BY 4.0
---
# Grpel2 maintains cardiomyocyte survival in diabetic cardiomyopathy through DLST-mediated mitochondrial dysfunction: a proof-of-concept study
## Abstract
### Background
Diabetic cardiomyopathy (DCM) has been considered as a major threat to health in individuals with diabetes. GrpE-like 2 (Grpel2), a nucleotide exchange factor, has been shown to regulate mitochondrial import process to maintain mitochondrial homeostasis. However, the effect and mechanism of Grpel2 in DCM remain unknown.
### Methods
The streptozotocin (STZ)-induced DCM mice model and high glucose (HG)-treated cardiomyocytes were established. Overexpression of cardiac-specific Grpel2 was performed by intramyocardial injection of adeno-associated virus serotype 9 (AAV9). Bioinformatics analysis, co-immunoprecipitation (co-IP), transcriptomics profiling and functional experiments were used to explore molecular mechanism of Grpel2 in DCM.
### Results
Here, we found that Grpel2 was decreased in DCM induced by STZ. Overexpression of cardiac-specific Grpel2 alleviated cardiac dysfunction and structural remodeling in DCM. In both diabetic hearts and HG-treated cardiomyocytes, Grpel2 overexpression attenuated apoptosis and mitochondrial dysfunction, including decreased mitochondrial ROS production, increased mitochondrial respiratory capacities and increased mitochondrial membrane potential. Mechanistically, Grpel2 interacted with dihydrolipoyl succinyltransferase (DLST), which positively mediated the import process of DLST into mitochondria under HG conditions. Furthermore, the protective effects of Grpel2 overexpression on mitochondrial function and cell survival were blocked by siRNA knockdown of DLST. Moreover, Nr2f6 bond to the Grpel2 promoter region and positively regulated its transcription.
### Conclusion
Our study provides for the first time evidence that Grpel2 overexpression exerts a protective effect against mitochondrial dysfunction and apoptosis in DCM by maintaining the import of DLST into mitochondria. These findings suggest that targeting Grpel2 might be a promising therapeutic strategy for the treatment of patients with DCM.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12967-023-04049-y.
## Introduction
Diabetic cardiomyopathy (DCM), which is a serious microvascular complication of diabetes, is a major cause of mortality in individuals with diabetes worldwide [1]. DCM is characterized by cardiac dysfunction and adverse structural remodeling, including decreased contractile function, decreased diastolic function, cardiac hypertrophy, interstitial fibrosis, mitochondrial dysfunction and cardiomyocyte apoptosis [2]. These characteristics of DCM eventually result in heart failure, increasing cardiovascular morbidity and mortality in patients with diabetes [2, 3]. However, the initiation and progression of DCM have not been thoroughly investigated, and effective therapies to prevent the development of DCM remain extremely limited.
Mitochondria are a major source of cellular energy production, and mitochondrial dysfunction is a central event in DCM [4]. Although mitochondria have a dedicated genome that encodes 13 mitochondrial proteins, more than $90\%$ of mitochondrial proteins are encoded by the nuclear genome and require import into mitochondria in mammals [5]. Increasing evidence suggests that the mitochondrial import process plays a vital role in the maintenance of mitochondrial homeostasis in the heart under physiological and pathological conditions [6]. Disruption of the mitochondrial import process leads to impaired proteostasis, decreased ATP production, harmful ROS accumulation and excessive mitochondrial fission, ultimately causing cellular apoptosis [6, 7]. Mitochondrial heat shock protein 70 (mtHSP70) is considered to be a core component of the presequence translocase-associated motor (PAM) [8]. GrpE-like 2 (Grpel2), a nucleotide exchange factor, interacts with mtHSP70 to facilitate the transport of peptide-containing proteins from the inner membrane into the mitochondrial matrix [9]. Previous studies have reported that Grpel2 is involved in mitochondrial stress responses in many tissues [10, 11]. For example, Grpel2 senses oxidative stress and forms dimers that enhance the activity of mtHSP70 to ensure proper mitochondrial protein import and folding [11]. In addition, in glioma cells, Grpel2 ablation stimulates autophagy and senescence, which suppress cell growth [10]. Importantly, our previous study reported that Grpel2 could alleviate myocardial ischaemia/reperfusion injury by inhibiting MCU-mediated mitochondrial calcium ([Ca2+]m) overload [12]. Given that Grpel2 is a crucial regulator of mitochondrial homeostasis in the heart, it is important to explore the role of Grpel2 in cardiac dysfunction and adverse structural remodeling in DCM.
In the present study, we investigated the role of Grpel2 in the initiation and progression of DCM. We observed that the downregulation of Grpel2 might be a critical factor contributing to heart dysfunction and structural remodeling in DCM. We identified Grpel2 overexpression as a potential therapeutic strategy for DCM. Furthermore, we explored the underlying transcriptional and molecular mechanisms by which Grpel2 regulates normal mitochondrial bioenergetics, oxidative stress and cardiomyocyte survival in the heart during diabetes.
## Animals
All animal experiments were carried out according to the National Institutes of Health Guidelines on the Use of Laboratory Animals (NIH Publication, 8th Edition, 2011) and were approved by the Animal Care Committee of the Second Affiliated Hospital of Xi'an Medical University. All male C57BL/6 J mice were maintained in a temperature-controlled environment at 22 ± 2 °C with a 12 h light–dark cycle and free access to food and water. A total number of 108 mice were included in this study.
## Experimental diabetes model
Diabetes mellitus was induced in male 8-week-old C57BL/6 J mice with consecutive daily intraperitoneal injections of streptozotocin (STZ, 50 mg/kg dissolved in 0.1 mol/l citrate buffer, pH 4.5, S0130, Sigma‒Aldrich, USA) for 5 days, as previously described [13]. On the 7th day after the STZ injection, the mice were considered diabetic mice in subsequent experiments if their fasting blood glucose levels were > 11.1 mmol/l. The control mice were intraperitoneally injected with citrate buffer (vehicle). Mice were used for subsequent experiments for 12 weeks post-induction.
## Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted with TRIzol reagent (15596026, Thermo Fisher Scientific, USA) and was reverse-transcribed to cDNA via a PrimeScript™ RT Reagent Kit with gDNA Eraser (RR047A, Takara, Japan). qRT-PCR was performed using SYBR® Premix Ex Taq™ II (RR820A, Takara, Japan) on an Applied Biosystems ABI Prism machine according to the manufacturer’s protocols. Gene expression was detected by using the standard comparative CT method. β-actin was used as an internal control to normalize gene expression. The primer sequences were as follows:GeneForwardReverseGrpel2GAGGACCCTCCTGATGGACTTAATGGTGCGGCCATGAAGTβ-actinAACAGTCCGCCTAGAAGCACCGTTGACATCCGTAAAGACCAnpGATTTCAAGAACCTGCTAGACCAGTTTGCTTTTCAAGAGGGCBnpCACCCAAAAAGAGTCCTTCGCAACAACTTCAGTGCGTTACCol1a1GATGGATTCCCGTTCGAGTACGCTGTTCTTGCAGTGATAGCol3a1ATGGTGGTTTTCAGTTCAGCGCCTTGAATTCGCCTTCATTTgfb1GACCTCAAGAGCTCTAACATCCGTCATCCACAGACAGAGTAGGNr2f6TCCAGGATGGAGGGTCCAATCCCACCATCCCACAAGTTCA
## Western blotting analysis
Cells or tissues were homogenized in RIPA lysis buffer supplemented with protease and phosphatase inhibitors (78440, Thermo Fisher Scientific, USA). After centrifugation at 4 °C for 10 min at 10,000 × g, the protein concentration of the supernatant was determined with a BCA Protein Assay Kit (23225, Thermo Fisher Scientific, USA). A total of 15–30 μg of protein from each sample was subjected to electrophoresis on $12\%$ SDS‒PAGE gels and then transferred to PVDF membranes. The membranes were blocked in $5\%$ nonfat milk for 1 h and then incubated with primary antibodies overnight at 4 °C. After incubation with HRP-conjugated IgG secondary antibodies for 1 h at room temperature, protein bands were detected using an enhanced chemiluminescence detection kit (Millipore, USA). ImageLab software version 5.1 (Bio-Rad, USA) was used to quantify the protein band intensity. The following primary antibodies were used as followed: β-actin (1:10,000, 4970S, CST, USA); DLST (1:10,000, Ab177934, Abcam, USA); Grpel2 (1:1000, PA5-54723, Invitrogen, USA); Nr2f6 (1:10,000, Ab137496, Abcam, USA); TOMM20 (1:10,000, 42406, CST, USA).
## Intramyocardial adeno-associated virus injection
Adeno-associated virus serotype 9 (AAV9), under the control of a cardiac troponin T promoter (cTnT), was designed and constructed by Tsingke Biotechnology Co., Ltd. (Beijing, China). The AAV9 vectors were suspended in PBS at approximately 3.0*1011 PFU/ml. All male mice were anaesthetized with inhaled $2.5\%$ isoflurane and maintained with $2.0\%$ isoflurane during the surgery. After exposing the heart, 10 μl of AAV9 vector was injected into four different sites of the free wall of the left ventricle, as previously described [12]. 12 weeks after AAV injection, transfection efficiency was measured by Western blotting and qRT-PCR analysis. Then, the mice underwent STZ or vehicle injection as described above.
## Echocardiography and hemodynamics
Echocardiography was performed in M-mode with a VEVO 2100 echocardiography system (VisualSonics Inc., Canada) as previously described [14]. During the procedure, mice were anaesthetized using $2.5\%$ isoflurane and maintained with $2\%$ isoflurane. Heart rate was measured by a continuous electrocardiogram monitoring system under conscious or anesthetic condition. Heart images were viewed in the long-axis and short-axis between the two papillary muscles to determine the left ventricular systolic and diastolic motion profile. The left ventricular ejection fraction (LVEF), fraction shortening (LVFS), systolic internal dimension (LVISD) and left ventricular diastolic internal dimension (LVIDD) were automatically calculated with echocardiography software. During M-mode echocardiography analysis in the short axis, LV end-diastolic posterior wall thickness (LVPWd) were obtained at the time of the apparent maximal LV diastolic dimension, and LV end-systolic posterior wall thickness (LVPWs) was obtained at the time of the most anterior systolic excursion of the posterior wall. The end-systolic posterior wall thickening (LVESWT) = (LVPWs − LVPWd)/LVPWs × 100, a marker of regional left ventricle systolic function. The early mitral diastolic wave/late mitral diastolic wave ratio (E/A ratio) were measured by Doppler echocardiography using VEVO 2100 software to evaluate diastolic function as previous described [15].
Haemodynamic parameters were measured by a Millar Mikro-tip-pressure catheter (Millar Instruments, Houston, TX, USA) as previously described [16]. After anaesthetized using $2.5\%$ isoflurane, the right carotid artery was cannulated with a high-fidelity catheter transducer. Haemodynamic parameters were recorded and analysed. The LV end diastolic pressure–volume relationship (EDPVR) and LV end-diastolic pressure (LVEDP) were measured to evaluated cardiac diastolic function.
All measurements were performed by a researcher who was blinded to the experimental groupings.
## Biochemical and histological analysis
Blood glucose levels were measured with an ACCU-CHEK Active Blood Glucose Meter using blood from the tail vein. Plasma insulin levels were determined using a commercial ELISA kit (EZRMI-13K, Millipore, USA) according to the manufacturer’s instructions. After excision, the hearts were fixed with $10\%$ neutral buffered formalin for 24 h, embedded in paraffin, and cut into 5 μm slices. Standard haematoxylin and eosin (H&E) staining (G1120, Solarbio, China) was performed according to the manufacturers’ protocol. Cardiomyocyte number were measured on light microscopy images as previously described [17]. FITC-labelled wheat germ agglutinin staining (WGA staining, W11261, Thermo Fisher, USA) was used to detect the cross-sectional area of the cardiomyocytes per the manufacturer’s protocol. Masson’s trichrome staining (G1346, Solarbio, China) was used for quantification of myocardial interstitial fibrosis. The size of the fibrotic area was calculated as the ratio of the fibrotic area to the left ventricle area. The results were quantified using ImageJ software (NIH, USA).
## Transmission electron microscopy (TEM)
The left ventricular wall was cut into 1–2 mm wide strips perpendicular to its long axis. These strips were fixed with $2.5\%$ glutaraldehyde in 0.1 mol/l phosphate buffer (pH 7.4, 4 °C) for 24 h and postfixed with $1\%$ osmium tetroxide in 0.1 mol/l sodium cacodylate buffer (pH 7.4) for 1 h. After dehydration and embedding in spur resin, the samples were cut into 80 nm-thick sections. All images were obtained with a transmission electron microscope (JEM-1230, JEOL Ltd., Japan) at 80 kV. Mitochondrial images were obtained at magnifications of 28,000 and 98,000 and analysed with ImageJ software in a blinded fashion.
## Cardiomyocyte isolation, culture and transfection
Primary neonatal mouse cardiomyocytes (NCMs) were isolated as previously described [18]. Briefly, the left ventricular tissue was removed from 1 to 3-day-old C57/BL6J mice. The ventricular tissues were minced thoroughly and digested with collagenase type II (1 mg/ml, 17101015, Gibco, Thermo Fisher) for 3 min at 37 °C (4–6 times). After differential plating to remove fibroblasts, the NCMs were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing $20\%$ foetal bovine serum (FBS) and $1\%$ penicillin–streptomycin for 48 h.
After plating, the NCMs were incubated with high-glucose medium (30 mmol/l glucose, HG) for 48 h to mimic diabetic cardiomyopathy in vivo. The NCMs were treated with normal-glucose medium (5.5 mmol/l glucose, NG) as a control group.
The NCMs were infected with adenovirus Ad-vector (Ad-EV, MOI: 1:50), Ad-Nr2f6 (MOI: 1:50) or Ad-Grpel2 (MOI: 1:50) in complete DMEM for 8 h, and the medium with adenovirus was replaced with fresh complete DMEM. After 48 h of adenoviral transfection, the efficiency of gene overexpression and knockdown was determined by Western blotting and qRT-PCR analysis. Then, the NCMs were subjected to NG or HG treatment for another 48 h.
All recombinant adenovirus and siRNA were constructed by Tsingke Biotechnology Co., Ltd. (Beijing, China). The sequences of siRNAs used in this study are listed below:siRNASenseAntisensesiCtrl:5′-TTCTCCGAACGTGTCACGT-3′5′-ACGTGACACGTTCGGAGAA-3′siGrpel2:5′-GCGGCTCTTTGATGCAAAT-3′5′-ATTTGCATCAAAGAGCCGC-3′siDLST5′-GAUAUUGAACGGACCAUUA-3′5′-UAAUGGUCCGUUCAAUAUC-3′siNr2f65′-GGUCCAACCGUGACUGUCA-3′5′-UGACAGUCACGGUUGGACC-3′
## Measurement of the mitochondrial oxygen consumption rate (OCR)
The mitochondrial OCR of NCMs was measured with an XF24 Extracellular Flux Analyser (Agilent Seahorse Bioscience, USA) as previously described [19]. In brief, NCMs were plated into XF24 Seahorse plates at 160,000 cells/well for 36 h and then infected with adenoviruses for 48 h. After the NCMs were exposed to NG or HG conditions, the mitochondrial OCR was determined according to the manufacturer’s protocols. The employed working concentrations of the inhibitors were as follows: oligomycin, 0.6 μM; trifluoromethoxy carbonyl cyanide phenylhydrazone (FCCP), 0.75 μM; antimycin A, 2 μM; and rotenone 1 μM. Basal respiration, maximal respiration, ATP production and spare respiration capacity were calculated by using XF Cell Mito Stress Test Generator software (Agilent Seahorse Bioscience, USA). All OCR measurements were normalized to protein concentrations.
## Measurement of mitochondrial membrane potential
After NG or HG treatment, cardiomyocytes were harvested and incubated with JC-1 (C2003S, Beyotime, China) at 37 °C for 30 min in the dark. The results were analysed within 1 h by flow cytometry. When the mitochondrial membrane potential is high, JC-1 accumulates in the mitochondrial matrix and forms J-aggregates, which can emit red fluorescence. When the mitochondrial membrane potential is low, JC-1 is a monomer that can produce green fluorescence. The analysis of mitochondrial membrane potential is presented as the aggregate (red fluorescence) ratio as previously described [20].
## Detection of ROS content in cardiomyocytes
The total cellular and mitochondrial ROS contents in frozen heart sections were evaluated by DHE staining (10 μM, S0063, Beyotime, China) and staining with the fluorescent probe MitoSOX (100 mM, M36008, Thermo Fisher Scientific, USA), respectively, as previously described [21]. Images were captured with a confocal laser scanning microscope (Nikon, Japan) and analysed with ImageJ software (NIH, USA).
Total cellular ROS in NCMs were detected by a ROS/Superoxide Detection Assay Kit (Ab139476, Abcam, USA) as previously described via a microplate reader [22]. The mitochondrial ROS in NCMs were detected by staining with the fluorescent probe MitoSOX (100 mM, M36008, Thermo Fisher Scientific, USA) following the manufacturers’ protocols via flow cytometry analysis with a BD FACS Aria II flow cytometer.
The Total Glutathione Peroxidase (GPx) Assay Kit with NADPH (S0058, Beyotime, China) and Lipid Peroxidation Malondialdehyde (MDA) Assay Kit (Ab118970, Abcam, USA) were used according to the manufacturer’s instructions to detect cardiomyocyte GPx activity and MDA levels and further assess oxidative stress levels.
## Cell apoptosis assay
The apoptosis rate in primary cardiomyocytes was determined by flow cytometry analysis using an Annexin V-FITC/propidium iodide (PI) apoptosis detection kit (C1062S, Beyotime, China). Apoptosis in heart tissues was measured by terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) staining (C1090, Beyotime, China). All procedures were performed as previously described [23]. Apoptosis levels were calculated by dividing the number of TUNEL-positive nuclei by the total number of 4′,6-diamidino-2-phenylindole (DAPI)-positive nuclei. All images were obtained with a confocal laser scanning microscope (Nikon, Japan) and analysed in a blinded fashion. A caspase 3 activity kit (BC3830, Solarbio, China) was also used to detect myocardial apoptosis.
## ATP detection
Total ATP production was detected with an ATP Assay Kit (S0027, Beyotime, China) following the protocols provided by the manufacturer as previously described [24]. Briefly, NCMs and heart tissues were homogenized and lysed in ice lysis buffer. After centrifugation at 4 °C at 12,000 × g for 20 min, 10 μl of the supernatant or standard ATP solution was incubated with the ATP probe in 96-well plates at room temperature in the dark. The results were detected with a luminometer (BioTek Epoch, USA).
## Measurement of cell viability
Cell viability was evaluated with a CCK-8 assay kit (C0005, Topscience, China) as previously described [25]. Briefly, after NG or HG treatment, NCMs were incubated with fresh complete medium containing $10\%$ CCK-8 for 2 h. The absorbance at 450 nm was measured to indicate cell viability.
## Isolation of mitochondria
The complete process of mitochondria isolation was performed at 4 °C using a Cell Mitochondria Isolation Kit (C3601, Beyotime, China) as previously described [26]. NCMs were washed twice with ice-cold PBS and homogenized in ice-cold isolation buffer for 10 min. The homogenates were centrifuged at 1000 × g for 10 min. The supernatant was centrifuged at 11,000 × g for 10 min to pellet mitochondria. Mitochondria-enriched fractions were washed with PBS at 3000 × g for 10 min and resuspended in mitochondrial storage buffer for storage at − 80 °C.
## Co-immunoprecipitation (Co-IP)
NCMs were homogenized with ice-cold IP lysis buffer (25 mM Tris–HCl pH 7.4, 150 mM NaCl, $1\%$ NP-40, 1 mM EDTA and $5\%$ glycerol, 87787, Thermo Fisher Scientific, USA) supplemented with a protease and phosphatase inhibitor cocktail (5872, CST, USA). Subsequent procedures were performed as previously described [27]. The homogenates were centrifuged at 10,000 × g for 10 min. Then, 1 mg of the supernatant was incubated with 1 μg of antibodies for sufficient immunoprecipitation at 4 °C overnight and incubated with protein A/G magnetic beads for another 3 h at 4 °C. After the beads were discarded, the supernatant was used for Western blotting analysis. Forty micrograms of cell lysate were used for the positive control and loading control.
## Chromatin immunoprecipitation (ChIP) assay
ChIP assays were performed using a ChIP Plus Enzymatic Chromatin IP Kit (9003, CST, USA) following the manufacturer’s protocols [21]. Briefly, NCMs were fixed with formaldehyde, and the chromatin was sheared into fragments. Then, the fragmented chromatin was incubated and precipitated with an Nr2f6 antibody (#NBP1-04676, Novus, USA) and protein G magnetic beads at 4 °C. DNA released from the precipitates was analysed by qRT-PCR. IgG was used as the negative control.
## Statistical analysis
All values are presented as the mean ± standard error (SD). All data were obtained in three or more independent experiments. The normality of the distribution for all data was examined using the Shapiro–Wilk normality test. If the data passed the normality assumption, statistical significance was determined by an unpaired, 2-tailed Student’s t test (two groups) or one-way ANOVA, followed by Tukey’s post hoc test (> 2 groups); otherwise, the data were analysed with the Mann–Whitney U test (two groups) or Kruskal–Wallis test with Dunn’s post hoc test (> 2 groups). Survival analysis was performed by Log-rank Mantel-Cox testing. Correlation analysis was performed using the Pearson correlation test. All statistical analyses were carried out by using GraphPad Prism 8.0 software (GraphPad Software, La Jolla, USA). A value of $p \leq 0.05$ was considered statistically significant.
## Cardiac-specific overexpression of Grpel2 effectively alleviates diabetes-induced cardiac dysfunction
To gain insight into the role of Grpel2 in the diabetic cardiomyopathy (DCM), we first assessed Grpel2 expression by Western blotting and qRT-PCR analysis in a STZ-induced diabetic mouse model. We found that Grpel2 protein expression and mRNA levels were both decreased in the hearts of mice 6 weeks after STZ injection compared to vehicle injected group. The hearts showed an even greater downregulation of Grpel2 protein and mRNA levels at 12 weeks after STZ injection (Fig. 1a–c). To characterize the effect of diabetes on cardiomyocyte Grpel2 levels in vitro, we also isolated primary neonatal mouse cardiomyocytes (NCMs) and subjected them to high-glucose (30 mmol/l glucose, HG) conditions to mimic diabetes in vivo. We found that Grpel2 protein expression and mRNA levels were significantly decreased 48 h after HG treatment (Additional file 1: Fig. S1A–C). An adeno-associated virus serotype 9 (AAV9) was designed and intramyocardially injected to overexpress cardiac Grpel2 expression to investigate the relationship between Grpel2 and DCM. A set of experimental analyses was carried out as shown in Additional file 1: Fig. S2A. As a result, intramyocardial injection of AAV9-Grpel2 successfully increased Grpel2 protein expression and mRNA levels in the hearts of the control or diabetic mice (Additional file 1: Fig. S2B–D). Compared with the control mice, the diabetic mice had decreased serum insulin, increased blood glucose levels and lower survival rate (Fig. 1d and Additional file 1: Fig. S3A, B). Although Cardiac-specific overexpression of Grpel2 had no significant effects on serum insulin or blood glucose levels, the diabetic mice injected with AAV9-Grpel2 showed a higher survival rate compared with mice injected with AAV9-Ctrl (Fig. 1d and Additional file 1: Fig. S3A, B). To evaluate the effect of Grpel2 on cardiac systolic function in diabetic mice, the left ventricular ejection fraction (LVEF), left ventricular fractional shorting (LVFS), left ventricular systolic internal diameter (LVISD) and left ventricular diastolic internal diameter (LVIDD) were measured using echocardiography from short-axis and long-axis views 12 weeks after vehicle or STZ injection. Our data revealed that diabetes-induced cardiac systolic dysfunction, including decreased LVEF and LVFS as well as increased LVISD, was ameliorated in AAV9-Grpel2-injected diabetic mice (Fig. 1e–i and Additional file 1: Fig. S4A–C). Furthermore, LV end-diastolic posterior wall thickness (LVPWd) was significantly increased in diabetic mice, which was attenuated by AAV9-Grpel2 injection, whereas no significant differences were observed in LV end-systolic posterior wall thickness (LVPWs) among the groups (Fig. 1j, k). And the LV end-systolic posterior wall thickening (LVESWT) was markedly decreased in diabetic mice, which was improved by cardiac-specific Grpel2 overexpression (Fig. 1l). Moreover, haemodynamic parameters and Doppler echocardiography were performed to evaluated the LV diastolic function. LV end diastolic pressure–volume relationship (EDPVR) and LV end-diastolic pressure (LVEDP) were significantly increased in diabetic mice compared with control mice. Grpel2 overexpression markedly decreased EDPVR and LVEDP in diabetic mice (Table 1). Moreover, the diabetic mice injected with AAV9-Grpel2 showed improved diastolic function as evidenced by increased early mitral diastolic wave/late mitral diastolic wave ratio (E/A ratio) (Fig. 1m). And Grpel2 overexpression had no evident effects on heart rate under conscious or anesthetic condition (Additional file 1: Fig. S5). In summary, our data indicates that Grpel2 expression is downregulated in the diabetic heart and that cardiac-specific overexpression of Grpel2 alleviates diabetes-induced cardiac contractile and diastolic dysfunction. Fig. 1Grpel2 was downregulated in the diabetic heart, and cardiac overexpression of Grpel2 attenuated the contractile dysfunction induced by STZ injection. A, B Western blotting and quantitative analysis of Grpel2 protein expression in the heart tissues of mice at 0, 6 and 12 weeks after vehicle or STZ injection ($$n = 4$$/group). C Quantitative real-time PCR of Grpel2 mRNA levels in heart tissues from mice at 0, 6 and 12 weeks after vehicle or STZ injection ($$n = 4$$/group). D Survival curve of mice intramyocardially injected with AAV9-Ctrl or AAV9-Grpel2 after vehicle or STZ injection ($$n = 16$$/group). E Representative short-axis M-mode echocardiographic images from mice intramyocardially injected with AAV9-Ctrl or AAV9-Grpel2 12 weeks after vehicle or STZ injection. F–L Quantification of left ventricular ejection fraction (LVEF, F), fraction shortening (LVFS, G), systolic internal dimension (LVISD, H), diastolic internal dimension (LVIDD, I), end-diastolic posterior wall thickness (LVPWd, J), end-systolic posterior wall thickness (LVPWs, K), and end-systolic posterior wall thickening (LVESWT, L) by short-axis M-mode echocardiography ($$n = 6$$–8/group). M *Quantitative analysis* of the early mitral diastolic wave/late mitral diastolic wave ratio (E/A ratio) by Doppler echocardiography ($$n = 6$$/group). Data are presented as the mean ± SD. Data in B and C were analysed with an unpaired, 2-tailed Student’s t test. Data in D were analysed with Log-rank Mantel-Cox testing. Other data were analysed by one-way ANOVA, followed by Tukey’s post hoc test. * $p \leq 0.05$Table 1Pressure–volume loop dataCon + AAV9-Ctrl $$n = 6$$Con + AAV9-Grpel2 $$n = 6$$DCM + AAV9-Ctrl $$n = 6$$DCM + AAV9-Grpel2 $$n = 6$$ESPVR, mmHg/ml23.46 ± 2.3922.19 ± 2.4515.54 ± 1.39*18.79 ± 2.04#EDPVR, mmHg/ml0.041 ± 0.0050.054 ± 0.0060.156 ± 0.004*0.106 ± 0.003#Tau, ms7.76 ± 0.298.19 ± 0. 5410.78 ± 1.09*9.49 ± 1.01# + dP/dt, mmHg/s5930 ± 1315971 ± 1432569 ± 214*3742 ± 325#− dP/dt, mmHg/s3562 ± 1763687 ± 2172196 ± 235*2957 ± 167#LVESP, mmHg128.56 ± 11.71130.33 ± 9.8185.33 ± 8.59*104.17 ± 6.24#LVEDP, mmHg4.67 ± 0.835.35 ± 1.0712.12 ± 1.06*7.48 ± 1.15#Data are mean ± SDESPVR, end systolic pressure–volume relationship; EDPVR, end diastolic pressure–volume relationship; LV, left ventricle; ESP, end-systolic pressure; EDP, end-diastolic pressure*$p \leq 0.05$ versus Con + AAV9-Ctrl, #$p \leq 0.05$ versus DCM + AAV9-Ctrl, using one-way ANOVA, followed by Tukey’s post-hoc test
## Cardiac-specific Grpel2 overexpression alleviated adverse cardiac remodeling in the diabetic heart
Since pathological myocardial hypertrophy and excess cardiac fibrosis are the most prominent features of DCM [28, 29], we next explored the potential role of Grpel2 in pathological cardiac remodeling induced by diabetes. There were no significant differences in heart pathology among the control mice. The hearts of diabetic mice exhibited significant cardiac remodeling, which was exemplified by enlarged hearts, elevated ratios of heart weight to tibia length, reduced cardiomyocyte number, increased cardiomyocyte cross-sectional area and increased interstitial fibrosis (Fig. 2a–f). Importantly, these pathological changes were efficiently ameliorated in the hearts of diabetic mice receiving AAV9-Grpel2 injection (Fig. 2a–f). Moreover, we detected the mRNA levels of common hypertrophic markers, such as Anp and Bnp, and fibrotic markers, such as Tgfb1, Col1a1 and Col3a1. There were no differences in the mRNA levels of common hypertrophic markers or fibrotic markers among control mice. However, the mRNA levels of these markers were significantly decreased in Grpel2-overexpressing mice compared to mice intramyocardially injected with AAV9-Ctrl after STZ treatment (Fig. 2g, h). In addition, we detected cardiomyocyte apoptosis, a common feature of the diabetic heart. As we expected, cardiomyocyte apoptosis, as determined by TUNEL staining and caspase 3 activity assays, was significantly decreased in diabetic mice injected with AAV9-Grpel2 compared with diabetic mice injected with AAV9-Ctrl (Fig. 2i–k). Collectively, our data indicated that cardiac-specific Grpel2 overexpression significantly ameliorates cardiac structure remodeling and cardiomyocyte apoptosis in diabetic mice. Fig. 2Overexpression of cardiac Grpel2 alleviated heart remodeling in DCM. A, B Representative images of haematoxylin and eosin (HE) staining of cardiac sections ($$n = 6$$/group). Scale bar = 2 mm. B Quantification of the heart weight/tibia length (HW/TL) ratio ($$n = 6$$/group). C Representative images of wheat germ agglutinin (WGA) staining (top) and Masson trichrome staining (bottom) of cardiac sections. Scale bars: 25 μm (top) and 50 μm (bottom). D, E *Quantitative analysis* of the cross-sectional area of cardiomyocytes (D) and number of cardiomyocytes (E) by WGA staining. F *Quantitative analysis* of interstitial fibrosis by Masson trichrome staining ($$n = 6$$/group). G, H Quantitative real-time PCR (qRT-PCR) of the mRNA levels of hypertrophy-associated genes (G) and fibrosis-associated genes (H) in the left ventricle of mice ($$n = 6$$/group). I, J Representative images and quantitative analysis of TUNEL staining of cardiac sections ($$n = 6$$/group). Scale bar: 50 μm. K Quantification of relative caspase 3 activity in the left ventricle of mice ($$n = 6$$/group). Data are presented as the mean ± SD. Data were analysed by one-way ANOVA, followed by Tukey’s post hoc test. * $p \leq 0.05$
## Grpel2 overexpression attenuated mitochondrial morphological disorder and oxidative stress in the diabetic heart
Excessive mitochondrial fission and oxidative stress lead to cardiomyocyte apoptosis, contributing to diabetic cardiomyopathy [15, 30]. As shown in the transmission electron microscopy (TEM) images, cardiomyocytes of the diabetic heart presented excessive mitochondrial fission, as indicated by a significantly larger mean mitochondrial size, an increased number of mitochondria per μm2 and a decreased number and area of mitochondrial cristae per mitochondrial area. No significant differences in mitochondrial morphology were observed in control mice (Fig. 3a–f). Grpel2 overexpression alleviated diabetes-induced mitochondrial morphological abnormalities, as reflected by the larger mean mitochondrial size, decreased number of mitochondria per μm2, increased number of mitochondrial cristae per mitochondrial area and increased cristae area per mitochondrial area (Fig. 3a–f). Moreover, of the mitochondrial fission proteins, Phospho-Ser-616-Drp1 expression was markedly increased and Phospho-Ser-637-Drp1 was significantly decreased, while Drp1 and Fis1 remained unchanged in diabetic hearts compared with control hearts. Of the mitochondrial fusion proteins, Mfn1 and Mfn2 remained unchanged, but the Opa1 expression was significantly decreased in the diabetic hearts compared with control hearts. Importantly, cardiac-specific Grpel2 overexpression markedly increased Opa1 and Phospho-Ser-637-Drp1 expressions, and significantly decreased Phospho-Ser-616-Drp1 expression in the diabetic hearts (Fig. 3f–h). We next investigated the effect of Grpel2 on oxidative stress in the diabetic heart. Mitochondrial ROS levels and total ROS levels were detected by MitoSOX staining and DHE staining, respectively. As anticipated, total cellular and mitochondrial ROS levels were markedly increased in the diabetic heart compared to the control heart (Fig. 3i–k). Moreover, Grpel2 overexpression significantly decreased the total and mitochondrial ROS contents compared to those of diabetic mice intramyocardially injected with AAV9-Ctrl (Fig. 3i–k). Furthermore, we also determined the GPx and MDA contents to evaluate cardiac oxidative stress. Similarly, there were no evident differences in control mice, while Grpel2 overexpression significantly increased the GPx content and decreased the MDA content in diabetic mice (Fig. 3l, m). Excessive mitochondrial fission and oxidative stress cause defects in cardiomyocyte mitochondrial ATP bioenergetics in the diabetic heart [31]. We found that Grpel2 overexpression significantly increased ATP levels in the diabetic heart (Fig. 3n). Therefore, these data indicate that Grpel2 overexpression attenuates diabetes-induced excessive mitochondrial fission and oxidative stress in the diabetic heart. Fig. 3Grpel2 overexpression protected against cardiomyocyte mitochondrial dysfunction in the diabetic heart. A Representative transmission electron microscopy images of the left ventricle of mice (magnification: 28,000 × [28 K] and 98,000 × [98 K]). Scale bars: 1 μm (top) and 0.5 μm (bottom). B–E *Quantitative analysis* of the mean mitochondrial size (B), number of mitochondria per μm2 (C), number of mitochondrial cristae per mitochondrial area (D) and cristae area per mitochondrial area (E) ($$n = 6$$/group). F–H Western blotting and quantitative analysis of mitochondrial fission related proteins (Drp1and Fis1) and fusion-related proteins (Opa1, Mfn1 and Mfn2) in the left ventricle of mice ($$n = 4$$/group). I Representative images of MitoSOX staining (top) and DHE staining (bottom) of cardiac sections. Scale bars: 50 μm (up and down). J, K *Quantitative analysis* of relative mitoROS fluorescence (J) by MitoSOX staining and relative DHE fluorescence (K) by DHE staining ($$n = 6$$/group). L, M *Quantitative analysis* of glutathione peroxidase (GPx) content (L) and malondialdehyde (MDA) content (M) in the left ventricle of mice ($$n = 6$$/group). N *Quantitative analysis* of the relative ATP content in heart lysates ($$n = 6$$/group). Data are presented as the mean ± SD. Data were analysed by one-way ANOVA, followed by Tukey’s post hoc test. * $p \leq 0.05$
## Grpel2 overexpression alleviated HG-induced apoptosis and mitochondrial dysfunction in NCMs
We also designed a recombinant adenovirus encoding Grpel2 (Ad-Grpel2) to overexpress Grpel2 in NCMs. Transfection with Ad-Grpel2 resulted in successful overexpression of Grpel2 protein and mRNA in NCMs (Additional file 1: Fig. S6A–C). We further detected the effects of Grpel2 on mitochondrial function in vitro. We first assessed the effect of Grpel2 on mitochondrial respiratory capacity by measuring the oxygen consumption rate (OCR, a classical indicator of mitochondrial function). Compared to the NG-treated NCMs, NCMs treated with HG exhibited decreased mitochondrial respiratory capacities, including basal respiration, maximal respiration, ATP production and spare respiration capacity (Fig. 4a, b). There was no change in NCMs infected with Ad-EV or Ad-Grpel2 under NG conditions. Notably, Grpel2 overexpression significantly enhanced the mitochondrial respiratory capacities of HG-treated NCMs (Fig. 4a, b). Decreased mitochondrial membrane potential is also a key feature of mitochondrial dysfunction. As expected, Grpel2 overexpression markedly increased the mitochondrial membrane potential in HG-treated NCMs (Fig. 4c, d). We next determined the contents of total and mitochondrial cellular ROS to further evaluate mitochondrial oxidative stress. Similar to the results above, Grpel2 overexpression did not change the ROS content under NG conditions but markedly decreased both mitochondrial ROS and total cellular ROS in HG-treated NCMs (Fig. 4e–g). Furthermore, the GPx content was increased in NCMs infected with Ad-Grpel2 under HG conditions (Fig. 4h). Importantly, Grpel2 overexpression markedly increased ATP content under HG conditions (Fig. 4i). And Grpel2 overexpression significantly decreased cardiomyocyte apoptosis, as detected by Annexin V/PI staining and caspase 3 activity assays, under HG conditions (Fig. 4j–l). Collectively, these results suggest that Grpel2 is involved in maintaining cardiomyocyte apoptosis and mitochondrial homeostasis under HG conditions. Fig. 4Grpel2 overexpression attenuated HG-induced mitochondrial dysfunction in vitro. A, B Oxygen consumption rate (OCR) and associated quantitative analysis of mitochondrial respiratory function in NCMs ($$n = 4$$/group). C, D Flow cytometry analysis and quantification of mitochondrial membrane potential by JC-1 staining in NCMs ($$n = 4$$/group). E, F Flow cytometry analysis and quantification of mitochondrial ROS content by MitoSOX staining in NCMs ($$n = 4$$/group). G Quantification of total intracellular ROS intensity in NCMs ($$n = 4$$/group). H *Quantitative analysis* of glutathione peroxidase (GPx) content in NCMs ($$n = 4$$/group). I *Quantitative analysis* of the relative ATP content in NCMs ($$n = 4$$/group). J Quantification of relative caspase 3 activity in control or Grpel2-overexpressing NCMs under NG or HG conditions ($$n = 4$$/group). K, L Flow cytometry analysis and quantification of apoptotic cells by Annexin V-FITC and propidium iodide (PI) staining ($$n = 4$$/group). Data are presented as the mean ± SD. Data were analysed by one-way ANOVA, followed by Tukey’s post hoc test. * $p \leq 0.05$
## Grpel2 positively mediated the process of DLST import into mitochondria
Given the important roles of Grpel2 in maintaining cardiac function and structure in DCM, we next investigated the underlying molecular mechanism responsible for the regulation of Grpel2-mediated mitochondrial function and cardiomyocyte survival. Previous studies revealed that Grpel2 facilitated mitochondrial protein import, and most of the possible interaction partners of Grpel2 were metabolic enzymes, including dehydrogenases of the tricarboxylic acid (TCA) cycle [11]. Among those potential interactors, dihydrolipoyl succinyltransferase (DLST) is one of unique interactors with high confidence [11]. We first found that there were interactions between Grpel2 and DLST, as detected by co-immunoprecipitation (IP) (Fig. 5a, b). Furthermore, knockdown of Grpel2 by siRNA decreased the protein expression of DLST under NG conditions (Fig. 5c, d). Considering that DLST is a nucleus-encoded protein and requires import into mitochondria, we suggested that Grpel2 mediates the import process of DLST into mitochondria. We found that Grpel2 overexpression had no effect on DLST expression in mitochondrial or cytoplasmic lysates under NG condition. However, Grpel2 overexpression increased DLST expression in mitochondrial lysates, but did not affect DLST expression in cytoplasmic lysates under HG conditions (Fig. 5e, f). And knockdown of Grpel2 had no effect on DLST expression in cytoplasmic lysates but decreased DLST expression in mitochondrial lysates under HG conditions (Fig. 5g, h). In summary, these data revealed that Grpel2 interacted with DLST and may be involved in the process of DLST import into mitochondria under HG conditions. Fig. 5Grpel2 promotes the process of DLST import into mitochondria. A Co-immunoprecipitation (IP) analysis of DLST with Grpel2 in whole-cell lysates of NCMs. B Co-IP analysis of Grpel2 with DLST in whole cell lysates of NCMs. C, D Western blotting and quantitative analysis of Grpel2 and DLST protein expression in NCMs infected with siCtrl or siGrpel2 under NG conditions ($$n = 4$$/group). E, F Western blotting and quantitative analysis of Grpel2 and DLST protein expression in the whole cell, cytoplasm or mitochondrial lysates of NCMs infected with Ad-EV or Ad-Grpel2 under NG or HG conditions ($$n = 4$$/group). G, H Western blotting and quantitative analysis of Grpel2 and DLST protein expression in the whole cell, cytoplasm or mitochondrial lysates of NCMs infected with siCtrl or siGrpel2 under HG conditions. Data are presented as the mean ± SD. Data in H were analysed by an unpaired, 2-tailed Student’s t test. Other data were analysed by one-way ANOVA, followed by Tukey’s post hoc test. * $p \leq 0.05$; ns, not significant
## Grpel2 overexpression protected mitochondrial function by maintaining mitochondrial import of DLST
To verify whether DLST is essential for the mitochondria-protective effects of Grpel2 in DCM, NCMs infected with Ad-EV or Ad-Grpel2 were also subjected to siRNA to block DLST expression under HG conditions. siRNA significantly decreased DLST expression (Additional file 1: Fig. S7A–C). As expected, Grpel2 overexpression significantly attenuated HG-induced mitochondrial dysfunction, as indicated by elevated mitochondrial respiratory capacities, increased mitochondrial membrane potential, decreased mitochondrial ROS production and decreased cellular ROS production (Fig. 6a–g). Interestingly, the protective effects of Grpel2 overexpression on mitochondrial function were almost blocked by DLST knockdown in NCMs under HG conditions. Even when Grpel2 was overexpressed, NCMs with DLST knockdown still exhibited more severe impaired mitochondrial function under HG conditions, including decreased mitochondrial respiratory capacities, decreased mitochondrial membrane potential and increased mitochondrial ROS production (Fig. 6a–g). Similarly, downregulation of DLST also markedly decreased cell viability and ATP contents, and increased cardiomyocyte apoptosis under HG conditions (Fig. 6i–l). All the protective effects of Grpel2 overexpression on cell survival were eliminated by DLST knockdown in NCMs under HG conditions (Fig. 6i–l). Overall, the protective effects of Grpel2 overexpression on mitochondrial function and cell survival were dependent on the expression of DLST.Fig. 6Grpel2 overexpression alleviates HG-induced mitochondrial dysfunction and apoptosis via mitochondrial DLST expression. A, B OCR and associated quantitative analysis of mitochondrial respiratory function in NCMs ($$n = 4$$/group). C, D Flow cytometry analysis and quantification of mitochondrial membrane potential by JC-1 staining in NCMs ($$n = 4$$/group). E, F Flow cytometry analysis and quantification of mitochondrial ROS content by MitoSOX staining in NCMs ($$n = 4$$/group). G Quantification of total intracellular ROS intensity in NCMs ($$n = 4$$/group). H *Quantitative analysis* of relative cell viability in NCMs ($$n = 4$$/group). I *Quantitative analysis* of the relative ATP content in NCMs ($$n = 4$$/group). J Quantification of relative caspase 3 activity in control or Grpel2-overexpressing NCMs under NG or HG conditions ($$n = 4$$/group). K, L Flow cytometry analysis and quantification of apoptotic cells by Annexin V-FITC and propidium iodide (PI) staining ($$n = 4$$/group). Data are presented as the mean ± SD. Data were analysed by one-way ANOVA, followed by Tukey’s post hoc test. * $p \leq 0.05$; ns, not significant
## Nr2f6 regulates the transcription of Grpel2 by directly binding the promoter of Grpel2
We next investigated the underlying transcriptional mechanism responsible for regulating Grpel2-mediated mitochondrial function and cardiomyocyte survival. We used the bioinformatics databases JASPAR and GENECARDS to predict candidate transcription factors of human Grpel2 (Fig. 6a). Four potential transcription factors (Nr2f6, RARA, IRF2 and ETV4) appeared in both databases and had significant correlations with Grpel2 in human left ventricular (LV) samples from the GEPIA database (Fig. 7a, b). Furthermore, we analysed a public microarray dataset from the Gene Expression Omnibus (GEO) database to evaluate the correlations between the mRNA expression of Grpel2 and those potential transcription factors in LV samples from normal control mice and diabetic mice. Pearson correlation analysis revealed that only Nr2f6 had a significant positive correlation with Grpel2 (Fig. 7c and Additional file 1: Fig. S8A–C). Importantly, this correlation became more significant in LV samples from only diabetic mice (Fig. 7c). Moreover, both the protein and mRNA levels of Nr2f6 were significantly decreased in diabetic hearts in vivo and in HG-treated NCMs in vitro (Fig. 7d–f and Additional file 1: Fig. S9A–C). We found a direct binding site of Nr2f6 in the Grpel2 promoter region of NCMs by ChIP-qRT-PCR assays (Fig. 6g and Additional file 1: Fig. S9D). In addition, Nr2f6 knockdown via siRNA markedly decreased the mitochondrial Grpel2 protein expression, whereas Nr2f6 overexpression via Ad-Nr2f6 significantly increased the mitochondrial Grpel2 protein expression under NG or HG conditions (Fig. 7h–k). And the mitochondrial DLST expression was consistent with the results in NCMs infected with siGrpel2 or Ad-Grpel2 under NG or HG conditions (Fig. 7h–k). Collectively, our results reveal that Nr2f6 directly binds to the Grpel2 promoter region and positively regulates its transcription. Fig. 7Nr2f6 directly binds to the promoter of Grpel2 and regulates Grpel2 expression. A Venn diagram showing the candidate transcription factors identified in the JASPAR and GENECARDS databases. B Scatter plots showing the Pearson correlations for the mRNA levels of Grpel2 and candidate transcription factors in human left ventricular samples from the GTEx database through the GEPIA website. C Scatter plots showing Pearson correlation analysis of mRNA levels of Nr2f6 and Grpel2 in control or diabetic heart tissue based on the GEO database (GSE123975). D, E Western blotting and quantitative analysis of heart Nr2f6 protein expression in the heart tissues of mice at 0, 6 and 12 weeks after STZ injury ($$n = 4$$/group). F qRT-PCR of Grpel2 mRNA levels in heart tissues of mice at 0, 6 and 12 weeks after STZ injury ($$n = 4$$/group). G Chromatin immunoprecipitation (ChIP) and qRT-PCR analysis of the binding of Nr2f6 to the Grpel2 promoter in NCMs ($$n = 4$$/group). H, I Western blotting and quantitative analysis of Nr2f6, mitochondrial Grpel2 and mitochondrial DLST protein expression in NCMs infected with siCtrl or siNr2f6 under NG or HG conditions ($$n = 4$$/group). J, K Western blotting and quantitative analysis of Nr2f6, and mitochondrial Grpel2 protein expression in NCMs infected with Ad-EV or Ad-Nr2f6 under NG or HG conditions ($$n = 4$$/group). Data are presented as the mean ± SD. Data in B, C were analysed by Pearson correlation analysis. Data in I, K were analysed by one-way ANOVA, followed by Tukey’s post hoc test. Other data were analysed by an unpaired, 2-tailed Student’s t test. * $p \leq 0.05$; ns, not significant
## Discussion
In the present study, we reported for the first time that Grpel2 is downregulated in DCM induced by STZ. Furthermore, Grpel2 overexpression significantly alleviated heart dysfunction and cardiac remodeling in DCM, including cardiac contractile dysfunction, cardiac diastolic function cardiac hypertrophy, interstitial fibrosis and cardiomyocyte apoptosis. Specifically, Grpel2 overexpression markedly attenuated mitochondrial dysfunction. These results in vivo were also verified in NCMs treated with HG in vitro. Moreover, we discovered that upregulated Grpel2 expression exerted protective effects against HG-induced mitochondrial dysfunction by binding to DLST and modulating its mitochondrial import process. Finally, our study demonstrated that the downregulation of Grpel2 was partially due to a decrease in Nr2f6 expression in HG-treated NCMs. Taken together, these findings suggest that Grpel2 plays a positive role in DCM and that targeting Grpel2 may represent a new therapeutic method for DCM.
The prevalence of diabetes mellitus has been rapidly increasing, and diabetes mellitus is a major threat to human health worldwide [32]. DCM has a high incidence and high mortality in diabetic patients [31]. However, the underlying pathological mechanism of DCM remains unclear. An increasing number of studies have reported that the mitochondrial protein import process is involved in mitochondrial bioenergetics, mitochondrial dynamics and protein quality control. Recent studies revealed that defects in the mitochondrial import process of nucleus-encoded protein subunits result in inhibition of mitochondrial respiratory capacities, decreased mitochondrial membrane potential and reduction of ATP synthesis in many cardiovascular diseases, such as heart failure, ischaemic cardiomyopathy, DCM and hypertension [6]. It is reported that mtHSP70, a central subunit of the presequence translocase-associated motor complex, was decreased in DCM, which coincides with decreased protein import in the diabetic interfibrillar mitochondria subpopulation [33]. However, the mechanisms involved in import process dysfunction have been not entirely clear.
Grpel2, a member of the mtHSP70 chaperone family, is involved in sensing oxidative stress to import precursor proteins in the mitochondrial matrix, thereby maintaining protein quality control and mitochondrial homeostasis [11, 34]. Grpel2 can form disulfide bonds and mediate metabolic adaptation to redox stress in a high-fat-feeding-induced oxidative environment [11]. A previous study reported that Grpel2 ablation significantly increased ROS production and promoted apoptosis by inhibiting the NF-κB pathway in hepatocellular carcinoma [34]. Our previous work indicated that Grpel2 alleviates myocardial ischaemia/reperfusion injury by inhibiting MCU-mediated mitochondrial calcium overload [12]. In this study, we explored the expression of Grpel2 in the diabetic heart and its role in DCM. For the first time, we observed that Grpel2 expression was decreased in the diabetic heart and in HG-treated NCMs. Then, to achieve prolonged overexpression of cardiac-specific Grpel2 in vivo, we designed adeno-associated virus serotype 9 under the control of a specific cTnT promoter. Cardiac Grpel2 overexpression markedly alleviated heart contractile and diastolic dysfunction induced by STZ injection. Moreover, Grpel2 overexpression attenuated cardiac hypertrophy, interstitial fibrosis and apoptosis in STZ-induced DCM. Considering that AAV-targeted gene therapies have low immunogenicity and have been explored in many preclinical and clinical studies, AAV9-Grpel2 treatment to alleviate heart dysfunction might be further validated in diabetic patients, which could be beneficial and provide hope for patients with diabetes.
Mitochondrial dysfunction and apoptosis orchestrate cardiomyocyte survival and death in the pathogenesis of DCM [15]. In the heart under pathological conditions, damaged mitochondria produce excessive harmful ROS, which eventually leads to apoptosis [35]. Mitochondrial homeostasis plays a central role in cardiomyocyte survival in DCM. Mitochondria depend on their necessary protein import machinery to respond to cellular stress. To date, mouse models for studying Grpel2-mediated mitochondrial function in vivo have not been reported, and little is known regarding the effects of Grpel2 on mitochondrial function in DCM. In the present study, we revealed that Grpel2 overexpression by AAV9-Grpel2 injection alleviated diabetes-induced mitochondrial morphological abnormalities and decreased oxidative stress levels in the diabetic heart in vivo. AAV transduction required a relatively high MOI in the cardiomyocytes, generally up to 1000 [36–38]. The MOI of adenoviruses used for the transfection of cardiomyocytes is much lower, generally 50–100 [39]. Moreover, adenoviruses have been widely used to transfect primary neonatal cardiomyocytes in our previous studies [31, 40] Thus, we continue to use adenoviruses in vitro in this study. Transfecting with AAV in primary cardiomyocytes is an alternative way for us in future studies. Our results indicated that Grpel2 overexpression by Ad-Grpel2 alleviated HG-induced mitochondrial dysfunction and apoptosis in NCMs. Overall, we demonstrated a protective role of Grpel2 in regulating mitochondrial oxidative stress in DCM.
To explore the detailed molecular mechanism by which Grpel2 regulates mitochondrial function, NCMs were isolated and infected with adenovirus in vitro, Compared with cell lines, primary neonatal mouse cardiomyocytes (NCMs) retain more biological characteristics of the original tissues in vivo. Due to ethical reasons and highly similar functional properties, isolated NCMs have been the most widely used models to study cardiac biology in vitro. We focused on the unique interactors of Grpel2 in an available public database [11]. We found that Grpel2 could bind to DLST by co-IP assays. As reported, DLST assists in converting αketoglutarate (αKG) into succinyl-CoA in the TCA cycle, which acts as an important entry point for glutamine anaplerosis [41]. Loss of DLST inhibits the electron transport chain via the reduction of NADH levels in human neuroblastoma cells [42]. DLST depletion induces ROS production in triple-negative breast cancer cells [43]. Overexpression of DLST in cardiomyocytes protects the heart against cardiac dysfunction upon pressure overload in vivo [44]. Our data revealed that Grpel2 knockdown decreased mitochondrial DLST expression, and upregulation of Grpel2 could maintain mitochondrial DLST expression under HG conditions. Importantly, the protective effects of Grpel2 overexpression on mitochondrial homeostasis were blocked by DLST knockdown. These results indicated that Grpel2 interacted with DLST and may be involved in facilitating DLST import into mitochondria to maintain mitochondrial function under HG conditions.
We further explored the transcriptional mechanism of *Grpel2* gene expression in NCMs. By performing bioinformatics analysis and ChIP assays, we found that Nr2f6 directly bound to the Grpel2 promoter region and regulated the transcription and expression of Grpel2. Nr2f6 belongs to the nuclear receptor (NR) superfamily, whose members directly bind to DNA loci as transcription factors [45]. Many NRs are critical for the development of the nervous system and heart, such as Nr2f1 and Nr2f2 [46]. Nr2f6 plays a crucial role in the regulation of hepatic lipid metabolism by directly binding to the CD36 promoter [47]. Previous studies have suggested that Nr2f6 may positively regulate tumour cell survival and induce cancer progression [48, 49]. An increasing number of studies have reported that Nr2f6 acts as a checkpoint that limits inflammatory tissue damage [50]. Our results revealed that Nr2f6 was downregulated in DCM. If Nr2f6 was knocked down by siRNA or overexpressed by adenovirus, Grpel2 was downregulated or upregulated accordingly. Our results provide the first evidence for understanding the regulatory mechanisms upstream of Grpel2 in the diabetic heart. Based on its effect on Grpel2 expression, Nr2f6 may also serve as a potential therapeutic target for DCM.
There are some limitations of our study. Firstly, since women with DCM also need effective therapies, female mice were not included in the in vivo studies. Secondly, the specific and dynamic process by which Grpel2 facilitates the import of DLST into mitochondria in DCM is still unclear. Thirdly, even though we identified Nr2f6 as a Grpel2 promoter, we cannot exclude other possible transcriptional mechanisms that contribute to Grpel2 expression. Finally, although many conclusions have been drawn on the basis of mouse models in vivo and in vitro, further preclinical and clinical studies are needed to confirm these results. Despite these limitations, we believe that our findings provide important novel insights for understanding the protective roles of Grpel2 and the underlying regulatory mechanisms in DCM.
## Conclusion
In summary, we uncovered for the first time that Grpel2 markedly attenuated heart dysfunction and cardiac remodeling in DCM by suppressing mitochondrial dysfunction, oxidative stress and aopotosis through Nr2f6 mediation the import of DLST into mitochondria in cardiomyocytes (Fig. 8). These findings suggest that targeting Grpel2 might be a promising therapeutic application for the treatment of patients with DCM. Fig. 8Schematic illustration of a novel molecular mechanism. Schematic illustration of a novel molecular mechanism by which the reduction of Nr2f6-regulated Grpel2 expression results in mitochondrial dysfunction in diabetic heart. Nr2f6 binds to the Grpel2 promoter to promote the Grpel2 expression. Diabetes-induced Nr2f6 reduction decreased the Grpe2 expression, which inhibited the import process of DLST into mitochondria. Afterward, the deficiency of DLST in mitochondria results in mitochondrial dysfunction, including increased ROS content, decreased ATP contents and decreased mitochondrial membrane potential, eventually aggravating DCM
## Supplementary Information
Additional file 1. Supplementary materials
## References
1. Dewanjee S, Vallamkondu J, Kalra RS, John A, Reddy PH, Kandimalla R. **Autophagy in the diabetic heart: a potential pharmacotherapeutic target in diabetic cardiomyopathy**. *Ageing Res Rev* (2021.0) **68** 101338. DOI: 10.1016/j.arr.2021.101338
2. Tan Y, Zhang Z, Zheng C, Wintergerst KA, Keller BB, Cai L. **Mechanisms of diabetic cardiomyopathy and potential therapeutic strategies: preclinical and clinical evidence**. *Nat Rev Cardiol* (2020.0) **17** 585-607. DOI: 10.1038/s41569-020-0339-2
3. Jankauskas SS, Kansakar U, Varzideh F, Wilson S, Mone P, Lombardi A. **Heart failure in diabetes. Metabolism: clinical and experimental**. *J Pharmacol Exp Therap* (2021.0) **125** 154910
4. Dabravolski SA, Sadykhov NK, Kartuesov AG, Borisov EE, Sukhorukov VN, Orekhov AN. **The role of mitochondrial abnormalities in diabetic cardiomyopathy**. *Int J Mol Sci* (2022.0) **23** 1. DOI: 10.3390/ijms23147863
5. Wiedemann N, Pfanner N. **Mitochondrial machineries for protein import and assembly**. *Annu Rev Biochem* (2017.0) **20** 685-714. DOI: 10.1146/annurev-biochem-060815-014352
6. Zhao F, Zou MH. **Role of the mitochondrial protein import machinery and protein processing in heart disease**. *Front Cardiovasc Med* (2021.0) **8** 749756. DOI: 10.3389/fcvm.2021.749756
7. Song J, Herrmann JM, Becker T. **Quality control of the mitochondrial proteome**. *Nat Rev Mol Cell Biol* (2021.0) **22** 54-70. DOI: 10.1038/s41580-020-00300-2
8. Shepherd DL, Hathaway QA, Nichols CE, Durr AJ, Pinti MV, Hughes KM. **Mitochondrial proteome disruption in the diabetic heart through targeted epigenetic regulation at the mitochondrial heat shock protein 70 (mtHsp70) nuclear locus**. *J Mol Cell Cardiol* (2018.0) **119** 104-115. DOI: 10.1016/j.yjmcc.2018.04.016
9. 9.Srivastava S, Savanur M, Sinha D, Birje A, Vigneshwaran R, Saha P, et al. Regulation of mitochondrial protein import by the nucleotide exchange factors GrpEL1 and GrpEL2 in human cells. J Biol Chem. 2017;292(44):18075–90.
10. 10.Tang C, Li Y, Chou C, Huang L, Huang S, Hueng D, et al. GRPEL2 knockdown exerts redox regulation in glioblastoma. Int J Mol Sci. 2021;22(23):1.
11. Konovalova S, Liu X, Manjunath P, Baral S, Neupane N, Hilander T. **Redox regulation of GRPEL2 nucleotide exchange factor for mitochondrial HSP70 chaperone**. *Redox Biol* (2018.0) **19** 37-45. DOI: 10.1016/j.redox.2018.07.024
12. Yang R, Zhang X, Xing P, Zhang S, Zhang F, Wang J. **Grpel2 alleviates myocardial ischemia/reperfusion injury by inhibiting MCU-mediated mitochondrial calcium overload**. *Biochem Biophys Res Commun* (2022.0) **609** 169-175. DOI: 10.1016/j.bbrc.2022.04.014
13. Cao T, Ni R, Ding W, Ji X, Li L, Liao G. **MLKL-mediated necroptosis is a target for cardiac protection in mouse models of type-1 diabetes**. *Cardiovasc Diabetol* (2022.0) **21** 165. DOI: 10.1186/s12933-022-01602-9
14. Yan W, Guo Y, Tao L, Lau WB, Gan L, Yan Z. **C1q/Tumor Necrosis Factor-Related Protein-9 Regulates the Fate of Implanted Mesenchymal Stem Cells and Mobilizes Their Protective Effects Against Ischemic Heart Injury via Multiple Novel Signaling Pathways**. *Circulation* (2017.0) **136** 2162-2177. DOI: 10.1161/CIRCULATIONAHA.117.029557
15. Liu C, Han Y, Gu X, Li M, Du Y, Feng N. **Paeonol promotes Opa1-mediated mitochondrial fusion via activating the CK2α-Stat3 pathway in diabetic cardiomyopathy**. *Redox Biol* (2021.0) **46** 102098. DOI: 10.1016/j.redox.2021.102098
16. Tong M, Saito T, Zhai P, Oka SI, Mizushima W, Nakamura M. **Mitophagy is essential for maintaining cardiac function during high fat diet-induced diabetic cardiomyopathy**. *Circ Res* (2019.0) **124** 1360-1371. DOI: 10.1161/CIRCRESAHA.118.314607
17. Cassis P, Cerullo D, Zanchi C, Corna D, Lionetti V, Giordano F. **ADAMTS13 deficiency shortens the life span of mice with experimental diabetes**. *Diabetes* (2018.0) **67** 2069-2083. DOI: 10.2337/db17-1508
18. 18.Zhang Y, Sun M, Wang D, Hu Y, Wang R, Diao H, et al. FTZ protects against cardiac hypertrophy and oxidative injury via microRNA-214/SIRT3 signaling pathway. Biomed Pharmacother Biomed Pharmacother. 2022;148:112696.
19. Song R, Dasgupta C, Mulder C, Zhang L. **MicroRNA-210 controls mitochondrial metabolism and protects heart function in myocardial infarction**. *Circulation* (2022.0) **145** 1140-1153. DOI: 10.1161/CIRCULATIONAHA.121.056929
20. Qi B, He L, Zhao Y, Zhang L, He Y, Li J. **Akap1 deficiency exacerbates diabetic cardiomyopathy in mice by NDUFS1-mediated mitochondrial dysfunction and apoptosis**. *Diabetologia* (2020.0) **63** 1072-1087. DOI: 10.1007/s00125-020-05103-w
21. Qi B, Song L, Hu L, Guo D, Ren G, Peng T. **Cardiac-specific overexpression of Ndufs1 ameliorates cardiac dysfunction after myocardial infarction by alleviating mitochondrial dysfunction and apoptosis**. *Exp Mol Med* (2022.0) **54** 946-960. DOI: 10.1038/s12276-022-00800-5
22. Buchholz M, Majchrzak-Stiller B, Hahn S, Vangala D, Pfirrmann RW, Uhl W. **Innovative substance 2250 as a highly promising anti-neoplastic agent in malignant pancreatic carcinoma—in vitro and in vivo**. *BMC Cancer* (2017.0) **17** 216. DOI: 10.1186/s12885-017-3204-x
23. Lin C, Guo Y, Xia Y, Li C, Xu X, Qi T. **FNDC5/Irisin attenuates diabetic cardiomyopathy in a type 2 diabetes mouse model by activation of integrin αV/β5-AKT signaling and reduction of oxidative/nitrosative stress**. *J Mol Cell Cardiol* (2021.0) **160** 27-41. DOI: 10.1016/j.yjmcc.2021.06.013
24. Li B, Chen K, Liu F, Zhang J, Chen X, Chen T. **Developmental angiogenesis requires the mitochondrial phenylalanyl-tRNA synthetase**. *Front Cardiovasc Med* (2021.0) **8** 724846. DOI: 10.3389/fcvm.2021.724846
25. Pan JA, Tang Y, Yu JY, Zhang H, Zhang JF, Wang CQ. **miR-146a attenuates apoptosis and modulates autophagy by targeting TAF9b/P53 pathway in doxorubicin-induced cardiotoxicity**. *Cell Death Dis* (2019.0) **10** 668. DOI: 10.1038/s41419-019-1901-x
26. 26.Hu J, Liu T, Fu F, Cui Z, Lai Q, Zhang Y, et al. Omentin1 ameliorates myocardial ischemia-induced heart failure via SIRT3/FOXO3a-dependent mitochondrial dynamical homeostasis and mitophagy. J Transl Med. 2022;20(1):447.
27. Chen Y, Wu G, Li M, Hesse M, Ma Y, Chen W. **LDHA-mediated metabolic reprogramming promoted cardiomyocyte proliferation by alleviating ROS and inducing M2 macrophage polarization**. *Redox Biol* (2022.0) **56** 102446. DOI: 10.1016/j.redox.2022.102446
28. Zhu H, Zhang L, Zhai M, Xia L, Cao Y, Xu L. **GDF11 alleviates pathological myocardial remodeling in diabetic cardiomyopathy through SIRT1-dependent regulation of oxidative stress and apoptosis**. *Front Cell Develop Biol* (2021.0) **9** 686848. DOI: 10.3389/fcell.2021.686848
29. Pezel T, Viallon M, Croisille P, Sebbag L, Bochaton T, Garot J. **Imaging interstitial fibrosis, left ventricular remodeling, and function in stage A and B heart failure**. *JACC Cardiovasc Imaging* (2021.0) **14** 1038-1052. DOI: 10.1016/j.jcmg.2020.05.036
30. Bhatt N, Aon M, Tocchetti C, Shen X, Dey S, Ramirez-Correa G. **Restoring redox balance enhances contractility in heart trabeculae from type 2 diabetic rats exposed to high glucose**. *Am J Physiol Heart Circ Physiol* (2015.0) **308** H291-302. DOI: 10.1152/ajpheart.00378.2014
31. Hu L, Ding M, Tang D, Gao E, Li C, Wang K. **Targeting mitochondrial dynamics by regulating Mfn2 for therapeutic intervention in diabetic cardiomyopathy**. *Theranostics* (2019.0) **9** 3687-3706. DOI: 10.7150/thno.33684
32. Nair GG, Tzanakakis ES, Hebrok M. **Emerging routes to the generation of functional β-cells for diabetes mellitus cell therapy**. *Nat Rev Endocrinol* (2020.0) **16** 506-518. DOI: 10.1038/s41574-020-0375-3
33. Baseler WA, Dabkowski ER, Williamson CL, Croston TL, Thapa D, Powell MJ. **Proteomic alterations of distinct mitochondrial subpopulations in the type 1 diabetic heart: contribution of protein import dysfunction**. *Am J Physiol Regul Integr Comp Physiol* (2011.0) **300** R186-200. DOI: 10.1152/ajpregu.00423.2010
34. Lai M, Zhu Q, Xu J, Zhang W. **Experimental and clinical evidence suggests that GRPEL2 plays an oncogenic role in HCC development**. *Am J Cancer Res* (2021.0) **11** 4175-4198. PMID: 34659882
35. Wu PY, Lai SY, Su YT, Yang KC, Chau YP, Don MJ. **β-Lapachone, an NQO1 activator, alleviates diabetic cardiomyopathy by regulating antioxidant ability and mitochondrial function**. *Phytomedicine* (2022.0) **104** 154255. DOI: 10.1016/j.phymed.2022.154255
36. Djurovic S, Iversen N, Jeansson S, Hoover F, Christensen G. **Comparison of nonviral transfection and adeno-associated viral transduction on cardiomyocytes**. *Mol Biotechnol* (2004.0) **28** 21-32. DOI: 10.1385/MB:28:1:21
37. Dandapat A, Hu CP, Li D, Liu Y, Chen H, Hermonat PL. **Overexpression of TGFbeta1 by adeno-associated virus type-2 vector protects myocardium from ischemia-reperfusion injury**. *Gene Ther* (2008.0) **15** 415-423. DOI: 10.1038/sj.gt.3303071
38. Ambrosi CM, Sadananda G, Han JL, Entcheva E. **Adeno-associated virus mediated gene delivery: implications for scalable in vitro and in vivo cardiac optogenetic models**. *Front Physiol* (2019.0) **10** 168. DOI: 10.3389/fphys.2019.00168
39. Louch WE, Sheehan KA, Wolska BM. **Methods in cardiomyocyte isolation, culture, and gene transfer**. *J Mol Cell Cardiol* (2011.0) **51** 288-298. DOI: 10.1016/j.yjmcc.2011.06.012
40. Ding M, Shi R, Cheng S, Li M, De D, Liu C. **Mfn2-mediated mitochondrial fusion alleviates doxorubicin-induced cardiotoxicity with enhancing its anticancer activity through metabolic switch**. *Redox Biol* (2022.0) **52** 102311. DOI: 10.1016/j.redox.2022.102311
41. Anderson NM, Li D, Peng HL, Laroche FJ, Mansour MR, Gjini E. **The TCA cycle transferase DLST is important for MYC-mediated leukemogenesis**. *Leukemia* (2016.0) **30** 1365-1374. DOI: 10.1038/leu.2016.26
42. Anderson NM, Qin X, Finan JM, Lam A, Athoe J, Missiaen R. **Metabolic enzyme DLST promotes tumor aggression and reveals a vulnerability to OXPHOS inhibition in high-risk neuroblastoma**. *Can Res* (2021.0) **81** 4417-4430. DOI: 10.1158/0008-5472.CAN-20-2153
43. Shen N, Korm S, Karantanos T, Li D, Zhang X, Ritou E. **DLST-dependence dictates metabolic heterogeneity in TCA-cycle usage among triple-negative breast cancer**. *Commun Biol* (2021.0) **4** 1289. DOI: 10.1038/s42003-021-02805-8
44. Heggermont WA, Papageorgiou AP, Quaegebeur A, Deckx S, Carai P, Verhesen W. **Inhibition of MicroRNA-146a and overexpression of its target dihydrolipoyl succinyltransferase protect against pressure overload-induced cardiac hypertrophy and dysfunction**. *Circulation* (2017.0) **136** 747-761. DOI: 10.1161/CIRCULATIONAHA.116.024171
45. Hermann-Kleiter N, Meisel M, Fresser F, Thuille N, Müller M, Roth L. **Nuclear orphan receptor NR2F6 directly antagonizes NFAT and RORγt binding to the Il17a promoter**. *J Autoimmun* (2012.0) **39** 428-440. DOI: 10.1016/j.jaut.2012.07.007
46. Wang T, Wang Z, de Fabritus L, Tao J, Saied EM, Lee HJ. **1-deoxysphingolipids bind to COUP-TF to modulate lymphatic and cardiac cell development**. *Dev Cell* (2021.0) **56** 3128-45.e15. DOI: 10.1016/j.devcel.2021.10.018
47. 47.Zhou B, Jia L, Zhang Z, Xiang L, Yuan Y, Zheng P, et al. The nuclear orphan receptor NR2F6 promotes hepatic steatosis through upregulation of fatty acid transporter CD36. Adv Sci (Weinheim, Baden-Wurttemberg, Germany). 2020;7(21):2002273.
48. Li H, Zhang W, Niu C, Lin C, Wu X, Jian Y. **Nuclear orphan receptor NR2F6 confers cisplatin resistance in epithelial ovarian cancer cells by activating the Notch3 signaling pathway**. *Int J Cancer* (2019.0) **145** 1921-1934. PMID: 30895619
49. Klepsch V, Gerner RR, Klepsch S, Olson WJ, Tilg H, Moschen AR. **Nuclear orphan receptor NR2F6 as a safeguard against experimental murine colitis**. *Gut* (2018.0) **67** 1434-1444. DOI: 10.1136/gutjnl-2016-313466
50. Hermann-Kleiter N, Baier G. **Orphan nuclear receptor NR2F6 acts as an essential gatekeeper of Th17 CD4+ T cell effector functions**. *Cell Commun Signal* (2014.0) **12** 38. DOI: 10.1186/1478-811X-12-38
|
---
title: 'Risk factors for SARS-CoV-2 related mortality and hospitalization before vaccination:
A meta-analysis'
authors:
- Hannah N. Marmor
- Mindy Pike
- Zhiguo (Alex) Zhao
- Fei Ye
- Stephen A. Deppen
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021978
doi: 10.1371/journal.pgph.0001187
license: CC BY 4.0
---
# Risk factors for SARS-CoV-2 related mortality and hospitalization before vaccination: A meta-analysis
## Abstract
The literature remains scarce regarding the varying point estimates of risk factors for COVID-19 associated mortality and hospitalization. This meta-analysis investigates risk factors for mortality and hospitalization, estimates individual risk factor contribution, and determines drivers of published estimate variances. We conducted a systematic review and meta-analysis of COVID-19 related mortality and hospitalization risk factors using PRISMA guidelines. Random effects models estimated pooled risks and meta-regression analyses estimated the impact of geographic region and study type. Studies conducted in North America and Europe were more likely to have lower effect sizes of mortality attributed to chronic kidney disease (OR: 0.21, $95\%$ CI: 0.09–0.52 and OR: 0.25, $95\%$ CI: 0.10–0.63, respectively). Retrospective studies were more likely to have decreased effect sizes of mortality attributed to chronic heart failure compared to prospective studies (OR: 0.65, $95\%$ CI: 0.44–0.95). Studies from Europe and Asia (OR: 0.42, $95\%$ CI: 0.30–0.57 and OR: 0.49, $95\%$ CI: 0.28–0.84, respectively) and retrospective studies (OR: 0.58, $95\%$ CI: 0.47–0.73) reported lower hospitalization risk attributed to male sex. Significant geographic population-based variation was observed in published comorbidity related mortality risks while male sex had less of an impact on hospitalization among European and Asian populations or in retrospective studies.
## Introduction
Coronavirus Disease 2019 (COVID-19) has quickly become a global pandemic with over 230 million confirmed cases and over 4 million deaths [1]. Clinical manifestations of COVID-19 have ranged from mild or no symptoms to death. While much remains to be learned about the virus, worse outcomes including hospitalization, intensive care admission, mechanical ventilation, and death have all been linked to older age, male sex, and medical comorbidities [2–5].
However, the literature remains scarce regarding potential reasons for varying point estimates of risk factors. For example, studies investigating smoking as a risk factor for worse outcomes have found vastly conflicting conclusions. While some have found smoking to be associated with an increased risk of death from COVID-19 [6], others have failed to find a similar association [7, 8]. Comorbidities such as chronic liver disease have also displayed substantial point estimate variance as risk factors for worse outcomes. Estimates for mortality risk for chronic liver disease have ranged from 0.55 to 5.88 with varying degrees of precision [6, 9].
In order to effectively treat patients, manage resources, and learn from this pandemic, we must not only recognize risk factors for COVID-19 associated mortality and hospitalization, but consider what drives the published variance of effect sizes among these risk factors as well. Our meta-analysis aims to investigate risk factors for mortality and hospitalization, estimate individual risk factor contribution, and determine likely drivers of point estimate effect size variances.
## Data sources and searches
We conducted a systematic review and meta-analysis of COVID-19 related mortality and hospitalization risk factors using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. PubMed and grey literature were searched, aided by a reference librarian using a predetermined search algorithm (Text in S1 Text). Reference lists of included papers were also reviewed for relevant studies.
We searched for studies published from May 2020 to January 2021 investigating chronic medical conditions and demographic characteristics as potential risk factors for hospitalization and mortality from COVID-19. The included risk factors were determined based on those present in the articles reviewed. Studies published after January 2021 were not included in order to avoid potential confounding from vaccine deployment, the arrival of additional variants, and the availability of new treatments.
We included preprints in our initial literature review. If the article was later published, we included the published version and excluded the original preprint. We excluded case reports, case series, meta-analyses, systematic reviews, editorials, guidelines, comments, letters to the editor, abstracts, studies which looked only at subpopulations (such as critically ill patients, those with cancer, or studies including only healthcare workers), were written in languages other than English, in which mortality or hospitalization was not a reported outcome, contained only descriptive statistics, those which were retracted, and those studies we were unable to access. We also excluded studies that only included lab values as risk factors, as lab values may not correspond to a medical diagnosis. We did not include features related to social determinants of health given the large amount of variation in inclusion, measurement, and categorization within the studies. Additionally, we excluded studies containing duplicate populations (either based on geographic location or hospital system) for the same outcome (Fig 1).
**Fig 1:** *Flow diagram for article inclusion and exclusion.*
## Study selection and data collection
HM and MP independently reviewed all articles for potential eligibility by screening titles, abstracts, and full-texts. Disagreements about eligibility were resolved through discussion with a third reviewer (SD).
## Data extraction
HM and MP independently extracted mortality and hospitalization data from each included study. Data relating to study characteristics, outcome (mortality or hospitalization), method of SARS-CoV-2 diagnosis, risk factors for outcome, associated point estimates with precision for risk factors, and study quality characteristics were collected and managed using REDCap electronic data capture tools hosted at Vanderbilt University. Specific risk factors collected for mortality included age, sex, coronary artery disease (CAD), chronic heart failure (CHF), chronic kidney disease (CKD), cardiovascular disease (CVD), diabetes, hypertension (HTN), lung disease, coronary heart disease (CHD), chronic obstructive pulmonary disease (COPD), cancer, immunosuppression, obesity, history of stroke, neurologic disease, and smoking history (former, current, or never). Specific risk factors collected for hospitalization included CVD, HTN, insurance status, sex, smoking history (former, current, or never), CHF, CKD, COPD, cancer, diabetes, and obesity.
## Quality assessment
The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines were utilized to assess quality of the included observational studies. Study quality characteristics included peer review status, description of the study setting, description of the study population, presence of a clearly defined outcome, discussion of statistical methods, whether missing data was addressed, inclusion of descriptive data, inclusion of precision, and discussion of study limitations.
## Outcomes
Outcomes of our meta-analysis included mortality or hospitalization secondary to COVID-19 infection as defined by individual studies. We excluded composite outcomes for both death and hospitalization, where the second outcome was mechanical ventilation, ICU, severe illness, or the outcome was unknown.
## Statistical methods
Descriptive characteristics including date of publication, study time frame, study design, country, population size, and method of COVID-19 diagnosis, were abstracted and described as median (interquartile range) or number (percentage).
We calculated pooled estimates of overall effect for each risk factor using random effects models with the inverse variance method (R packages “meta” and “mada”). The random effects model assumes that each study estimates a different underlying true effect, whereas the fixed-effect model assumes that all studies share the same, one common effect. The fixed-effect model assumption did not seem reasonable given the heterogeneity detected among studies.
The random effect was study, while the fixed effect was risk factor. We used the compound symmetry structure for the variance-covariance matrix. We did not consider other correlation structures as we assumed the matrix is diagonal and that it’s reasonable to assume the random effects are independent (i.e., the studies are independent of each other).
Forest plots for each risk factor were created for both mortality and hospitalization to visualize the point estimate (supplemented by $95\%$ confidence intervals) for each study involved and the overall pooled effect. Funnel plots were created to detect potential reporting bias, heterogeneity, and other bias in meta-analysis. Publication bias was additionally quantitatively measured using Egger’s test [10].
The effect of age was only estimated in two studies (one using logistic regression and the other Cox regression) with mortality as an outcome, and so a pooled estimate was not calculated [11, 12]. Additionally, coronary heart disease, COPD, lung disease, obesity, and stroke were only collected in 2 or fewer studies using Cox regression [13–17], while presence of one or more comorbidities and smoking were only collected in two studies using logistic regression [18, 19].
Likewise, insurance status was only collected in two studies with hospitalization as an outcome, and so the pooled estimate was not calculated [20, 21].
Further meta-regression analysis was conducted for studies with significant heterogeneity (I2>$50\%$) [22] to assess the impact of study-level covariates (moderators) on the estimation of effect sizes of the risk factors. Potential moderators investigated were geographic region (Europe vs. North America) and study type (retrospective vs prospective). All analyses were performed with R version 4.1.0 (R Foundation for Statistical Computing).
## Characteristics of included studies
We identified 2,539 studies in our database search and screened 2,505 articles after excluding for language ($$n = 30$$), abstract only ($$n = 2$$), and unavailability ($$n = 2$$) (Fig 1). After exclusions based on factors such as relevance, duplicate population, subpopulations, and retractions, 135 full-text articles were assessed for eligibility. We further excluded 51 articles with descriptive statistics only and included 71 mortality studies and 22 hospitalization studies in our final meta-analysis (Fig 1).
A total of 2,505 articles were screened for eligibility and a sample of 595 articles were reviewed by two reviewers to determine eligibility agreement. The overall agreement for study eligibility between reviewers (MP and HM) was $94.3\%$ and the k was 0.76, showing moderate agreement between reviewers. Consensus and discussion with a third reviewer (SD) were used when reviewers disagreed.
For studies with mortality as an outcome, the majority were published in December of 2020 with data collection from March to April 2020 (Table 1). Most studies were from various countries other than the United States. The most common study type was the retrospective cohort study, and the most common population type was patients admitted to the hospital. Notably, while most studies utilized laboratory test results to formulate a diagnosis of COVID-19, several included those cases diagnosed by clinical suspicion, radiographic findings, or diagnostic coding [13, 23–30]. While testing method (such as real time-polymerase chain reaction (RT-PCR) or antibody test) was included for most studies, there were several in which it was not [13, 14, 23, 31–39]. Furthermore, about half of the studies did not list the sample source (such as nasopharyngeal swab) for diagnostic methods [11, 13–16, 23–25, 27, 28, 31–36, 38–55].
**Table 1**
| Unnamed: 0 | Mortality | Hospitalization |
| --- | --- | --- |
| Study Characteristics | Study Characteristics | Study Characteristics |
| Number of studies | 71 | 22 |
| Median population size | 911 (397, 6916) | 8621 (2820, 20293) |
| Month of publication | | |
| May 2020 | 3 (4.2) | 2 (9.1) |
| June 2020 | 6 (8.5) | 4 (18.2) |
| July 2020 | 10 (14.1) | 4 (18.2) |
| August 2020 | 5 (7.0) | 4 (18.2) |
| September 2020 | 5 (7.0) | 2 (9.1) |
| October 2020 | 11 (15.5) | 2 (9.1) |
| November 2020 | 9 (12.7) | 1 (4.5) |
| December 2020 | 14 (19.7) | 2 (9.1) |
| January 2021 | 8 (11.3) | 1 (4.5) |
| Country | | |
| United States | 26 (36.6) | 10 (45.5) |
| Europe | 18 (25.4) | 7 (31.8) |
| Asia | 17 (23.9) | 1 (4.5) |
| Africa | 5 (7.0) | - |
| North America (not United States) | 1 (1.4) | 3 (13.6) |
| South America | 4 (5.6) | 1 (4.5) |
| Population type | | |
| General | 26 (36.6) | 22 (100) |
| Hospitalized | 44 (62.0) | - |
| Other | 1 (1.4) | - |
| Study design | | |
| Prospective Cohort | 6 (8.5) | 5 (22.7) |
| Retrospective Cohort | 45 (63.4) | 10 (45.5) |
| Other | 12 (16.9) | 3 (13.6) |
| Not listed | 8 (11.3) | 4 (18.2) |
| Multi-site | 37 (52.1) | 19 (86.4) |
| Type of analysis | | |
| Cox Proportional Hazards | 19 (26.8) | 1 (4.5) |
| Logistic Regression | 46 (64.8) | 20 (90.9) |
| Other | 6 (8.5) | 1 (4.5) |
| Adjusted | 53 (74.7) | 17 (77.3) |
| Median number of covariates | 4 (2, 10) | 7 (3, 11) |
| COVID Diagnosis | COVID Diagnosis | COVID Diagnosis |
| Case identification method | | |
| Clinical suspicion | 1 (1.4) | - |
| Lab confirmed | 62 (87.3) | 21 (95.5) |
| Lab or clinical | 1 (1.4) | 1 (4.5) |
| Lab or radiographic | 2 (2.8) | - |
| Lab or radiographic and clinical | 2 (2.8) | - |
| Lab or clinical or radiographic | 1 (1.4) | - |
| Lab and clinical and radiographic | 1 (1.4) | - |
| Diagnostic code or lab | 1 (1.4) | - |
| Testing Method | | |
| Diagnostic | 57 (80.3) | 18 (81.8) |
| Not listed | 12 (16.9) | 3 (13.6) |
| Diagnostic or antibody | 2 (2.8) | 1 (4.5) |
| Source of Sample | | |
| Nasopharyngeal | 29 (40.8) | 8 (36.4) |
| Not listed | 34 (47.9) | 14 (63.6) |
| Saliva or nasopharyngeal | 8 (11.3) | - |
| Study Quality Characteristics | Study Quality Characteristics | Study Quality Characteristics |
| Peer reviewed | 68 (95.8) | 18 (81.8) |
| Setting is described | 71 (100.0) | 22 (100.0) |
| Population defined | 71 (100.0) | 22 (100.0) |
| Outcome defined | 65 (91.5) | 20 (90.9) |
| Methods described | 71 (100.0) | 22 (100.0) |
| Addressed missing data | 37 (52.1) | 11 (50.0) |
| Included sensitivity analyses | 7 (9.9) | 5 (22.7) |
| Descriptive data included | 71 (100.0) | 22 (100.0) |
| Precision included | 71 (100.0) | 21 (95.5) |
| Discussed limitations | 62 (87.3) | 22 (100.0) |
In terms of statistical analyses, most studies used logistic regressions and included adjustments for risk estimates [12, 18, 24, 29, 36, 38–40, 42, 43, 46, 49, 51, 52, 54–70]. Information regarding statistical analysis for individual studies as well as the raw data extracted from each study can be found in the supplement. Finally, most studies were peer reviewed and few included sensitivity analyses [16, 37, 40, 71–74] (Table 1).
For studies with hospitalization as an outcome, the majority of studies were published in the summer of 2020 with data collection in the spring (Table 1). $54.5\%$ of the studies were from countries other than the United States. The majority of studies were retrospective cohort studies, and all were conducted in the general population (Table 1). All studies except for one utilized laboratory test results to diagnose COVID-19 [75], specifically RT-PCR. Fewer than half of studies included the sample source [12, 18, 19, 60, 76–79].
For statistical analyses, the majority of studies used logistic regressions with adjustments to determine risk estimates (S3 and S4 Tables). Peer review occurred in 18 of 22 ($81.8\%$) hospitalization studies. All studies discussed limitations and few included sensitivity analyses [19, 20, 40, 76, 79] (Table 1).
For both mortality and hospitalization, all included studies clearly described the setting, study participants, and methods. Reporting of risk factors was more heterogeneous, with some studies describing co-morbidities in greater detail according to guidelines.
## Mortality
After meta-analysis of studies that used logistic regression, pooled estimates for male sex, COPD, obesity, CHF, lung disease, neurologic disease, cancer, diabetes, and CKD were significant (Fig 2A–2L). In the meta-analysis of studies using Cox regression, pooled estimates for male sex, CKD, cancer, and diabetes were significant risk predictors (Fig 3A–3G).
**Fig 2:** *a-l. Forest plots for individual study estimates and pooled estimates of risk factors associated with COVID-19 related mortality in studies using logistic regression.* **Fig 3:** *a-g. Forest plots for individual study estimates and pooled estimates of risk factors associated with COVID-19 related mortality in studies using Cox regression.*
The presence of CKD was associated with poorer survival (meta-HR:1.57, $95\%$ CI: 1.25–1.97) and higher mortality (meta-OR: 2.13, $95\%$ CI: 1.69–2.67). Cancer was associated with poorer survival (meta-HR: 1.27, $95\%$ CI: 1.05–1.53) and higher mortality (meta-OR: 1.42, $95\%$ CI: 1.17–1.72) as was the presence of diabetes (meta-HR: 1.32, $95\%$ CI: 1.18–1.48) (meta-OR: 1.41, $95\%$ CI: 1.24–1.62). The odds of death for men were 1.48 times higher the odds for women ($95\%$ CI: 1.25–1.76) and the risk of death for men was 1.58 times higher the risk for women ($95\%$ CI: 1.13–2.20).
The presence of COPD was associated with higher odds of mortality (meta-OR: 1.59, $95\%$ CI: 1.25–2.03) and the presence of obesity was also associated with higher odds of death (meta-OR: 1.28, $95\%$ CI: 1.00–1.65). CHF, lung disease, and neurologic disease were all associated with death with meta-ORs of 1.83 ($95\%$ CI: 1.20–2.78), 1.16 ($95\%$ CI: 1.07–1.25), and 1.65 ($95\%$ CI: 1.00–2.74), respectively.
CVD, hypertension, CAD, and immunosuppression were not statistically associated with increased risk of death (Figs 2B, 2D and 2J and 3C, 3F and 3G), however with lower bound confidence intervals near 1.0, this does not necessarily exclude clinical significance.
Meta- regression analysis of mortality demonstrated study region (continent) to be a significant effect moderator for CKD, and study type to be a significant effect moderator for CHF. North American based population studies were more likely to have a lower estimate for CKD mortality risk (OR: 0.21, $95\%$ CI: 0.09–0.52) than Asian, South American, and African based population studies. Similarly, European populations were also more likely to have lower mortality estimates for CKD (OR: 0.25, $95\%$ CI: 0.10–0.63) than Asian, South American, and African populations. The stratified mortality analyses for CKD by study region can be found in Fig 4, showing an estimated pooled OR of 4.16 ($95\%$ CI: 2.51–6.89) for Asia, 1.69 ($95\%$ CI: 1.13–2.52) for North America, and 1.87 ($95\%$ CI: 1.47–2.39) for Europe. Retrospective studies were also more likely to have decreased mortality risks attributed to CHF compared to prospective studies (OR: 0.65, $95\%$ CI: 0.44–0.95). For meta-regression analysis of time-to-death, no significant moderators were detected at the 0.05 significance level. Results of the meta-regression analysis can be found in Table 2.
**Fig 4:** *Stratified mortality meta-analysis for chronic kidney disease by study region.* TABLE_PLACEHOLDER:Table 2
## Hospitalization
After meta-analysis, the risk factors significantly associated with hospitalization included male sex, CKD, CHF, CVD, hypertension, COPD, diabetes, and obesity (Fig 5A–5K). Cancer, past history of smoking, and current history of smoking were not significant risk factors for hospitalization with meta-ORs of 1.05 ($95\%$ CI: 0.76–1.45), 0.99 ($95\%$ CI: 0.66–1.50), and 1.14 ($95\%$ CI: 0.68–1.93), respectively. The analysis of sex demonstrated that male sex was associated with 1.62 higher odds of hospitalization ($95\%$ CI: 1.37–1.92). The presence of CKD, CHF, CVD, hypertension, COPD, diabetes, and obesity were all associated with hospitalization with meta-ORs of 2.90 ($95\%$ CI: 1.96–4.30), 2.18 ($95\%$ CI: 1.14–4.18), 1.37 ($95\%$ CI: 1.10–1.69), 1.46 ($95\%$ CI: 1.29–1.67), 1.35 ($95\%$ CI: 1.13–1.62), 2.08 ($95\%$ CI: 1.71–2.53), and 1.88 ($95\%$ CI: 1.44–2.45), respectively.
**Fig 5:** *a-k. Forest plots for individual study estimates and pooled estimates of risk factors associated with COVID-19 related hospitalization.*
Meta-regression analysis of hospitalization risk revealed study region and type to be significant effect moderators for sex. European based (OR: 0.42, $95\%$ CI: 0.30–0.57) and Asian based (OR: 0.49, $95\%$ CI: 0.28–0.84) study populations had decreased risks of hospitalization attributed to male sex compared to North American based populations. Similarly, retrospective studies were more likely to report lower effect sizes of male sex as a risk factor for hospitalization compared to prospective studies (OR: 0.58, $95\%$ CI: 0.47–0.73). Results of the meta-regression analysis can be found in Table 2.
## Publication bias analysis
In our analysis of the funnel plots for this study, most were symmetrical indicating minimal bias and between-study heterogeneity. There was some degree of asymmetry in the funnel plot for the association of CVD with hospitalization suggesting potential bias. However, Egger’s test was not significant for publication bias ($$p \leq 0.63$$) and only a few studies were included in this meta-analysis so the power is low to distinguish chance from real asymmetry that would indicate bias or heterogeneity (S1 Fig).
The funnel plot for the association between cancer and mortality in Cox regression studies showed asymmetry and Egger’s test was significant for publication bias ($$p \leq 0.02$$) (S2 Fig). Egger’s test additionally showed significant publication bias for cancer ($p \leq 0.01$), CAD ($$p \leq 0.01$$), and CVD ($$p \leq 0.02$$) in logistic regression mortality studies. The funnel plots had fewer studies in the left lower corner and this asymmetry indicates that smaller studies with results closer to the null were less likely to be reported or included in the study. ( S3–S5 Figs).
## Discussion
We conducted a meta-analysis of risk factors for COVID-19 associated mortality and hospitalization in the era from early pandemic through the arrival of new variants such as Delta, further advances in treatment modalities, and the initiation of directed risk group vaccination in 2021. We found that male sex, COPD, obesity, CHF, lung disease, neurologic disease, cancer, diabetes, and CKD were significantly associated with mortality. For hospitalization analyses, male sex, CHF, CKD, CVD, hypertension, COPD, diabetes, and obesity were found to be significant risk factors. We also examined possible sources of heterogeneity and found that geographic region and study type were significant for CKD, CHF, and sex.
Our literature review revealed a lack of published hospitalization specific meta-analyses. This could be due to variable criteria or differing thresholds for hospitalization. While mortality, mechanical ventilation, and vital signs are definite singular outcomes, hospitalization could be the result of a wide range of factors (especially given the dynamic shifts of healthcare resources seen in this pandemic). More data and additional studies are needed to assess the impact of hospitalization during this pandemic (i.e., on morbidity and mortality) and determine additional sources of variation.
Our meta-analysis is consistent with prior studies demonstrating an association between male sex and the presence of comorbidities with mortality [80–82]. However, CAD, CVD, hypertension, and immunosuppression were not associated with poorer survival or increased mortality in our analyses. Other meta-analyses have reported an association between CVD, coronary heart disease, immunosuppression, HTN and death [82–87]. However, among observational studies, the association between HTN and death has been more varied with some failing to demonstrate an association (aHR: 0.98, $95\%$ CI: 0.78–1.23) [19], (aHR: 0.89, $95\%$ CI: 0.85–0.93) [13], and (aRR: 1.07, $95\%$ CI: 0.79–1.45) [87].
Meta-analysis revealed CHF, CKD, CVD, hypertension, COPD, diabetes, obesity, and male sex were significantly associated with hospitalization due to COVID-19, which is consistent with previous literature [88–90]. Smoking status (both current and former) and cancer were not associated with increased odds of hospitalization in our meta-analysis. Additionally, our meta-regression of smoking status demonstrated no significant moderators of variance at the 0.05 significance level. Other observational studies have found associations between cancer and hospitalization with risk estimates varying depending on cancer type and treatment status [91, 92]. Similarly, some studies found an association between smoking and worse outcomes while others failed to find similar results [93–96].
Each of the above differences in reported point estimates for these risk factors likely represent the multifaceted variation seen during the COVID-19 pandemic. Geographic variation contributes to different prevalence of risk factors, disease, disease severity, and disease treatment in the underlying populations around the world. For example, diet and lifestyle habits vary among countries, and these factors play a role in the development and treatment of disease. Our meta-regression demonstrates the impact study region can have on risk factor effect size, specifically the risk factors CKD and sex for mortality and hospitalization. We have shown studies from Europe and Asia tend to have a higher likelihood of reporting smaller effect sizes of male sex on hospitalization. We also demonstrated that studies conducted in North America and Europe are more likely to report smaller effect sizes of CKD as a risk factor for mortality (Table 2). These differences in effect sizes could be related to lifestyle or treatment modalities.
There are several strengths in our study. We conducted an extensive literature review using two independent reviewers which contributed to a wide variety and number of studies included in the meta-analysis. These studies cover a wide geographic region both within and among different countries (S1 and S2 Tables). Furthermore, by excluding narrow subpopulations, we were able to increase the generalizability of our findings. We excluded duplicate populations so as not to include people twice in our analysis.
This study also has some limitations. We were not able to stratify all meta-analyses based on the study-level factors used in the meta-regression (i.e., study region and type), since after stratification, we were left with very few studies in each stratum. For example, there were not enough studies after stratification for hospitalization and time-to-death outcomes.
While there was a sufficient number of studies available after stratification for diabetes and hypertension, there was still remaining heterogeneity in the data for these risk factors that was not explained by study type and region, or anything else available for analysis. This is the reason we performed random-effect models instead of fixed-effect in the meta-analyses.
Additionally, we did not exclude studies based on method of COVID-19 diagnosis. While the majority of included studies used laboratory methods to diagnose infection, there were some which relied on clinical suspicion and/or radiographic findings for diagnosis. This likely represents different stages of testing during the pandemic, however some of the findings in these studies may have been incorrectly associated with COVID-19. Likewise, much of the data pertaining to risk factors was collected through electronic medical records which can often be incomplete, outdated, or inaccurate. Additionally, risk factor definitions varied from study to study. Therefore, we could not separate data based risk factor severity, treatment, or duration.
Finally, given the extensive amount of published literature related to COVID-19, reviewers could have missed relevant studies in the literature review process.
## Conclusions
In conclusion, geographic region and study type were associated with observed variances in risk point estimate. Men and those with certain medical conditions such as kidney, heart, and lung disease are at significantly increased risk of mortality or hospitalization due to COVID-19, however cancer and smoking status were not significant risk factors for hospitalization. We demonstrated that studies conducted in North America and Europe are more likely to report smaller effect sizes of CKD as a risk factor for mortality, similarly to retrospective studies being more likely to report smaller effect sizes attributed to CHF. Additionally, we demonstrated retrospective studies from Europe and Asia are more likely to show lower effect sizes of male sex as a risk factor for hospitalization. Our meta-analysis highlights this rapidly changing pandemic with high geographic variation. This variation drives the heterogeneity we see in published literature, increasing the difficulty for a consistent, unified public health message.
## References
1. Dong E, Du H, Gardner L. **An interactive web-based dashboard to track COVID-19 in real time**. *Lancet Infect Dis* (2020) **20** 533-4. DOI: 10.1016/S1473-3099(20)30120-1
2. Murillo-Zamora E, Hernandez-Suarez CM. **Survival in adult inpatients with COVID-19**. *Public Health* (2021) **190** 1-3. DOI: 10.1016/j.puhe.2020.10.029
3. Wang X, Fang X, Cai Z, Wu X, Gao X, Min J. **Comorbid Chronic Diseases and Acute Organ Injuries Are Strongly Correlated with Disease Severity and Mortality among COVID-19 Patients: A Systemic Review and Meta-Analysis**. *Research (Wash D C)* (2020) **2020** 2402961. DOI: 10.34133/2020/2402961
4. **Severe Outcomes Among Patients with Coronavirus Disease 2019 (COVID-19)—United States, February 12-March 16, 2020**. *MMWR Morb Mortal Wkly Rep* (2020) **69** 343-6. DOI: 10.15585/mmwr.mm6912e2
5. Izurieta HS, Graham DJ, Jiao Y, Hu M, Lu Y, Wu Y. **Natural History of Coronavirus Disease 2019: Risk Factors for Hospitalizations and Deaths Among >26 Million US Medicare Beneficiaries**. *J Infect Dis* (2021) **223** 945-56. PMID: 33325510
6. Dorjee K, Kim H, Bonomo E, Dolma R. **Prevalence and predictors of death and severe disease in patients hospitalized due to COVID-19: A comprehensive systematic review and meta-analysis of 77 studies and 38,000 patients**. *PLoS One* (2020) **15** e0243191. DOI: 10.1371/journal.pone.0243191
7. Ho KS, Narasimhan B, Sheehan J, Wu L, Fung JY. **Controversy over smoking in COVID-19-A real world experience in New York city**. *J Med Virol* (2021) **93** 4537-43. DOI: 10.1002/jmv.26738
8. Lippi G, Henry BM. **Active smoking is not associated with severity of coronavirus disease 2019 (COVID-19)**. *Eur J Intern Med* (2020) **75** 107-8. DOI: 10.1016/j.ejim.2020.03.014
9. Galiero R, Pafundi PC, Simeon V, Rinaldi L, Perrella A, Vetrano E. **Impact of chronic liver disease upon admission on COVID-19 in-hospital mortality: Findings from COVOCA study**. *PLoS One* (2020) **15** e0243700. DOI: 10.1371/journal.pone.0243700
10. Egger M, Davey Smith G, Schneider M, Minder C. **Bias in meta-analysis detected by a simple, graphical test**. *BMJ* (1997) **315** 629-34. DOI: 10.1136/bmj.315.7109.629
11. Chen L, Yu J, He W, Chen L, Yuan G, Dong F. **Risk factors for death in 1859 subjects with COVID-19**. *Leukemia* (2020) **34** 2173-83. DOI: 10.1038/s41375-020-0911-0
12. Mendy A, Apewokin S, Wells AA, Morrow AL. **Factors Associated with Hospitalization and Disease Severity in a Racially and Ethnically Diverse Population of COVID-19 Patients**. *medRxiv* (2020)
13. Williamson EJ, Walker AJ, Bhaskaran K, Bacon S, Bates C, Morton CE. **Factors associated with COVID-19-related death using OpenSAFELY**. *Nature* (2020) **584** 430-6. DOI: 10.1038/s41586-020-2521-4
14. Bello-Chavolla OY, Bahena-López JP, Antonio-Villa NE, Vargas-Vázquez A, González-Díaz A, Márquez-Salinas A. **Predicting Mortality Due to SARS-CoV-2: A Mechanistic Score Relating Obesity and Diabetes to COVID-19 Outcomes in Mexico**. *J Clin Endocrinol Metab* (2020) **105**. DOI: 10.1210/clinem/dgaa346
15. Giorgi Rossi P, Marino M, Formisano D, Venturelli F, Vicentini M, Grilli R. **Characteristics and outcomes of a cohort of COVID-19 patients in the Province of Reggio Emilia, Italy**. *PLoS One* (2020) **15** e0238281. DOI: 10.1371/journal.pone.0238281
16. Kabarriti R, Brodin NP, Maron MI, Guha C, Kalnicki S, Garg MK. **Association of Race and Ethnicity With Comorbidities and Survival Among Patients With COVID-19 at an Urban Medical Center in New York**. *JAMA Netw Open* (2020) **3** e2019795. DOI: 10.1001/jamanetworkopen.2020.19795
17. Alguwaihes AM, Al-Sofiani ME, Megdad M, Albader SS, Alsari MH, Alelayan A. **Diabetes and Covid-19 among hospitalized patients in Saudi Arabia: a single-centre retrospective study**. *Cardiovasc Diabetol* (2020) **19** 205. DOI: 10.1186/s12933-020-01184-4
18. Reilev M, Kristensen KB, Pottegård A, Lund LC, Hallas J, Ernst MT. **Characteristics and predictors of hospitalization and death in the first 11 122 cases with a positive RT-PCR test for SARS-CoV-2 in Denmark: a nationwide cohort**. *Int J Epidemiol* (2020) **49** 1468-81. DOI: 10.1093/ije/dyaa140
19. Petrilli CM, Jones SA, Yang J, Rajagopalan H, O’Donnell L, Chernyak Y. **Factors associated with hospital admission and critical illness among 5279 people with coronavirus disease 2019 in New York City: prospective cohort study**. *BMJ* (2020) **369** m1966. DOI: 10.1136/bmj.m1966
20. Azar KMJ, Shen Z, Romanelli RJ, Lockhart SH, Smits K, Robinson S. **Disparities In Outcomes Among COVID-19 Patients In A Large Health Care System In California**. *Health Aff (Millwood)* (2020) **39** 1253-62. PMID: 32437224
21. Price-Haywood EG, Burton J, Fort D, Seoane L. **Hospitalization and Mortality among Black Patients and White Patients with Covid-19**. *N Engl J Med* (2020) **382** 2534-43. DOI: 10.1056/NEJMsa2011686
22. Higgins JPT, Thompson SG. **Quantifying heterogeneity in a meta-analysis**. *Stat Med* (2002) **21** 1539-58. DOI: 10.1002/sim.1186
23. Tartof SY, Qian L, Hong V, Wei R, Nadjafi RF, Fischer H. **Obesity and Mortality Among Patients Diagnosed With COVID-19: Results From an Integrated Health Care Organization**. *Ann Intern Med* (2020) **173** 773-81. DOI: 10.7326/M20-3742
24. Munblit D, Nekliudov NA, Bugaeva P, Blyuss O, Kislova M, Listovskaya E. **Stop COVID Cohort: An Observational Study of 3480 Patients Admitted to the Sechenov University Hospital Network in Moscow City for Suspected Coronavirus Disease 2019 (COVID-19) Infection**. *Clin Infect Dis* (2021) **73** 1-11. DOI: 10.1093/cid/ciaa1535
25. Zali A, Gholamzadeh S, Mohammadi G, Azizmohammad Looha M, Akrami F, Zarean E. **Baseline Characteristics and Associated Factors of Mortality in COVID-19 Patients; an Analysis of 16000 Cases in Tehran, Iran**. *Arch Acad Emerg Med* (2020) **8** e70. PMID: 33134966
26. Loffi M, Piccolo R, Regazzoni V, Di Tano G, Moschini L, Robba D. **Coronary artery disease in patients hospitalised with Coronavirus disease 2019 (COVID-19) infection**. *Open Heart* (2020) **7** e001428. DOI: 10.1136/openhrt-2020-001428
27. Caliskan T, Saylan B. **Smoking and comorbidities are associated with COVID-19 severity and mortality in 565 patients treated in Turkey: a retrospective observational study**. *Rev Assoc Med Bras (1992)* (2020) **66** 1679-84. DOI: 10.1590/1806-9282.66.12.1679
28. Mejía F, Medina C, Cornejo E, Morello E, Vásquez S, Alave J. **Oxygen saturation as a predictor of mortality in hospitalized adult patients with COVID-19 in a public hospital in Lima, Peru**. *PLoS One* (2020) **15** e0244171. DOI: 10.1371/journal.pone.0244171
29. Salari A, Mahdavi-Roshan M, Ghorbani Z, Mortazavi SS, Naghshbandi M, Faraghnia F. **An investigation of risk factors of in-hospital death due to COVID-19: a case-control study in Rasht, Iran**. *Ir J Med Sci* (2021). DOI: 10.1007/s11845-020-02455-5
30. Rastad H, Ejtahed HS, Shafiee G, Safari A, Shahrestanaki E, Khodaparast Z. **The risk factors associated with COVID-19-Related death among patients with end-stage renal disease**. *BMC Nephrol* (2021) **22** 33. DOI: 10.1186/s12882-020-02221-w
31. Mishra V, Burma AD, Das SK, Parivallal MB, Amudhan S, Rao GN. **COVID-19-Hospitalized Patients in Karnataka: Survival and Stay Characteristics**. *Indian J Public Health* (2020) **64** S221-4. DOI: 10.4103/ijph.IJPH_486_20
32. Harmouch F, Shah K, Hippen JT, Kumar A, Goel H. **Is it all in the heart? Myocardial injury as major predictor of mortality among hospitalized COVID-19 patients**. *J Med Virol* (2021) **93** 973-82. DOI: 10.1002/jmv.26347
33. Smith AA, Fridling J, Ibrahim D, Porter PS. **Identifying Patients at Greatest Risk of Mortality due to COVID-19: A New England Perspective**. *West J Emerg Med* (2020) **21** 785-9. DOI: 10.5811/westjem.2020.6.47957
34. Mohammed M, Muhammad S, Mohammed FZ, Mustapha S, Sha’aban A, Sani NY. **Risk Factors Associated with Mortality Among Patients with Novel Coronavirus Disease (COVID-19) in Africa**. *J Racial Ethn Health Disparities* (2021) **8** 1267-72. DOI: 10.1007/s40615-020-00888-3
35. Doganci S, Ince ME, Ors N, Yildirim AK, Sir E, Karabacak K. **A new COVID-19 prediction scoring model for in-hospital mortality: experiences from Turkey, single center retrospective cohort analysis**. *Eur Rev Med Pharmacol Sci* (2020) **24** 10247-57. DOI: 10.26355/eurrev_202010_23249
36. Shah C, Grando DJ, Rainess RA, Ayad L, Gobran E, Benson P. **Factors associated with increased mortality in hospitalized COVID-19 patients**. *Ann Med Surg (Lond)* (2020) **60** 308-13. DOI: 10.1016/j.amsu.2020.10.071
37. Rustgi V, Makar M, Minacapelli CD, Gupta K, Bhurwal A, Li Y. **In-Hospital Mortality and Prediction in an Urban U.S. Population With COVID-19**. *Cureus* (2020) **12** e11786. DOI: 10.7759/cureus.11786
38. Kristić I, Pehlić M, Pavlović M, Kolarić B, Kolčić I, Polašek O. **Coronavirus epidemic in Croatia: case fatality decline during summer?**. *Croat Med J* (2020) **61** 501-7. DOI: 10.3325/cmj.2020.61.501
39. Mash RJ, Presence-Vollenhoven M, Adeniji A, Christoffels R, Doubell K, Eksteen L. **Evaluation of patient characteristics, management and outcomes for COVID-19 at district hospitals in the Western Cape, South Africa: descriptive observational study**. *BMJ Open* (2021) **11** e047016. DOI: 10.1136/bmjopen-2020-047016
40. Chishinga N, Gandhi NR, Onwubiko UN, Telford C, Prieto J, Smith S. **Characteristics and Risk Factors for Hospitalization and Mortality among Persons with COVID-19 in Atlanta Metropolitan Area**. *medRxiv* (2020). DOI: 10.1101/2020.12.15.20248214
41. Escalera-Antezana JP, Lizon-Ferrufino NF, Maldonado-Alanoca A, Alarcon-De-la-Vega G, Alvarado-Arnez LE, Balderrama-Saavedra MA. **Risk factors for mortality in patients with Coronavirus Disease 2019 (COVID-19) in Bolivia: An analysis of the first 107 confirmed cases**. *Infez Med* (2020) **28** 238-42. PMID: 32487789
42. Pettit NN, MacKenzie EL, Ridgway JP, Pursell K, Ash D, Patel B. **Obesity is Associated with Increased Risk for Mortality Among Hospitalized Patients with COVID-19**. *Obesity (Silver Spring)* (2020) **28** 1806-10. DOI: 10.1002/oby.22941
43. Nogueira PJ, de Araújo Nobre M, Costa A, Ribeiro RM, Furtado C, Bacelar Nicolau L. **The Role of Health Preconditions on COVID-19 Deaths in Portugal: Evidence from Surveillance Data of the First 20293 Infection Cases**. *J Clin Med* (2020) **9** E2368. DOI: 10.3390/jcm9082368
44. Zhao Z, Chen A, Hou W, Graham JM, Li H, Richman PS. **Prediction model and risk scores of ICU admission and mortality in COVID-19**. *PLoS One* (2020) **15** e0236618. DOI: 10.1371/journal.pone.0236618
45. McPadden J, Warner F, Young HP, Hurley NC, Pulk RA, Singh A. **Clinical Characteristics and Outcomes for 7,995 Patients with SARS-CoV-2 Infection**. *medRxiv* (2020). DOI: 10.1101/2020.07.19.20157305
46. Czernichow S, Beeker N, Rives-Lange C, Guerot E, Diehl JL, Katsahian S. **Obesity Doubles Mortality in Patients Hospitalized for Severe Acute Respiratory Syndrome Coronavirus 2 in Paris Hospitals, France: A Cohort Study on 5,795 Patients**. *Obesity (Silver Spring)* (2020) **28** 2282-9. DOI: 10.1002/oby.23014
47. Santos MM, Lucena EES, Lima KC, Brito AAC, Bay MB, Bonfada D. **Survival and predictors of deaths of patients hospitalised due to COVID-19 from a retrospective and multicentre cohort study in Brazil**. *Epidemiol Infect* (2020) **148** e198. DOI: 10.1017/S0950268820002034
48. Ramachandran P, Kathirvelu B, Chakraborti A, Gajendran M, Zhahid U, Ghanta S. **COVID-19 in Cancer Patients From New York City: A Comparative Single Center Retrospective Analysis**. *Cancer Control* (2020) **27**. DOI: 10.1177/1073274820960457
49. Omar SM, Musa IR, Salah SE, Elnur MM, Al-Wutayd O, Adam I. **High Mortality Rate in Adult COVID-19 Inpatients in Eastern Sudan: A Retrospective Study**. *J Multidiscip Healthc* (2020) **13** 1887-93. DOI: 10.2147/JMDH.S283900
50. Vafadar Moradi E, Teimouri A, Rezaee R, Morovatdar N, Foroughian M, Layegh P. **and white blood cells count are associated with higher COVID-19 mortality**. *Am J Emerg Med* (2021) **40** 11-4. PMID: 33333477
51. **The first wave of the COVID-19 pandemic in Spain: characterisation of cases and risk factors for severe outcomes, as at 27 April 2020**. *Euro Surveill* (2020) **25**. DOI: 10.2807/1560-7917.ES.2020.25.50.2001431
52. Ortiz-Prado E, Simbaña-Rivera K, Barreno LG, Diaz AM, Barreto A, Moyano C. **Epidemiological, socio-demographic and clinical features of the early phase of the COVID-19 epidemic in Ecuador**. *PLoS Negl Trop Dis* (2021) **15** e0008958. DOI: 10.1371/journal.pntd.0008958
53. Kvåle R, Bønaa KH, Forster R, Gravningen K, Júlíusson PB, Myklebust TÅ. **Does a history of cardiovascular disease or cancer affect mortality after SARS-CoV-2 infection?**. *Tidsskr Nor Laegeforen* (2021) **140**
54. Lee SG, Park GU, Moon YR, Sung K. **Clinical Characteristics and Risk Factors for Fatality and Severity in Patients with Coronavirus Disease in Korea: A Nationwide Population-Based Retrospective Study Using the Korean Health Insurance Review and Assessment Service (HIRA) Database**. *Int J Environ Res Public Health* (2020) **17** E8559. DOI: 10.3390/ijerph17228559
55. McNeill JN, Lau ES, Paniagua SM, Liu EE, Wang JK, Bassett IV. **The role of obesity in inflammatory markers in COVID-19 patients**. *Obes Res Clin Pract* (2021) **15** 96-9. DOI: 10.1016/j.orcp.2020.12.004
56. De Vito A, Geremia N, Fiore V, Princic E, Babudieri S, Madeddu G. **Clinical features, laboratory findings and predictors of death in hospitalized patients with COVID-19 in Sardinia, Italy**. *Eur Rev Med Pharmacol Sci* (2020) **24** 7861-8. DOI: 10.26355/eurrev_202007_22291
57. Almazeedi S, Al-Youha S, Jamal MH, Al-Haddad M, Al-Muhaini A, Al-Ghimlas F. **Characteristics, risk factors and outcomes among the first consecutive 1096 patients diagnosed with COVID-19 in Kuwait**. *EClinicalMedicine* (2020) **24** 100448. DOI: 10.1016/j.eclinm.2020.100448
58. Vena A, Giacobbe DR, Di Biagio A, Mikulska M, Taramasso L, De Maria A. **Clinical characteristics, management and in-hospital mortality of patients with coronavirus disease 2019 in Genoa, Italy**. *Clin Microbiol Infect* (2020) **26** 1537-44. DOI: 10.1016/j.cmi.2020.07.049
59. Miller J, Fadel RA, Tang A, Perrotta G, Herc E, Soman S. **The Impact of Sociodemographic Factors, Comorbidities, and Physiologic Responses on 30-Day Mortality in Coronavirus Disease 2019 (COVID-19) Patients in Metropolitan Detroit**. *Clin Infect Dis* (2021) **72** e704-10. DOI: 10.1093/cid/ciaa1420
60. Muñoz-Price LS, Nattinger AB, Rivera F, Hanson R, Gmehlin CG, Perez A. **Racial Disparities in Incidence and Outcomes Among Patients With COVID-19**. *JAMA Netw Open* (2020) **3** e2021892. DOI: 10.1001/jamanetworkopen.2020.21892
61. Lunski MJ, Burton J, Tawagi K, Maslov D, Simenson V, Barr D. **Multivariate mortality analyses in COVID-19: Comparing patients with cancer and patients without cancer in Louisiana**. *Cancer* (2021) **127** 266-74. DOI: 10.1002/cncr.33243
62. Islam MZ, Riaz BK, Islam ANMS, Khanam F, Akhter J, Choudhury R. **Risk factors associated with morbidity and mortality outcomes of COVID-19 patients on the 28th day of the disease course: a retrospective cohort study in Bangladesh**. *Epidemiol Infect* (2020) **148** e263. DOI: 10.1017/S0950268820002630
63. Tehrani S, Killander A, Åstrand P, Jakobsson J, Gille-Johnson P. **Risk factors for death in adult COVID-19 patients: Frailty predicts fatal outcome in older patients**. *Int J Infect Dis* (2021) **102** 415-21. DOI: 10.1016/j.ijid.2020.10.071
64. Kim TS, Roslin M, Wang JJ, Kane J, Hirsch JS, Kim EJ. **BMI as a Risk Factor for Clinical Outcomes in Patients Hospitalized with COVID-19 in New York**. *Obesity (Silver Spring)* (2021) **29** 279-84. DOI: 10.1002/oby.23076
65. Shah P, Owens J, Franklin J, Mehta A, Heymann W, Sewell W. **Demographics, comorbidities and outcomes in hospitalized Covid-19 patients in rural southwest Georgia**. *Ann Med* (2020) **52** 354-60. DOI: 10.1080/07853890.2020.1791356
66. Ayaz A, Arshad A, Malik H, Ali H, Hussain E, Jamil B. **Risk factors for intensive care unit admission and mortality in hospitalized COVID-19 patients**. *Acute Crit Care* (2020) **35** 249-54. DOI: 10.4266/acc.2020.00381
67. van Halem K, Bruyndonckx R, van der Hilst J, Cox J, Driesen P, Opsomer M. **Risk factors for mortality in hospitalized patients with COVID-19 at the start of the pandemic in Belgium: a retrospective cohort study**. *BMC Infect Dis* (2020) **20** 897. DOI: 10.1186/s12879-020-05605-3
68. Kaeuffer C, Le Hyaric C, Fabacher T, Mootien J, Dervieux B, Ruch Y. **Clinical characteristics and risk factors associated with severe COVID-19: prospective analysis of 1,045 hospitalised cases in North-Eastern France, March 2020**. *Euro Surveill* (2020) **25**
69. Elimian KO, Ochu CL, Ebhodaghe B, Myles P, Crawford EE, Igumbor E. **Patient characteristics associated with COVID-19 positivity and fatality in Nigeria: retrospective cohort study**. *BMJ Open* (2020) **10** e044079. DOI: 10.1136/bmjopen-2020-044079
70. Crouse AB, Grimes T, Li P, Might M, Ovalle F, Shalev A. **Metformin Use Is Associated With Reduced Mortality in a Diverse Population With COVID-19 and Diabetes**. *Front Endocrinol (Lausanne)* (2020) **11** 600439. PMID: 33519709
71. Hajifathalian K, Kumar S, Newberry C, Shah S, Fortune B, Krisko T. **Obesity is Associated with Worse Outcomes in COVID-19: Analysis of Early Data from New York City**. *Obesity (Silver Spring)* (2020) **28** 1606-12. DOI: 10.1002/oby.22923
72. Garibaldi BT, Fiksel J, Muschelli J, Robinson ML, Rouhizadeh M, Perin J. **Patient Trajectories Among Persons Hospitalized for COVID-19: A Cohort Study**. *Ann Intern Med* (2021) **174** 33-41. DOI: 10.7326/M20-3905
73. Farrell RJ, O’Regan R, O’Neill E, Bowens G, Maclellan A, Gileece A. **Sociodemographic variables as predictors of adverse outcome in SARS-CoV-2 infection: an Irish hospital experience**. *Ir J Med Sci* (2021) **190** 893-903. DOI: 10.1007/s11845-020-02407-z
74. Rodriguez-Nava G, Yanez-Bello MA, Trelles-Garcia DP, Chung CW, Chaudry S, Khan AS. **Clinical Characteristics and Risk Factors for Death of Hospitalized Patients With COVID-19 in a Community Hospital: A Retrospective Cohort Study**. *Mayo Clin Proc Innov Qual Outcomes* (2021) **5** 1-10. DOI: 10.1016/j.mayocpiqo.2020.10.007
75. Soares R de CM, Mattos LR, Raposo LM. **Risk Factors for Hospitalization and Mortality due to COVID-19 in Espírito Santo State, Brazil**. *Am J Trop Med Hyg* (2020) **103** 1184-90. DOI: 10.4269/ajtmh.20-0483
76. Lassale C, Gaye B, Hamer M, Gale CR, Batty GD. **Ethnic disparities in hospitalisation for COVID-19 in England: The role of socioeconomic factors, mental health, and inflammatory and pro-inflammatory factors in a community-based cohort study**. *Brain Behav Immun* (2020) **88** 44-9. DOI: 10.1016/j.bbi.2020.05.074
77. Zhu Z, Hasegawa K, Ma B, Fujiogi M, Camargo CA, Liang L. **Association of asthma and its genetic predisposition with the risk of severe COVID-19**. *J Allergy Clin Immunol* (2020) **146**. DOI: 10.1016/j.jaci.2020.06.001
78. Gottlieb M, Sansom S, Frankenberger C, Ward E, Hota B. **Clinical Course and Factors Associated With Hospitalization and Critical Illness Among COVID-19 Patients in Chicago, Illinois**. *Acad Emerg Med* (2020) **27** 963-73. DOI: 10.1111/acem.14104
79. Gu T, Mack JA, Salvatore M, Prabhu Sankar S, Valley TS, Singh K. **Characteristics Associated With Racial/Ethnic Disparities in COVID-19 Outcomes in an Academic Health Care System**. *JAMA Netw Open* (2020) **3** e2025197. DOI: 10.1001/jamanetworkopen.2020.25197
80. Noor FM, Islam MM. **Prevalence and Associated Risk Factors of Mortality Among COVID-19 Patients: A Meta-Analysis**. *J Community Health* (2020) **45** 1270-82. DOI: 10.1007/s10900-020-00920-x
81. Singh AK, Gillies CL, Singh R, Singh A, Chudasama Y, Coles B. **Prevalence of co-morbidities and their association with mortality in patients with COVID-19: A systematic review and meta-analysis**. *Diabetes Obes Metab* (2020) **22** 1915-24. DOI: 10.1111/dom.14124
82. Luo L, Fu M, Li Y, Hu S, Luo J, Chen Z. **The potential association between common comorbidities and severity and mortality of coronavirus disease 2019: A pooled analysis**. *Clin Cardiol* (2020) **43** 1478-93. DOI: 10.1002/clc.23465
83. Parohan M, Yaghoubi S, Seraji A, Javanbakht MH, Sarraf P, Djalali M. **Risk factors for mortality in patients with Coronavirus disease 2019 (COVID-19) infection: a systematic review and meta-analysis of observational studies**. *Aging Male* (2020) **23** 1416-24. DOI: 10.1080/13685538.2020.1774748
84. Zhang L, Hou J, Ma FZ, Li J, Xue S, Xu ZG. **The common risk factors for progression and mortality in COVID-19 patients: a meta-analysis**. *Arch Virol* (2021) **166** 2071-87. DOI: 10.1007/s00705-021-05012-2
85. Fang X, Li S, Yu H, Wang P, Zhang Y, Chen Z. **Epidemiological, comorbidity factors with severity and prognosis of COVID-19: a systematic review and meta-analysis**. *Aging (Albany NY)* (2020) **12** 12493-503. DOI: 10.18632/aging.103579
86. Pranata R, Lim MA, Huang I, Raharjo SB, Lukito AA. **Hypertension is associated with increased mortality and severity of disease in COVID-19 pneumonia: A systematic review, meta-analysis and meta-regression**. *J Renin Angiotensin Aldosterone Syst* (2020) **21**
87. Kim L, Garg S, O’Halloran A, Whitaker M, Pham H, Anderson EJ. **Risk Factors for Intensive Care Unit Admission and In-hospital Mortality Among Hospitalized Adults Identified through the US Coronavirus Disease 2019 (COVID-19)-Associated Hospitalization Surveillance Network (COVID-NET)**. *Clin Infect Dis* (2021) **72** e206-14. DOI: 10.1093/cid/ciaa1012
88. Ioannou GN, Locke E, Green P, Berry K, O’Hare AM, Shah JA. **Risk Factors for Hospitalization, Mechanical Ventilation, or Death Among 10 131 US Veterans With SARS-CoV-2 Infection**. *JAMA Netw Open* (2020) **3** e2022310. DOI: 10.1001/jamanetworkopen.2020.22310
89. Vahey GM, McDonald E, Marshall K, Martin SW, Chun H, Herlihy R. **Risk factors for hospitalization among persons with COVID-19-Colorado**. *PLoS One* (2021) **16** e0256917. DOI: 10.1371/journal.pone.0256917
90. Ko JY, Danielson ML, Town M, Derado G, Greenlund KJ, Kirley PD. **Risk Factors for Coronavirus Disease 2019 (COVID-19)-Associated Hospitalization: COVID-19-Associated Hospitalization Surveillance Network and Behavioral Risk Factor Surveillance System**. *Clin Infect Dis* (2021) **72** e695-703. DOI: 10.1093/cid/ciaa1419
91. Chakravarty D, Ratnani P, Sobotka S, Lundon D, Wiklund P, Nair SS. **Increased Hospitalization and Mortality from COVID-19 in Prostate Cancer Patients**. *Cancers (Basel)* (2021) **13** 1630. DOI: 10.3390/cancers13071630
92. Lee KA, Ma W, Sikavi DR, Drew DA, Nguyen LH, Bowyer RCE. **Cancer and Risk of COVID-19 Through a General Community Survey**. *Oncologist* (2021) **26**. DOI: 10.1634/theoncologist.2020-0572
93. Simons D, Shahab L, Brown J, Perski O. **The association of smoking status with SARS-CoV-2 infection, hospitalization and mortality from COVID-19: a living rapid evidence review with Bayesian meta-analyses (version 7)**. *Addiction* (2021) **116** 1319-68. DOI: 10.1111/add.15276
94. Hamer M, Kivimäki M, Gale CR, Batty GD. **Lifestyle risk factors, inflammatory mechanisms, and COVID-19 hospitalization: A community-based cohort study of 387,109 adults in UK**. *Brain Behav Immun* (2020) **87** 184-7. DOI: 10.1016/j.bbi.2020.05.059
95. Jehi L, Ji X, Milinovich A, Erzurum S, Merlino A, Gordon S. **Development and validation of a model for individualized prediction of hospitalization risk in 4,536 patients with COVID-19**. *PLoS One* (2020) **15** e0237419. DOI: 10.1371/journal.pone.0237419
96. Grundy EJ, Suddek T, Filippidis FT, Majeed A, Coronini-Cronberg S. **Smoking, SARS-CoV-2 and COVID-19: A review of reviews considering implications for public health policy and practice**. *Tob Induc Dis* (2020) **18** 58. DOI: 10.18332/tid/124788
|
---
title: A qualitative exploration of Chinese rural older adults’ adaption experience
to disability in Henan Province
authors:
- Mengke Gao
- Yan Zhang
- Yutong Tian
- Yue Gao
- Xiaohua Li
- Yixin Lu
journal: BMC Public Health
year: 2023
pmcid: PMC10021979
doi: 10.1186/s12889-023-15425-0
license: CC BY 4.0
---
# A qualitative exploration of Chinese rural older adults’ adaption experience to disability in Henan Province
## Abstract
### Background
The global population is ageing in a serious way and the number of disabled elderly people is increasing. Disability is a combination of physical and functional impairments, activity limitations, and social participation restrictions that significantly affect the quality of life of older adults. This study used the Roy adaptation model to examine the adaptive strategies of rural disabled elderly.
### Methods
An interview outline was prepared based on the Roy Adaptation Model, in-depth interviews were conducted with eligible rural elderly with disabilities using purposive sampling. Interview data were analyzed using the colaizzi method to obtain relevant themes and sub-themes of the adaptation experience.
### Results
Fifteen eligible disabled elderly participated in the interview, with an average age of 73.7 years old, showing different adaptation experiences in different aspects, a total of 5 themes and 18 sub-themes were extracted: (a)physiological function adaptation: learning to monitor physiological indicators, active medical compliance behavior, active rehabilitation exercise, adjusting lifestyle and coping with failure, (b) self-concept adaptation: adjustment of gratitude mentality, self-consolation, transferring the attention, seeking emotional comfort, and negative emotional response, (c) role function adaptation: positive self-care role, negative family role and escape of social role, (d) interdependence adaptation: actively seeking support and complex social coping, and (e) adaptation influencing factors: personal factors, caregiver factors and the policy factors.
### Conclusions
The disabled elderly show different adaptation strategies in four ways, and are affected by personal factors, caregiver factors and policy factors. A multi-faceted support system for the disabled elderly is recommended, and the caregivers should be trained in all-round care knowledge and skills.
## Introduction
The increase in the aging global population is the most challenging social problem in the world [1].The number of people aged 60 years or older will rise from 900 million to 2 billion between 2015 and 2050 (moving from 12 to $22\%$ of the total global population) [2]. As life expectancy increases, the number of older people with disabilities at risk of chronic illness or injury also inevitably increases [3]. Disability poses an important challenge to countries all over the world since it affects more than $15\%$ of the global population [2019] [4]. In 2020, a study conducted in eight low- and middle-income countries (China, Cuba, Dominican Republic, India, Mexico, Peru, Puerto Rico, and Venezuela) found that the proportion of remaining life spent disability-free at age 65 ranged from the lowest in Peru ($76\%$ for men and $69\%$ for women) to the highest in China ($92\%$ for men and $89\%$ for women) [5].At the end of the last century, China entered an aging society, with a rapidly growing, aging, disabled and empty-nest population. China has the largest population of older people with partial or total disabilities in the world [6]. In 2021, the results of the seventh national census showed that 18.70 percent of the population aged 60 and over in China, with over 200 million rural older adults [7].*The serious* population aging trend aggravates the physiological decline of the older adults, there are about 44 million disabled and semi-disabled elderly in China, and the total number of rural disabled people exceeds 8 million. It is expected that older disabled adults population in China will reach 77 million by 2030[2021] [8], with rural disabled people accounting for the largest proportion.
The Chinese government is very concerned about the disabled elderly, having proposed that improving and safeguarding their living conditions is an essential part of promoting the well-being of the people [9].In the Strategic Plan for Rural Revitalization (2018–2022), it is mentioned to support the construction of rural pension service systems for the disabled and semi-disabled elderly and to improve the level of mutual pension services [10]. Therefore, paying attention to the quality of life of rural disabled elderly is an important link to realize ‘healthy China’ and ‘ healthy countryside ‘. It will also be useful for research on disabled elderly in other countries/regions of the world.
The disability adaptation of the rural older adults needs to be paid attention to. Adaptability is a kind of ability that individuals adjust themselves to the new human environment, new communication groups, new behavior and target requirements [11]. According to the ICF (International Classification of Functioning, Disability and Health), an disabled person is one who, due to old age, illness or disability, has to be assisted by others in activities of daily living such as eating, bathing, dressing, toileting, continence control and indoor activities, or older people who are completely dependent on the assistance of others. Of the six activities listed above, one to two “can not do” is defined as “mild disability”, three to four “can not do” is defined as “moderate disability”, five to six “can not do” is defined as “severe disability” [12]. Disability is a combination of physical structural and functional impairment, activity limitation and social participation restriction in interaction with health conditions, resulting from the individual’s functional status, physical environment, cultural environment and policy environment [13]. Functional impairment is positively correlated with the increasing age of older people [14]. Therefore, the adaptability of disabled older people should be taken into account. Due to the loss of productivity resulting from partial or complete loss of self-care, coupled with low rural income levels, low medical resources and part of the traditional thinking of rural elderly people, rural disabled elderly people are unable to adapt to the physical changes caused by their disability and are more prone to physical, psychological, emotional and social maladjustment [15].Therefore, the adaptability of the elderly disabled in rural areas is an issue that needs urgent attention. However, research on the disabled elderly has mainly focused on care models, and studies on their adaptability have only been conducted on myocardial infarction patients and cancer patients in China [16, 17], with less research on the level of adaptability and factors affecting the disabled elderly in rural areas, and it is urgent to carry out relevant studies.
At present, much of the research on the rural older adults with disabilities has been based on the quantitative method [18]; however, a fixed questionnaire can often ignore the indelible individual experience of the interviewees. To obtain more vivid and richer knowledge, we adopt an inductive approach in this paper, using a qualitative method to search the subjective disability adaptation experiences of the disabled elderly in rural China through the collection of in-depth interviews. An in-depth interview is a type of unstructured, direct, deep, and one-to-one interview, which is an appropriate way to collect data on the potential motives, experiences, attitudes, and emotions of the respondents regarding a certain issue [19].
Based on the Roy Adaptation Model, this study interviewed the disabled elderly in a rural county in Henan Province to analyse this population’s experiences of adaptation to disability.
## Study setting and participants
Guided by the qualitative research method, this study analyzed the adaptation experience of rural disabled elderly through in-depth interviews. And this study referred to the qualitative research report standard prepared by O’ Brien et al. [ 20] as the guidance for research design and development. The principles of informed consent, no harm, and participant confidentiality were strictly followed in our study during the in-depth interviews. The research protocol was approved by the school’s ethics committee. ( ZZUIRB-2022–20).
We recruited participants through the local village committees, an administrative institution in rural China responsible for the provision of social welfare and social services for the rural disabled and rural older adults, including cash allowances, health rehabilitation,care services, and medical reimbursement support [21]. Henan *Province is* the most populous province in China and the third most populous province in terms of resident population, and the evolution of the stages of population ageing is in line with national trends. In 2020, the province’s population aged 60 and above reaches 17.964 million, accounting for $18.08\%$ of the province’s resident population [22]. In recent years, the level of population ageing in rural Henan *Province is* higher than that in urban areas, and the number of disabled and semi-disabled elderly people is on the rise [23]. Therefore, we chose rural areas in Henan Province as our study site. From June2022 to July 2022, the researcher used a purpose sampling method to investigate the disabled elderly in a administrative village in a county of Henan Province($16.91\%$ of older people in the area), where villagers are mainly engaged in farming and some young people work outside the home, representing most rural conditions in China [23].
The participants were recruited through purposive sampling, considering specific criteria related to the research objectives. The criteria were as follows:Over 60 years old;The degree of disability reaches at least “mild disability” (in accordance with ICF, eating, dressing, getting out of bed, toileting, indoor walking, bathing six indicators, one to two “can not do” is defined as “mild disability”);having certain language expression ability(ability to complete listening, speaking, reading and writing independently);*No previous* cognitive impairment or psychosis (one or more of the functions of memory, spatial ability, judgment and calculation are impaired);Voluntary participation in this study;Not in the acute phase of the disease(the elderly with unstable condition within two weeks of illness, such as fever, acute respiratory infection, asthma attack, etc.); or chronic infectious diseases (such as tuberculosis, hepatitis B that can spread between people).
The study began with an introduction to the purpose and methods of the study, an explanation of the study, and a promise to participants that all personal information and conversations would be confidential and secure, and that participants could withdraw from the study at any time without consequence. The sample size for this study was referenced to the principles of qualitative research and the inclusion of new subjects was stopped when the data analysis was saturated and no new themes emerged [24]. Ultimately, a total of 15 eligible rural elderly people with disabilities participated and completed the study, and all signed an informed consent form.
## Conceptual framework
Roy’s adaptation model was proposed by the renowned theorist Roy in the 1960s [25], argues that stimulus input affects the cognitive/regulatory system of an individual and leads to the cognitive adaptation process. The perception of illness in the internal environment interacts with this process, leading to dynamic adjustments in the body’s physiological functioning, self-concept, role functioning and interdependence, with the latest red manifestations being behavioural adaptation or maladaptation. These four adaptation strategies suggest that an outline of an interview on the adaptation experiences of rural older people with disabilities can be developed from these four areas. The conceptual framework is shown in Fig. 1.Fig. 1Conceptual framework
## Procedure
All researchers were trained in qualitative methods and interview techniques. Firstly, the researchers developed an interview guide based on Roy’s adaptation model [25], combined with expert opinion—focusing on asking the elderly about their reactions and coping styles before and after disability—and used language that was easily understood by the rural elderly, focusing on four aspects of the adaptation experiences of the disabled elderly. Through interviews with two cases of rural disabled older people, we found that different older people had different experiences of adaptation influenced by different factors. Based on this insight, the interview outline was adapted after discussion and influencing factors were added after each question. The final version of the interview guide is shown in Table 1. Demographic information was collected at the beginning of the formal interview, including information on gender, age, marital status, and educational attainment. Table 1Interview guidelineFinal editionQuestion1:What difficulties have been encountered in your life after illness? How do you deal with these difficulties? What factors affect you to deal with these difficulties?Question2: How your mood changes since the illness? How do you adjust your mood? What factors affect your emotion regulation?Question3: What are your changes in self-care roles, family roles, and social roles after your illness? How do you deal with the role changes before and after illness? What factors affect your response to role change?Question4:What social support do you perceive after illness? What changes have taken place in interpersonal relationship? How do you deal with these changes? What factors affect your response to these changes?
## Data collection
The interviews were conducted using a face-to-face semi-structured interview method. Prior to the interviews, the researcher made appointments by phone and text messages. The location and time of the interviews were chosen to suit the convenience of the interviewees. The interviews were all conducted in the patient’s home to ensure a relaxed and comfortable environment, with no other people present at the interview. The entire interviews wss recorded with the consent of the interviewee; the interview was asked in an open-ended questioning style, allowing the interviewee to freely express his or her true experiences and feelings. The interview questions and the interviewer’s attitude were neutral to avoid bias against the participants. During the interviews, the researcher carefully observed and recorded the subtle facial expressions and movements of the interviewees. The average interview time was 35 min.
## Data analysis
After each interview, the researcher transcribed the interview transcripts verbatim within 24 h, collated and documented each interviewee’s interview transcripts, and conducted data analysis and collection. The interview data were analysed using the Colaizzi seven-step analysis method [26]: a) detailed transcription and careful reading of all interview materials; b) selection of summaries and meaningful expressions that corresponded to the research phenomenon; c) distillation of meaning from the meaningful expressions; d) search for common conceptual or meaningful features to form themes, thematic groups and categories; e) linking themes to the research phenomenon to a complete narrative; f) stating the essential structure of the phenomenon; and g) returning the results to the interviewee to verify the authenticity of the content.
## Participants
The characteristics of the 15 participants are presented in Table 2. The adaptation experience of rural disabled elderly is finally summarized as the following 5 themes and 18 sub-themes in Table 3.Table 2Characteristics of the participantsCharacteristicsNumber of participants (%)CharacteristicsNumber of participants (%)Age groupsEducation level 60–694(26.67) Non-formal education2(13.33) 70–798(53.33) Elementary education4(26.67) 80–892(13.33) Middle school education7(46.67) 90–1001(6.67) High school education2(13.33)GenderMarital status Male8(53.33) Married8(53.33) Female7 (46.67) Widowed7(46.67)Degree of disabilityLength of Disability/years Mild Disability8(53.33) 0–58(53.33) Moderate Disability4(26.67) 6–104(26.67) Severe Disability3(20.00) 11–153(20.00)In accordance with ICF, eating, dressing, getting out of bed, toileting, indoor walking, bathing six indicators, one to two “can not do” is defined as “mild disability” ‘, three to four “can not do” is defined as “moderate disability”, five to six “can not do” is defined as “severe disability”Table 3Rural disabled elderly adaptation experience themeThemeSub-themePhysiological function adaptationa) learning to monitor physiological indicatorsb) active medical compliance behaviorc) active rehabilitation exercised) adjusting lifestylee) coping with failureSelf-concept adaptationa) adjustment of gratitude mentalityb) self-consolationc) transferring the attentiond) seeking emotional comforte) negative emotional responseRole function adaptationa) positive self-care roleb) negative family rolec) escape of social roleInterdependence adaptationa) actively seeking supportb) complex social copingadaptation influencing factorsa) personal factorsb) caregiver factorsc) policy factors Five themes were generated in the interview. The first four themes were the adaptive experiences of disabled elderly people in physiological function, self-concept, role function and interdependent mode, and the sub-themes were specific behaviors of adaptive mode or maladaptive outcomes. The fifth theme is the reasons that influence disabled elderly people to have these different adaptive experiences. The three sub-themes are explained in detail from personal reasons, caregiver reasons and policy reasons.
## Theme1: Physiological function adaptation
Physiological function adaptation refers to the ways in which rural disabled older adults cope with changes in their physical functioning, including physiological index response and physiological symptom response. Five sub-themes emerged in this theme.
## Learning to monitor physiological indicators
Seven rural disabled elderly mentioned that they learned to measure their blood pressure, blood sugar and other indicators after illness, and tried to maintain these physiological indicators related to disability at normal levels. They were mostly mildly disabled elderly, or highly educated severely disabled elderly. “I have an annual medical check-up and usually take my own blood pressure and blood sugar at home. If there are any physical symptoms, a call to the doctor can help.” ( female,85 years old, widowed, severe disability, high school education)
## Active medical compliance behavior
Ten respondents with disabilities due to chronic illnesses reported that they always followed medical advice, adhered to the principles of medication and coped with changes in physical function through specialist treatment. Most of these people were widowed, and the degree of disability was mild or moderate. “I take my medication regularly as required by my doctor, and because of my high blood pressure and thick blood lipids, I have to take several kinds of medication, all of which I take as required by my doctor. I had a stroke mainly because I didn’t pay attention to my high blood pressure before, so I am now very careful to take my blood pressure medication as required.” ( female,75 years old, married, mild disability, non-formal education)
## Active rehabilitation exercise
Seven out of 15 disabled elderly in rural areas reported that physical changes such as reduced mobility and swallowing disorders that affect self-care can occur after illness. They would choose to actively exercise their physical activity and swallowing ability, reflecting the rural elderly’s quest for a healthy physique. These people were mostly men, as well as married elderly people with mild disability, or highly educated elderly people with severe disability. “Doctors said that my swallowing function is affected. I often practice at home, repeatedly do pharyngeal movements, do tongue activities... And I can’t move this hand, now better, because I grabbed my own hands every day bed exercise, clench fists twenty times, pull arms twenty times, and then lying in bed, legs up in the air to learn to walk fifty times (while demonstration)...” (male,87 years old, widowed, severe disability, high school education)
## Adjusting lifestyle
Eight out of 15 respondents reported changing their bad habits after becoming ill. The quality of sleep decreased for some people. Some older adults with moderate to severe disabilities have severe activity limitations, resulting in increased bedtime and slower bowel movements, often leading to constipation. Some elderly families with disabilities cautiously adjusted their dietary strategies. These people showed different degrees of disability, with mild disability and male majority. “After getting sick, I quit smoking and drinking, and I don’t stay up late. ” ( male, 61 years old, married, mild disability, elementary education)“I usually don’t eat vegetables. For a period of time, I always constipation. My daughter always makes vegetables in different ways, squeezes them into juice and makes them into soup... and tries to make me eat more vegetables and regulate them.” ( male,87 years old, widowed, severe disability, high school education)
## Coping with failure
Ten out of 15 disabled elderly said they feel difficult to adapt to the decline of thinking ability, unable to cope with and choose to compromise. Among these people, 3out of 4 are women with low educational level. “I used to play chess with my friends, since the surgery, the brain can’t think about things, a thought is blind, I don’t want to think about anything...” (female, 70 years old, widowed, mild disability, elementary education)
## Theme 2: Self-concept adaptation
Self-concept is people’s emotion, confidence and evaluation of themselves at a certain time. Self-concept adaptation reflects the individual’s response to psychological and mental health changes. This theme includes five sub-themes.
## Adjustment of gratitude mentality
We found that 9 respondents can adjust their negative emotions with a heart of satisfaction and gratitude. 6 respondents felt happy and moved because they felt the care of their relatives in life after disability. These people were mostly married elderly people, or widowed elderly people with higher education levels. “I felt that my daughter-in-law was very kind to me, and I was moved... I can’t walk, my grandson hugged me downstairs. Without elevators, he held me in his hands and was tired to breathe. I said, having a strong man, I was so moved.” ( female,85 years old, widowed, severe disability, high school education)
## Self-consolation
The interview found that 8 rural disabled elderly often persuade themself to keep positive after illness, maintain an optimistic attitude, make their usual treatment of disease. Most of these people were married, and had a shorter duration of disability. “I sometimes think that it won’t be worse, so just rest and recuperate. The kids grow up and don’t need me to care. Anyway, life is much better than when I was a kid.” ( male,72 years old, married, moderate disability, middle school education)
## Transferring the attention
Seven disabled elderly in rural areas showed that they can relieve negative emotions by developing new interests and learning new things to transfer their attention. These people were mostly mildly disabled elderly. “It’s a little hard to talk after a stroke. And my hearing is poor. It’s very difficult to communicate with others. So I usually spending time painting alone. I paint on the blackboard with chalk, which make me feel happy..” (female,75 years old, widowed, mild disability, elementary education)
## Seeking emotional comfort
The disabled elderly will have feelings of loneliness, inferiority and self-accusation. They tell their friends about their feelings to alleviate their emotions. These people were widowed and disabled for a long time. “I have a friend, who often came home to chat with me, I often tell her my feelings which I don’t want to say to the family members, after that she will comfort and understand me, let me feel not lonely.” ( male, 92 years old, widowed, moderate disability, middle school education)
## Negative emotional response
Eleven out of 15 respondents had long been in negative emotions since their disability, including guilt and self-accusation, pessimistic depression and so on. And they cannot release these emotions successfully. These people were widowed elderly, mostly moderate disability and severe disability.
Ten rural disabled elderly felt remorseful for their family economic burden due to long illness. “I feel like I’m a burden in my family. Every day I exist adds a lot of burden to children.” ( male, 87 years old, widowed, severe disability, high school education) Seven respondents were disappointed and depressed because of limited activity or prolonged disease. “I lie in bed day by day because I can’t move, and I can’t sleep night by night. It’s better to die to think about what...” (male, 87 years old, widowed, severe disability, high school education)
## Theme 3: Role function adaptation
Role function adaptation includes coping with self-care role, family role and social role change of disabled elderly. This theme contains three sub-themes.
## Positive self-care role
Rural older people have an increased ability to care for themselves after illness. Most older people with disabilities love life and show positive self-care skills. This is demonstrated by paying close attention to their own health, actively cooperating with treatment, seeking help in a timely manner, learning about relevant health issues in a variety of ways, and developing self-care skills. These people were mostly married, mildly disabled elderly, or widowed elderly with junior school or high school education. “I often remind myself as a patient, the heart is not good, so don’t angry with others”. ( female,76 years old, widowed, mild disability, middle school education)“I usually pay attention to health knowledge on TV, and the blood glucose was well controlled because I don’t eat sweet food.” ( female, 85 years old, widowed, severe disability, high school education)“I especially care about my body, a little uncomfortable I will ask the people around me for help.” ( male, 61 years old, married, mild disability, elementary education)
## Negative family role coping
Eight older adults were reluctant to participate in the management and decision-making of family affairs after disability, and their concern for family members was also reduced. Only 4 out of 15 older adults were willing to do some housework that they can.
Seven out of 15 older adults lost their ability to work due to limited activities. They needed to be care all time at home and completely lost the ability to help families share responsibilities. “I’m in a bad mood after I get sick. I have no mood to do anything for my family.” ( female, 61 years old, married, mild disability, elementary education) There were also 8 out of 15 respondents who trid their own affairs within the scope of ability after disability, reflecting their family participation. “I can’ t do most of the housework after I get sick. I can only do some small things, such as sweeping the floor.” ( male, 60 years old, widowed, mild disability, elementary education)
## Escape of social roles
Rural disabled elderly rarely participate in social activities due to cultural constraints and physical reasons after illness. 5 disabled elderly indicated that they were not concerned about national affairs, and they had no TV or radio at home, unwilling to know the latest pension policy and unwilling to participate in collective activities. These elderly people were primary school educated or uneducated. “I only watch TV drama, not want to watch news at all,and I am just a sick rural old man, I do not care about national affairs, just want to have a good life.” ( male, 78 years old, married, mild disability, elementary education)“I don’t have much experience sharing with young people either. Now young people have their own opinions.” ( female, 61 years old, married, mild disability, elementary education)
## Theme 4: Interdependence adaptation
Interdependence refers to individuals develop relationships with others to meet emotional, developmental and resource needs. Interdependence adaptation refers to the situation of rural older adults who obtain social support and maintain interpersonal relationships after disability. There are two sub-themes in this theme.
## Actively seeking support
Actively seeking support is manifested in many aspects: seeking family support, seeking relatives and friends support, seeking village committee support, seeking village doctor support, perceived policy support. Those with these presentations tended to have mild or moderate disabilities, and there were more widowed older people than married older people, showing different levels of education. “After being ill, I could not engage in physical labour. Relatives came to help us harvest corn and grow wheat, thanks to the help of relatives, or the crops were wasted.” ( male, 92 years old, widowed, moderate disability, middle school education)“Our neighbor has a van, and he drives us to the hospital for several times, which is so convenient.” ( female, 77 years old, widowed, mild disability, elementary education)“The first thing I thought of when I was sick was going to our village clinic, just take some medicine, I will be fine. Sometimes I can’t go, but I call the village doctor, he rides a bike with a medical bag and comes to my house.” ( female, 76 years old, widowed, mild disability, middle school education)“Whenever I have trouble at home I call the village cadres. Everyone is very good to me, take care of me as family members.” ( male, 61years old, married, mild disability, high school education)
## Complex social response
Complex social response is manifested as social interaction with peers, communication with relatives and friends, and social interaction with strange sick friends. “I always feel uncomfortable communicating with my peers since I was sick, I wouldn’t like to talk with them, sometimes I envy their physical condition.” ( male, 78years old, married, mild disability, elementary education)“There are more contacts with relatives after illness, because I often need their care. At the beginning, everyone was very positive, they knew I was sick and often came to see me, asking me if I had enough money, but I was too embarrassed to bother them, thinking that everyone was too busy to visit me every day.” ( female, 77years old, widowed, mild disability, elementary education)
## Theme 5: Factors influencing for adaptation
This theme includes three sub-themes.
## Personal factors
Personal factors include disability duration, psychological resilience, self-efficacy and family income.
In the interview, 5 out of 15 disabled elderly said that the longer their illness, the more able to adapt to the life after illness. “When I have just checked out this disease, I am very irritable, often angry, and become very irritable, and I cannot control myself, but now I have accepted the fact and am very peaceful to the disease.” ( male, 65years old, married, moderate disability, elementary education) Psychological resilience is an individual’s psychological capacity to cope with difficulties or stress. Rural elderly people with disabilities who adapt to different experiences present different psychological resilience. “I adapt to this disease so well due to my personality. I used to be the village head, handled a lot of tricky things, experienced a variety of problems in the village, developed the strong psychological endurance.” ( male, 72years old, married, moderate disability, middle school education) We found that 6 out of 15 older adults showed positive self-efficacy after disability, whom were willing to actively share their own experience and actively participate in social interaction. “I was a soldier, then a worker, and now I’m a civilian, still a disabled old man, but it doesn’t matter, my friends and I talk a lot about what happened in the past, about society now, about what I’ve learned about my life, and about this illness.” ( male, 61years old, married, mild disability, high school education) We found that rural disabled elderly with high family income showed good adaptability. “My husband was a primary school teacher when she was younger and now receives a monthly pension. Sometimes the children give some change and we receive a pension every month. I bought my electric wheelchair with the money we both saved.” ( female,70years old, widowed, mild disability, elementary education)
## Caregiver factors
The caregiver factor is mainly whether the caregiver has received training. The trained caregivers make the disabled elderly show good adaptability. “My daughter-in-law has ever trained in the town especially for caring me, and she have been taking care of me since I am sick, feeding, scrubbing, taking medicine, also teaching me some rehabilitation exercises. With her careful care, I recovered quickly.” ( male, 65years old, married, moderate disability, elementary education)
## The policy factors
The policy factors include social security policy and medical insurance policy. The Chinese government attaches great importance to social security for the vulnerable elderly, which facilitates the adaptation of the disabled elderly in rural areas. “The national policy is very good, especially for us elderly people, we get a pension when we are over 60 years old, those in our village who are over 60 years old receive more than 100 yuan per month and those over 70 years old receive 200 yaun per month, this policy is very good and protects the livelihood of our elderly people.” ( female, 75years old, married, mild disability, non-formal education)“I was afraid of getting sick and spending money so I didn’t want to go to the hospital, but we have the New Agricultural Cooperative Medical Insurance and we can get $80\%$ reimbursement for hospitalisation at the county hospital, which saves us a lot of money and makes me feel a lot more relieved.” ( male, 75years old, widowed, severe disability, elementary education)
## Discussion
This study explored the adaptation experience of rural disabled elderly. The disabled elderly in rural areas showed different experiences in different aspects of adaptation.
## Different adaptation strategies
In the physiological functioning adaptation experience, our study mainly found that rural disabled elderly took various measures to cope with physiological changes, including monitoring of physiological indicators, active compliance behaviours, active rehabilitation exercises and lifestyle adjustments, which is consistent with the findings of a similar study by Bockwoldt. Bockwoldt et al. [ 27] described adaptation strategies for elderly with diabetes in terms of physiological functioning by incorporating diabetes-related activities into existing lifestyles, such as monitoring blood glucose, observing symptoms, and taking and/or adjusting oral or injectable medications; studies by Li H [28] also found that rural elderly with disability experience self-lifestyle adjustments, positive health beliefs, and increased self-management of health conditions. This suggests that rural elderly people with disabilities are actively adapting themselves to cope with physical changes, in line with the concept of "active ageing" advocated by the Chinese government [29]. From this survey, we learned that the elderly in rural areas have purchased blood pressure monitors to measure their own blood pressure, reflecting the strong health awareness of the elderly in rural areas, but this is inseparable from the country’s strategic plan for economic development, achieving comprehensive poverty eradication and improving the living standards of the grassroots [30]. Our secondary finding is some people said they are unable to cope with physiological changes, whom were mostly women or elderly people with low levels of education, suggesting that gender and education are relevant to the physical adaptability of the disabled elderly. The main findings and secondary findings are complementary to each other, which together constitute the diversity of physiological function adaptation and the different outcomes of adaptive behavior and ineffective adaptation, which is consistent with the two adaptation results in Roy ‘s adaptation model. Caregivers should pay more attention to these people with ineffective adaptation, analyze the reasons for their inability to cope, and help them create adaptive environments and opportunities.
In self-concept adaptation, our interviews mainly found that in the face of psycho-emotional changes, rural disabled elderly coped through self-adjustment, distraction and seeking help from others. This is consistent with the findings of He et al. [ 31], which showed positive self-efficacy and stress coping abilities of some of the disabled elderly. Our secondary finding was that other disabled older people were unable to demonstrate good self-adaptation, to mitigate the psycho-emotional changes associated with their disability, and to get along well with their surroundings. The majority of this group are widowed or severely disabled, showing that marital status and degree of disability are related to the self-concept of adaptation of the disabled elderly. Some studies have pointed out that when negative emotions such as guilt and low self-esteem, depression and depression are not resolved, as well as morbid shame such as remorse and guilt, fear and despair cannot be relieved, and stress is not managed properly, rural elderly people may engage in suicidal behaviour [32], which may be related to their lack of mental health knowledge and less attention to this issue of their own mental health. It is suggested that the government should pay attention to the mental health problems of rural disabled elderly people, especially for the elderly who are widowed or severely disabled, strengthen the training of medical personnel in rural health centres and village clinics on mental health knowledge, as well as increase the popularisation of mental health knowledge among rural residents, set up a psychological counselling room in each village, etc., to pay attention to the mental health status of each disabled person attending the clinic and increase their adaptability to mental health.
In the role-functional adaptation experience, our study mainly found that rural disabled elderly showed positive self-care roles and active adjustment of themselves to disability in terms of health management, which is consistent with Bo [33].This may be due to the improvement of living standards so they can pay more attention to their own health than before, also reflects the desire for health of the disabled elderly in rural areas. In addition, our study also found that regarding family roles and social roles, the disabled elderly showed negative avoidance, which is consistent with the results of similar studies [34], which may because the limitation of resources in rural areas and the age of the older adults make them at the edge of society and are unwilling to participate in social things. The education level of this segment of the population is usually low, limiting their ability to learn new things. Guan’s research showed that with the progress of disease, patients will be aware of their responsibilities and obligations to their families so as to actively face life and self-growth, which is related to their self-efficacy [35].Patients will actively self-efficacy can due their responsibilities and obligations to the family, suggesting that caregivers should pay attention to the psychological characteristics of the disabled elderly, mobilize their enthusiasm for life, encourage them to realize their self-efficacy and actively participate in family affairs and social activities.
In the interdependent adaptation experience, our study mainly found that most rural disabled elderly received care and financial support from family, relatives and friends, and society, which is consistent with the study of Huo’s [36]. Social support plays an important role in the health promotion of patients, which can reduce the impact of various negative stresses, alleviate physical and psychological barriers, and enhance patients’ social adaptation [37, 38]. We also found that there are different situations when the rural older adults dealing with the relationship with different people such as friends, family members, peers and village doctors, and their social adaptation is terrible. Some studies showed the same view [39],some disabled elderly reduce social activities or even social willingness due to decreased activity ability or psychological reasons. This may also be caused by the physiological and psychological changes caused by disability. The disabled elderly are unable to participate in social activities due to malnutrition, inability to sit long, need to be accompanied outside, visual and hearing loss, urinary incontinence and other reasons, resulting in a decline in enthusiasm for participation, thereby reducing the demand for social participation [40].Leisure activities were important for the promotion of physical and mental health [41, 42],it is necessary to emphasize the importance of socialization with family and friends and leisure activities in assisting the older adults in rural areas with disability issues. Leisure activities for older adults rural people with mobility difficulties and living in areas with limited infrastructure development could be directed towards mental activities [43]. In addition, studies have shown that peer support can improve the self-management level of the older adults [44]. Similar cultural backgrounds and rehabilitation experiences enable rural disabled elderly neighbours to understand each other’s inner feelings and real needs [45], which is the principle of the ‘peer effect’ [46, 47]. This suggests that patients from neighbouring villages with the same illness can share their experiences, do rehabilitation exercises together, monitor each other, and improve the health management of rural disabled older people, as well as help them to establish new social circles, participate in new social activities, regain social confidence and improve their social adaptability.
The factors that effect these adaptation strategies.
People are in a state of constant response to environmental changes, and the result of interaction and integration between people and the environment is adaptation. The adaptation process of the disabled elderly is affected by some environmental factors such as personal factors, caregiver factors or policy factors. When analyzing the factors that affect the adaptation strategies of disabled elderly, we mainly found that disability duration, family income, psychological resilience and self-efficacy were the personal factors affecting the adaptation experience of rural disabled elderly.
The interviews revealed that some of the disabled older people had become more comfortable with their disabilities as they grew older, possibly because they had been exposed to more knowledge about illness and coping skills as they grew older, and were more comfortable with their surroundings. The elderly with higher family incomes were more able to afford health products, especially assistive living devices such as electric wheelchairs, which can help them get along better with their surroundings. Psychological resilience refers to an individual’s ability to adapt positively and cope effectively in the face of adverse circumstances such as stress, pressure and adversity, and is conducive to an individual’s health and positive adaptation [48]. Disabled elderly people with different psychological resilience in this study showed different adaptive abilities. Song [49] showed that disabled elderly people have difficulty coping effectively with various stresses and pressures due to their high level of physical illness and low ability to perform daily activities, as evidenced by a lack of confidence and patience, poor mental capacity and increased negative emotions and maladjustment. This shows that psychological resilience is crucial to the level of adaptation of elderly people with disabilities, suggesting that attention should be paid to the psychological resilience of elderly people with disabilities in rural areas and timely psychological care guidance should be provided to caregivers. Self-efficacy refers to the intensity or degree of individual confidence in their ability to complete tasks and achieve goals [50, 51], which has a positive predictive effect on interpersonal adaptability, role adaptability and life self-care adaptability [52]. It is suggested that caregivers are prompted to help disabled older people to identify their self-worth, assess their abilities, give full play to their self-efficacy and facilitate their role adaptation, based on their own occupations and personality traits when they were younger.
In addition, our study also found that gender, education level, disability level and marital status can also affect the adaptation. In physiological function adaptation, we found that male disabled elderly were more willing to take active rehabilitation exercise and were willing to adjust their lifestyles to cope with the physiological changes caused by disability, while women are more likely to cope with failure. Studies have shown that older rural women are less physically fit, at higher risk of disability and relatively less resilient to disability due to a variety of physical or psychological factors [53]. In our study, the mildly disabled elderly showed positive adaptive responses in four adaptation methods, such as more positive compliance behavior, distraction, positive self-care behavior, and active pursuit of social support, which may be due to the fact that the self-care ability of the mildly disabled elderly is not completely lost and there are significant differences in physiological function, vitality and emotional function among the older adults with different levels of disability. As the degree of disability increases, the physiological adaptations of older people decline [54]. In addition, high levels of education had a positive impact on the adjustment of disabled older people in our study. Some older people with moderate disabilities who had a high level of education performed better in monitoring physical indicators, gratitude adjustment and positive self-care. It may be because the older adults with high educational level are good at self-conform, have correct disease cognition, and show high self-efficacy, so they can better adapt to the changes caused by disability. In our study, the elderly with spouses performed better in positive compliance behavior and self-consolation, while the widowed elderly performed worse in self-concept adaptation.
Caregiver factors were the knowledge and skills of the caregiver and whether or not they have received training. This interview revealed that the influence of different caregivers on the disabled elderly is completely different. The trained caregivers of the disabled elderly pay close attention to their health and mental conditions, and make reasonable arrangements for their outings, nutritional intake, adequate companionship and psychological support, therefore, these disabled elderly often feel happy and moved. This suggested that the caregivers play an important role in the psychological adaptation process of the disabled elderly, which is consistent with Liu’s [55]. Rural disabled elderly depend on caregivers for long-term care because of their poor self-care ability; meantime, they spend most of time at home due to limited mobility, and loneliness caused by long-term lack of peer communication has resulted in strong psychological dependence on caregivers [36]. Therefore, caregivers are crucial for the physical and mental health development of rural disabled elderly. The study pointed out that the family caregivers of the disabled elderly have become the mainstay of the long-term care system [55], and played the roles of daily caregivers, collaborative treatment, safety defenders and psychological comforters of the disabled elderly. However, the care needs of rural elderly with disabilities are special, long-term and complex, while caregivers are mostly non-professionals such as spouses and children, and are old and of low professional level. Therefore, their ability to receive new knowledge is limited, and their caregiving ability cannot meet the care needs. Mao pointed out that comfortable nursing training for caregivers of disabled elderly can improve their care skills and self-confidence, and also improve the quality of life of disabled elderly [48]. It is suggested that the caregivers should be trained in comprehensive care knowledge and skills, such as blood pressure monitoring, feeding skills, medication precautions, rehabilitation training skills, skin care, oral care, and psychological care. It can also empower the caregivers to have more knowledge and ability to be competent for the role of caregivers, so as to improve the adaptability of the disabled elderly.
Policy support was an important factor for the adaptation of rural disabled elderly, and other scholars also proves this view. Szanton noted that the elderly who receive social security and services have a higher level of disability adaptation, which is reflected in both psychological adaptation and disease adaptation [56]. Rural elderly with disability have a certain degree of financial burden due to the long duration of chronic diseases. Government-provided health insurance and Medicaid policies will reduce their disease burden and reduce their negative emotions such as guilt and self-blame due to the financial burden [57]. Choi suggested disability welfare programs should be provided to older adults who present with frailty [58]. It can be seen that the social security and services provided by the government play a positive role in supporting the improvement of the disability adaptability of the rural older adults.
## Suggestions on improving the adaptability of disabled elderly
Here are some suggestions to improve the adaptability of disabled elderly. At the national level, a multi-faceted social support system for the disabled elderly should be established on the existing basis. Firstly, laws and regulations relating to the disabled elderly should be improved, including regulations on long-term care insurance and discrimination against the elderly with disabilities; secondly, a mechanism should be established to promote rehabilitation activities for the elderly with multiple support from families, doctors and healthy peers, and an artificial intelligence-based “integrated network information security platform” should be improved for both urban and rural areas. Thirdly, the training of carers in all aspects of their skills should be strengthened. At the family and community levels, caregivers of the disabled elderly and rural grassroots health workers should pay more attention to women, the elderly with low education levels and the elderly with severe disabilities, using easy-to-understand language to popularise their health knowledge, help them build up a correct understanding of the disease, and enhance their confidence and courage to adapt. At the same time, we should pay attention to their psychological and emotional changes, help them better regulate their negative emotions in the process of adaptation and strengthen their courage to cope with difficulties, so that they can adapt better.
## Limitations
Firstly, the sample source has regional limitations. This study only interviewed rural disabled elderly in one city in Henan Province, and more research is needed on rural disabled elderly in other cities in Henan Province; secondly, qualitative research cannot analyse all the influencing factors. The analysis of factors influencing the adaptation of rural elderly with disabilities should be combined with quantitative surveys. Currently, there is no quantitative research on the adaptation of rural older people with disabilities and there is a lack of relevant assessment tools. As a next step, we plan to develop an adaptive capacity questionnaire for rural disabled older people in conjunction with Roy’s adaptive model as an assessment tool for quantitative research. Future research could incorporate progressive quantitative research to collect data and analyse influencing factors. Thirdly, this interview only interviewed older people with disabilities who have some verbal skills. Those with limited expressive language skills may have different experiences. The next step could be to explore the adaptation experiences and influencing factors of these individuals by interviewing their carers.
## Conclusion
The results of the interviews revealed that rural disabled elderly people present a multifaceted adaptation experience. In terms of physiological functional adaptation, medication and non-pharmacological treatments were chosen to adapt to physiological changes. About self-concept adaptation, negative emotions were alleviated through self-persuasion, distraction and help-seeking, but inability to cope with physical and psychological changes also occurred. For role function adaptation, positive self-care roles, negative family roles and avoidance of social roles are present. For interdependency adaptation, rural disabled elderly show active seeking of social support and complex interpersonal changes. Personal self-efficacy and psychological resilience, as well as the caregiving skills of caregivers, are important factors influencing the adjustment of older people with disabilities. It is recommended that a multi-dimensional social support system be established for the rural elderly with disabilities, and that training of carers in relevant caregiving knowledge and skills, with particular attention to the development of psychological caregiving skills, be strengthened, while promoting the reasonable development of the elderly with disabilities’ own self-efficacy.
## References
1. Li J, Han X, Zhang X, Wang S. **Spatiotemporal evolution of global population ageing from 1960 to 2017**. *BMC Public Health* (2019.0) **19** 127. DOI: 10.1186/s12889-019-6465-2
2. 2.WHO. 10 facts on ageing and health. Retrieved from https://www.who.int/news-room/fact-sheets/detail/10-facts-on-ageing-and-health. Accessed 15 Oct 2022.
3. Wang Y, Qi C. **Multi-dimensional accessibility barriers in care services for the rural elderly with disabilities: a qualitative study in China**. *Int J Environ Res Public Health* (2021.0) **18** 6373. DOI: 10.3390/ijerph18126373
4. Harsha N, Ziq L, Giacaman R. **Disability among Palestinian elderly in the occupied Palestinian territory (oPt): prevalence and associated factors**. *BMC Public Health* (2019.0) **19** 432. DOI: 10.1186/s12889-019-6758-5
5. Prina AM, Wu YT, Kralj C, Acosta D, Acosta I, Guerra M. **Dependence- and disability-free life expectancy across eight low- and middle-income countries: a 10/66 study**. *J Aging Health* (2020.0) **32** 401-409. DOI: 10.1177/0898264319825767
6. Chen S, Zheng J, Chen C, Xing Y, Cui Y, Ding Y. **Unmet needs of activities of daily living among a community-based sample of disabled elderly people in Eastern China: a cross-sectional study**. *BMC Geriatr* (2018.0) **18** 160. DOI: 10.1186/s12877-018-0856-6
7. 7.National Bureau of Statistics. Main data of the seventh National Census. National Bureau of Statistics. Retrieved from http://www.stats.gov.cn/tjsj/zxfb/202105/t20210510_1817176.htm. Accessed 15 Oct 2022.
8. Luo Y, Su B, Zheng X. **Trends and challenges for population and health during population aging - China, 2015–2050**. *China CDC Wkly* (2021.0) **3** 593-598. DOI: 10.46234/ccdcw2021.158
9. Chen N, Deng M, Xu H. **Intensity of informal care demand for disabled older adults and its influencing factors**. *Health Manag China* (2021.0) **38** 386-389
10. 10.Ministry of Agriculture and Rural Affairs, PRC. The CPC Central Committee and The State Council issued the Strategic Plan for Rural revitalization, 2018–2020. China Government. Revised from http://www.moa.gov.cn/nybgb/2018/201810/201812/t20181218_6165130.htm. Accessed 15 Oct 2022.
11. 11.Yu LL. Research on learning adaptability of undergraduates in ordinary universities. Master's thesis, Nanchang University, Education Economics and Management Department, 2020.
12. Li Z, Tang Z, Wang R. **A survey on the current status of disability among the elderly in seven cities in China**. *Chin J Epidemiol* (2016.0) **37** 1561-1564. DOI: 10.3760/cma.j.issn.0254-6450.2016.12.001
13. Zhao Y, Yang R, Hu D, Yan J, Han F, Li J. **Disability status of the older adults-disability assessment based on long-term care insurance**. *Chin J Gerontol* (2020.0) **40** 416-419. DOI: 10.3969/j.issn.1005-9202.2020.02.057
14. Tareque MI, Tiedt AD, Islam TM, Begum S, Saito Y. **Gender differences in functional disability and self-care among seniors in Bangladesh**. *BMC Geriatr* (2017.0) **17** 177. DOI: 10.1186/s12877-017-0577-2
15. 15.Zhao X. Research on the protection of the rights and interests of rural disabled older adults. Master's thesis, Northeast University of Finance and Economics, Political Science Theory Department, 2016.
16. 16.Wang P. Study on psychosocial adaptation and intervention model of young and middle-aged patients with acute myocardial infarction with type D personality. PhD thesis, Zhengzhou University, Nursing Department, 2019.
17. Wang Q, Zhang C, Zhang Z, Chen R, Liu L, Zhang X. **Effect of positive psychological intervention on stigma and adaptation level of gynecological cancer patients**. *J Nurs* (2020.0) **35** 71-73. DOI: 10.3870/j.issn.1001-4152.2020.13.071
18. Wang Q, Xiao H, Chen M, Xiao Y, Zhang Y. **Current status of research on the burden and influencing factors of caregivers of the disabled elderly**. *Occupation Health* (2021.0) **37** 2289-2292. DOI: 10.13329/j.cnki.zyyjk.2021.0560
19. Jimenez ME, Hudson SV, Lima D, Crabtree BF. **Engaging a community leader to enhance preparation for in-depth interviews with community members**. *Qual Health Res* (2019.0) **29** 270-278. DOI: 10.1177/1049732318792848
20. O’Brien BC, Harris IB, Beckman TJ, Reed DA, Cook DA. **Standards for reporting qualitative research: a synthesis of recommendations**. *Acad Med* (2014.0) **89** 1245-1251. DOI: 10.1097/ACM.0000000000000388
21. Patton MQ. *Research and Evaluation Methods* (2002.0)
22. Gao YB. *The trend characteristics of population aging in Henan Province and its countermeasures* (2021.0)
23. Wang P, Tong C, He J, Li Y. **Analysis of influencing factors of family structure change of rural elderly**. *Nurs Res* (2022.0) **36** 8-15. DOI: 10.12102/j.issn.1009-6493.2022.01.002
24. Sandelowski M. **Sample size in qualitative research**. *Res Nurs Health* (1995.0) **18** 179-183. DOI: 10.1002/nur.4770180211
25. Roy C. **Research based on the Roy adaptation model: last 25 years**. *Nurs Sci Q* (2011.0). DOI: 10.1177/0894318411419218
26. Coliazzi P, Valle RS, King M. **Psychological research as the phenomenologist views it**. *Existential phenomenological alternatives for psychology* (1978.0) 48-71
27. Bockwoldt D, Staffileno BA, Coke L, Quinn L. **Perceptions of insulin treatment among african americans with uncontrolled type 2 diabetes**. *J Transcult Nurs* (2016.0) **27** 172-180. DOI: 10.1177/1043659614543477
28. Li H, Zhang Y, Wang R, Yu J, Du C, Zhao J. **Study on the current situation of active aging of rural disabled older adults in Zhengzhou**. *Medical and social* (2019.0) **32** 56-58. DOI: 10.13723/j.yxysh.2019.10.014
29. Li LY, Li L. **Building a positive aging policy system to release the dividend of China’s elderly population**. *China Adm Manag.* (2020.0) **8** 21-25. DOI: 10.19735/j.issn.1006-0863.2020.08.03
30. Li Y, Liu H. **Financial support for poverty eradication has achieved remarkable results**. *China Finance* (2020.0) **24** 11-13. DOI: 10.14115/j.cnki.zgcz.2020.24.005
31. He S, Zhang Y, Zhang H, Duan Y, Li H. **Study on happiness status of 350 rural disabled older adults in Henan province**. *Gen Nur* (2019.0) **17** 1520-1523. DOI: 10.12104/j.issn.1674-4748.2019.12.045
32. Wu Y, Zhang T, Hao W, Su L, Sun W. **Analysis of the status quo and influencing factors of depressive symptoms among rural older adults in Shenyang in 2013**. *Pract Prev Med* (2018.0) **25** 773-776. DOI: 10.3969/j.issn.1006-3110.2018.07.002
33. Bo Z, Huang A, Li Q, Xue M, Du S, Meng T. **Meta-analysis of long-term care needs of disabled older adults in China**. *Chin J Gerontol* (2020.0) **40** 1013-1017
34. Li L, Sun J, Wu H. **Study on influencing factors of stigma among disabled older adults with chronic diseases in pension institutions**. *Prev Med* (2020.0) **32** 1030-1033. DOI: 10.19485/j.cnki.issn2096-5087.2020.10.014
35. Guan Z, Chen D, Yang Y, Song W, Wang Y, Dou L. **Qualitative research on the causes of self-growth behavior in maintenance hemodialysis patients**. *Nurs J* (2020.0) **35** 78-80. DOI: 10.3870/j.issn.1001-4152.2020.12.078
36. Huo A, Liu Y, Li H. **Effect of social support and empowerment of primary caregivers on care dependence and moderating effect of frailty in community disabled older adults**. *Nurs Res* (2021.0) **35** 2445-2451. DOI: 10.12102/j.issn.1009-6493.2021.14.001
37. Kiajamali M, Hosseini M, Estebsari F, Nasiri M, Ashktorab T, Abdi A. **Correlation between social support, self-efficacy and health-promoting behavior in hemodialysis patients hospitalized in Karaj in 2015**. *Electron Physician* (2017.0) **9** 4820-4827. DOI: 10.19082/4820
38. Sajadi SA, Ebadi A, Moradian ST. **Quality of life among family caregivers of patients on hemodialysis and its relevant factors: a systematic review**. *Int J Community Based Nurs Midwifery* (2017.0) **5** 206-218. PMID: 28670583
39. Huang WJ. **Research on the medical and care needs of the older adults from a sociological perspective**. *Gen Pract China* (2017.0) **20** 842-851. DOI: 10.3969/j.issn.1007-9572.2017.07.017
40. Lane NE, Wodchis WP, Boyd CM, Stukel TA. **Disability in long-term care residents explained by prevalent geriatric syndromes, not long-term care home characteristics: a cross-sectional study**. *BMC Geriatr* (2017.0) **17** 49. DOI: 10.1186/s12877-017-0444-1
41. Fimland MS, Vie G, Holtermann A, Krokstad S, Nilsen TIL. **Occupational and leisure-time physical activity and risk of disability pension: prospective data from the HUNT Study, Norway**. *Occup Environ Med* (2018.0) **75** 23-28. DOI: 10.1136/oemed-2017-104320
42. Lamb SE, Sheehan B, Atherton N, Nichols V, Collins H, Mistry D. **Dementia And Physical Activity (DAPA) trial of moderate to high intensity exercise training for people with dementia: randomised controlled trial**. *BMJ* (2018.0) **361** k1675. DOI: 10.1136/bmj.k1675
43. Puspitasari MD, Rahardja MB, Gayatri M, Kurniawan A. **The vulnerability of rural older adults Indonesian people to disability: an analysis of the national socioeconomic survey**. *Rural Remote Health* (2021.0) **21** 6695. DOI: 10.22605/RRH6695
44. Ji H, Song J, Peng T. **Implementation of peer education for older adults diabetics in pension institutions**. *J Nurs* (2017.0) **32** 11-13. DOI: 10.3870/j.issn.1001-4152.2017.21.011
45. 45.Zhao J, Zhang Y, Yu Z, Wang R, Li H, Du C, et al. Study on home rehabilitation experience of rural disabled older adults in a county of Henan province. Nurs Res. 2019;33(17):3063–6. 10.12102/j.issn.1009-6493.2019.17.036.
46. 46.Toija AS, Kettunen TH, Leidenius MHK, Vainiola THK,Roine RPA. Effectiveness of peer support on health-related quality of life in recently diagnosed breast cancer patients: a randomized controlled trial. Supportive care in cancer: official journal of the Multinational Association of Supportive Care in Cancer. 2019;27(1):123–30. 10.1007/s00520-018-4499-0.
47. Cherrington AL, Khodneva Y, Richman JS, Andreae SJ, Gamboa C, Safford MM. **Impact of peer support on acute care visits and hospitalizations for individuals with diabetes and depressive symptoms: a cluster-randomized controlled trial**. *Diabetes Care* (2018.0) **41** 2463-2470. DOI: 10.2337/dc18-0550
48. Song S, Guo J. **Effect of resilience and self-efficacy on self-perceived burden in older adults patients with acute myocardial infarction**. *Chin J Gerontol* (2021.0) **41** 646-649. DOI: 10.3969/j.issn.1005-9202.2021.03.056
49. Szanton SL, Alfonso YN, Leff B, Guralnik J, Wolff JL, Stockwell I. **Medicaid cost savings of a preventive home visit program for disabled older adults**. *J Am Geriatr Soc* (2018.0) **66** 614-620. DOI: 10.1111/jgs.15143
50. Thom B, Benedict C. **The impact of financial toxicity on psychological well-being, coping self-efficacy, and cost-coping behaviors in young adults with cancer**. *J Adolesc Young Adult Oncol* (2018.0) **8** 236-242. DOI: 10.1089/jayao.2018.0143
51. Santos PR, Lima Neto JA, Carneiro RAA, Soares AITD, Oliveira WR, Figueiredo JO. **Variables associated with lung congestion as assessed by chest ultrasound in diabetics undergoing hemodialysis**. *J Bras Nefrol* (2017.0) **39** 406-412. DOI: 10.5935/0101-2800.20170073
52. Liao J. **The impact of positive psychological capital of freshmen on school adaptation**. *Career Health* (2019.0) **35** 666-670, 675. DOI: 10.13329/j.cnki.zyyjk.2019.0181
53. Zhang J, Lou D, Ma H, Yu C, Chen L, Li Y. **Development of a validated Chinese version of the inflammatory bowel disease disability index**. *J Dig Dis* (2020.0) **21** 52-58. DOI: 10.1111/1751-2980.12836
54. Ramadass S, Rai SK, Gupta SK, Kant S, Wadhwa S, Sood M. **Prevalence of disability and its association with sociodemographic factors and quality of life in a rural adult population of northern India**. *Natl Med J India* (2018.0) **31** 268-273. DOI: 10.4103/0970-258X.261179
55. Liu Y, Guo H, Liu J, Gong S, Yi X. **Research progress on social ecosystem of disabled older adults family caregivers in China**. *Nurs Res* (2020.0) **34** 1764-1767
56. Mao Z, Zhang H, Sun X, Liu X. **Research progress on comfortable nursing and its influencing factors**. *Nurs Res* (2017.0) **31** 513-517. DOI: 10.12102/j.issn.1009-6493.2020.10.016
57. Wang H, Ning M. **The impact of urban-rural medical insurance integration policy on the medical burden of rural middle-aged and elderly people**. *China Health Policy Research* (2022.0) **15** 9-16. DOI: 10.3969/j.issn.1674-2982.2022.06.002
58. Choi YS, Kim MJ, Lee GY, Seo YM, Seo AR, Kim B. **The Association between Frailty and Disability among the older adults in Rural Areas of Korea**. *Int J Environ Res Public Health* (2019.0) **16** 2481. DOI: 10.3390/ijerph16142481
|
---
title: 'A spatial obesity risk score for describing the obesogenic environment using
kernel density estimation: development and parameter variation'
authors:
- Maximilian Präger
- Christoph Kurz
- Rolf Holle
- Werner Maier
- Michael Laxy
journal: BMC Medical Research Methodology
year: 2023
pmcid: PMC10021981
doi: 10.1186/s12874-023-01883-y
license: CC BY 4.0
---
# A spatial obesity risk score for describing the obesogenic environment using kernel density estimation: development and parameter variation
## Abstract
### Background
Overweight and obesity are severe public health problems worldwide. Obesity can lead to chronic diseases such as type 2 diabetes mellitus. Environmental factors may affect lifestyle aspects and are therefore expected to influence people’s weight status. To assess environmental risks, several methods have been tested using geographic information systems. Freely available data from online geocoding services such as OpenStreetMap (OSM) can be used to determine the spatial distribution of these obesogenic factors. The aim of our study was to develop and test a spatial obesity risk score (SORS) based on data from OSM and using kernel density estimation (KDE).
### Methods
Obesity-related factors were downloaded from OSM for two municipalities in Bavaria, Germany. We visualized obesogenic and protective risk factors on maps and tested the spatial heterogeneity via Ripley’s K function. Subsequently, we developed the SORS based on positive and negative KDE surfaces. Risk score values were estimated at 50 random spatial data points. We examined the bandwidth, edge correction, weighting, interpolation method, and numbers of grid points. To account for uncertainty, a spatial bootstrap (1000 samples) was integrated, which was used to evaluate the parameter selection via the ANOVA F statistic.
### Results
We found significantly clustered patterns of the obesogenic and protective environmental factors according to Ripley’s K function. Separate density maps enabled ex ante visualization of the positive and negative density layers. Furthermore, visual inspection of the final risk score values made it possible to identify overall high- and low-risk areas within our two study areas. Parameter choice for the bandwidth and the edge correction had the highest impact on the SORS results.
### Discussion
The SORS made it possible to visualize risk patterns across our study areas. Our score and parameter testing approach has been proven to be geographically scalable and can be applied to other geographic areas and in other contexts. Parameter choice played a major role in the score results and therefore needs careful consideration in future applications.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s12874-023-01883-y.
## Background
Overweight and obesity are severe problems worldwide, causing a number of diseases such as type 2 diabetes, and thus reducing expected life years and quality of life [1, 2]. In Germany, for example, the prevalence of being overweight or obese among adults was $54.0\%$ according to the GEDA (GEDA, German Health Update) study from the Robert Koch Institute (a national public health institute in Germany) in $\frac{2014}{2015}$, with men being affected more often than women [3]. Other personal aspects affecting obesity besides gender were low education and higher age according to Mader et al. [ 4]. In addition, several German cohort studies have shown that the average weight in middle-aged populations increased slightly during recent years [5].
Obesity has become a major public health concern, and recent studies describe regional heterogeneity [6, 7]. In obesity-related research, the term “obesogenic environment” describes environmental influences such as green space or fast food restaurants on the development of obesity [8, 9], which has been investigated intensively in the past [10]. Several approaches have been developed in order to analyze the effect of the environment on the risk of obesity. Examples include obesogeneity assessment via questionnaires [11] and via data visualization tools for obesity policy [12].
Some geographic modeling approaches were used to characterize the accumulation of environmental factors. Common techniques include kernel density estimation (KDE), a density method that allows for the estimation of a continuous risk surface [13, 14], as well as hot spot mapping [15] and further geographic information system (GIS) methods [16]. These methods can be used to develop obesity risk scores [17].
Online geocoding services offer low-cost geographic data for researchers that can be downloaded and used for spatial statistical analyses. Their validity has been investigated in the past with reasonable results regarding completeness of environmental factors and positional accuracy of their coordinates [18, 19]. Therefore, they offer a rich database on which geographic tools can be built. However, geocoding services such as Google Maps offer data only in a limited way. In contrast, geodata from OpenStreetMap (OSM) contain geographic information provided by volunteers and thus are less restricted [20]. In a recent study, we performed an extensive literature search to identify obesity-related environmental factors [18]. Furthermore, we operationalized and downloaded corresponding points of interest (POIs).
The aim of our study extends this approach by developing and testing the spatial obesity risk score (SORS) based on data from OSM. The SORS calculates the obesity risk for each geographic point in a given region based on the local density of positive and/or negative obesity-related environmental factors. In our study, we developed a methodological framework for risk score estimation using KDE and tested the influence of five KDE parameters on the SORS values: [1] bandwidth, [2] edge correction via the size of the download area, [3] number of grid points, [4] risk interpolation method, and [5] weighting scheme of the environmental factors.
## Overview of the study approach
We describe the general strategy of obesity risk score estimation, which consists of several steps, by applying it to two regions in Bavaria, Germany. First, we chose our study area and downloaded POIs related to obesogenic environmental factors (cf. [ 18]). Second, the data were processed to adjust for some imprecisions and minimize redundancy, e.g., we adapted the coordinate units to represent the same length. Third, we analyzed the spatial heterogeneity of the study area and inspected the study area visually to get first insights regarding the distribution of POIs. Fourth, we presented the basic risk score estimation approach, as well as a deterministic sensitivity analysis, in which a spatial POI resampling approach was integrated. This approach made it finally possible in a last step to evaluate the deterministic parameter selection via the ANOVA F statistic. An overview of the risk score estimation process is shown in Fig. 1. Further details regarding each step are given below. Fig. 1Modelling process for obesity risk score estimation. ANOVA = analysis of variances, KDE = kernel density estimation
## Area selection and data download
We based our analysis on the overall obesogenic and protective environmental factors identified in our previous study [18]. The list of chosen variables for the score was derived via an extensive literature review, i.e., previous work by Mackenbach et al. [ 10], Jia et al. [ 21], and enriched with our own further searches.
We chose two areas in the south of Bavaria, Germany, to illustrate our approach and develop the SORS. Our aim was to base the geographical analysis on different levels of urbanity. The first area was the city of Augsburg with about 300,000 inhabitants covering an area of 147 km2. A more rural region, the town of Meitingen with 11,000 inhabitants and a size of 30 km2 lying 20 km north of Augsburg, was chosen as the second area. Information on these regions was available from the German Federal Statistical Office [22]. We selected the region around the city of Augsburg, as it is well known within obesity- and diabetes-related research [23].
Spatial POIs related to the selected variables, such as fast food restaurants and parks (compare Additional file 1), were downloaded from OSM using the online data filtering tool Overpass turbo [24]. Data regarding the area borders of Augsburg and Meitingen were downloaded in shape file format from a geographic online portal provided by the Bavarian government [25]. We directly downloaded maps intended for graphical visualization of results from the OSM web page [26]. Additional file 2 contains further information regarding data download.
## Data processing
The downloaded GeoJSON files contained information on each POI regarding name, type of obesity-related environmental factor, spatial coordinates in latitude and longitude format, and other characteristics such as opening hours and street addresses. The relevant information, i.e., name, coordinates, and type of environmental factor, was extracted for each spatial POI and processed into lists and data frames. Table A1 of Additional file 1 gives an overview of the downloaded and processed variables from OSM.
As a further pre-processing step, we introduced a synthetic origin to the south-west of both study areas and adapted the length of a longitudinal coordinate unit to the length of a latitudinal coordinate unit. The schematic structure after the introduction of the origin and the coordinate adaptation is shown in Table 1. The single bus stops were reduced from POIs to centroids of dense regions of bus stops. Further details regarding these additional pre-processing steps can be found within Additional file 2.Table 1Schematic example of six processed POIsCategorizationCoordinatesa Type of environmental factorCategory of the POIb Subcategory of the POIc LongitudeLatitudeObesogenicUnhealthy_foodPastry0.25950.3189ObesogenicUnhealthy_foodPastry0.24880.3161ObesogenicUnhealthy_foodSweets0.23030.2657ProtectivePhysical_activityCanoe0.27890.2966ProtectivePhysical_activityClimbing0.24040.2867ProtectivePhysical_activityClimbing0.30090.2084 POI Point of interest aCoordinates after equidistant transformation and relative to the synthetic origin bCategories were derived from the literature [10, 18] cThe subcategories were derived from OpenStreetMap map features [27]
## Spatial inhomogeneity and ex ante KDE visualization
Ripley’s K function was calculated for the unweighted obesogenic and protective POIs separately to describe the spatial inhomogeneity of the two study areas. The K function is a second-order moment function that is based on the variance of the radial interpoint distance r around each POI [28]. It compares the cumulative number of actual POIs with the number of expected POIs under random distribution assumption [29]. This random comparison process is realized via the Poisson process, which has a K value of r2π [30]. K functions lying above the K values of a random process therefore represent clustered patterns, whereas smaller values indicate regular processes [31]. We examined the spatial inhomogeneity to investigate ex ante whether clusters are expected to occur in our subsequent POI analysis. The isotropic edge correction, which is a method based on weighting of the POIs according to the probability of their next neighbors being within the study area, was applied to the K function [32]. In order to test whether the K functions of the POIs were significantly different from the K function of a random point pattern, we estimated bootstrap confidence bands around the K functions of the POIs based on the method of Loh with 1000 simulations [33].
In order to visualize the spatial distribution of the POIs, we created KDEs separately for positive and negative spatial data points [34]. These density layers were superimposed and shown together on a single map. Within this process, we defined certain parameter choices as the base case, which were then changed as part of the sensitivity analysis in a later step.
## Risk score estimation
We estimated the SORS based on the integration of obesogenic and protective kernel densities into a combined score. The aim of KDE is to provide a smooth and continuous estimation of the accumulation of spatial data points based on a sliding window technique. The geographic plane is represented by two dimensions and the estimated densities account for a third dimension, which thus leads to a three-dimensional mountainous structure, a so-called “risk surface”. To visualize this structure on a map, the level of the density coordinate can be plotted by contour lines or via coloring [35]. An overview of this method is provided by Hastie et al., for example, as well as by King et al. [ 36, 37].
To estimate risk score values, several steps have been performed. First, the positive spatial data points were included into a single positive spatial data layer. Analogously, the negative spatial data points were integrated into a data layer. Second, an observation window together with a grid of suitable size was set up and laid on the respective study area. For each of the following calculations, the same grid was used. Third, KDEs were performed to generate a risk surface based only on positive environmental factors and a second risk surface based only on negative environmental factors. Following this process, a density value based on positive factors and a value based on negative factors were generated at each grid point. Fourth, these positive and negative estimates were set against each other by taking the difference, which results in the final score values at the grid points [34]. Finally, to determine the risk value at the exact desired spatial location, interpolation methods were applied.
## Deterministic analysis
The procedure described above implies several parameter choices within its different steps. We tested alternative values for five of these parameters: bandwidth, edge correction, number of grid points, interpolation method, and an alternative weighting scheme (see overview in Table 2). For each given parameter variation, the remaining KDE parameters were set to their base case values (see also Table 2). Further explanations regarding the parameters are provided below. Table 2Overview on the sensitivity parametersParameterBase caseDeterministic sensitivity analysis1) BandwidthMethod of Terrell [38]$\frac{1}{3}$, $\frac{2}{3}$, $\frac{4}{3}$, $\frac{5}{3}$ of the base case bandwidth2) Size of download area (edge correction)1.4 * side length of the exact boxx * side length of the exact box,with x ϵ {1.8, 2.2, 2.6, 3.0}3) Number of grid points35 × 35 grid points70 × 70 grid points, 105 × 105 grid points4) Interpolation methodAutomatic interpolation with the R function “interp.surface”Inverse distance weightingdensity of the nearest grid cellordinary Kriging5) Weighting scheme of the environmental factorsEqual weighting with unityDouble weighting of supermarkets and gyms
## Bandwidth selection
The bandwidth is an important parameter in KDE that determines the degree of smoothing. An increased bandwidth results in a higher smoothing level. For the base case, the oversmoothing bandwidth proposed by Terrell et al. [ 38] was chosen, which can be described as the maximum smoothing degree that can be suitably applied to a given data set. Within the deterministic scenarios, alternative higher and lower values in steps of $\frac{1}{3}$ of the base case bandwidth were tested. We used a pooled bandwidth as described by Davies and colleagues [39].
## Size of the download area (edge correction)
Restricting the observational window for KDE to the area boundaries would lead to an underestimation of the true densities at the borders. These effects are especially high if the observation window has a complex structure. To correct for edge effects, we defined a rectangular observation window around the area borders, which simultaneously served as the POI download area and as the KDE area. As a first step, a window was determined from the maximum latitude, maximum longitude, minimum latitude, and minimum longitude of the city boundaries. This minimum bounding rectangle around the study area will be called the “exact box”. For the base case, each side length of the rectangle was increased by $40\%$, and the resulting rectangle was held in a concentric orientation to the smaller rectangle from the step before. The whole observation window was used for the creation of the risk surface. However, risk score values were only evaluated at locations that lay within the boundaries of the respective study area. To determine the effects of the window size, we gradually increased the download area in steps of $40\%$ of the exact box.
## Number of grid points
A further central parameter of KDE is the number of grid points. These points were distributed equally on the estimation rectangle. Therefore, KDEs were generated for grid points lying both inside and outside the study area, and both types of grid points were used to estimate the risk score inside the borders of the study areas. A higher number of grid points means that the amount of interpolation is reduced, as more spatial points exist at which an exact estimation is known. To preserve the location of and distance between the inner grid points within the edge correction scenarios, we increased the number of grid points in $40\%$ steps according to the increase in the side length of the download area. Setting the number of grid points to 25 × 25 for the exact box, this led to a base case grid of 35 × 35. For each of the following $40\%$ steps of edge correction, again 10 additional grid points were added to each grid point dimension. Within the remaining grid point sensitivity analyses, increased numbers of 70 × 70 and 105 × 105 grid points were tested.
## Interpolation method
The automatic interpolation function “interp.surface” of the R package “fields” [40] was applied within the base case. As an alternative interpolation approach, we used inverse distance weighting using the four grid points surrounding a targeted evaluation point. Furthermore, extraction of the score value of the grid point with the minimum distance to the targeted evaluation point was implemented as the third interpolation method [41]. As a fourth scenario, we chose ordinary Kriging which has proven its reliability for interpolating surfaces [42].
## Weighting scheme of the environmental factors
Several approaches exist for the design of proper weighting schemes. One approach, for example, would be to weight factors according to the strength of evidence for a positive or negative correlation. Regarding our analysis, we chose an equal weighting scheme as a base case scenario. To test an alternative weighting scenario, we followed the approach of Jones-Smith et al. [ 34]. These authors gave a higher weighting to factors that generally reach more people because of their longer opening hours or size. For the deterministic sensitivity analysis, we therefore tested a double weighting of supermarkets and gyms.
## Resampling approach to account for uncertainty in the distribution of the POIs
Uncertainty concerning POIs was integrated into density score estimations via a spatial bootstrapping method. *We* generated 1000 bootstrap samples for the positive and 1000 bootstrap samples for the negative environmental POIs at random with replacement. For each of the samples, the parameter variation described above was executed. This made it possible to integrate probabilistic variation of POIs into SORS value estimation for each deterministic scenario. These uncertainty estimates were used for the calculation of the ANOVA F statistic, which we applied to compare the deterministic sensitivity scenarios for a given parameter. We describe further details regarding the sampling process in Additional file 2.
## Evaluation of the risk score estimates
A random sample ($$n = 50$$) of spatial data points (evaluation points, EPs) was drawn from each of the two areas for which we calculated and compared risk score results across the scenarios defined above. Our aim was to develop a robust and stable score that accounts for uncertainty and exhibits discriminatory power. To describe how much the choice of parameter values influences the discrimination of data points with low and high risk, we performed analysis of variance (ANOVA) between all 50 EPs based on their bootstrap replications. Thus, for each EP, 1000 bootstrap replicates of the SORS constituted an ANOVA group of estimates for a given parameter scenario. We calculated the F statistic in order to determine the degree of separation between the 50 groups of estimates with higher F values indicating higher degrees of separation. Therefore, the highest value of the F statistic within a given parameter variation analysis in this sense indicates the best result. The relative influence of the parameters on model results is estimated based on a normalization of the F statistic values. The algorithm used to implement deterministic and probabilistic analysis is shown in Fig. 2. Finally, we created heat maps based on the risk score estimates of the base case. For this purpose, the KDE values were again transferred back to the original map dimensions that were used in the pre-visualization step. To compare this base case risk map to an alternative visualization using a common geographical methodology, we estimated an inhomogeneous cluster point process with polynomial trend of degree two for positive and negative POIs that is designed to provide similar cluster structures compared to our KDE approach. Subsequently, we derived intensities for risk score plotting in a comparable way as it was done for the kernel density approach, i.e., via setting off the surfaces. We tested several point process types, such as “Thomas” and MatClust”, and chose the model with the best fit based on the Akaike Information Criterion [43]. The code for data download, data processing, and analysis of the scenarios defined above can be found in the supplementary information. Fig. 2Algorithm describing the combination of deterministic and probabilistic analysis, BS = bootstrap sample, EC = edge correction
## Software
The spatial POIs were processed and analyzed using R version 4.0.2 [44]. For graphical visualization, R packages “ggplot2” [45], “graphics” [44], and “fields” [40] were used. We applied the “spatstat” [46] package to estimate Ripley’s K function and the bootstrap confidence bands around it. For KDEs, the packages “MASS” [47] and “sparr” [39] were chosen. Spatial data objects were built and handled via the packages “rgdal” [48], “geojsonR” [49], “prob” [50], and “spatstat” [46]. For interpolation and for the generation of risk score maps, the packages “fields” [40] and “gstat” [51] were used. To plot spatial objects on maps, we used the package “png” [52]. Finally, the DBSCAN algorithm was applied using the package “dbscan” [53].
## Spatial inhomogeneity and visual inspection of the study area
The upper part of Fig. 3 shows estimates of Ripley’s K function for Augsburg, separately for the obesogenic and the protective POIs. The point pattern for Augsburg was significantly clustered, as the lower confidence bands of the K functions lay above the random Poisson processes for each interpoint distance r, which means that the actual number of POIs within a distance r was greater than the number of expected POIs under random distribution assumptions [31]. This underlined the importance of subsequent KDE analysis, as the spatial pattern was suitable for clustering tasks. Estimates of the obesogenic and the protective K functions for Meitingen also revealed significantly clustered patterns, as the lower confidence bands lay above the random pattern for all or at least for several interpoint distances r, which is shown in the bottom part of Fig. 3.Fig. 3Spatial inhomogeneity measured via Ripley’s K function for Augsburg (top) and Meitingen (bottom), r = interpoint distance, K(r) = Ripley’s K function, iso = isotropic edge correction, pois = Poisson point process The separate obesogenic and protective risk surfaces are shown in Fig. 4 for Augsburg and Meitingen. The obesogenic and protective kernel densities for Augsburg accumulated within a region lying to the northwest within the city boundaries, whereas the eastern and the southern areas showed no dense region. In contrast, kernel densities for Meitingen showed several dense regions outside the town borders. Fig. 4Contour lines of obesogenic and protective kernel densities in Augsburg (left) and Meitingen (right). Values of KDE are constant on a contour line with increasing values toward the respective KDE center
## Randomly drawn sample points
The final set of randomly chosen EPs for the evaluation of the risk score for Augsburg ($$n = 50$$) and Meitingen ($$n = 50$$) is presented in Fig. 5. As seen within the graphics, the sample points generated via the random drawing process inside the study areas widely covered the respective regions under consideration. Fig. 5Final set of randomly chosen spatial data points for Augsburg (left, $$n = 50$$) and Meitingen (right, $$n = 50$$)
## Effect of parameter variation within KDE estimation process
Table 3 summarizes the values of the ANOVA F statistic for each scenario of base case and deterministic sensitivity analysis. The higher the F statistic for a given parameter variation, the higher the degree of separation between the sample point groups, and thus higher values were more preferable. Values of the F statistic for Augsburg and Meitingen increased with the amount of bandwidth for the first three scenarios. This trend continued for Meitingen with decreasing slope, whereas it showed a rather inverted u-shaped functional behavior for Augsburg. Regarding edge correction for Augsburg and Meitingen, increasing the study area to some degree led to the highest F statistic, but this effect was not permanently observed with increasing amounts of edge correction. For Augsburg, the second grid point scenario was preferred according to the ANOVA value, in contrast to Meitingen, for which the base case was preferred. Furthermore, the inverse distance weighting had the highest F value in Meitingen and Augsburg,. Finally, the second weighting scenario had a higher F value than the equal weighting scenario in Augsburg, whereas the opposite was seen for Meitingen. Overall, the bandwidth and the edge correction had the highest influence on the values of the F statistic. Table 3ANOVA F statistics for Augsburg and MeitingenAugsburgMeitingenScenarioa F statisticb Scenarioa F statisticb BW1871BW11460BW21095BW23014BW3 (BC)1135BW3 (BC)4531BW4848BW44779BW5657BW54079EC1 (BC)1135EC1 (BC)4531EC2956EC28372EC3724EC39158EC4854EC47047EC51156EC57912GP1 (BC)1135GP1 (BC)4531GP21495GP24293GP31398GP34530INT1 (BC)1135INT1 (BC)4531INT21274INT24628INT31121INT34306INT4875INT46WT1 (BC)1135WT1 (BC)4531WT21180WT24335 BC Base case, BW Bandwidth, EC Edge correction, GP Grid points, INT Interpolation, WT *Weighting a* bandwidth, edge correction, and number of grid points presented in ascending order, i.e., with the first scenario describing the least amount of bandwidth, edge correction, and number of grid points respectively b The F statistic refers to the ANOVA F statistic between the groups of estimated data points at the 50 evaluation points for a given area (1000 data points at each evaluation point), calculated as follows: F = between-group variability/within-group variability; higher values of the F statistic reflect more preferable parameter values for a given scenario. The groups are generated based on the POI bootstrap replications
## Obesity risk score map
Figure 6 depicts the risk score maps for Augsburg and Meitingen in which the five parameters bandwidth, edge correction, number of grid points, interpolation method and weighting, were set to their base case. The score values of the SORS are depicted as incremental densities resulting from subtracting the negative surface from the positive surface. The risk score map shows a composite picture of the separate estimates illustrated in Fig. 4. There was little heterogeneity in risk scores for the region of Augsburg except for a small area with higher obesogenic scores at the northwestern city boundary. The risk score map for Meitingen showed high risk scores for the northwestern part of the town as well as for the area north of the town borders. The risk score map based on point processes for comparison purposes is shown in Fig. 7. A region with a low obesity level is present in Figs. 6 and 7 for both study areas partly at similar places, however, obesity hotspots could only be derived from the SORS KDE plot in Fig. 6.Fig. 6Risk score maps for Augsburg (left) and Meitingen (right) showing the base case with the following parameters: bandwidth: method of Terrell [38]; edge correction: area side length: 1.4 * side length of the exact box; number of grid points: 35 × 35; weighting scheme of the environmental factors: equal weighting with unity, interpolation here via the “image.plot” functionFig. 7Risk Score map based on incremental intensities derived from inhomogeneous cluster point processes
## Discussion
We developed the SORS based on KDE using freely available data from online geocoding services. We tested several parameters which could potentially influence the final score values. Our tests showed that the SORS depended on the choice of bandwidth and the amount of edge correction applied to the KDE; the latter, however, for only one of the two study areas. In contrast, the interpolation method, the numbers of grid points, and an alternative weighting scenario had a small influence on the results.
The SORS was calculated by taking the difference of the positive and the negative kernel density surface. We followed a similar approach to that in the work of Jones-Smith et al. [ 34]. They estimated correlations of their score with obesity. In contrast, the aim of our study was to investigate the effect of parameter variation on the robustness of our score. Furthermore, we covered an extensive list of obesogenic and protective environmental factors that expanded the approach of a food score to a more comprehensive measure, also including the physical activity environment. Some alternative approaches using KDE were based on the division of the kernel density surfaces [54]. A major drawback of these quotients is the issue with division by zero leading to values approaching infinity and thus leading to instability. Our approach to the SORS avoids this and also the need for some adjustments correcting for the instability.
Previous studies used various approaches to estimate risk scores based on kernel techniques, both in obesity-related research areas and elsewhere. Fitzpatrick and colleagues [55], for example, developed the keeping score based on KDE to characterize crime patterns, which has often been used by the police. Crime heat maps can be generated with this technique. This approach is based on the locations of past events instead of geolocated environmental factors, and the authors assumed that the pattern of these historic events would be maintained in the future.
Some studies created kernel density surfaces based on POIs and extracted density estimations from these surfaces in order to investigate the association with weight status. Rundle et al. [ 2009] analyzed the effects of environmental factors on body mass index (BMI). Results of KDE analysis concerning healthy and unhealthy food outlets were used to classify the neighborhood environment of each individual within the study based on a quintile approach [56]. Furthermore, walkability, land use mix, and population density were considered. These variables could not be implemented in our study based on the chosen POI approach with OSM data.
The five chosen SORS parameters, bandwidth, edge correction, grid points, interpolation, and weights, have also been investigated in the literature. Laraia et al. [ 2017] used a business software and ArcGIS to geocode the information from the study data [57]. As in our analysis, several bandwidths were tested within their KDE approach, which was found to be a sensitive model parameter. Similarly, we also found a fundamental influence of bandwidth on the results.
Effects at the edge of the study area were estimated in a simulation study concerning cluster models for food outlets [58]. Estimations at the boundaries were biased, and the authors came to the conclusion that edge effects should be corrected in studies considering measures of availability and accessibility. This underlined the importance of edge correction, which was also a major topic in our study. In addition, extending the study area has been proven to be a valuable edge correction method.
Finding the optimal number of grid points was also discussed in the literature. Some authors suggested that a choice between 100 and 500 grid points gives reasonable results [59]. In our analysis, we chose 25 × 25 points for the minimum bounding rectangle, i.e., 625 grid points, and chose some additional amount of edge correction for the base case. In addition, we performed some adjustments to preserve the distance between the grid points for the edge correction scenarios. In this case, the number of grid points was extended proportionally to the amount of edge correction applied, i.e., to the amount of study area extension. This made it possible to analyze grid point and edge effects separately. The choice of grid points in our base case and sensitivity analysis was chosen in accordance with default grid sizes implemented in KDE packages.
An inverse distance weighting method was applied in the past in KDE estimation regarding homicide locations as a parameter of area safety [60]. This method could be used to estimate effects at specific locations. We used such an inverse distance method in our model as an alternative to the automatic interpolation function of the base case. As a further common method, linear interpolation has been applied within the literature [61]. The “interp.surface” function applied to our SORS model was based on bilinear weights.
It was challenging to find a suitable weighting scheme applicable within our analysis. For the base case, we assumed that each factor has the same positive or negative weight, although this might look different in reality. Additionally, we tested an example from the literature [34]. We found that double weighting of supermarkets and physical activity facilities had little effect on the results. Owing to several possible weighting methods for spatial POIs, it is necessary to test further alternatives within future studies.
Finally, the SORS was graphically compared to a risk score that was derived from incremental intensities of inhomogeneous spatial point processes. Although the methodology applied here changed from KDE-based to intensity-based estimations, similar visual patterns could be derived from the two score approaches for protective patterns, which further underlines the robustness of our chosen algorithm.
## Implications of the SORS on obesity-related research and policy
The SORS is a helpful tool to understand the spatial distribution of health-related harmful environmental factors in relation to health-promoting environmental factors. Risk score maps allow for an overall intuitive view on summarized structures, which can be a valuable help in obesity-related research and also within policy. Although the actual use of those structures might look different in reality, it nevertheless gives a composite simplifying measure of the environment and can be further extended to a more comprehensive tool accounting for several health dimensions affecting individuals simultaneously.
## Strengths and limitations
Several strengths exist regarding our study. The automated processing of data and the automated testing of several important KDE parameters makes it possible to repeat the application of risk score estimation for other areas efficiently, given that the spatial data points and the shape files of the city or town boundaries have been downloaded before. This enables the user to describe, compare, and monitor (if done repeatedly) risk scores as well as the influence of relevant risk score parameters within several areas of interest, within other regions worldwide, and also on a larger geographic scale. For example, the analysis could be performed for a whole country in order to identify national inequalities regarding environmental obesity risks or to guide and prioritize prevention efforts that concentrate on the food and the physical activity environment. To achieve this on a regional scale, the data download area simply has to be increased to cover a larger area for the subsequent data download from OSM. The data files would be of a manageable size, as only a small number of features are important for this kind of analysis. For Augsburg, i.e., for the larger of our two study areas, the data file size was 8 MB. For larger areas, e.g., for Germany, other portals such as Geofabrik should be used. In this case, no query process is needed, and the data files are directly ready for download. The data size for Germany, for example, would be 3.1 gigabytes in this case [62]. Furthermore, using so-called planet OSM files, data disk space of around one terabyte (compressed 89 GB) or less is required [63].
We integrated uncertainty into our analysis by performing a spatial bootstrap. Subsequently, we used the samples directly for the evaluation of our method. This allowed us to assess the stability of the score values against POI variations and helped us to compare deterministic parameter scenarios based on the ANOVA F statistic. On the one hand, the impact of each parameter on score results could be assessed. In addition, the values of the F statistic could be used to find optimal parameter combinations for the SORS.
We checked the robustness of the score and repeated our analysis several times for a given area. Results were qualitatively equivalent, i.e., for each given parameter variation, the repeated analysis could be used to rank the scenarios in the same order.
However, there are also some limitations regarding the study. First, some of the environmental factors discovered during the literature search could not be implemented based on spatial POIs, especially complex constructs such as land use mix or walkability.
Second, the categorization of positive and negative obesogenic factors was based on data from pre-existing literature, and it is not known whether POIs categorized as “positive” or “negative” are really positively or negatively associated with obesogenic health (behavior). Further studies could compare the SORS with external data sources, such as walk scores in a given region, in order to test these associations [64].
As the content of OSM is generated by users, it is necessary to assess the data quality within validation studies. Within our previous work, we calculated sensitivity, specificity, and positive predictive values for OSM and compared the results with the corresponding values for Google Maps [18]. It became evident that both geocoding services performed adequately. OSM had higher positive predictive value but, in contrast, lower sensitivities than Google Maps.
## Conclusion
KDE has been proven to be a useful methodology in the development of an obesity risk score, predominantly on account of the nature of the continuous estimation approach enabling efficient generation of risk score maps. However, some parameters of KDE have a large effect on score results. Parameter optimization should therefore play a major role during score model development.
## Supplementary Information
Additional file 1. Complete list of chosen variables. Additional file 2. Methodological details. Additional file 3.
## References
1. Bluher M. **Obesity: global epidemiology and pathogenesis**. *Nat Rev Endocrinol* (2019.0) **15** 288-298. DOI: 10.1038/s41574-019-0176-8
2. Elbe AM, Elsborg P, Dandanell S, Helge JW. **Correlates and predictors of obesity-specific quality of life of former participants of a residential intensive lifestyle intervention**. *Obes Sci Pract* (2018.0) **4** 188-193. DOI: 10.1002/osp4.163
3. Schienkiewitz A, Mensink G, Kuhnert R, Lange C. *Overweight and obesity among adults in Germany* (2017.0)
4. Mader S, Rubach M, Schaecke W, Röger C, Feldhoffer I, Thalmeier E-M. **Healthy nutrition in Germany: a survey analysis of social causes, obesity and socioeconomic status**. *Public Health Nutr* (2020.0) **23** 2109-2123. DOI: 10.1017/S1368980019004877
5. Haftenberger M, Mensink GB, Herzog B, Kluttig A, Greiser KH, Merz B. **Changes in body weight and obesity status in German adults: results of seven population-based prospective studies**. *Eur J Clin Nutr* (2016.0) **70** 300-305. DOI: 10.1038/ejcn.2015.179
6. Merlo J, Wagner P, Leckie G. **A simple multilevel approach for analysing geographical inequalities in public health reports: the case of municipality differences in obesity**. *Health Place* (2019.0) **58** 102145. DOI: 10.1016/j.healthplace.2019.102145
7. Zhou Y, Wu K, Shen H, Zhang J, Deng H-W, Zhao L-J. **Geographical differences in osteoporosis, obesity, and sarcopenia related traits in white American cohorts**. *Sci Rep* (2019.0) **9** 12311. DOI: 10.1038/s41598-019-48734-9
8. Elinder LS, Jansson M. **Obesogenic environments--aspects on measurement and indicators**. *Public Health Nutr* (2009.0) **12** 307-315. PMID: 18498677
9. Hobbs M, Griffiths C, Green MA, Jordan H, Saunders J, McKenna J. **Associations between the combined physical activity environment, socioeconomic status, and obesity: a cross-sectional study**. *Perspect Public Health* (2018.0) **138** 169-172. DOI: 10.1177/1757913917748353
10. Mackenbach JD, Rutter H, Compernolle S, Glonti K, Oppert JM, Charreire H. **Obesogenic environments: a systematic review of the association between the physical environment and adult weight status, the SPOTLIGHT project**. *BMC Public Health* (2014.0) **14** 233. DOI: 10.1186/1471-2458-14-233
11. Rendina D, Campanozzi A, De Filippo G. **Methodological approach to the assessment of the obesogenic environment in children and adolescents: a review of the literature**. *Nutr Metab Cardiovasc Dis* (2019.0) **29** 561-571. DOI: 10.1016/j.numecd.2019.02.009
12. Monsivais P, Francis O, Lovelace R, Chang M, Strachan E, Burgoine T. **Data visualisation to support obesity policy: case studies of data tools for planning and transport policy in the UK**. *Int J Obes* (2018.0) **42** 1977-1986. DOI: 10.1038/s41366-018-0243-6
13. Daley D, Bachmann M, Bachmann BA, Pedigo C, Bui MT, Coffman J. **Risk terrain modeling predicts child maltreatment**. *Child Abuse Negl* (2016.0) **62** 29-38. DOI: 10.1016/j.chiabu.2016.09.014
14. Caplan JM, Kennedy LW, Barnum JD, Piza EL. **Risk terrain modeling for spatial risk assessment**. *Cityscape.* (2015.0) **17** 7-16
15. Ripoche M, Lindsay LR, Ludwig A, Ogden NH, Thivierge K, Leighton PA. **Multi-scale clustering of Lyme disease risk at the expanding leading edge of the range of Ixodes scapularis in Canada**. *Int J Environ Res Public Health* (2018.0) **15** 603. DOI: 10.3390/ijerph15040603
16. Lafontaine SJV, Sawada M, Kristjansson E. **A direct observation method for auditing large urban centers using stratified sampling, mobile GIS technology and virtual environments**. *Int J Health Geogr* (2017.0) **16** 6. DOI: 10.1186/s12942-017-0079-7
17. Cebrecos A, Diez J, Gullon P, Bilal U, Franco M, Escobar F. **Characterizing physical activity and food urban environments: a GIS-based multicomponent proposal**. *Int J Health Geogr* (2016.0) **15** 35. DOI: 10.1186/s12942-016-0065-5
18. Präger M, Kurz C, Bohm J, Laxy M, Maier W. **Using data from online geocoding services for the assessment of environmental obesogenic factors: a feasibility study**. *Int J Health Geogr* (2019.0) **18** 13. DOI: 10.1186/s12942-019-0177-9
19. Lemke D, Mattauch V, Heidinger O, Hense HW. **Who Hits the Mark? A Comparative Study of the Free Geocoding Services of Google and OpenStreetMap**. *Gesundheitswesen (Bundesverband der Arzte des Offentlichen Gesundheitsdienstes (Germany))* (2015.0) **77** e160-e165. PMID: 26154258
20. Mocnik F-B, Mobasheri A, Zipf A. **Open source data mining infrastructure for exploring and analysing OpenStreetMap**. *Open Geospatial Data, Software and Standards* (2018.0) **3** 7. DOI: 10.1186/s40965-018-0047-6
21. Jia P, Cheng X, Xue H, Wang Y. **Applications of geographic information systems (GIS) data and methods in obesity-related research**. *Obes Rev* (2017.0) **18** 400-411. DOI: 10.1111/obr.12495
22. 22.Statistische Ämter des Bundes und der Länder. Gemeinsames Statistikportal. Gemeindeverzeichnis-Online. https://www.statistikportal.de/de/produkte/gemeindeverzeichnis. Accessed 10 Jun 2020.
23. Meisinger C, Döring A, Thorand B, Heier M, Löwel H. **Body fat distribution and risk of type 2 diabetes in the general population: are there differences between men and women? The MONICA/KORA Augsburg cohort study**. *Am J Clin Nutr* (2006.0) **84** 483-489. DOI: 10.1093/ajcn/84.3.483
24. 24.Overpass turbo. https://overpass-turbo.eu/. Accessed 11 Aug 2021.
25. 25.Bayerische Vermessungsverwaltung. OpenData. https://www.ldbv.bayern.de/produkte/weitere/opendata.html. Accessed 11 Aug 2021.
26. 26.Export | OpenStreetMap. https://www.openstreetmap.org/export. Accessed 11 Aug 2021.
27. 27.Map features. https://wiki.openstreetmap.org/wiki/Map_features. Accessed 20 Jul 2021.
28. Ripley BD. **Modelling spatial patterns**. *J R Stat Soc Series B Stat Methodol* (1977.0) **39** 172-192
29. Dai D, Taquechel E, Steward J, Strasser S. **The impact of built environment on pedestrian crashes and the identification of crash clusters on an urban university campus**. *West J Emerg Med* (2010.0) **11** 294-301. PMID: 20882153
30. Haase P. **Spatial pattern analysis in ecology based on Ripley’s K-function: introduction and methods of edge correction**. *J Veg Sci* (1995.0) **6** 575-582. DOI: 10.2307/3236356
31. 31.ExplainKplot. http://spatstat.org/explainKplot.html. Accessed 11 Aug 2021.
32. Gabriel E. **Estimating second-order characteristics of inhomogeneous Spatio-temporal point processes**. *Methodol Comput Appl Probab* (2014.0) **16** 411-431. DOI: 10.1007/s11009-013-9358-3
33. Loh JM. **A valid and fast spatial bootstrap for correlation functions**. *Astrophys J* (2008.0) **681** 726. DOI: 10.1086/588631
34. Jones-Smith JC, Karter AJ, Warton EM, Kelly M, Kersten E, Moffet HH. **Obesity and the food environment: income and ethnicity differences among people with diabetes: the diabetes study of northern California (DISTANCE)**. *Diabetes Care* (2013.0) **36** 2697-2705. DOI: 10.2337/dc12-2190
35. Klemelä JS. *Smoothing of multivariate data: density estimation and visualization* (2009.0)
36. Hastie T, Tibshirani R, Friedman JH. *The elements of statistical learning: data mining, inference, and prediction* (2009.0)
37. King TL, Bentley RJ, Thornton LE, Kavanagh AM. **Using kernel density estimation to understand the influence of neighbourhood destinations on BMI**. *BMJ Open* (2016.0) **6** e008878. DOI: 10.1136/bmjopen-2015-008878
38. Terrell GR. **The maximal smoothing principle in density estimation**. *J Am Stat Assoc* (1990.0) **85** 470-477. DOI: 10.1080/01621459.1990.10476223
39. Davies TM, Marshall JC, Hazelton ML. **Tutorial on kernel estimation of continuous spatial and spatiotemporal relative risk**. *Stat Med* (2018.0) **37** 1191-1221. DOI: 10.1002/sim.7577
40. Nychka D, Furrer R, Paige J, Sain S. *fields: Tools for spatial data* (2017.0)
41. Wong DW, Yuan L, Perlin SA. **Comparison of spatial interpolation methods for the estimation of air quality data**. *J Expo Anal Environ Epidemiol* (2004.0) **14** 404-415. DOI: 10.1038/sj.jea.7500338
42. Kethireddy SR, Tchounwou PB, Ahmad HA, Yerramilli A, Young JH. **Geospatial interpolation and mapping of tropospheric ozone pollution using geostatistics**. *Int J Environ Res Public Health* (2014.0) **11** 983-1000. DOI: 10.3390/ijerph110100983
43. Wang X, Wiegand T, Wolf A, Howe R, Davies SJ, Hao Z. **Spatial patterns of tree species richness in two temperate forests**. *J Ecol* (2011.0) **99** 1382-1393. DOI: 10.1111/j.1365-2745.2011.01857.x
44. 44.R Core TeamR: a language and environment for statistical computing2016ViennaAustria. *R: a language and environment for statistical computing* (2016.0)
45. Wickham H. *ggplot2: elegant graphics for data analysis* (2009.0)
46. Baddeley A, Rubak E, Turner R. *Spatial point patterns: methodology and applications with R* (2015.0)
47. 47.Venables WN, Ripley BD. Modern Applied Statistics with
S. Fourth Edition. New York: Springer; 2002. ISBN 0-387-95457-0.
48. Bivand R, Keitt T, Rowlingson B. *Rgdal: bindings for the geospatial data abstraction library. R package version 1* (2016.0) 2-4
49. Mouselimis L. *geojsonR: a GeoJson processing toolkit. R package version 1.0.0* (2017.0)
50. Jay Kerns G. *Prob: elementary probability on finite sample spaces. R package version 1.0–1* (2018.0)
51. Pebesma EJ. **Multivariable geostatistics in S: the gstat package**. *Comput Geosci* (2004.0) **30** 683-691. DOI: 10.1016/j.cageo.2004.03.012
52. Urbanek S. *Png: read and write PNG images. R package version 0.1–7* (2013.0)
53. Hahsler M, Piekenbrock M. *Dbscan: density based clustering of applications with noise (DBSCAN) and related algorithms. R package version 1.1–1* (2017.0)
54. 54.Loo BP, Yao S, Wu J. Spatial point analysis of road crashes in Shanghai: a GIS-based network kernel density method. In: 2011 19th international conference on geoinformatics IEEE. 2011. p. 1–6.
55. Fitzpatrick DJ, Gorr WL, Neill DB. **Keeping score: predictive analytics in policing**. *Annu Rev Criminol* (2019.0) **2** 473-491. DOI: 10.1146/annurev-criminol-011518-024534
56. Rundle A, Neckerman KM, Freeman L, Lovasi GS, Purciel M, Quinn J. **Neighborhood food environment and walkability predict obesity in new York City**. *Environ Health Perspect* (2009.0) **117** 442-447. DOI: 10.1289/ehp.11590
57. Laraia BA, Downing JM, Zhang YT, Dow WH, Kelly M, Blanchard SD. **Food environment and weight change: does residential mobility matter?: the diabetes study of northern California (DISTANCE)**. *Am J Epidemiol* (2017.0) **185** 743-750. DOI: 10.1093/aje/kww167
58. Van Meter EM, Lawson AB, Colabianchi N, Nichols M, Hibbert J, Porter DE. **An evaluation of edge effects in nutritional accessibility and availability measures: a simulation study**. *Int J Health Geogr* (2010.0) **9** 40. DOI: 10.1186/1476-072X-9-40
59. Raykar VC, Duraiswami R, Zhao LH. **Fast computation of kernel estimators**. *J Comput Graph Stat* (2010.0) **19** 205-220. DOI: 10.1198/jcgs.2010.09046
60. Weiss CC, Purciel M, Bader M, Quinn JW, Lovasi G, Neckerman KM. **Reconsidering access: park facilities and neighborhood disamenities in new York City**. *J Urban Health* (2011.0) **88** 297-310. DOI: 10.1007/s11524-011-9551-z
61. Van Kerm P. **Adaptive kernel density estimation**. *Stata J* (2003.0) **3** 148-156. DOI: 10.1177/1536867X0300300204
62. 62.Geofabrik downloads. Germany. https://download.geofabrik.de/europe/germany.html. Accessed 10 May 2021.
63. 63.Planet.osm. https://wiki.openstreetmap.org/wiki/Planet.osm. Accessed 10 May 2021.
64. Carr LJ, Dunsiger SI, Marcus BH. **Walk score as a global estimate of neighborhood walkability**. *Am J Prev Med* (2010.0) **39** 460-463. DOI: 10.1016/j.amepre.2010.07.007
|
---
title: 'Determinants and drivers of young children’s diets in Latin America and the
Caribbean: Findings from a regional analysis'
authors:
- Franziska Gassmann
- Richard de Groot
- Stephan Dietrich
- Eszter Timar
- Florencia Jaccoud
- Lorena Giuberti
- Giulio Bordon
- Yvette Fautsch-Macías
- Paula Veliz
- Aashima Garg
- Maaike Arts
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021987
doi: 10.1371/journal.pgph.0000260
license: CC BY 4.0
---
# Determinants and drivers of young children’s diets in Latin America and the Caribbean: Findings from a regional analysis
## Abstract
The Latin America and Caribbean region exhibit some of the lowest undernutrition rates globally. Yet, disparities exist between and within countries and countries in the region increasingly face other pressing nutritional concerns, including overweight, micronutrient deficiencies and inadequate child feeding practices. This paper reports findings from a regional analysis to identify the determinants and drivers of children’s diets, with a focus on the complementary feeding window between the age of 6–23 months. The analysis consists of a narrative review and descriptive data analysis, complemented with qualitative interviews with key informants in four countries: Guatemala, Paraguay, Peru and Uruguay. Findings indicate that poverty and inequality (disparities within countries by wealth and residence), unequal access to services, inadequate coverage of social programmes and lack of awareness on appropriate feeding practices are important drivers for inadequate diets. We conclude that countries in the region need to invest in policies to tackle overweight and micronutrient deficiencies in young children, considering inequalities between and within countries, enhance coverage of social protection programmes, improve coordination between sectors to improve children’s diets and expand the coverage and intensity of awareness campaigns on feeding practices, using iterative programme designs.
## 1. Introduction
Nutritious, safe and diverse food is one of the most essential building blocks of human life. Hunger and malnutrition not only violate the basic human right to food but can also have severe consequences that last over one’s lifetime. Malnutrition in young children has detrimental and long-lasting consequences for their physical and cognitive development [1]. Throughout the last decades, countries have made remarkable progress towards eradicating child malnutrition, but many children around the world are still at risk of not meeting their dietary needs. While the global share of children under five with stunting (being too short for one’s age) has declined steadily since 1990, one in five children still experienced impaired growth due to poor nutrition in 2020 [2].
Although the Latin American and Caribbean (LAC) region registers one of the lowest rates of stunting and wasting (being too thin for one’s height) globally, regional aggregates hide large disparities across countries [3]. In Guatemala, $47\%$ of children under five are stunted, and in Ecuador, Haiti and Honduras more than $20\%$ of the children suffer from wasting [3]. Meanwhile, adequate maternal, infant and young child nutrition remains an exception rather than the norm. According to UNICEF’s latest estimates, a little over one out of three ($38\%$) infants under 6 months old are exclusively breastfed [3]. Although the vast majority ($84\%$) of infants are introduced to solid, semi-solid or soft foods at the recommended age between 6–8 months, many children are consuming solid foods before the recommended 6 months [4]. There is also growing evidence that a dietary transition is taking place and the consumption of foods high in sugar, salt and fat is increasing [5,6], resulting in increasing rates of overweight and its severe form obesity [7]. As a result, $7.5\%$ of children under 5 in the LAC region are affected by overweight, ranging from $3.7\%$ in Haiti to $12.9\%$ in Argentina [2]. Given the large share of children who suffer from micronutrient deficiencies (e.g. iron and vitamin A) [3], LAC faces a ‘triple burden of malnutrition’ (the co-existence of undernutrition, micronutrient deficiencies and overweight).
Against this background, it is imperative to understand determinants and drivers of children’s diets and to identify effective interventions that improve the diets of young children. This is particularly relevant since progress against the key nutritional Sustainable Development Goals (Target 2.1 on ending hunger and Target 2.2 on eliminating malnutrition) has been slow [8]. The main aim of this paper is therefore to analyze the determinants and drivers of young children’s diets in LAC. In doing so, we start by providing an overview of current rates of malnutrition and complementary feeding practices in the region. In addition, we aim to identify existing strategic actions that improve young children’s diets in the region. Using multiple research methods, including a narrative review, descriptive data analysis and interviews with key stakeholders, the analysis in this paper identifies barriers and opportunities on how to improve the diets of young children in LAC. The focus is on children from 6 to 23 months of age, which is the age when children transition from exclusive breastfeeding to age-appropriate complementary feeding. This period is a critical window in a child’s development and if complementary foods and feeding practices are inappropriate, there is increased risk of undernutrition (stunting, wasting), micronutrient deficiencies and overweight [9]. For instance, complementary foods high in sugar and fats can lead to overweight while also failing to meet a child’s micronutrient needs. This window is not only critical in meeting children’s dietary needs for their growth and development, but it is also when children’s food preferences and dietary habits are shaped for the rest of their lives. This is the time when they learn to listen and respond to cues of hunger and satiety, which are essential for upholding healthy diets and weight throughout life.
The analysis is based on the UNICEF action framework to improve the diets of young children during the complementary feeding period [10] (Fig 1). The framework recognizes the role of a situational analysis of the determinants of children’s diets. The determinants of young children’s diets during the complementary feeding period include adequate complementary foods, adequate complementary feeding practices, and adequate services. Access to adequate food includes physical and economic access to nutritious, safe and affordable foods. These determinants are shaped by context-specific factors–referred to as drivers. Together, they determine children’s ability to enjoy nutritious, safe, affordable and sustainable diets that protect, promote and support survival, growth and development.
**Fig 1:** *The UNICEF action framework to improve the diets of young children during the complementary feeding period.Source: UNICEF [11].*
The action framework reinforces the need to deliver context-specific strategic actions through multiple systems that have the potential to deliver nutrition interventions: the food system, the health system, the water and sanitation system and the social protection system. The food system includes all actors and steps to get food from farm to mouth. It therefore has a critical role in the availability, accessibility and affordability of nutritious food [3]. The health system too plays a key role in child nutrition from pregnancy through the first years of life through access to affordable and high-quality preventive and curative health care. For example, contact with proper health services through antenatal care monitoring can assist women in keeping proper nutrition during their pregnancy. Having received antenatal care (ANC) services in low- and middle-income countries is associated with improved birth outcomes and longer-term reductions of child mortality and malnourishment [12]. Furthermore, delivery in a health facility with a skilled provider—particularly care delivered in public sector facilities—appears to be positively correlated with favorable breastfeeding practices, providing a good foundation for nutrition at birth [13]. After birth, growth monitoring and nutrition counselling through the health system can be effective ways to improve nutritional status as well as providing an entry point to preventive and curative health care [14]. Water, sanitation and hygiene (WASH) systems are particularly important for safe food preparation, drinking water, and the reduction of infectious disease (like diarrhea) that would limit children’s absorption of micronutrients. The lack of access to proper WASH services can affect child nutrition through diarrheal diseases, intestinal parasite infections and environmental enteropathy [15]. Social protection systems can support households in meeting their dietary needs in different ways. Direct assistance in the form of cash transfers can help caregivers to buy healthy complementary foods for their children, thereby playing an essential role in facilitating access to appropriate diets [16]. Alternatively, transfers can help in seeking health care or upgrading housing conditions, affecting other determinants of malnutrition [17].
## 2.1. Ethics statement
This study is primarily based on secondary data analysis and a desk review of relevant documents and journal articles. Primary data collection for this study involved in-depth interviews with selected key informants. Participation in the interview was voluntary and consensual. Prior to the interview, written consent was sought and this was reaffirmed at the beginning of the interview. Although ethical approval was not required for this type of study, as employees or affiliates of Maastricht University, the authors were bound by the Netherlands Code of Conduct for Research Integrity [18].
## 2.2. Methods
This regional analysis is based on a narrative review and a descriptive analysis of secondary nutritional data, complemented by qualitative interviews with key informants in four selected countries (Guatemala, Paraguay, Peru and Uruguay).
We conducted a descriptive analysis of malnutrition rates (stunting, wasting and overweight) and key complementary feeding indicators for assessing infant and young child feeding practices following WHO-UNICEF infant and young child feeding (IYCF) indicators guidelines using publicly available data, as shown in Table 1. The data for malnutrition rates was extracted from the UNICEF/WHO/World Bank Joint Malnutrition Estimates Expanded Databases, which is the prominent source for malnutrition data to monitor progress towards the Sustainable Development Goal, Target 2.2 to end all forms of malnutrition [2]. Infant and young child feeding indicators were extracted from the UNICEF Global Database on Infant and Young Child feeding, the key source for nutritional data on children [19]. Data visualizations were prepared with Microsoft Excel.
**Table 1**
| Indicator | Definition |
| --- | --- |
| Exclusive breastfeeding | The % of children aged 0–5 months who are exclusively breastfed. |
| Introduction of solid foods | The % of children aged 6–8 months who receive solid, semi-solid or soft foods. |
| Minimum dietary diversity (MDD) | The % of children aged 6–23 months who received foods from at least five food groups from the following eight: 1. Breastmilk, 2. Grains, roots, tubers, 3. Legumes and nuts, 4. Dairy products (formula, milk, yogurt, cheese), 5. Flesh foods (meat, fish, poultry, liver, organ meats), 6. Eggs, 7. Vitamin-A rich fruits and vegetables, 8. Other fruits and vegetables |
| Minimum meal frequency (MMF) | The % of children aged 6–23 months who receive solid (solid, semi-solid or soft) foods the minimum number of times. The minimum number of times of complementary feeding for breastfed children is: 2 times for those aged 6–8 months, 3 times for those aged 9–23 months. The minimum number of feeding for non-breastfed children is 4 times solid foods and/or milk feeds for 6–23 months of age. |
| Minimum acceptable diet (MAD) | The % of children aged 6–23 months who receive both the minimum dietary diversity and the minimum meal frequency. |
The narrative review was organized around two elements of the UNICEF Action Framework: 1) The situation and status of young children’s diets, including what, when and how they are fed and the determinants and drivers of young children’s diets, especially those related to food, services and practices; and 2) Strategic actions to improve young children’s diets delivered through the food, WASH, health and social protection systems. Documents were identified via several approaches.
The objective of the narrative review was to 1) identify patterns and determinants of young children’s diets in the region, 2) understand the economic and socio-cultural contexts and drivers relevant to child feeding in LAC, 3) map out policy and existing approaches to improving young children’s diets, and 4) identify knowledge gaps and directions for future research and policy action in the region. The review involved three stages and was conducted in January and February 2020.
First, several key research reports and grey literature were reviewed. This included scoping and review reports on the topic from international organizations such as UNICEF, the FAO, and WFP. The purpose of these reports was to provide a preliminary understanding of recommendations, indicators, and best practices related to the complementary feeding of young children. Key words to be used as search strings in the second stage of the scoping review were also identified based on these reports.
The objective of the second stage of the review was to identify academic and grey literature on the state, determinants and drivers of young children’s diets in the region. This stage relied on a protocol for searching and including literature. The team searched for academic literature via four academic databases/search engines: Google Scholar, PubMed, LATINDEX and Elsevier by using search terms (including BOOLEAN terms) such as “complementary feeding”, “complementary food”, “child diet”, “child nutrition”, “infant nutrition”, “exclusive breastfeeding”, “responsive feeding”, “food system” etc. in combination with “Latin America”, “Caribbean”, and names of countries in the region. The search terms were entered in English, Spanish and Portuguese. We also searched on websites of relevant organizations and initiatives (e.g. Food and Agricultural Organization of the United Nations, the World Food Programme, UNICEF, the International Food Policy Research Institute, the Inter-American Development Bank, the Institute of Nutrition of Central America and Panama, the Global Alliance for Improved Nutrition, Scaling up Nutrition etc.). Finally, we used snowballing techniques starting from key documents identified through the desk review, in which the reference list in these key papers was traced and relevant papers reviewed. These reference lists were subjected to the same key search terms as above to identify potentially relevant papers. To provide the most up-to-date information, the literature search was limited to documents published after 2010. The search protocol did not include restrictions on methodology or sample size: both quantitative and qualitative research, as well as review papers were permitted if they aligned with the study objective. A geographic restriction was applied: publications had to pertain to the region (either as a whole or to individual countries in the region). See S1 Text for details on the search protocol.
The literature identified through these stages were first subjected to title- and abstract-screening. The information of those that appeared relevant were entered into an Excel sheet, and the study team conducted a full-text screening to decide which publications to include (based on their relevance and quality).
To gain a more in-depth understanding of the determinants and drivers of children’s diets as well as institutional contexts, key informant interviews were conducted with stakeholders in four countries: Guatemala, Paraguay, Peru and Uruguay. Countries were selected purposefully to include a broad range of dietary and nutrition situations in the region, as well as different socio-economic and policy situations. For example, Guatemala has the highest stunting rates in the region and a high cultural diversity. Paraguay also has a diverse population. Peru is considered a success story in reducing stunting. And while *Uruguay is* a high-income country; the others are all upper-middle income according to the World Bank classification [21]. In each of the four countries, four or five key informants from different sectors and institutional levels were selected, including staff from NGOs, government officials, researchers and international development practitioners. Recruitment into the study was based on purposeful selection in consultation with local UNICEF offices. By purposefully selecting informants based on their expertise, the study was able to generate the most salient information for each country within a relatively short timeframe.
Interviews followed a semi-structured approach and respondents were asked about the overall situation of child nutrition in their countries, about the barriers and facilitators of improving children’s diets, and the legal and policy frameworks. Interview questions were adjusted for the specific type of respondent. For example, interviews with central government bodies were more concerned with the overall situation and the government’s priorities and vision. NGO workers were asked about their concrete experiences on the ground, descriptions of children’s diets and needs, as well as potential barriers and lessons learned. Interviews were conducted in July and August 2020 via telephone or over the internet, in the language preferred by the respondent (Spanish or English). Interviews were completed by three members of the research team, while the analysis was conducted by two other researchers to avoid bias in the interpretation of the interviews. All interviewers had no relation to the key informants prior to the interview. The interviews adhered to ethical standards as set out by the Netherlands Code of Conduct for Research Integrity [18]. Participation was voluntary and consensual. Prior to the interview consent was sought and this was reaffirmed at the beginning of the interview. In addition, provided the participant consented, the interviews were audio-recorded to ensure the statements were captured in the correct context. These audio recordings were stored in a password protected online server and were only accessible to the interviewers to clarify any statements during the analysis stage. Furthermore, the data was treated confidentially and anonymously, and hence no direct reference to the key informant is made in this paper. The interviewers, who were also fluent in English, took detailed notes during the interview and then translated the notes into English for the analysis. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S4 Text).
The qualitative analysis followed a deductive content analysis process [22], focusing on extracting predetermined key emerging themes across six broad topics: 1) General context in the country regarding the food environment and recent developments, 2) Institutional context and legal frameworks with respect to complementary feeding; 3) Food and agriculture, 4) Health system; 5) Water, Sanitation and Hygiene (WASH); 6) Social protection. The latter four sectors were included following the UNICEF conceptual framework as sectors that have the potential to deliver nutrition interventions at scale. The interviews were analyzed per country and answers were categorized under each relevant topic using a predefined analysis sheet. For each topical area, it was determined to what extent the various informants for each country agreed on the major themes, as a triangulation measure. Based on this categorization, a summary was developed per country, which formed the basis for a more extensive reporting and comparisons on the emerging findings across the four countries. We integrated findings from the narrative review and the descriptive analysis into the country qualitative analysis to triangulate findings from the expert interviews with the available data and country-specific literature.
## 3. Results
On average, the LAC region registers one of the lowest rates of stunting ($11\%$) and wasting ($1.3\%$) globally [2]. However, notable exceptions remain hidden by regional aggregates: in Guatemala, $47\%$ of children under five are affected by stunting, and Ecuador, Haiti and Honduras have stunting rates above $20\%$. Meanwhile, overweight affects more and more of the region’s population, including children. The magnitude of the problem of overweight is shown in Fig 2. In most countries with available data (collected after 2010), rates of overweight among children under 5 are $5\%$ or higher, with the largest burden in Argentina, Paraguay and Barbados. Malnutrition in children under 5 also differs within countries, with stunting and wasting more prevalent among children in the lowest wealth quintiles and (for stunting) in rural areas (S2 Text, Fig A, B, D and E). In contrast, overweight is more concentrated among children in urban areas and in the highest wealth quintile (S2 Text, Fig C and F). Recently, micronutrient deficiencies have become a more pressing public health concern than stunting. The rate of micronutrient deficiency, in terms of vitamin A and iron deficiency, in children under 5 is $36\%$, $37\%$ and $46\%$ in South America, Central America and the Caribbean, respectively [3].
**Fig 2:** *The prevalence of stunting, wasting and overweight in Latin American and the Caribbean countries.Note: The figure shows the prevalence of stunting, wasting and overweight for each country using the latest available data since 2010. Abbreviations: ARG: Argentina, BLZ: Belize, BOL: Bolivia, BRB: Barbados, CHL: Chili, COL: Colombia, DOM: Dominican Republic, ECU: Ecuador, ELS: El Salvador, GTM: Guatemala, GUY: Guyana, HND: Honduras, HTI: Haiti, JAM: Jamaica, MEX: Mexico, NIC: Nicaragua, PER: Peru, PRY: Paraguay, SUR: Suriname, TAT: Trinidad and Tobago, URY: Uruguay. Source: authors’ calculations based on UNICEF Global Database.*
An overview of the key complementary feeding indicators paints the picture of a very heterogenous region (Fig 3). Countries with the highest rates of exclusive breastfeeding during the first six months of life are Peru ($66\%$), Bolivia ($58\%$), Guatemala ($53\%$), El Salvador ($47\%$) and Haiti ($40\%$). Among countries with available data, Suriname ($3\%$) and the Dominican Republic ($5\%$) lag the most in terms of exclusive breastfeeding. Within countries, exclusive breastfeeding is more common among households in the lowest wealth quintile and those in rural areas (S2 Text, Fig G and I).
**Fig 3:** *Exclusive breastfeeding and introduction of solid foods in Latin American and the Caribbean countries.Note: The figure shows the prevalence of exclusive breastfeeding among children < 6 months and the share of infants aged 6–8 months who are introduced to solid foods for each country using the latest available data since 2010. Abbreviations: ARG: Argentina, BLZ: Belize, BOL: Bolivia, BRB: Barbados, COL: Colombia, CRI: Costa Rica, CUB: Cuba, DOM: Dominican Republic, ECU: Ecuador, ELS: El Salvador, GTM: Guatemala, GUY: Guyana, HND: Honduras, HTI: Haiti, JAM: Jamaica, MEX: Mexico, NIC: Nicaragua, PAN: Panama, PER: Peru, PRY: Paraguay, SUR: Suriname, TAT: Trinidad and Tobago. Source: authors’ calculations based on UNICEF Global Database.*
Most infants between six and eight months of age received solid, semi-solid or soft foods in all countries with available data (Fig 3). In Argentina, Cuba, El Salvador, Haiti and Peru, over $90\%$ of infants have been introduced to complementary foods during this period. While on average, most infants are receiving solid, semi-solid or soft foods at the recommended time, there is quite some regional variation within the region. For example, in Belize, Ecuador, Guatemala, Jamaica, Panama, Suriname and Trinidad and Tobago, the corresponding rate was below $80\%$. There is also variation within countries, with a larger share of urban children and those in the highest wealth quintile receiving solid, semi-solid or soft foods at the recommended time (S2 Text, Fig H and J).
Information about the indicators on dietary adequacy (MDD, MMF and MAD) is limited in the region (Fig 4). Some countries have data on one indicator but not the other, which makes it difficult to measure compliance with minimum acceptable diet recommendations. In the countries with available data, the rates of young children meeting the MDD requirements are alarming. Most countries reported rates below $60\%$, with the most alarming situation in Haiti, where only one out of five infants ($19\%$) met the requirement for a diverse diet. Peru ($82.9\%$) has the highest share of children receiving a sufficiently diverse diet. For those countries for which multiple years of data are available, trends are not positive. In the Dominican Republic, the rate of MDD decreased between 2007 and 2014 from $58\%$ to $51\%$. In Haiti, the rate decreased from $23\%$ in 2005 to $19\%$ in 2016, and in Guyana, the rate decreased from $48\%$ in 2009 to $40\%$ in 2014. In Honduras, the rate of MDD increased only marginally from $58\%$ to $61\%$ between 2005 and 2011. The exception is Peru, which registered an increase from $73\%$ in 2007 to $83\%$ in 2016, though the trend stagnated over the last three data periods.
**Fig 4:** *Adequacy of children’s diets in Latin American and the Caribbean countries.Note: The figure shows the prevalence of minimum meal frequency, minimum diet diversity and minimum acceptable diet among children 6–23 months for each country using the latest available data since 2010. Abbreviations: ARG: Argentina, BLZ: Belize, BOL: Bolivia, BRB: Barbados, COL: Colombia, CRI: Costa Rica, CUB: Cuba, DOM: Dominican Republic, ELS: El Salvador, GTM: Guatemala, GUY: Guyana, HND: Honduras, HTI: Haiti, JAM: Jamaica, MEX: Mexico, PAN: Panama, PER: Peru, PRY: Paraguay, SUR: Suriname, TAT: Trinidad and Tobago. Source: Authors’ calculations based on UNICEF Global Database.*
Compliance with MMF recommendations is less of a challenge in the region. Two Caribbean countries, Haiti ($39\%$) and Jamaica ($42\%$) report the lowest prevalence of MMF. In all other countries, at least half of young children are fed the minimally recommended number of times, reaching at least $80\%$ in the Central American countries of El Salvador, Guatemala, Honduras and Mexico.
The computation of MAD requires data on both MDD and MMF, it can therefore only be reported for 10 countries (Fig 4). There are large differences in the prevalence of MAD. In El Salvador nearly two-thirds of children 6–23 months receive a minimal acceptable diet. In Cuba, Guatemala and Honduras, every second infant receives both enough quantity and types of food, while in Haiti only one in nine does.
Indicators of dietary adequacy vary considerably with wealth and residence. A larger share of children in wealthier households and those residing in urban areas meet dietary recommendations compared to children in the lowest wealth quintile or those in rural areas (S2 Text, Fig K-P).
## 3.1.1. Access to adequate foods
The LAC region produces more than enough quantity of food to meet the energy needs of its population [23]. Home production in the form of family farming plays a major role, accounting for about $81\%$ of agricultural activities in LAC [24], and supplies between $27\%$ and $67\%$ of food production [5]. The growth of food production exceeds population growth and the region is blessed with favorable environmental conditions and capacity for production. Marked differences exist between and within sub-regions, which makes trade and collaboration between countries important for ensuring equitable food availability [23].
At the household-level, a major obstacle to consuming diverse and healthy diets is the inability to afford them. The cost of a simple, healthy plate of food (a stew made of beans or other pulses, paired with a carbohydrate component that matches local preferences amounting to 600 kcal) is estimated to cost $4.0\%$, $5.4\%$, $6.3\%$ and $35\%$ of an average daily income in Guatemala, El Salvador, Nicaragua and Haiti, respectively. A similar plate of food would cost only $0.6\%$ of the average daily income of a citizen in New York. Since these estimates are averages, costs are likely higher for poorer households [25]. Another study on the relative costs of nutritious food showed that dark green leafy vegetables and vitamin-A rich fruits and vegetables are a relatively expensive source of calories in LAC. On the other hand, with the exception of fish, animal-sourced foods like diary and meat are relatively inexpensive [26]. A cost-of-diet analysis in El Salvador showed that between $24\%$ and $30\%$ of households would not be able to purchase a minimum cost nutritious diet [27].
At the child-level, grains, roots and tubers are the most prevalent complementary foods in the region, consumed by $88\%$ of infants [4]. Animal source foods are also relatively widespread, with dairy products being the most common, followed by flesh foods and eggs. Overall, less than $10\%$ of infants were reported to not consume any type of animal source food [4].
There are considerable disparities in the diets of infants across areas of residence and quintiles of wealth, with animal source food consumption being more common in urban and rich households [4]. An understanding of this inequality is important because animal source foods are the richest sources of iron and zinc [28], and meat intake is associated with better growth and development outcomes [29].
As countries move forward on the path of socio-economic development, formula-feeding becomes more prevalent [3]. Recent aggregate sales data confirms that commercial milk-based formulas are increasingly becoming a part of young children’s diets in the region. Between 2008 and 2013, Brazil and Peru were among the countries with the fastest growing sales of formulas globally by $133\%$ and $160\%$, respectively. In contrast, the sales of infant/child milk-based formulas grew by only $11\%$ in Mexico [30].
Overweight, obesity and the shift towards unhealthy diets can also be explained by the increased availability of energy-rich but micronutrient poor, processed foods in the region. Recently, there has been a major shift in the availability and consumption of ready-to-eat, ready-to-heat, processed, and packaged foods and beverages in the LAC region [23,31]. For example, sales of ultra-processed food and drink products increased by $8.3\%$ between 2009 and 2014, with these trends projected to continue to grow [23]. In addition, the consumption of ultra-processed products grew by more than $25\%$ between 2000 and 2013, and fast-food consumption increased by almost $40\%$ over the same period. In Latin America, most ultra-processed products are increasingly sold in convenience stores, supermarkets, and hypermarkets [6].
## 3.1.2. Adequate services
Some countries in the region still have a long way to go to achieve adequate WASH services for all, especially Bolivia and Haiti. While in 2015, $95\%$ of the region’s population used an improved drinking water source, the gap between the poor and the better-off, and between rural and urban areas, remains wide [32,33].
The significance of inappropriate WASH services for children’s diets was highlighted by a qualitative study of 16 low- and middle-income countries (including Bolivia, Colombia, Guatemala, Haiti and Mexico), which examined the relationship between water security and infant feeding [34]. Water insecurity was associated with poor quality and quantity of drinking water available for young children. Moreover, caregivers echoed that unsafe water meant that they could not prepare complementary foods as safely and hygienically as they would have wished to. Some of them substituted preferred complementary foods with other, less preferred ones to reduce the risk of contamination. Some respondents also reported delaying infant feeding until clean water is available. Respondents also observed a link between poor water quality and infant morbidity (such as infectious diseases, diarrhea, stomach and skin irritation, undernutrition and dehydration).
Overall, $34\%$ of all households in the region are covered by social assistance programmes, including $60\%$ of all households in the bottom quintile of the income distribution [35]. The LAC region has been a global leader in demonstrating the potential of cash transfer programmes in supporting the health and nutritional needs of young children, with different studies identifying 15 to 30 national conditional cash transfers (CCTs) in $\frac{2015}{16}$ [36,37]. In 2016, $16.9\%$ of all households (a total of nearly 30 million families) in the LAC region participated in at least one programme [36]. A review of three CCTs in Brazil, Colombia and Mexico found consistent evidence that these programmes had a positive effect on child nutrition [38].
In terms of health services, a regional study showed that out of 15 countries in LAC, 13 have growth monitoring policies in place [39], but it is unclear at a regional level how many children comply with monitoring visits. Moreover, Community Health Workers (CHWs) can play a key role in nutrition service delivery through the health system. For example, a case study in Haiti showed that CHWs performed 24 of the recommended 38 nutrition services [40]. Yet, little information is available across the region on the numbers of CHWs, and their responsibilities [41].
## 3.1.3. Adequate feeding practices
Indigenous populations are often more at risk of malnutrition and poor health outcomes [42,43]. A study of infant feeding practices in the Peruvian Amazon found that traditional complementary foods in the Ajawún community largely met WHO recommendations, and the foods consumed were more nutrient dense and higher quality than marketed foods. However, the intake of some micronutrients (mostly zinc, iron and Vitamin A) was below the adequate level [44]. Another study, also conducted in the Peruvian Amazon, concluded that the period of exclusive breastfeeding in the community was too short and that meal frequency and diversity did not meet standards [45].
Caregivers’ perceptions and preferences for complementary foods may diverge from those of nutritionists’ and policymakers [46]. Researchers in Mexico found that mothers working outside the home value the convenience, flavor and vitamin content of marketed and/or processed complementary foods but are also concerned about their sugar and chemical contents. The shift towards formula feeding is also related to their convenience, especially in the absence of daycare, maternity leave and flexible work policies [30].
## 3.1.4. Existing strategic actions to improve children’s diets
In LAC, the supplementation of food with vital micronutrients, such as vitamin A, iron, zinc and others is widespread [47]. Many countries also provide universal fortification of staple foods, for example iodine in salt, vitamin A in sugar, and varying fortifications including iron, zinc, vitamin B12, folic acid and more in wheat and maize flour and rice [47]. Twelve countries target households with infants 6–23 months with multiple micronutrients powders so that complementary foods can be fortified by caregivers themselves [47].
Several countries have adopted regulatory strategies to reduce the accessibility, availability and desirability of sugar-sweetened beverages and unhealthy or ultra-processed foods. These strategies include ‘sugar taxes’, advertising restrictions and requirements for transparency in food labeling and facilitating purchasing choices by caregivers. An overview of Latin American countries found 39 such regulatory strategies aimed at reducing the consumption of unhealthy foods and beverages, with Chile, Ecuador and Mexico having made the most comprehensive efforts [48].
Nutrition actions are most commonly delivered through food security and nutrition policies, but some countries have incorporated nutrition goals in other sectors such as agriculture, education, environment, development and employment policies [39]. A review of nutrition policies showed that most of the 18 countries reviewed have one or more nutrition-related and/or sectoral policy in line with the WHO recommendations for Comprehensive Implementation Plan on Maternal, Infant and Young Child Nutrition [39]. Interdisciplinary evidence on what types of national strategies are the most effective, however, are lacking [48].
## 3.2. A view from the experts
Interviews with key informants provided further insights into the determinants and drivers of infant and young child nutrition in four countries: Guatemala, Paraguay, Peru and Uruguay. The affiliation of key informants per country is provided in Table 2. Names and positions of the informants are excluded to protect their identity. Key findings from the interviews are summarized in Table 3 and described below.
Child undernutrition remains a widespread issue in Guatemala: the country has the highest rate of stunting in the LAC region, affecting nearly half of all children under five, and almost half of all children under six suffer from anemia. There appears to be inequality of access to healthy diets between different groups of the population, evidenced by the concentration of nutritional deficiencies among children in low-income households, in rural areas and from indigenous groups. In terms of adequate food, low dietary diversity is one of the main drivers of nutritional deficiencies, according to key informants, driven by a lack of resources, limited caregiver awareness and lack of access to iron-rich foods Adequate services are limited by unequal access to quality health care due to geographical barriers and traditional gender roles. In addition, funding and management of WASH services is hampered by structural issues such as financing and administrating the technical services. Moreover, social protection programmes currently lack nutrition sensitive goals.
Adequate practices in Guatemala are constrained by cultural beliefs, for example on the use of chlorine to treat water. In addition, traditional gender roles around hygiene persist and men do not follow standard hygiene practices.
Despite a strong policy framework, in the form of the recently established Gran Cruzada Nacional por la Nutrición, there is room for improved inter-institutional collaboration and continuity over time in governments’ actions, according to the key informants.
In Paraguay, overweight among young children is a more pressing public health problem than undernutrition. Only three out of ten children under six months of age are exclusively breastfed and children’s diets are low in diversity. Major barriers to adequate food include poverty, the price of healthy products versus processed foods and the high availability of ultra-processed foods (Table 3).
In terms of adequate services, key informants noted barriers to access healthcare services, including geographic distance as well as a lack of trained health staff. Moreover, social protection programmes, like PANI and Tekopora only provide limited coverage, due to budgetary restrictions. The country also exhibits inequalities in access to WASH services. While $95\%$ of households in Paraguay have access to an improved source of drinking water, and $83\%$ has access to an improved sanitation facility, there are large inequalities by area of residence, income level and for indigenous groups [49]. There is also limited support in the country regarding infant and young child feeding according to key informants, resulting in poor caregivers’ knowledge and practices about how to feed young children.
Furthermore, key informants noted that adequate practices in Paraguay are limited by a food culture that is high in fat and carbohydrates, the (too) early introduction of complementary foods and early introduction of sweet beverages, like juices, at very early ages.
There is also limited attention to the issue of overweight and accompanying policy measures on the political agenda, for example by regulating ultra-processed foods, according to key informants. Experts also noted the need for strong coordination between different State and non-State actors, as the issue of child nutrition extends across sectors: Peru made significant progress to reduce stunting, and it outperforms the other countries in the region when it comes to complementary feeding practices. Yet, overweight and micronutrient deficiencies, notably anemia, continue to be public health problems. Respondents stated the low levels of caregiver’s knowledge on appropriate feeding practices, insufficiently trained health workers and ineffective coordination mechanisms between relevant line ministries and other stakeholders as the key bottlenecks for adequate children’s diets. Furthermore, animal-sourced foods tend to be expensive, especially in rural areas, and pharmaceutical companies heavily promote the use of breastmilk substitutes. Iron-rich, processed foods are also promoted in response to concerns over anemia, yet this compromises resources for fresh and healthy foods.
In terms of health services, key informants also expressed concerns about the inadequate flow of budgetary resources and the lack of coordination to support regional directorates of health. In addition, according to key informants, health sector personnel have a ‘vertical relationship’ with clients, giving orders rather than providing tailored recommendations for complementary feeding practices. The social protection landscape is characterized by poor coverage in urban areas, while access to WASH, like potable water and sanitation, is more restricted in rural areas.
Besides lack of nutritional knowledge among caregivers, convenience to purchase processed food was cited as a barrier for adequate practices. In urban areas, many mothers work in informal jobs, creating a barrier for exclusive breastfeeding and adequate complementary feeding. Due to convenience and lack of time to prepare meals, use of processed foods is high among working caregivers in urban areas.
Experts expressed that Peru has a strong legislative and policy framework, in part due to the progress made when the nutrition agenda was under the prime minister’s responsibility.
A key emerging theme was the need for integrated and cross-sectoral collaboration, as informants recognized that child nutrition is a complex issue reaching across sectors. In the words of one informant: Uruguay’s malnutrition rates are in line with the regional average, but the country lacks recent data on complementary feeding practices and micronutrient deficiencies at the time of the study. Key barriers for healthy diets include monetary poverty, the high intake of ultra-processed foods, and the lack of credibility of health workers. Another barrier in the health sector is outdated health professionals’ knowledge on infant and young child feeding. According to key informants, specialists often provide conflicting advice, together with a perceived lack of empathy from their side, which undermines their credibility. As for social protection services, including Uruguay Crece Contigo, they are often limited in terms of coverage. According to the key informants, while UCC targets the entire population of pregnant women and children under four years, there is more demand than the programme can meet.
Adequate feeding practices are further limited by a high consumption of ultra-processed foods, especially among children, driven by the strong influence of marketing. In addition, certain beliefs and social representations around food create barriers. For example, the introduction of meat is believed to cause choking, which leads some parents to introduce it at a later age than appropriate. Finally, experts stated that families have a limited amount of time for food preparation and to devote for eating.
Overall, Uruguay’s policy landscape to support progress in terms of healthy diets is adequate, according to key informants. However, there is room for the expansion of integrated social programmes and a stronger policy response to unhealthy eating habits.
Experts expressed that regular data collection and monitoring of complementary feeding indicators should be prioritized to monitor progress. In addition, there needs to be a better vision on how the food system can support health diets.
## 4. Discussion
This regional analysis sought to identify determinants and drivers of young children’s diets in LAC. It applied the UNICEF Action Framework for improving young children’s diets to examine determinants and drivers in terms of adequate food, adequate services and adequate practices. The study also provided a detailed look at nutritional indicators across the region. In addition, the analysis identified strategic actions to improve children’s diets through the food system, health system, WASH system and social protection system.
The descriptive analysis showed that the region performs rather well in child undernutrition prevalence, but the dietary transition of the past decades has brought about new challenges. Wasting and stunting have reached low aggregate levels, but sub-regional differences and notable exceptions exist. In addition, disparities within countries exist by residence and wealth status, with those in rural areas and in the poorest quintile in general having poorer nutritional outcomes and complementary feeding practices. Despite progress in stunting reduction, undernutrition (stunting and wasting) and micronutrient deficiencies of young children remain a concern. Moreover, overweight and its severe form obesity are becoming a regional challenge. This triple burden of malnutrition (the co-existence of undernutrition, micronutrient deficiencies and overweight) is perhaps the most pressing issue for countries in LAC to tackle, in order to realize good nutrition and health for all children.
Complementary feeding indicators paint a mixed picture. In countries with low rates of exclusive breastfeeding, too early introduction of complementary foods or beverages is likely to be a concern, particularly if those foods or beverages are high in sugar, sodium or unhealthy fats. Many children are consuming solid foods during the 6-8-month window, but many are introduced to them too early as demonstrated by varying rates of exclusive breastfeeding in the first six months of life. Too early introduction of complementary foods puts the child at risk of early weaning and compromises the essential nutrients from breastmilk. Indeed, nearly half ($48\%$) of all children aged 4–5 months in LAC received solid foods, suggesting that too early introduction of complementary food is common [50]. Most infants are fed an appropriate number of times a day, but their diets are lacking in diversity. Overall, the concern is the quality rather than quantity of diets.
The low quality of diets is linked to the availability and convenience of packaged, ultra-processed, and sugary foods and drinks. According to key informants, healthy food is often more expensive and more difficult to access than pre-packaged and processed products, especially in urban areas. Access is compounded by a lack of economic means. Research shows that such foods are associated with overweight/obesity and many nutrition-related noncommunicable diseases [6,51]. Evidence from Brazil, Mexico, and Peru suggests that infants’ diets are rich in sugar-sweetened beverages and other unhealthy products [44,52,53]. Intense lobbying of the food industry may prevent efficient regulation to prevent high intakes of such ultra-processed foods. The quality of diets also varies by the purchasing power of the household and the area of residence. Infants in poor and rural households are fed less animal source foods than their peers in better-off and urban families. However, there are considerable knowledge gaps about how changing dietary patterns in Latin American societies are affecting what young children are fed. There is also a need to better understand potential disparities of infant feeding among indigenous and non-indigenous populations. In nearly all case study countries, key informants indicated that children from indigenous communities are generally worse off in terms of dietary quality and their households have difficulties in accessing a diverse diet. Local customs, beliefs and traditions also pose a key barrier for healthy diets.
Services in the dimensions of social protection, health care and WASH can and do contribute to better nutrition for children, but further efforts in these sectors are needed. For example, there is a wide gap in access to safe water and sanitation between urban and rural households and among different countries [32], with Bolivia and Haiti having particularly worrying conditions. According to key informants, households can easily access health services, although geographic distance remains a barrier, especially for those in rural and remote areas. Another barrier in the health system appears to be the lack of empathy or the vertical relation between health care personnel and parents. Due to these issues, parents may have less trust in the formal health system and turn to relatives or other acquaintances for advice and guidance on feeding practices. The role of health services will become more important as LAC embarks on health systems reforms to achieve universal health coverage. The region is currently characterized by large inequalities in access to quality health services [54], and health systems in the region need to strengthen their ties with other sectors to act on the social determinants of health and the risk factors associated with non-communicable diseases, including undernutrition and overweight [55]. The main bottleneck for social protection services is coverage. Most programmes are targeted to rural populations, while urbanization and urban poverty are increasing. For example, the coverage rates for all types of social assistance programmes among the poorest $20\%$ rural households was $72\%$, $86\%$ $93\%$ and $87\%$ in Guatemala, Paraguay, Peru and Uruguay respectively. For urban households in the bottom quintile, these rates were $63\%$, $85\%$, $80\%$ and $90\%$ in the respective countries. Coverage gaps are wider for the total population [35].
Large-scale national policies to reduce micronutrient deficiencies and underweight are implemented in the majority of LAC countries, with supplementation and staple food fortification programmes being the most popular. Despite national implementation, the effectiveness of these programmes depends on the coverage level, which in many countries leaves room for improvement. In addition, several countries in the region could benefit from fortifying oil with vitamin A as a new strategy, including Argentina, Brazil, Dominican Republic, Ecuador, Guatemala, Honduras, Jamaica, Mexico, Paraguay and Suriname [56,57].
New efforts are emerging to reduce the triple burden of malnutrition, for instance in the form of sugar taxes, marketing restrictions and labeling requirements. However, inter-sectoral and comprehensive action against overweight and obesity remains a rarity, and more evidence is needed about the effectiveness of individual policies as well as cross-cutting strategies. This was echoed in the qualitative interviews, as key informants called for more integrated and cross-sectoral programming to address nutritional problems. A key theme was the issue of the level of coordination, as for example in Peru, major advancements were made when the authority for the nutrition agenda was at the prime minister’s level, rather than at the level of a line ministry. Perhaps the most pressing gap is related to tackling overweight and obesity. Since these are emerging health challenges in the region, recent reviews found relevant strategic action to be weak or uncoordinated [37,39]. While the initiatives to reduce the desirability of unhealthy foods is a good start, more comprehensive, intersectoral action may be necessary to overcome the triple burden of malnutrition.
The knowledge, attitudes and practices of caregivers vary, and programmes aimed at improving infant feeding practices should consider the local preferences, beliefs, and socio-cultural contexts, as well as the interactions between service providers and caregivers. In the qualitative interviews, several informants noted the lack of awareness on adequate feeding practices among caregivers. They called for additional programming to counsel caregivers in complementary feeding practices, as well as improve the interaction between service providers and caregivers. A notable example is the development and implementation of the food guides in Paraguay, which are used to demonstrate the composition of a healthy diet for children under two years. Understanding the strengths, opportunities as well as the challenges of different population groups’ infant feeding preferences may prove useful in designing infant and young child feeding promotion strategies that caregivers can comply with. Undertaking formative research in the community before implementation was found to be a cornerstone of culturally relevant nutrition education programmes in Peru [58]. Such an iterative process in programme design was also crucial to the strengthening and successful scale-up of the Integrated Strategy for Attention to Nutrition element of Mexico’s flagship CCT programme [59].
This study has several limitations. The literature review was not a systematic review and hence, some relevant studies could have been overlooked. However, the included studies paint a comprehensive picture for the LAC region and for some countries in particular. The triangulation with key informants for selected countries further strengthened the completeness of the information. As for the qualitative part of the study, the number of key informants per country was limited, which may have led to relatively narrow views. The relatively small sample of four countries may prevent major generalizations from these findings, although care was taken to select countries with varying nutritional challenges and socio-economic conditions. In addition, informants were purposefully selected through local UNICEF offices, which may have also influenced the results. Nevertheless, the interviewers were independent researchers and key themes per country were triangulated between interviews to draw out recurrent insights.
Overall, the regional analysis showed that children’s diets in LAC are shaped by several important drivers. Most notably, poverty and inequality, unequal access to services, inadequate coverage of social programmes and lack of awareness on appropriate feeding practices. In addition, disparities within countries by wealth and residence are important factors that influence children’s diets. Countries in the region need to invest in policies to tackle all forms of malnutrition in young children, considering inequalities between and within countries, enhance coverage of social protection programmes, improve coordination between sectors to improve children’s diets and expand coverage and intensity of awareness campaigns on feeding practices, using iterative programme designs.
## References
1. Dewey KG, Begum K. **Long-term consequences of stunting in early life.**. *Maternal & child nutrition.* (2011.0) **7** 5-18. DOI: 10.1111/j.1740-8709.2011.00349.x
2. 2UNICEF, WHO, World Bank. Joint child malnutrition estimates, 2021 edition [Internet]. 2021 [cited 2020 Feb 1]. Available from: https://www.who.int/news/item/06-05-2021-the-unicef-who-wb-joint-child-malnutrition-estimates-group-released-new-data-for-2021
3. 3UNICEF. State of the World’s Children 2019. New York: UNICEF; 2019.. *State of the World’s Children 2019* (2019.0)
4. White JM, Bégin F, Kumapley R, Murray C, Krasevec J. **Complementary feeding practices: Current global and regional estimates.**. *Matern Child Nutr.* (2017.0) **13** e12505. DOI: 10.1111/mcn.12505
5. 5Food and Agriculture Organization of the United Nations/Pan American Health, Pan American Health Organization. Panorama of food and nutritional security in Latin America and the Caribbean 2017. Santiago, Chile; 2017.
6. 6Pan American Health Organization. Ultra-processed food and drink products in Latin America: Trends, impact on obesity, policy implications.
Washington, D.C.: PAHO; 2015.. *Ultra-processed food and drink products in Latin America: Trends, impact on obesity, policy implications.* (2015.0)
7. Popkin BM, Reardon T. **Obesity and the food system transformation in Latin America.**. *Obesity Reviews.* (2018.0) **19** 1028-64. DOI: 10.1111/obr.12694
8. 8Regional progress towards the SDG targets [Internet]. [cited 2022 Jan 28]. Available from: https://agenda2030lac.org/estadisticas/regional-progress-sdg-targets.html.
9. 9UNICEF. Infant and young child feeding database [Internet]. Infant and young child feeding database. 2019 [cited 2020 Feb 1]. Available from: https://data.unicef.org/topic/nutrition/infant-and-young-child-feeding/.
10. 10UNICEF. Action framework on improving young children’s diets. New York: UNICEF; 2020.. *Action framework on improving young children’s diets* (2020.0)
11. 11UNICEF. Improving Young Children’s Diets During the Complementary Feeding Period.
UNICEF Programming Guidance. [Internet]. New York: UNICEF; 2020 [cited 2022 Jun 29]. Available from: https://www.unscn.org/en/news-events/recent-news?idnews=2030.. *Improving Young Children’s Diets During the Complementary Feeding Period.* (2020.0)
12. Kuhnt J, Vollmer S. **Antenatal care services and its implications for vital and health outcomes of children: evidence from 193 surveys in 69 low-income and middle-income countries**. *BMJ Open* (2017.0) **7**. DOI: 10.1136/bmjopen-2017-017122
13. Oakley L, Benova L, Macleod D, Lynch CA, Campbell OMR. **Early breastfeeding practices: Descriptive analysis of recent Demographic and Health Surveys.**. *Matern Child Nutr* (2018.0) **14** e12535. DOI: 10.1111/mcn.12535
14. Ashworth A, Shrimpton R, Jamil K. **Growth monitoring and promotion: review of evidence of impact.**. *Matern Child Nutr.* (2008.0) **4** 86-117. DOI: 10.1111/j.1740-8709.2007.00125.x
15. 15World Health Organization, UNICEF, USAID. Improving nutrition outcomes with better water, sanitation and hygiene: Practical solutions for policy and programmes [Internet]. Geneva: World Health Organization; 2015 [cited 2020 Feb 20]. Available from: http://www.who.int/water_sanitation_health/publications/washandnutrition/en/.. *Improving nutrition outcomes with better water, sanitation and hygiene: Practical solutions for policy and programmes* (2015.0)
16. 16Owens J. Social Protection [Internet]. Rome: Food and Agriculture Organization of the United Nations/Pan American Health …; 2017 [cited 2020 Feb 20]. (Strengthening sector policies for better food security and nutrition results). Report No.: 4. Available from: http://www.fao.org/publications/card/en/c/fbb74efa-2601-4fd2-999a-d521767a51de.
17. de Groot R, Palermo T, Handa S, Ragno LP, Peterman A. **Cash transfers and child nutrition: pathways and impacts**. *Development Policy Review* (2017.0) **35** 621-43. DOI: 10.1111/dpr.12255
18. 18KNAW, NFU, NWO, TO2-Federatie, Vereniging Hogescholen, VSNU. Nederlandse gedragscode wetenschappelijke integriteit [Internet]. Data Archiving and Networked Services (DANS); 2018 [cited 2022 Feb 14]. Available from: https://easy.dans.knaw.nl/ui/datasets/id/easy-dataset:110600.
19. 19United Nations Children’s Fund, Division of Data, Analysis, Planning and Monitoring. UNICEF Global Databases: Infant and Young Child Feeding: Exclusive breastfeeding, Introduction of solid, semi-solid or soft foods, Minimum dietary diversity, Minimum meal frequency, Minimum acceptable diet.
New York: UNICEF; 2021.. *UNICEF Global Databases: Infant and Young Child Feeding: Exclusive breastfeeding, Introduction of solid, semi-solid or soft foods, Minimum dietary diversity, Minimum meal frequency, Minimum acceptable diet.* (2021.0)
20. 20World Health Organization, UNICEF. Indicators for assessing infant and young child feeding practices: definitions and measurement methods.
Geneva: World Health Organization and the United Nations Children’s Fund (UNICEF); 2021.. *Indicators for assessing infant and young child feeding practices: definitions and measurement methods.* (2021.0)
21. 21World Bank. World Bank Country and Lending Groups [Internet]. 2021 [cited 2022 Jan 18]. Available from: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups.
22. Elo S, Kyngäs H. **The qualitative content analysis process**. *Journal of advanced nursing* (2008.0) **62** 107-15. DOI: 10.1111/j.1365-2648.2007.04569.x
23. 23Food and Agriculture Organization of the United Nations/Pan American Health. Panorama of food and nutritional security in Latin America and the Caribbean 2019. Santiago, Chile; 2019.
24. Leporati M, Salcedo S, Jara B, Boero V, Muñoz M, Salcedo S, Guzmán L. *La agricultura familiar en cifras.* (2014.0)
25. 25World Food Programme. Counting the Beans—the true cost of a plate of food around the world.
Rome: World Food Programme; 2017.. *Counting the Beans—the true cost of a plate of food around the world.* (2017.0)
26. Headey DD, Alderman HH. **The Relative Caloric Prices of Healthy and Unhealthy Foods Differ Systematically across Income Levels and Continents**. *J Nutr* (2019.0) **149** 2020-33. DOI: 10.1093/jn/nxz158
27. 27World Food Programme. Fill the Nutrient Gap—El Salvador. Rome: World Food Programme; 2017.. *Fill the Nutrient Gap—El Salvador* (2017.0)
28. Dewey KG. **The Challenge of Meeting Nutrient Needs of Infants and Young Children during the Period of Complementary Feeding: An Evolutionary Perspective**. *The Journal of Nutrition* (2013.0) **143** 2050-4. DOI: 10.3945/jn.113.182527
29. Allen LH. **Micronutrient Research, Programs, and Policy: From Meta-analyses to Metabolomics.**. *Advances in Nutrition.* (2014.0) **5** 344S-351S. DOI: 10.3945/an.113.005421
30. Baker P, Smith J, Salmon L, Friel S, Kent G, Iellamo A. **Global trends and patterns of commercial milk-based formula sales: is an unprecedented infant and young child feeding transition underway?**. *Public Health Nutr.* (2016.0) **19** 2540-50. DOI: 10.1017/S1368980016001117
31. 31Pan American Health Organization. Ultra-processed food and drink products in Latin America: Trends, impact on obesity, policy implications.
Washington, DC: PAHO; 2015.. *Ultra-processed food and drink products in Latin America: Trends, impact on obesity, policy implications.* (2015.0)
32. 32WHO & UNICEF. Inequalities in sanitation and drinking water in Latin America and the Caribbean [Internet]. Panama: WHO and UNICEF; 2016. Available from: https://washdata.org/report/lac-snapshot-wash-2016-en.. *Inequalities in sanitation and drinking water in Latin America and the Caribbean* (2016.0)
33. Mújica OJ, Haeberer M, Teague J, Santos-Burgoa C, Galvão LAC. **Health inequalities by gradients of access to water and sanitation between countries in the Americas, 1990 and 2010.**. *Revista Panamericana de Salud Pública* (2015.0) **38** 347-54. PMID: 26837519
34. Schuster RC, Butler MS, Wutich A, Miller JD, Young SL. **“If there is no water, we cannot feed our children”: The far‐reaching consequences of water insecurity on infant feeding practices and infant health across 16 low‐ and middle‐income countries.**. *Am J Hum Biol* (2020.0) **32**. PMID: 31868269
35. 35World Bank. The Atlas of Social Protection: Indicators of Resilience and Equity [Internet]. World Bank; 2020 [cited 2021 Mar 11]. Available from: https://databank.worldbank.org/source/1229#.. *The Atlas of Social Protection: Indicators of Resilience and Equity* (2020.0)
36. Cecchini S, Atuesta B. **Conditional Cash Transfer Programmes in Latin America and the Caribbean: Coverage and Investment Trends.**. *SSRN Journal* (2017.0)
37. Galicia L, de Romana D, Hardling K, De-Regil R, Grajeda R. **Tackling malnutrition in Latin America and the Caribbean: challenges and opportunities.**. *Rev Panam Salud Publica.* (2016.0) **40** 138-46. PMID: 27982371
38. Segura-Pérez S, Grajeda R, Pérez-Escamilla R. **Conditional cash transfer programs and the health and nutrition of Latin American children.**. *Rev Panam Salud Publica* (2016.0) **40** 124-37. PMID: 27982370
39. Tirado MC, Galicia L, Husby HM, Lopez J, Olamendi S, Pia Chaparro M. **Mapping of nutrition and sectoral policies addressing malnutrition in Latin America.**. *Revista Panamericana de Salud Pública* (2016.0) **40** 114-23. PMID: 27982369
40. SPRING APC. *How Do Community Health Workers Contribute to Better Nutrition? Haiti.* (2016.0)
41. Perry HB. *Health for the People: National Community Health Worker Programs from Afghanistan to Zimbabwe* (2020.0)
42. Anderson I, Robson B, Connolly M, Al-Yaman F, Bjertness E, King A. **Indigenous and tribal peoples’ health (The Lancet–Lowitja Institute Global Collaboration): a population study.**. *The Lancet* (2016.0) **388** 131-57. DOI: 10.1016/S0140-6736(16)00345-7
43. 43High Level Panel of Experts,. HLPE Report # 12—Nutrition and food systems.
Rome: HLPE; 2017 p. 152.. *HLPE Report # 12—Nutrition and food systems.* (2017.0) 152
44. Roche ML, Creed‐Kanashiro HM, Tuesta I, Kuhnlein HV. **Infant and young child feeding in the Peruvian Amazon: the need to promote exclusive breastfeeding and nutrient-dense traditional complementary foods.**. *Maternal & Child Nutrition.* (2011.0) **7** 284-94. DOI: 10.1111/j.1740-8709.2009.00234.x
45. Lee G, Paredes Olortegui M, Rengifo Pinedo S, Ambikapathi R, Peñataro Yori P, Kosek M. **Infant feeding practices in the Peruvian Amazon: implications for programs to improve feeding.**. *Rev Panam Salud Publica.* (2014.0) **36** 150-7. PMID: 25418764
46. Rodriguez-Oliveros MG, Bisogni CA, Frongillo EA. **Knowledge about food classification systems and value attributes provides insight for understanding complementary food choices in Mexican working mothers**. *Appetite* (2014.0) **83** 144-52. DOI: 10.1016/j.appet.2014.08.022
47. Lopez de Romana D., Cediel G.. *Current situation of micronutrients in Latin America: Prevalence of deficienciesand national micronutrient delivery programs* (2016.0)
48. Bergallo P, Castagnari V, Fernández A, Mejía R, Romer D. **Regulatory initiatives to reduce sugar-sweetened beverages (SSBs) in Latin America.**. *PLoS ONE.* (2018.0) **13** e0205694. DOI: 10.1371/journal.pone.0205694
49. 49Republic of Paraguay, UNICEF. Paraguay Multiple Indicator Cluster Survey 2016 [Internet]. Asuncion: Republic of Paraguay; 2016 [cited 2021 Mar 1]. Available from: https://mics-surveys-prod.s3.amazonaws.com/MICS5/Latin%20America%20and%20Caribbean/Paraguay/2016/Final/Paraguay%202016%20MICS_Spanish.pdf
50. 50UNICEF. From the first hour of life: Making the case for improved infant and young child feeding everywhere. New York: UNICEF; 2016.. *From the first hour of life: Making the case for improved infant and young child feeding everywhere* (2016.0)
51. Popkin B.. *Ultra-processed foods’ impacts on health* (2019.0)
52. Karnopp EVN, Vaz J dos S, Schafer AA, Muniz LC. **Food consumption of children younger than 6 years according to the degree of food processing.**. *Jornal de Pediatria.* (2017.0) **93** 70-8. DOI: 10.1016/j.jped.2016.04.007
53. Rodríguez-Ramírez S, Muñoz-Espinosa A, Rivera JA, González-Castell D, González de Cosío T. **Mexican Children under 2 Years of Age Consume Food Groups High in Energy and Low in Micronutrients**. *J Nutr* (2016.0) **146** 1916S-1923S. DOI: 10.3945/jn.115.220145
54. Frenk J.. **Leading the way towards universal health coverage: a call to action**. *The Lancet* (2015.0) **385** 1352-8. DOI: 10.1016/S0140-6736(14)61467-7
55. de Andrade LOM, Filho AP, Solar O, Rígoli F, de Salazar LM, Serrate PCF. **Social determinants of health, universal health coverage, and sustainable development: case studies from Latin American countries**. *The Lancet* (2015.0) **385** 1343-51
56. Mkambula P, Mbuya MNN, Rowe LA, Sablah M, Friesen VM, Chadha M. **The Unfinished Agenda for Food Fortification in Low- and Middle-Income Countries: Quantifying Progress, Gaps and Potential Opportunities.**. *Nutrients* (2020.0) **12** 354. DOI: 10.3390/nu12020354
57. Aaron GJ, Friesen VM, Jungjohann S, Garrett GS, Neufeld LM, Myatt M. **Coverage of Large-Scale Food Fortification of Edible Oil, Wheat Flour, and Maize Flour Varies Greatly by Vehicle and Country but Is Consistently Lower among the Most Vulnerable: Results from Coverage Surveys in 8 Countries**. *The Journal of Nutrition* (2017.0) **147** 984S-994S. DOI: 10.3945/jn.116.245753
58. Robert RC, Creed-Kanashiro HM, Villasante R, Narro MR, Penny ME. **Strengthening health services to deliver nutrition education to promote complementary feeding and healthy growth of infants and young children: formative research for a successful intervention in peri-urban Trujillo, Peru.**. *Matern Child Nutr.* (2017.0) **13**. DOI: 10.1111/mcn.12264
59. Bonvecchio Arenas A, González W, Théodore FL, Lozada-Tequeanes AL, Garcia-Guerra A, Alvarado R. **Translating Evidence-Based Program Recommendations into Action: The Design, Testing, and Scaling Up of the Behavior Change Strategy EsIAN in Mexico**. *J Nutr* (2019.0) **149** 2310S-2322S. PMID: 31793647
|
---
title: Maternal hyperglycemia inhibits pulmonary vasculogenesis during mouse fetal
lung development by promoting GβL Ubiquitination-dependent mammalian target of Rapamycin
assembly
authors:
- Qingqing Luo
- Xinqun Chai
- Xiaoyan Xin
- Weixiang Ouyang
- Feitao Deng
journal: Diabetology & Metabolic Syndrome
year: 2023
pmcid: PMC10021989
doi: 10.1186/s13098-022-00974-y
license: CC BY 4.0
---
# Maternal hyperglycemia inhibits pulmonary vasculogenesis during mouse fetal lung development by promoting GβL Ubiquitination-dependent mammalian target of Rapamycin assembly
## Abstract
### Background
Gestational diabetes mellitus (GDM) is associated with retarded lung development and poor lung health in offspring. Mammalian target of rapamycin (mTOR) is a key regulator of vasculogenesis and angiogenesis. The aim of this study was to investigate the role mTOR plays in pulmonary vasculogenesis during fetal lung development under maternal hyperglycemia.
### Methods
First, GDM was induced via streptozotocin injection in pregnant C57BL/6 mice before the radial alveolar count (RAC) in the fetal lungs was assessed using hematoxylin and eosin staining. The angiogenic ability of the cultured primary mouse fetal lung endothelial cells (MFLECs) was then assessed using the tube formation assay technique, while western blot and real-time polymerase chain reaction were performed to determine the expression of mTOR, regulatory-associated protein of mTOR (Raptor), rapamycin-insensitive companion of mTOR (Rictor), stress-activated protein kinase interacting protein 1 (Sin1), G protein beta subunit-like protein (GβL), Akt, tumor necrosis receptor associated factor-2 (TRAF2), and OTU deubiquitinase 7B (OTUD7B) in both the fetal lung tissues and the cultured MFLECs. Immunoprecipitation assays were conducted to evaluate the status of GβL-ubiquitination and the association between GβL and mTOR, Raptor, Rictor, and Sin1 in the cultured MFLECs.
### Results
The GDM fetal lungs exhibited a decreased RAC and reduced expression of von Willebrand factor, CD31, and microvessel density. The high glucose level reduced the tube formation ability in the MFLECs, with the mTOR, p-mTOR, p-Raptor, and TRAF2 expression upregulated and the p-Rictor, p-Sin1, p-Akt, and OTUD7B expression downregulated in both the GDM fetal lungs and the high-glucose-treated MFLECs. Meanwhile, GβL-ubiquitination was upregulated in the high-glucose-treated MFLECs along with an increased GβL/Raptor association and decreased GβL/Rictor and GβL/Sin1 association. Furthermore, TRAF2 knockdown inhibited the high-glucose-induced GβL-ubiquitination and GβL/Raptor association and restored the tube formation ability of the MFLECs.
### Conclusion
Maternal hyperglycemia inhibits pulmonary vasculogenesis during fetal lung development by promoting GβL-ubiquitination-dependent mTORC1 assembly.
## Introduction
Gestational diabetes mellitus (GDM), defined as any degree of glucose intolerance with onset or first recognition during pregnancy, occurs in around $12\%$ of pregnant women. In fact, GDM is associated with various birth defects, especially in the cardiac and central nervous systems [1–4]. In addition, GDM is a risk factor for poor lung health in offspring, which is manifested as retarded lung development and/or an increased incidence of acute and chronic respiratory diseases [5]. Maternal hyperglycemia is the typical clinical manifestation of GDM and is used for the screening and diagnosis of the disease [6]. In a GDM rat model, maternal hyperglycemia led to the formation of smaller fetal lungs along with decreased pulmonary vascular density [7]. Pulmonary vasculogenesis and angiogenesis are fundamental components of fetal lung development [8]. Understanding the molecular mechanisms through which maternal hyperglycemia affects fetal pulmonary vasculogenesis may help with the development of strategies aimed at preventing GDM-associated pulmonary impairment in newborns.
Mammalian target of rapamycin (mTOR) is a key regulator of embryogenesis and its development, serving as the catalytic core of two distinct multi-protein complexes, mTORC1 and mTORC2. The former is composed of mTOR, regulatory-associated protein of mTOR (Raptor), G protein beta subunit-like protein (GβL), and DEP domain-containing mTOR-interacting protein, while the latter is composed of mTOR, rapamycin-insensitive companion of mTOR (Rictor), GβL, and mammalian stress-activated protein kinase interacting protein 1 (Sin1). The polyubiquitination of GβL by the tumor necrosis factor receptor-associated factor 2 (TRAF2) E3 ubiquitin ligase prevents its assembly into mTORC2, consequently favoring mTORC1 formation. In contrast, the removal of polyubiquitin chains from the GβL by the OTU deubiquitinase 7B (OTUD7B) facilitates mTORC2 formation [9]. It is known that mTOR plays an important regulatory role in vasculogenesis and angiogenesis [10, 11], with mTORC1 and mTORC2 reported to have different impacts on the regulation of these processes. Specifically, mTORC1 represents an anti-angiogenic property, while mTORC2 appears to present a pro-angiogenic function [12, 13]. However, the specific influence of mTORC1 and mTORC2 under maternal hyperglycemia on fetal lung vasculogenesis and angiogenesis remains unclear.
In this study, the effects of maternal hyperglycemia on pulmonary vasculogenesis during mouse fetal lung development are evaluated in vivo, while the effects of high glucose on mouse fetal lung endothelial cell (MFLEC) angiogenesis are evaluated in vitro. The process of TRAF2-mediated GβL-ubiquitination and the status of mTORC1 and mTORC2 assembly are also explored as possible mechanisms underlying these effects.
## Gestational diabetes mellitus mouse model
A GDM mouse model was established with reference to our previous report [14]. Here, C57BL/6 mice (eight weeks old) were purchased from the Hubei Provincial Center for Disease Control and Prevention (China). The temperature of the feeding environment was 25 °C, with alternative day and night environments set every 12 h. Each cage contained three mice, which were provided with adequate water and food, while the padding was replaced daily. The mice were randomly divided into the control group and the GDM group. For the purpose of breeding, the female mice were caged with the male mice overnight at a 1:2 ratio. The following day, samples of the vaginal secretion were collected using cotton swabs and smeared on slides before being examined under a microscope. If sperms were detected, the female mouse was recorded as undergoing the first day of pregnancy [15].
Twelve pregnant mice were divided into two groups: the control group ($$n = 6$$) and the GDM group ($$n = 6$$). The estimated food intake in control and GDM group was 4–5 g/24 h and 6-8 g/24 h, respectively. The daily water intake of control and GDM group was 5-6mL and 8-10mL, respectively. The mice in the GDM group were fasted overnight starting at 6 pm on the first day of gestation, while they had free access to water. The following morning, the mice received an intraperitoneal injection of streptozotocin (V900890, Sigma) ($2\%$ solution in 0.1 M, pH 4.4 citrate buffer) at a dose of 45 mg/kg, which was continued for three consecutive days. One hour after the injection, the mice were fed with chow. Meanwhile, the mice in the control group received similar treatment, except they were injected with the citrate buffer alone. Maternal blood samples were then collected from the tail vein to determine the maternal glucose levels on days 4, 10, 15, and 19 of gestation. All the mice in the GDM group exhibited a blood glucose level of > 11.1 mol/L, which was deemed to indicate the successful establishment of the GDM model [16]. Next, fetal lung tissues were collected on day 21 of gestation under general anesthesia (pentobarbital sodium, 50 mg/kg). Specifically, all of the lung tissues were removed after the mice were killed and were then divided into left and right lung tissue. The tissues were fixed with $4\%$ paraformaldehyde for at least 24 h. The study protocol was approved by the Ethics Committee of Union Hospital ([2012] IACUC Number: 607, Wuhan, China).
## Immunohistochemical and immunofluorescence staining
The lung tissue sections were de-paraffinized in xylene followed by heat-induced epitope retrieval in ethylenediaminetetraacetic acid (AS1016, ASPEN Biotechnology Co. Ltd., China). After incubation at room temperature with $3\%$ hydrogen peroxide in the dark for 10 min and then with $5\%$ bicinchoninic acid for 20 min, the sections were incubated with a corresponding primary antibody (1:100) at 4 ℃ overnight. After washing with phosphate buffered saline (PBS, AS1025, AspenTech), the sections were incubated with a horseradish peroxidase (HRP)- or fluorescein-labeled secondary antibody at 37 ℃ for 50 min. For immunohistochemical (IHC) detection, the sections were stained with 3, 3’-diaminobenzidine and then counterstained with hematoxylin [17], while for immunofluorescence (IF) detection, the sections were stained with DAPI [18]. The sections were then evaluated under a fluorescence microscope, with the data analyzed using the ImageJ image color analysis system. The primary and secondary antibodies used in the experiments were as follows: rabbit anti-von Willebrand factor (vWF) (1:200; 11778-1-AP, Proteintech Group Inc., China), rabbit anti-CD31 (1:200; ab28364, Abcam, USA), HRP-conjugated goat anti-rabbit IgG (1:200; AS-1107, AspenTech), and CY3-conjugated goat anti-rabbit IgG (1:200; AS-1109, AspenTech). The integrated optical density was calculated using the ImageJ software (National Institutes of Health, Bethesda, MD, USA).
## Radial alveolar count assessment
The fetal lung specimens were fixed in $10\%$ formalin before being paraffin-embedded and cut into 4-µm sections. After staining with hematoxylin and eosin (H&E), the radial alveolar count (RAC) was determined under a microscope as previously described in view of measuring the alveolar development and maturation of the fetal lungs [19]. Briefly, the number of alveoli was counted along the line drawn from the center of the respiratory tube to the nearest connective tissue septum.
## Microvascular density evaluation
The microvascular density (MVD) was measured based on the Weidner method [20]. In short, areas with the highest density of vessels were selected under a light microscope (×40), with the number of vessels in five different visual fields then counted under the light microscope (×400) to calculate the average, which was recorded as the MVD value.
## Isolation and culture of primary mouse fetal lung endothelial cells
The isolation and culture of primary MFLECs were performed with reference to a previous report [21]. On day 19 of gestation, the mice were anesthetized with 0.2 mL of $10\%$ chloral hydrate in $75\%$ alcohol for 5 min before the fetuses were collected via cesarean section under sterile conditions. The fresh fetal lungs were then harvested and washed in cold PBS, and the tracheal, connective, and other tissues that did not contain blood vessels were collected. The tissues were then cut into 1-mm3 pieces using iris scissors and treated with $0.25\%$ trypsin (GNM25200, Zhejiang Tianhang Biotechnology Co. Ltd., China) at 37 ℃ for 20 min in the presence of $0.01\%$ deoxyribonuclease I (10,104,159,001, Sigma). An equal volume of Dulbecco’s modified *Eagle medium* (DMEM/F12, SH30022.01, HyClone, USA) was added to stop the digestion. The resulting cell suspension was then filtered through a 200-mesh membrane and centrifuged for 5 min before the cell pellet was re-suspended in the DMEM/F12 medium containing $10\%$ fetal bovine serum (FBS; GNM20012, Zhejiang Tianhang Biotechnology Co. Ltd., China) and cultured at 37 ℃ in humidified air containing $5\%$ carbon dioxide (CO2). After approximately 45 min, any non-adherent cells were transferred into a new culture dish, with this process repeated three times. The purified non-adherent cells were maintained in a DMEM/F12–5 mM glucose medium (SH30022.01, HyClone, USA) containing $10\%$ FBS at 37℃ in humidified air containing $5\%$ CO2. The culture medium was changed every 24 h.
## Tube formation assay
The tube formation assay was performed as previously described with modification[22]. First, Matrigel (200 µL/well) (354,248, Corning, USA) was added into pre-cooled 96-well plates and incubated at 37 °C for 30 min. The primary MFLECs (2 × 105/mL, 100 µL) were then loaded onto the Matrigel in each well and cultured at 37 °C for 12 h. For high-glucose treatment, the cells were incubated using a DMEM/F12–30 mM glucose medium (SH30021.01, HyClone, USA) for 72 h before they were loaded onto the Matrigel. Images were captured immediately after loading and again following the 12-hour incubation. The tube formation was evaluated based on the number of tubes formed, the total number of branch points, and the total length of the tube branches.
## Western blot analysis
*The* general experimental procedure of Western blot analysis was described in our previous paper [14]. The protein concentrations were determined using the bicinchoninic acid (BCA) assay (P0012, Beyotime, China) after the mouse fetal lung tissues and cells were lysed. Here, the proteins (40 µg per sample) were separated on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (AS1012, AspenTech) and transferred to methanol-activated polyvinylidene difluoride membranes. After blocking in $5\%$ non-fat milk, the membranes were incubated with a corresponding primary antibody at 4 ℃ overnight. The membranes were then rinsed with Tween® 20 detergent (AS1100, AspenTech) and probed with an HRP-conjugated second antibody for 30 min at room temperature. The protein bands were visualized using an enhanced chemiluminescence reagent (AS1059, AspenTech) and quantified using the AlphaEaseFC image analysis software. The primary antibodies used in the experiments were as follows: rabbit anti-β-Actin (1:10000; TDY051, Beijing TDY Biotech Co., LTD., China), rabbit anti-TRAF2 (1:2000; #4712, Cell Signaling Technology, USA), rabbit anti-OTUD7B (1:1000; SAB2101695, Sigma-Aldrich, Germany), rabbit anti-p-mTOR(Ser2448) (1:1000; #2971, Cell Signaling Technology), rabbit anti-mTOR (1:1000; #2983, Cell Signaling Technology), rabbit anti-GβL (1:1000; #3274, Cell Signaling Technology), rabbit anti-p-Raptor(Ser863) (1:500; AF8506, Affinity Biosciences, USA), rabbit anti-Raptor (1:1000; #2280, Cell Signaling Technology), rabbit anti-p-Rictor(Ser1591) (1:500; AF7411, Affinity Biosciences), rabbit anti-Rictor (1:500; AF7530, Affinity Biosciences), rabbit anti-p-Sin1(Thr86) (1:500; AF2407, Affinity Biosciences), rabbit anti-Sin1 (1:1000; PRS4077, Sigma-Aldrich), rabbit anti-p-Akt(Ser473) (1:1000; #4060, Cell Signaling Technology), and rabbit anti-Akt (1:2000; #9272, Cell Signaling Technology). Meanwhile, the secondary antibody used in the experiments was a HRP-conjugated goat anti-rabbit antibody (1:10000; AS1107, AspenTech).
## Immunoprecipitation assay
The IP assay was performed using the SureBeads™ Starter Kit Protein A (1,614,813, BIO-RAD, USA) following the manufacturer’s instructions. In brief, the cells were lysed in an IP lysis buffer (AS1003, AspenTech) containing a protease inhibitor cocktail (AS1005C, AspenTech). The cell lysates were then incubated with protein A beads pre-bound with an anti-GβL antibody (1:500, sc-514,982, Santa Cruz Biotechnology, USA) at 4 °C overnight before the precipitated proteins were subjected to western blot analysis. The primary and secondary antibodies used in the western blotting procedure were as follows: rabbit anti-β-Actin (1:10000; TDY051, Beijing TDY Biotech Co., LTD.), rabbit anti-mTOR (1:1000; #2983, Cell Signaling Technology), rabbit anti-Raptor (1:1000; #2280, Cell Signaling Technology), rabbit anti-Rictor (1:500, DF7530, Affinity Biosciences), rabbit anti-Sin1 (1:1000; PRS4077, Sigma-Aldrich), rabbit anti-ubiquitin (1:500; ab134953, Abcam), mouse anti-GβL (1:500, sc-514,982, Santa Cruz Biotechnology), and HRP-conjugated goat anti-rabbit (1:10000; AS1107, AspenTech) and goat anti-mouse antibodies (1:10000, AspenTech).
## Small interfering ribonucleic acid transfection
To knockdown the TRAF2 gene, the cells were transfected with a TRAF2–small interfering ribonucleic acid (siRNA) solution (50 µM) or a scramble siRNA (si-NC, 50 µM) solution using the Lipofectamine 2000 reagent (12566014, Thermo Fisher Scientific, USA) for 12 h. Following this, the medium was changed, and the cells were cultured for a further 72 h. The cells were collected and then used in the subsequent experimental studies. The siRNA sequences were as follows: si-TRAF2-1 5′-CCTCAGGTGTGCATCCATT-3′, si-TRAF2-2 5′-GGAGACGTTTCAGGACCAT-3′, si-TRAF2-3, 5′-CCATAACAACCGGGAGCAT-3′, and si-NC 5′-CCTGTGGACGTCCTACATT-3′.
## Quantitative real-time polymerase chain reaction
The total RNA was extracted using the TRIpure Total RNA Extraction reagent (EP013, ELK Biotechnology, China). Complementary DNA (cDNA) was synthesized from 1 µg of RNA using the M-MLV Reverse Transcriptase kit (Eq. 002, ELK Biotechnology). Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using the StepOne™ Real-Time PCR System (Life Technologies, USA). The relative micro-RNA (mRNA) expression was calculated using the 2−ΔΔCt method and normalized to β-Actin. The primer information was listed in Table 1. A concentration of RNA larger than 20ng/ul was considered as qualified samples. The purity of RNA was evaluated spectrophotometrically by determining absorbance ratios of A260/A280. The A$\frac{260}{280}$ ratio ranged between 1.85 and 2.01.
Table 1Sequences of PrimersPrimerSequences(5′ − 3′)TmCG%Product sizeM-actinSenseCTGAGAGGGAAATCGTGCGT6055208AntisenseCCACAGGATTCCATACCCAAGA61.350M-TRAF2SenseGAAAGCGTCAGGAAGCCGTA60.555139AntisenseAGAGACAGATGAGTTCCCCGC60.557.1M-CEBPαSenseCCCCACTTGCAGTTCCAGAT59.655244AntisenseTGTCCACCGACTTCTTGGCT60.255M-CEBPβSenseCTACTACGAGCCCGACTGCC59.965120AntisenseAGGTAGGGGCTGAAGTCGATG60.757.1
## Statistical analysis
The data were expressed in terms of mean ± standard error of the mean, with the differences between the groups evaluated using a Student’s t-test along with SPSS19 software. Graphs were then created using the GraphPad Prism 5 Software Package (GraphPad, USA). Differences with a p-value of < 0.05 were considered to be statistically significant.
## Fetal pulmonary alveolar and vascular development in the gestational diabetes mellitus and control pregnant mice
As shown in Fig. 1A, the fetal lung tissue from the GDM group exhibited a reduced RAC compared with the control group (2.8 ± 0.84 vs. 6.0 ± 1.58, $p \leq 0.05$). The IHC and IF staining results indicated that the expression of vWF and CD31 were significantly lower in the GDM fetal lungs than that in the control group ($p \leq 0.001$) (Fig. 1B–C). The mean MVD in the GDM fetal lung group was 16.95 ± 5.65 mm2, compared with 35.78 ± 8.63 mm2 in the control group ($p \leq 0.05$) (Fig. 1D). These results indicated that the GDM fetal lungs suffered impaired alveolar and vascular development. Fig. 1The GDM fetal lungs exhibiting deficient alveolar and vascular development. A The RAC determined via H&E staining of the control and GDM fetal lung tissues. B The IHC staining of the control and GDM fetal lung tissues for vWF expression. C The IF staining of the control and GDM fetal lung tissues for CD31 expression. D The MVD evaluation in terms of the GDM and control groups ($$n = 6$$ in each group, *$p \leq 0.05$, **$p \leq 0.001$)
## Upregulation of mammalian target of rapamycin C1 and deregulation of mammalian target of rapamycin C2 in the gestational diabetes mellitus fetal lungs
The western blot analysis revealed significantly higher mTOR ($p \leq 0.01$) and p-mTOR (Ser2448) ($p \leq 0.001$) levels in the GDM fetal lungs than in the controls (Fig. 2A), indicating that maternal hyperglycemia stimulates the total mTOR signaling in fetal lungs. The status of the mTORC1 and mTORC2 in the fetal lungs was subsequently evaluated by determining the levels of proteins that are associated with their respective functionality. The western blot analysis data revealed that the p-Rictor (Ser1591), p-Sin1(Thr86), and p-Akt (Ser473) levels were all downregulated in the GDM fetal lungs ($p \leq 0.001$) (Fig. 2B), suggesting that the mTORC2 activity is suppressed under GDM. In contrast, the p-Raptor (Ser863) level was upregulated in the GDM lungs ($p \leq 0.01$) (Fig. 2B). Moreover, while similar GβL levels were detected in the GDM and control groups (Fig. 2A), the E3 ubiquitin ligase, TRAF2, was upregulated, while the deubiquitinase, OTUD7B, was downregulated in the GDM group ($p \leq 0.001$) (Fig. 2A), suggesting more extensive GβL-ubiquitination and, consequently, increased GβL incorporation into mTORC1 versus mTORC2 under GDM. Collectively, these data supported the notion that the mTORC1 is upregulated and the mTORC2 downregulated in fetal lungs exposed to maternal hyperglycemia. Fig. 2Upregulation of mTORC1 and downregulation of mTORC2 in the GDM fetal lungs. A The protein expression of TRAF2, OTUD7B, p-mTOR, mTOR, and GβL in the control and GDM fetal lung tissues. B The determined protein levels of p-Sin1, Sin1, p-Rictor, Rictor, p-Raptor, Raptor, p-Akt, and Akt in the control and GDM fetal lung tissues. C The relative protein levels of TRAF2, OTUD7B, p-mTOR, mTOR, and GL in the control and GDM fetal lung tissues. D The relative protein levels of p-Sin1, Sin1, p-Rictor, Rictor, p-Raptor, Raptor, p-Akt, and Akt in the control and GDM fetal lung tissues ($$n = 3$$ in triplicate, **$p \leq 0.01$, ***$p \leq 0.001$)
## High-glucose-inhibited mouse fetal lung endothelial cell angiogenesis through mTOR pathway
Primary cultured MFLECs were cobblestone-like morphology and assumed island-like pattern under light microscope. The effects of hyperglycemia on MFLEC angiogenesis were subsequently evaluated in vitro using the tube formation assay. As Fig. 3 showed, the MFLECs under high-glucose stress (30mM) exhibited fewer branch points (68.0 ± 12.6 vs. 181.8 ± 13.8, $p \leq 0.001$) compared with the control group while the angiogenesis capability was partly restored when mTOR pathway was inhibited by rapamycin (100ng/ml) (68.0 ± 12.6 vs. 149.8 ± 15.0, $p \leq 0.001$). These data indicated that high-glucose conditions impair MFLEC angiogenesis through mTOR pathway. Fig. 3High-glucose-inhibited MFLEC angiogenesis through mTOR pathway in vitro. The MFLECs were pre-incubated in a culture medium containing 5 mM glucose (control) or 30 mM glucose (high glucose) or 30 mM glucose with 100ng/ml rapamycin (HG + RAPA) for 72 h and loaded into Matrigel. The cells treated with high glucose levels exhibited a dramatic decrease in the total number of branch points while the number was almost restored to control level in HG + RAPA group. ( HG: high glucose, RAPA: rapamycin, $$n = 3$$ in triplicate, ***$p \leq 0.001$, * $p \leq 0.05$)
## High-glucose-stimulated G protein beta subunit-like protein ubiquitination and mammalian target of rapamycin C1 formation in cultured mouse fetal lung endothelial cells
To reveal the mechanisms through which high-glucose conditions impair MFLEC angiogenesis, the status of mTORC1 and mTORC2 assembly was examined under high- and low-glucose conditions. The western blot data revealed that, similar to the findings from the GDM mouse model, the MFLECs exposed to 30 mM glucose (high glucose) exhibited significantly higher mTOR and p-mTOR (Ser2448) protein levels than those exposed to 5 mM glucose (control) (Fig. 4A). The results also revealed increased p-Rictor (Ser1591) and p-Sin1 (Thr86) levels but decreased p-Akt(Ser473) and p-Raptor (Ser863) levels in the high-glucose-treated cells (Fig. 4B), suggesting that hyperglycemia promotes mTORC1 formation over mTORC2 formation. Similarly, it was found that the high-glucose levels increased the TRAF2 expression and decreased the OTUD7B expression (Fig. 4A), meaning hyperglycemia could also promote mTORC1 assembly over mTORC2 assembly through stimulating the GβL-ubiquitination and, consequently, its preferable incorporation into mTORC1. Meanwhile, the immunoprecipitation assay confirmed that, while the high-glucose levels did not change the GβL expression, they increased the level of ubiquitinated GβL, as a greater amount of ubiquitin was detected in the GβL pull-down under high-glucose conditions (Fig. 4C). The immunoprecipitation assay also revealed an increased amount of Raptor but decreased amounts of Rictor and Sin1 in the GβL pull-downs under similar conditions (Fig. 4C), further supporting the idea that high glucose shifts the dynamic assembly of mTORC1 and mTORC2 toward mTORC1 assembly. Fig. 4High-glucose-stimulated GβL-ubiquitination and mTORC1 assembly in the cultured MFLECs. A The protein levels of TRAF2, OTUD7B, p-mTOR, mTOR, and GβL. B The protein levels of p-Sin1, Sin1, p-Rictor, Rictor, p-Raptor, Raptor, p-Akt and Akt. C The cell lysates were incubated with anti-GβL antibody-coated beads. The mTOR, Raptor, Rictor, Sin1, and ubiquitin levels in the precipitates were determined via western blot analysis ($$n = 3$$ in triplicate, **$p \leq 0.01$, ***$p \leq 0.001$)
## Tumornecrosis factorreceptor-associated factor 2 knockdownrestored mouse fetal lung endothelial cell angiogenesis under high-glucoseconditions
To confirm the role of TRAF2 in high-glucose-induced GβL-ubiquitination and mTORC1 formation, the effects of TRAF2 knockdown were investigated. Compared with the cells transfected with si-NC, the TRAF2 was significantly downregulated in terms of mRNA and protein expression following siRNA treatment ($p \leq 0.001$) (Fig. 5A and B). The immunoprecipitation assay revealed that the TRAF2 knockdown inhibited the GβL-ubiquitination and GβL/Raptor association enhanced by the high-glucose conditions (Fig. 5C). In fact, the TRAF2 knockdown also restored the GβL/Rictor and GβL/Sin1 binding inhibited by the high-glucose conditions (Fig. 5C). These results confirmed that TRAF2 plays a critical role in high-glucose-induced GβL-ubiquitination and mTORC1 formation over mTORC2 formation, which impairs the angiogenesis capacity of MFLECs. Finally, the tube formation assay indicated that the impaired angiogenesis capacity of MFLECs under high-glucose stress was restored by TRAF2 knockdown (Fig. 5D). These in-vitro findings strongly support the idea that maternal hyperglycemia inhibits fetal lung vasculogenesis in vivo by upregulating TRAF2-mediated GβL-ubiquitination and consequently promoting mTORC1 formation over mTORC2 formation. Fig. 5The TRAF2 knockdown restored the MFLEC angiogenesis impaired by the high glucose. A Expression level of TRAF2 protein. B Expression level of TRAF2 mRNA. C The association between GβL and mTOR, Raptor, Rictor, Sin1, and ubiquitin was evaluated using the immunoprecipitation assay. D The cell angiogenesis ability under high-glucose conditions (HG: high glucose, $$n = 3$$ in triplicate, ***$p \leq 0.001$)
## Discussion
Delayed fetal lung maturation is a common adverse effect in GDM-affected pregnancy, posing a risk to the newborn in terms of contracting respiratory distress syndrome. Our previous study confirmed the histological changes of a reduced number of alveoli, thickened alveolar septum, and decreased alveolar area in the fetal lungs of GDM rats [14], indicating impaired lung vasculogenesis in the offspring. In this study, it was found that the impaired alveolar and vascular development in the fetal lungs of the pregnant GDM mice was characterized by an upregulated expression of mTORC1 and a downregulated expression of mTORC2. In addition, exposure to high-glucose conditions reduced the angiogenic ability of the cultured MFLECs in vitro. Our in-vitro and in-vivo mechanistic studies demonstrated that these detrimental effects of GDM on mouse fetal lung vasculogenesis are mediated by maternal-hyperglycemia-induced mTORC1 assembly in fetal lung tissues.
Vasculogenesis is the de novo formation of blood vessels from endothelial precursor cells to form blood islands and subsequently differentiate them into endothelial and hematopoietic cells, while angiogenesis refers to the new formation of vessels from pre-existing ones to remodel and expand the vasculature [23]. Both vasculogenesis and angiogenesis are necessary during early lung development [24]. Human fetal lung development starts from the fourth gestational week and is divided into five stages according to the morphologic changes: the embryonic stage, the pseudoglandular stage, the canalicular stage, the saccular stage, and the alveolar stage [25]. During the embryonic stage (4–7 gestational weeks), the first blood vessel of the lung arises from mesenchymal progenitors via vasculogenesis. During the pseudoglandular stage, epithelial cells line along the primary bronchial tree, while the number of lung capillaries undergoes a dramatic increase during the canalicular stage (16–26 gestational weeks), which continues to mature in the saccular stage (24–38 gestational weeks). At the last stage (36 gestational weeks–infancy), the fetal capillary network develops into a single capillary layer for efficient gas exchange [26].
Abnormal vascular development was found to be lethal in the embryos of gene-deficient animal models [27]. The impairment or immaturity of pulmonary vasculature is involved in the pathogenesis of various neonatal and pediatric pulmonary vascular diseases, such as respiratory disease syndrome and persistent pulmonary hypertension. Vasculogenesis and angiogenesis are regulated by several factors, including the glucose level. In adults, different vasculogenic and angiogenic responses have been reported under high-glucose exposure in different tissues [28]. The present study confirmed that the fetal lungs of GDM mice are characterized by deficient vascular development. This result is consistent with that obtained in a previous report [7] and is in accordance with the GDM clinical manifestation that the offspring of GDM mothers have a higher risk of respiratory diseases. While studies have demonstrated that high-glucose conditions contribute to the dysfunction in endothelial cells [29], little is known about the fetal lung endothelial function under such conditions. Thus, the biological function of MFLECs was further investigated under high-glucose conditions in vitro. The results indicated that the high-glucose conditions damaged the angiogenic ability in the MFLECs, indicating a possible mechanism of retarded lung development in GDM offspring.
As an evolutionary conserved serine/threonine protein kinase, mTOR is a master regulator in specific cell biological processes and metabolic states. The dysregulation of mTOR is involved in the progression of both cancer and diabetes [30]. Interestingly, the activity of mTOR is, in turn, regulated by a variety of extracellular and intracellular cues, including glucose. The in-vitro experiment confirmed that high-glucose conditions facilitate the development of endometrial cancer via mTOR [31].
Furthermore, mTOR plays a crucial role in glucose-induced angiogenesis. Zou et al. reported that their human umbilical vein endothelial cells exhibited an enhanced ability of angiogenesis under high-glucose and low-serum conditions via mTOR activation [32], while a recent study demonstrated that the promotion of mTOR might be involved in high-glucose-induced retinal angiogenesis [33]. To investigate the underlying mechanism of high-glucose-induced pulmonary vasculogenesis impairment, the expression of mTOR was assayed under high glucose conditions both in vivo and in vitro. The results indicated that the damaged vasculogenesis in the GDM fetal lungs was related to the increased expression and activation of mTOR, which would appear to be opposite to the pro-angiogenetic effect of mTOR in tumorgenesis [24, 34].
While mTORC1 and mTORC2 share the core structure of mTOR protein, the difference in other protein components determines their diverse biological functions. In fact, the two complexes may even represent opposite effects. The knockdown of mTORC1 leads to improved angiogenesis and vascular integrity in diabetic mice [35], while a higher mTORC2 activity is related to increased angiogenesis [36]. By detecting the expression of protein components of mTORC1 and mTORC2, the specific role of the two complexes in fetal lung and MFLECs under high-glucose stimulation was further demonstrated.
Ubiquitination is a major post-translational modification that regulates the stability of proteins and affects the biological functions. Ubiquitination is modulated by the ubiquitin-proteasome system, which includes E1 activating enzymes, E2 conjugating enzymes, and E3 ligases [37]. Meanwhile, TRAF2 belongs to the tumor necrosis factor receptor-associated factors family and has the capacity to serve as an adaptor protein in the assembly of receptor-associated signaling complexes [38]. High-glucose conditions induce a decreased expression of TRAF2 in neuroblastoma cells [39]. The present study demonstrated that the downregulation of TRAF2 in MFLECs hinders GβL-ubiquitination and restores the cell angiogenic ability under high-glucose conditions. Given that TRAF2 acts as an E3 ligase in the regulation of mTORC1 and mTORC2 formation [9], GβL-ubiquitination-dependent mTORC1 assembly may be a key mechanism for impaired vasculogenesis in GDM fetal lungs. According to our data, we summarized the underlying mechanism how does maternal high glucose condition affect fetal lung development (Fig. 6): maternal hyperglycemia shifts dynamic assembly of mTORC1 and mTORC2 toward mTORC1 assembly by upregulating TRAF2-mediated GβL ubiquitination, which inhibits mouse fetal lung angiogenesis, thus affecting fetal lung health.
**Fig. 6:** *The proposed mechanism of hyperglycemia inhibits pulmonary vasculogenesis. Maternal hyperglycemia can up-regulate the expression level of TRAF2 in mouse fetal lung, which acts as an E3 ligase in the regulation of mTORC1 and mTORC2 formation. By promoting the polyubiquitination of GβL, up-regulated TRAF2interferes with the interaction of GβL with SIN1 (a unique component of mTORC2), which promotes the formation of mTORC1, thus inhibiting mouse fetal lung angiogenesis.↑represents the up-regulated expression levels of TRAF2, GβL-ubiquitination, and mTORC1 assembly;↓represents down-regulation of mTORC2 assembly; indicates promotion of the events; ⊥ indicates inhibition of the events.*
## Conclusion
Overall, this study demonstrated that the GβL-ubiquitination-dependent dynamic assembly of mTORC1 and mTORC2 controls the pulmonary angiogenesis during lung development. The inhibition of mTORC1 formation may prevent GDM-associated poor lung health in newborns.
## References
1. Ramos-Arroyo MA, Rodriguez-Pinilla E, Cordero JF. **Maternal diabetes: the risk for specific birth defects**. *Eur J Epidemiol* (1992) **8** 503-8. DOI: 10.1007/BF00146367
2. Ferencz C, Rubin JD, McCarter RJ, Clark EB. **Maternal diabetes and cardiovascular malformations: predominance of double outlet right ventricle and truncus arteriosus**. *Teratology* (1990) **41** 319-26. DOI: 10.1002/tera.1420410309
3. Anderson JL, Waller DK, Canfield MA, Shaw GM, Watkins ML, Werler MM. **Maternal obesity, gestational diabetes, and central nervous system birth defects**. *Epidemiology* (2005) **16** 87-92. DOI: 10.1097/01.ede.0000147122.97061.bb
4. Yang GR, Dye TD, Li D. **Effects of pre-gestational diabetes mellitus and gestational diabetes mellitus on macrosomia and birth defects in Upstate New York**. *Diabetes Res Clin Pract* (2019) **155** 107811. DOI: 10.1016/j.diabres.2019.107811
5. Azad MB, Moyce BL, Guillemette L, Pascoe CD, Wicklow B, McGavock JM, Halayko AJ, Dolinsky VW. **Diabetes in pregnancy and lung health in offspring: developmental origins of respiratory disease**. *Paediatr Respir Rev* (2017) **21** 19-26. PMID: 27665512
6. **ACOG Practice Bulletin No. 190: gestational diabetes Mellitus**. *Obstet Gynecol* (2018) **131** e49-64. DOI: 10.1097/AOG.0000000000002501
7. Baack ML, Forred BJ, Larsen TD, Jensen DN, Wachal AL, Khan MA, Vitiello PF. **Consequences of a maternal High-Fat Diet and late Gestation diabetes on the developing rat lung**. *PLoS ONE* (2016) **11** e0160818. DOI: 10.1371/journal.pone.0160818
8. Mullassery D, Smith NP. **Lung development**. *Semin Pediatr Surg* (2015) **24** 152-5. DOI: 10.1053/j.sempedsurg.2015.01.011
9. Wang B, Jie Z, Joo D, Ordureau A, Liu P, Gan W, Guo J, Zhang J, North BJ, Dai X. **TRAF2 and OTUD7B govern a ubiquitin-dependent switch that regulates mTORC2 signalling**. *Nature* (2017) **545** 365-9. DOI: 10.1038/nature22344
10. Li JJ, Yan YY, Sun HM, Liu Y, Su CY, Chen HB, Zhang JY. **Anti-Cancer Effects of Pristimerin and the Mechanisms: a critical review**. *Front Pharmacol* (2019) **10** 746. DOI: 10.3389/fphar.2019.00746
11. Karar J, Maity A. **PI3K/AKT/mTOR pathway in Angiogenesis**. *Front Mol Neurosci* (2011) **4** 51. DOI: 10.3389/fnmol.2011.00051
12. Mohlin S, Hamidian A, von Stedingk K, Bridges E, Wigerup C, Bexell D, Pahlman S. **PI3K-mTORC2 but not PI3K-mTORC1 regulates transcription of HIF2A/EPAS1 and vascularization in neuroblastoma**. *Cancer Res* (2015) **75** 4617-28. DOI: 10.1158/0008-5472.CAN-15-0708
13. Ziegler ME, Hatch MM, Wu N, Muawad SA, Hughes CC. **mTORC2 mediates CXCL12-induced angiogenesis**. *Angiogenesis* (2016) **19** 359-71. DOI: 10.1007/s10456-016-9509-6
14. Zhang QM, Ouyang WX, Chai XQ, Deng FT. **Expression of lung surfactant proteins SP-B and SP-C and their Regulatory factors in fetal lung of GDM rats**. *Curr Med Sci* (2018) **38** 847-52. DOI: 10.1007/s11596-018-1952-8
15. Abdul Aziz SH, John CM, Mohamed Yusof NI, Nordin M, Ramasamy R, Adam A, Mohd Fauzi F. **Animal model of Gestational Diabetes Mellitus with Pathophysiological Resemblance to the Human Condition Induced by multiple factors (Nutritional, Pharmacological, and stress) in rats**. *Biomed Res Int* (2016) **2016** 9704607. DOI: 10.1155/2016/9704607
16. Pennington KA, van der Walt N, Pollock KE, Talton OO, Schulz LC. **Effects of acute exposure to a high-fat, high-sucrose diet on gestational glucose tolerance and subsequent maternal health in mice**. *Biol Reprod* (2017) **96** 435-45. DOI: 10.1095/biolreprod.116.144543
17. Zhang J, Zhou X, Zhu J. **Beauveria attenuates asthma by inhibiting inflammatory response and inducing lymphocytic cell apoptosis**. *Oncotarget* (2016) **7** 74557-68. DOI: 10.18632/oncotarget.12958
18. Zaqout S, Becker LL, Kaindl AM. **Immunofluorescence staining of paraffin sections step by step**. *Front Neuroanat* (2020) **14** 582218. DOI: 10.3389/fnana.2020.582218
19. Wallace B, Peisl A, Seedorf G, Nowlin T, Kim C, Bosco J, Kenniston J, Keefe D, Abman SH. **Anti-sflt-1 therapy preserves lung alveolar and vascular growth in Antenatal Models of Bronchopulmonary Dysplasia**. *Am J Respir Crit Care Med* (2018) **197** 776-87. DOI: 10.1164/rccm.201707-1371OC
20. Weidner N, Semple JP, Welch WR, Folkman J. **Tumor angiogenesis and metastasis–correlation in invasive breast carcinoma**. *N Engl J Med* (1991) **324** 1-8. DOI: 10.1056/NEJM199101033240101
21. Fu YQ, Fang F, Lu ZY, Kuang FW, Xu F. **N-acetylcysteine protects alveolar epithelial cells from hydrogen peroxide-induced apoptosis through scavenging reactive oxygen species and suppressing c-Jun N-terminal kinase**. *Exp Lung Res* (2010) **36** 352-61. DOI: 10.3109/01902141003678582
22. Luo Q, Liu X, Zheng Y, Zhao Y, Zhu J, Zou L. **Ephrin-B2 mediates trophoblast-dependent maternal spiral artery remodeling in first trimester**. *Placenta* (2015) **36** 567-74. DOI: 10.1016/j.placenta.2015.02.009
23. Gao Y, Cornfield DN, Stenmark KR, Thebaud B, Abman SH, Raj JU. **Unique aspects of the developing lung circulation: structural development and regulation of vasomotor tone**. *Pulm Circ* (2016) **6** 407-25. DOI: 10.1086/688890
24. Woik N, Kroll J. **Regulation of lung development and regeneration by the vascular system**. *Cell Mol Life Sci* (2015) **72** 2709-18. DOI: 10.1007/s00018-015-1907-1
25. Bush A. **Lung Development and Aging**. *Ann Am Thorac Soc* (2016) **13** 438-46. DOI: 10.1513/AnnalsATS.201602-112AW
26. Goss K. **Long-term pulmonary vascular consequences of perinatal insults**. *J Physiol* (2019) **597** 1175-84. DOI: 10.1113/JP275859
27. Woods L, Perez-Garcia V, Hemberger M. **Regulation of placental development and its impact on fetal growth-new Insights from mouse models**. *Front Endocrinol (Lausanne)* (2018) **9** 570. DOI: 10.3389/fendo.2018.00570
28. Fadini GP, Albiero M, Bonora BM, Avogaro A. **Angiogenic abnormalities in diabetes Mellitus: mechanistic and clinical aspects**. *J Clin Endocrinol Metab* (2019) **104** 5431-44. DOI: 10.1210/jc.2019-00980
29. Zhang Y, Lv X, Hu Z, Ye X, Zheng X, Ding Y, Xie P, Liu Q. **Protection of Mcc950 against high-glucose-induced human retinal endothelial cell dysfunction**. *Cell Death Dis* (2017) **8** e2941. DOI: 10.1038/cddis.2017.308
30. Saxton RA, Sabatini DM. **mTOR Signaling in Growth, Metabolism, and Disease**. *Cell* (2017) **169** 361-71. DOI: 10.1016/j.cell.2017.03.035
31. Han J, Zhang L, Guo H, Wysham WZ, Roque DR, Willson AK, Sheng X, Zhou C, Bae-Jump VL. **Glucose promotes cell proliferation, glucose uptake and invasion in endometrial cancer cells via AMPK/mTOR/S6 and MAPK signaling**. *Gynecol Oncol* (2015) **138** 668-75. DOI: 10.1016/j.ygyno.2015.06.036
32. Zou Y, Wu F, Liu Q, Deng X, Hai R, He X, Zhou X. **Downregulation of miRNA328 promotes the angiogenesis of HUVECs by regulating the PIM1 and AKT/mTOR signaling pathway under high glucose and low serum condition**. *Mol Med Rep* (2020) **22** 895-905. DOI: 10.3892/mmr.2020.11141
33. Li R, Du J, Yao Y, Yao G, Wang X. **Adiponectin inhibits high glucose-induced angiogenesis via inhibiting autophagy in RF/6A cells**. *J Cell Physiol* (2019) **234** 20566-76. DOI: 10.1002/jcp.28659
34. Conciatori F, Bazzichetto C, Falcone I, Pilotto S, Bria E, Cognetti F, Milella M, Ciuffreda L. **Role of mTOR Signaling in Tumor Microenvironment: an overview**. *Int J Mol Sci.* (2018) **19** 2453. DOI: 10.3390/ijms19082453
35. Fan W, Han D, Sun Z, Ma S, Gao L, Chen J, Li X, Li X, Fan M, Li C. **Endothelial deletion of mTORC1 protects against hindlimb ischemia in diabetic mice via activation of autophagy, attenuation of oxidative stress and alleviation of inflammation**. *Free Radic Biol Med* (2017) **108** 725-40. DOI: 10.1016/j.freeradbiomed.2017.05.001
36. Maiti S, Mondal S, Satyavarapu EM, Mandal C. **mTORC2 regulates hedgehog pathway activity by promoting stability to Gli2 protein and its nuclear translocation**. *Cell Death Dis* (2017) **8** e2926. DOI: 10.1038/cddis.2017.296
37. Morreale FE, Walden H. **Types of Ubiquitin Ligases**. *Cell* (2016) **165** 248-8 e241. DOI: 10.1016/j.cell.2016.03.003
38. Xie P. **TRAF molecules in cell signaling and in human diseases**. *J Mol Signal* (2013) **8** 7. DOI: 10.1186/1750-2187-8-7
39. Liu Y, Sun L, Ma Y, Wei B, Gao M, Shang L. **High glucose and bupivacaine-induced cytotoxicity is mediated by enhanced apoptosis and impaired autophagy via the PERK-ATF4-CHOP and IRE1-TRAF2 signaling pathways**. *Mol Med Rep* (2019) **20** 2832-42. PMID: 31524237
|
---
title: Yeast hydrolysate attenuates lipopolysaccharide-induced inflammatory responses
and intestinal barrier damage in weaned piglets
authors:
- Runqi Fu
- Chan Liang
- Daiwen Chen
- Gang Tian
- Ping Zheng
- Jun He
- Jie Yu
- Xiangbing Mao
- Yuheng Luo
- Junqiu Luo
- Bing Yu
journal: Journal of Animal Science and Biotechnology
year: 2023
pmcid: PMC10021991
doi: 10.1186/s40104-023-00835-2
license: CC BY 4.0
---
# Yeast hydrolysate attenuates lipopolysaccharide-induced inflammatory responses and intestinal barrier damage in weaned piglets
## Abstract
### Background
Intestinal inflammation is the main risk factor causing intestinal barrier dysfunction and lipopolysaccharide (LPS) can trigger inflammatory responses in various eukaryotic species. Yeast hydrolysate (YH) possesses multi-biological effects and is received remarkable attention as a functional ingredient for improving growth performance and promoting health in animals. However, there is still inconclusive on the protective effects of dietary YH supplementation on intestinal barrier of piglets. This study was conducted to investigate the attenuate effects of YH supplementation on inflammatory responses and intestinal barrier injury in piglets challenged with LPS.
### Methods
Twenty-four piglets (with an average body weight of 7.42 ± 0.34 kg) weaned at 21 days of age were randomly assigned to one of two dietary treatments (12 replications with one pig per pen): a basal diet or a basal diet containing YH (5 g/kg). On the 22nd d, 6 piglets in each treatment were intraperitoneally injected with LPS at 150 μg/kg BW, and the others were injected with the same amount of sterile normal saline. Four hours later, blood samples of each piglet were collected and then piglets were euthanized.
### Results
Dietary YH supplementation increased average daily feed intake and average daily gain ($P \leq 0.01$), decreased the ratio of feed intake to gain of piglets ($$P \leq 0.048$$). Lipopolysaccharide (LPS) injection induced systemic inflammatory response, evidenced by the increase of serum concentrations of haptoglobin (HP), adrenocorticotropic hormone (ACTH), cortisol, and interleukin-1β (IL-1β). Furthermore, LPS challenge resulted in inflammatory intestinal damage, by up-regulation of the protein or mRNA abundances of tumor necrosis factor-α (TNF-α), IL-1β, toll-like receptors 4 (TLR4) and phosphor-nuclear factor-κB-p65 (p-NFκB-p65) ($P \leq 0.01$), and down-regulation of the jejunal villus height, the protein and mRNA abundances of zonula occludens-1 (ZO-1) and occludin (OCC; $P \leq 0.05$) in jejunal mucosa. Dietary YH supplementation decreased the impaired effects of ACTH, cortisol, HP, IL-1β and diamine oxidase in serum ($P \leq 0.05$). Moreover, YH supplementation also up-regulated the jejunal villus height, protein and mRNA abundances of ZO-1 and OCC ($P \leq 0.05$), down-regulated the mRNA expressions of TNF-α and IL-1β and the protein abundances of TNF-α, IL-1β, TLR4 and p-NFκB-p65 in jejunal mucosa in LPS-challenged pigs ($P \leq 0.01$).
### Conclusion
Yeast hydrolysate could attenuate inflammatory response and intestinal barrier injury in weaned piglets challenged with LPS, which was associated with the inhibition of TLR4/NF-κB signaling pathway activation.
### Supplementary Information
The online version contains supplementary material available at 10.1186/s40104-023-00835-2.
## Introduction
Intestinal epithelial barrier, a structure of continuous monolayer enterocytes, is dominated by intercellular tight junction [1]. Under a normal state, it acts as a selective filter that enables the absorption of nutrients and ensures an effective defense against exogenous pathogens, luminal antigens, etc. [ 2]. For early weaned mammals, however, the gastrointestinal tract is immature and vulnerable to multitudinous stresses [3], and intestinal epithelial barrier is frequently defective in various pathological status, especially in bacterial infections induced by pathogenic bacteria [4, 5]. The damaged intestinal epithelial barrier mainly causes the increase of intestinal permeability, promotes the transfer of antigens in lumen to the subepithelial tissues, and further exacerbates the mucosal and systemic inflammatory reactions [1, 6, 7]. The secretion of pro-inflammatory cytokines during the inflammatory response is a major factor in triggering the disruption of the intestinal barrier [8]. In piglets, intestinal barrier dysfunction and accompanied by the enhancement of intestinal permeability are often observed, with subsequent diarrhea and growth retardation [9, 10]. Therefore, effective and safe preventive approaches to maintenance of intestinal barrier function are urgently needed for piglets. Notably, lipopolysaccharide (LPS), an intrinsic component of the cell wall of gram-negative bacteria, has been shown to be a key molecule in inducing the production of pro-inflammatory cytokines [7]. It often contributes to inflammatory intestinal damage in piglets. The molecular mechanism is manifested by the activation of inflammatory signaling pathway by LPS, which induces the expression of key proteins of pro-inflammatory factors and thus leads to intestinal barrier damage [11]. Previous studies have shown that LPS can be used to construct a well-established model of inflammation in pigs [12, 13].
Yeast (Saccharomyces cerevisiae) is widely distributed in nature and has been inseparable from human life [14]. Yeast hydrolysate (YH), also known as yeast autolysate, is obtained from *Saccharomyces cerevisiae* via protein hydrolysis enzyme [15, 16]. Several researches suggested that the autolytic yeast fractions or peptides from autolyzed yeasts revealed physiological effects on anti-obesity [17, 18], anti-fatigue [19], anti-stress [20, 21] and immuno-promotional activities [22, 23]. For these reasons, YH has attracted much attention as a functional material supplement and it is generally recognized as non-toxic, effective and safe [24]. Recently, several studies were focused on improving intestinal health and immune-potentiating activities with YH. Specifically, YH has multiple roles on promoting digestion and absorption of nutrients [25], improving intestinal microflora structure [26] and decreasing diarrhea of young animals [19], while also acts as an immunomodifier to prevent gut inflammation [27]. However, the protective effects of dietary supplementation with YH on intestinal barrier are limited and inconclusive. It is widely known that hyperinflammation in intestine is one of the most factors causing intestinal barrier dysfunction [1, 2, 10]. Consequently, considering the above, we postulated that YH has the potential to prevent LPS-induced intestinal barrier dysfunction in piglets by alleviating inflammation via the related signaling pathways. This study tested these hypotheses by assessing the effects of YH on the systemic inflammatory response, intestinal morphology, expression of tight junction-related proteins and gut anti-inflammatory capacity in piglets.
## Chemical analysis of yeast hydrolysate
Yeast hydrolysate was provided by Jiangmen Thealth Bioengineering Co., Ltd. (Guangzhou, China). The contents of moisture, crude protein, crude fat and crude ash were measured with the reference of AOAC [28]. Gross energy was determined by an oxygen bomb calorimeter (Parr instruments, Moline, IL, USA). The soluble protein in yeast hydrolysate was extracted according to the method of Wang et al. [ 29] and fractionated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) system according to the previous study [30]. The gel was stained with Coomassie Brilliant Blue (Beyotime, Shanghai, China) for 40 min and de-stained with deionized water for 8 h.
## Experimental animals, diet, and design
The animal procedures were reviewed and approved by the Animal Care and Use Committee at Sichuan Agricultural University with the approval number of SYXK (Sichuan)-2019–187. A total of twenty-four weaned piglets (21-day-old), with an average initial body weight (BW) of 7.42 ± 0.34 kg, were randomly allotted to two groups (12 replicates per treatment and one piglets per replicate) receiving either a basal diet or a basal diet with YH (5 g/kg) for 21 d. All piglets were fed individually in metabolic cages (1.5 m × 0.7 m × 1.0 m) and housed in an environmentally controlled room. Piglets were fed three times daily at 8:00, 14:00 and 20:00 and have free accessed to water and feed throughout the experiment. After 21-d feeding trial, immunological challenge was applied to the half of piglets in each treatment (Fig. 1A). That means 6 piglets in each treatment were intraperitoneally injected with LPS at 150 μg/kg BW, and the other 6 piglets were injected with an equal volume of sterile physiological saline. The basal diet (Table 1) was a corn-soybean meal-fish meal diet and was formulated to meet or exceed National Research Council (NRC 2012) [31] requirements for piglets from 7–11 kg and 11–25 kg stage. The YH diet was formulated by replacing soybean meal with 5 g/kg YH in equal amounts in the basal diet. The molecular weight of YH was less than 50 kDa and mostly clustered below 25 kDa (Fig. S1). YH mainly provided a rich source of crude protein ($45.50\%$), and contributed less to crude ash ($6.47\%$) and crude fat ($2.17\%$) (Table S1).Fig. 1Dietary yeast hydrolysate supplementation improved the growth performance of weaned piglets before LPS challenge. A Schematic of the feeding experiment and LPS challenge. B–D Effects of dietary YH supplementation on average daily feed intake (ADFI), average daily gain (ADG) and the ratio of feed intake to gain (F/G) of weaned piglets. Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. * $P \leq 0.05$, **$P \leq 0.01$, compared with control groupTable 1Composition and nutrients levels of the basal diet (air-dry basis, %)IngredientsContent, %Nutrient level3ContentCorn31.42Digestible energy, MJ/kg14.73Extruded corn30.00Crude protein, %18.50Soybean meal9.94Ca, %0.75Extruded soybean8.00Total P, %0.56Fish meal4.00Available P, %0.37Whey powder5.00Digestible Lys, %1.30Soybean protein concentrate5.00Digestible Met, %0.38Soybean oil1.90Digestible Thr, %0.77Sucrose2.00Digestible Trp, %0.21Limestone0.85Dicalcium phosphate0.45NaCl0.20L-Lys·HCl ($78\%$)0.33DL-Met ($99\%$)0.07L-Thr ($98.5\%$)0.03L-Trp ($98\%$)0.01Chloride choline0.15Vitamin premix10.05Mineral premix20.30Benzoic acid0.30Total100.001Vitamin premix provided the following per kg of diet: VA 9000 IU, VD3 3000 IU, VE 20.0 IU, VK3 3.0 mg, VB1 1.5 mg, VB2 4.0 mg, VB6 3.0 mg, VB12 0.02 mg, nicotinic acid 30.0 mg, pantothenic acid 15.0 mg, folic acid 0.75 mg, biotin 0.1 mg2Mineral premix provided the following per kg of diet: Fe (FeSO4·H2O) 100.0 mg, Cu (CuSO4·5H2O) 6.0 mg, Zn (ZnSO4·H2O) 100.0 mg, Mn (MnSO4·H2O) 4.0 mg, I (KI) 0.14 mg, Se (Na2SeO3) 0.3 mg3Nutrients levels were calculated values
## LPS injection
The challenged piglets were intraperitoneally injected with *Escherichia coli* LPS (E. coli serotype O55:B5, Sigma Chemical Inc., St. Louis, MO, USA) at 150 μg/kg BW, and unchallenged piglets were administrated the same volume of sterile physiological saline. The dose and serotype of LPS used in this study was consistent with the previous reports [12, 13]. Previous experiments have presented that LPS injection particularly caused dramatic inflammatory response and intestinal barrier dysfunction in pigs, rats and mice. And these negative effects generally occurred within 3–6 h after LPS injection [7, 32]. Therefore, blood and intestinal samples in this study were collected 4 h following LPS or saline injection.
## Growth performance
Yeast hydrolysate treatment was a main factor prior to the LPS challenge. Piglets were weighted individually on 1st d and 22nd d of the experiment. Daily feed consumption was recorded for each piglet. Average daily gain and the ratio of feed intake to gain were calculated as well.
## Blood sample collection and analysis
Four hours following LPS and saline injection, blood samples were collected in 10 mL vacutainer tubes via anterior vena cava. Blood was centrifugated at 3500 × g for 10 min at 4 ℃. Serum samples were stored at −20 ℃ until subsequent analysis for inflammatory markers and diamine oxidase (DAO) concentrations. Commercially available porcine ELISA kits (Chenglin Biological Technology Co., Ltd., Beijing, China) were performed according to the manufacturer’s instructions for the following indicators: adrenocorticotropic hormone (ACTH), cortisol, C-reactive protein (CRP), serum amyloid A (SAA), haptoglobin (HP), tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β) and DAO in serum.
## Intestinal samples collection and analysis
Piglets were euthanized with pentobarbital sodium (200 mg/kg) in a separate sampling room away from other animals. The intestine was immediately removed. A 2-cm segment was removed from mid-jejunum and fixed with $4\%$ paraformaldehyde solution. Paraffin embedding was used to cut into cross sects (5 μm thick). The jejunal morphology was determined by hematoxylin and eosin (H&E) stain. Intestinal morphological images were photographed with a Nikon TS100 microscope (40 × and 100 ×). Villus height and crypt depth were analyzed and calculated by Image Pro Plus 6.0 software (Media Cybernetics, Bethesda, MD, USA). Other sections were stained using immunofluorescence for TLR4 protein. Briefly, mouse anti-TLR4 monoclonal antibody (1:100, sc-293072, Santa Cruz, Dallas, TX, USA) was incubated overnight at 4 ℃. Corresponding secondary antibody (Cy3 conjugated Goat Anti-mouse IgG, 1:300, GB21301 from Servicebio, Wuhan, China) was incubated for 50 min at room temperature. The slides were washed three times with PBS, and then incubated with DAPI solution at room temperature for 10 min and stored in the dark. After immunofluorescence, microphotographs were acquired with an inverted microscope (Leica DMI400B, Wetzlar, Germany). In addition, the inner wall of the middle jejunum was washed with ice-cold saline and the mucosal samples were then scraped into a sterile tube. Mucosal samples were immediately placed into liquid nitrogen and stored at −80℃ until the analysis of genes and proteins expressions. About 0.5 g of frozen jejunal mucosal scrapings were homogenized in ice-cold saline and prepared into a $10\%$ homogenate, crushed using an ultrasonic cell crushing system at 4 °C and then centrifuged (3000 × g, 15 min, 4 °C). The collected supernatant was used to analyze TNF-α and IL-1β contents by ELISA kits according to the manufacturer’s instructions.
## mRNA abundance analysis
Total RNA was extracted from jejunal mucosa using the TRIZOL reagent (TaKaRa Biotechnology (Dalian) Co., Ltd., Dalian, China). RNA integrity was verified by agarose gel electrophoresis. cDNA was synthesized with PrimeScript RT kit (TaKaRa). Real-time PCR was performed using SYBR Premix Ex Taq reagents (TaKaRa) and CFX-96 RT-qPCR Detection System (Bio-Rad, Hercules, CA, USA). *The* genes of intestinal barrier and inflammatory markers related primer pairs were synthesized by Sangon Biotech (Shanghai) Co., Ltd. (Shanghai, China) and listed in Table 2. The mRNA expression of target gene relative to housekeeping gene (β-actin) was calculated by the method of Arce et al. [ 33].Table 2Primer sequences used for real-time PCRGenePrimer sequence (5’ →3’)Product length, bpAccession numberβ-actinF: TCTGGCACCACACCTTCT114DQ178122R: TGATCTGGGTCATCTTCTCACZO-1F: CAGCCCCCGTACATGGAGA114XM_005659811R: GCGCAGACGGTGTTCATAGTTOCCF: CTACTCGTCCAACGGGAAAG158NM_001163647.2R: ACGCCTCCAAGTTACCACTGClDN-1F: TCTTAGTTGCCACAGCATGG106NM001244539R: CCAGTGAAGAGAGCCTGACCMUC2F: GGTCATGCTGGAGCTGGACAGT181XM_003122394.1R: TGCCTCCTCGGGGTCGTCACIL-1βF: CAGCTGCAAATCTCTCACCA112NM_214055.1R: TCTTCATCGGCTTCTCCACTTNF-αF: CGTGAAGCTGAAAGACAACCAG121NM_214022.1R: GATGGTGTGAGTGAGGAAAACGR: CAGGCTTCCGTCATCTGGTTCLDN-1 Claudin-1, OCC Occludin, MUC2 Mucin2, IL-1β interleukin-1β, TNF-α *Tumor necrosis* factor-α, ZO-1 Zonula occludens-1
## Western blot analysis
Western blot analysis was performed as previously described [34]. Briefly, protein was extracted from jejunal mucosa using the lysis buffer (Beyotime, Shanghai, China). Protein concentration was measured with the BCA protein assay kit (Pierce, Rockford, IL, USA). Then, protein was transferred to polyvinylidene fluoride membranes using a wet Trans-Blot system (Bio-Rad). After blocking, membranes were incubated with primary antibodies: anti-TLR4 (sc-293072, Santa Cruz), anti-ZO-1 (61–7300, Invitrogen, MA, USA), anti-OCC (ab31721, abcam, Shanghai, China), anti-TNF-α (ab6671, abcam), anti-IL-1β (sc-12742, Santa Cruz), anti-NFκB-p65 (6956, CST, Cell signaling Technology, Beverly, USA), anti-p-NFκB-p65 (3033, CST), and anti-β-actin (sc-47778, Santa Cruz). After washing, the corresponding secondary antibodies, goat anti-rabbit/mouse IgG -HRP secondary antibody (sc-2030 and sc-2031, Santa Cruz), were incubated at room temperature for 1 h. Visualization of membranes was performed with the Clarity™ Western ECL substrate (Bio-Rad) and the ChemiDoc XRS imaging system (Bio-Rad). The β-actin was applied as a controller for the mean of protein load.
## Statistical analysis
Statistical analysis was performed using SAS software package (Version 9.4; S.A.S, Institute Inc., Cary, NC, USA) [35]. All data were expressed as mean values with their standard error and checked for normal distribution using the Shapiro–Wilk test of SAS. Each piglet served as the statistical unit. Specifically, data on growth performance prior to LPS challenge were analyzed by two-tailed Student’s t-test. After LPS injection, data from serum and jejunum samples were statistically analyzed by two-way ANOVA using the PROC MIXED procedure of SAS with the following model:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${y}_{ijk} = \mu + {\alpha }_{i} + {\beta }_{j} + {\alpha }_{i} \times {\beta }_{j} + {e}_{ijk}$$\end{document}yijk=μ+αi+βj+αi×βj+eijk where yij is an observed trait, μ is the overall mean, αi is the fixed effect of immunological challenge (i = saline or LPS), βj is the fixed effect of dietary YH ($j = 0$ or 5 g/kg YH), αi × βj is the interaction between LPS and YH and eijk is the random error. Differences between the different groups were analyzed by Duncan’s multiple comparison method. $P \leq 0.05$ was considered statistically significant, and 0.05 < $P \leq 0.10$ indicated a trend.
## Effects of YH on growth performance in piglets prior to LPS injection
As shown in Fig. 1, a 21-d feeding experiment was conducted to examine the effects of YH on growth performance of piglets under normal condition (Fig. 1A). Compared with control group, dietary YH supplementation increased ADFI ($P \leq 0.01$) and ADG ($P \leq 0.01$) (Fig. 1B and C), decreased F/G of piglets ($$P \leq 0.048$$) (Fig. 1D).
## Effects of YH on systemic inflammatory response and serum DAO concentration in piglets challenged with LPS
Results of serum acute phase protein, stress hormone and inflammatory cytokines concentrations in piglets challenged with LPS were showed in Fig. 2 and Fig. 3. As expected, LPS injection enhanced HP (Fig. 2B), cortisol (Fig. 2D), ACTH (Fig. 2E) and IL-1β (Fig. 3B) concentrations in serum ($P \leq 0.01$). However, LPS + YH group significantly decreased the concentrations of HP, cortisol, ACTH and IL-1β in serum compared with the LPS group ($P \leq 0.05$).Fig. 2Dietary yeast hydrolysate supplementation inhibited the over-production of acute phase protein and stress hormone in piglets challenged with LPS. A–C Effects of dietary YH supplementation on the serum concentrations of serum amyloid A (SAA), haptoglobin (HP) and C-reactive protein (CRP) in piglets challenged with LPS. D and E Effects of dietary YH supplementation on the serum levels of stress hormones (D) cortisol and (E) adrenocorticotropic hormone. Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. a,b,cMeans with different superscript letters in a row were significantly different ($P \leq 0.05$)Fig. 3Dietary yeast hydrolysate supplementation attenuated the enhancement of serum concentrations of (A) tumor necrosis factor-α (TNF-α) and (B) interleukin-1β (IL-1β) in piglets challenged with LPS. Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. a,bMeans with different superscript letters in a row were significantly different ($P \leq 0.05$) As shown in Fig. 4, compared with saline group, there was greater concentration of serum DAO in LPS group ($P \leq 0.01$), however, YH supplementation significantly inhibited the increase of the serum DAO concentration in LPS challenged piglets ($$P \leq 0.02$$).Fig. 4Dietary yeast hydrolysate supplementation attenuated the effects of LPS-injection on jejunal permeability and morphology. A Representative picture of the appearance of the intestinal tract of a piglet. B Hematoxylin and eosin section of jejunum at 40 times magnification (up) and 100 times magnification (down). C The concentrations of diamine oxidase (DAO) in serum of piglets. D–F *The villus* height, crypt depth and the villus height:crypt depth ratio of jejunum of piglets. Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. a,b,cMeans with different superscript letters in a row were significantly different ($P \leq 0.05$)
## Effects of YH on intestinal morphology in LPS- challenged piglets
The effects of YH on intestinal morphology in LPS-challenged piglets were shown in Fig. 4. LPS challenge caused dramatic intestinal hyperemia (Fig. 4A), induced intestinal mucosal injury reflected by villous atrophy and mucosal detachment (Fig. 4B), and decreased the villus height (Fig. 4D) ($$P \leq 0.02$$). Compared with LPS group, LPS + YH group attenuated the state of intestinal hyperemia, improved the morphology and significantly increased the villus height ($$P \leq 0.03$$).
## Effects of YH on the expression of intestinal barrier related genes in LPS-challenged piglets
Compared with saline treatment, LPS injection reduced the relative mRNA expressions of CLDN-1 (Fig. 5A), OCC (Fig. 5C) and MUC2 (Fig. 5D) ($P \leq 0.05$), as well as the protein expressions of OCC (Fig. 5E and 5F; $$P \leq 0.01$$) and ZO-1 (Fig. 5E and G; $P \leq 0.01$) in jejunal mucosa. In compared to LPS group, YH supplementation significantly inhibited the down-regulation of mRNA expression of OCC ($$P \leq 0.02$$) and MUC2 ($P \leq 0.01$) and protein abundances of OCC ($P \leq 0.01$) and ZO-1 ($$P \leq 0.04$$) in jejunal mucosa of LPS-challenged piglets. Fig. 5Dietary yeast hydrolysate supplementation improved the barrier function of jejunal mucosa in piglets challenged with LPS. A–D The relative mRNA expression of Claudin-1 (CLDN-1), Zonula occludens-1 (ZO-1), Occludin (OCC) and mucin2 (MUC2) in the jejunal mucosa of piglets. E–G The protein abundance of OCC ZO-1 in the jejunal mucosa of piglets. Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. a,b,cmeans with different superscript letters in a row were significantly different ($P \leq 0.05$)
## Effects of YH on the expression of TNF-α and IL-1β of jejunum in piglets challenged with LPS
As shown in Fig. 6, LPS injection enhanced the concentrations of TNF-α (Fig. 6A; $$P \leq 0.01$$) in jejunal mucosa, however, YH significantly reversed this change ($P \leq 0.01$). A further analysis by RT-qPCR and western blot revealed that LPS challenge significantly increased the mRNA expressions of TNF-α (Fig. 6C) and IL-1β (Fig. 6D) and the corresponding protein abundances (Fig. 6E–G; $P \leq 0.01$). Conversely, a down-regulation was observed in the mRNA expressions of IL-1β and the protein abundance of TNF-α and IL-1β in LPS + YH group ($P \leq 0.01$).Fig. 6Dietary yeast hydrolysate supplementation inhibited the inflammatory response of jejunal mucosa in piglets challenged with LPS. A and B The concentrations of tumor necrosis factor-α (TNF-α) and interleukin-1β (IL-1β) in the jejunal mucosa of piglets by ELISA method. C and D The mRNA expression of TNF-α and IL-1β in the jejunal mucosa of piglets. E–G The protein abundance of TNF-α and IL-1β in the jejunal mucosa of piglets. Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. a,b,cMeans with different superscript letters in a row were significantly different ($P \leq 0.05$)
## Effects of YH on the TLR4/NF-κB signaling pathway in LPS-challenged piglets
As shown in Fig. 7, we investigated the expression of TLR4, the major inflammation-associated receptor, by immunofluorescence analysis and discovered that the distribution of TLR4 in jejunum of piglets enhanced by LPS challenge (Fig. 7A). Nevertheless, the variability was reversed with YH supplementation when compared to LPS group. Moreover, we observed an up-regulation in the expressions of TLR4 protein (Fig. 7B and C) and p-NFκB-p65 protein (Fig. 7B and D) in LPS group compared with saline group ($P \leq 0.01$). LPS + YH group decreased the protein abundances of TLR4 ($P \leq 0.01$) and p-NFκB-p65 ($$P \leq 0.01$$) compared with LPS group. Fig. 7Dietary yeast hydrolysate supplementation decreased the protein abundance of TLR4 and p-NFκB-p65 in jejunal mucosa in piglets challenged with LPS. A Immunofluorescence staining (100 times magnification) of toll-like receptors 4 (TLR4) in jejunum of piglets. B–D Relative protein abundance of B and C TLR4 and B and D phosphor-Nuclear factor-κB-p65 (p-NFκB-p65). Control, piglets were fed with a basal diet; YH, piglets were fed with a YH containing diet, 5 g/kg. a,b,cMeans with different superscript letters in a row were significantly different ($P \leq 0.05$)
## Discussion
Yeast hydrolysate (YH), an autolysate of Saccharomyces cerevisiae, has attracted much attention as a nutritional additive, which involved in various biological activities including growth promotion and immune regulation. Throughout the 21-d feeding experiment (pre-LPS-challenge) in this study, we observed that dietary YH supplementation increased growth performance of piglets. This effect may be related to the molecular weight of YH. As shown in an animal model, the low molecular weight YH (below 30 kDa) exhibits a low toxicity for rats [24]. This hydrolysate is more accessible to animals due to its low molecular weight contributing to a higher solubility in aqueous media and a better digestibility and absorptivity [36]. Previous studies have found that YH (below 10 kDa) revealed physiological effects on anti-obesity [37, 38] and anti-stress [20, 21]. In this study, the molecular weight of YH was less than 50 kDa and mostly clustered below 25 kDa. Furthermore, the beneficial effects of YH on growth performance have been characterized in rats [39], piglets [40], growing-finishing pigs [25], chickens [41] and fish [27, 42]. Similar results have been found in our previous study that feeding YH significantly improved growth performance for piglets under the normal physiological condition and the reason might also be that YH could improve intestinal barrier function [22]. Nevertheless, it remains to be studied whether YH can improve intestinal health under pathological states. In the present study, we focused on whether dietary YH attenuated the intestinal barrier impairment through inhibiting the massive release of pro-inflammatory cytokines. Hence, we utilized an incontrovertible model for acute gut injure by injecting lipopolysaccharide (LPS). LPS is an intrinsic constituent of membranes in gram-negative bacteria and it is a powerful endotoxin [7]. It binds to TLR4 and subsequently activates downstream signaling pathways, triggering an inflammatory response [11].
Acute phase proteins (APPs) such as C-reactive protein (CRP), haptoglobin (HP) and serum amyloid A (SAA) were secreted by hepatocytes and served as a crucial role in the etiopathogenesis of immune diseases [43]. In addition, some stress hormones (e.g., ACTH and cortisol) were frequently used in response to infectious status in the body [44]. With the infection and injury of the host (such as LPS injection), the expressions or serum concentrations of APPs and stress hormones will be dramatically enhanced [45, 46]. Indeed, our results found that serum concentrations of HP, SAA, ACTH and cortisol in piglets were increased 4 h after LPS-injection. However, supplementation of YH to the LPS-infected piglets reduced the levels of HP, ACTH and cortisol, suggesting that YH attenuated LPS-induced stress response. To our knowledge, TNF-α and IL-1β were the critical inducers of APPs [47]. And it is a highly significant correlation between serum APPs and TNF-α levels in some infectious agents [48]. In the present investigation, LPS challenge resulted in an increase of TNF-α and IL-1β in serum. Undoubtedly, LPS caused an excessive activation of immune system. Nevertheless, YH supplementation decreased the concentrations of TNF-α and IL-1β in LPS-challenged piglets, which indicated dietary YH supplementation could decrease the systemic inflammation after LPS infection. Over-production of proinflammatory cytokines manufactured intestinal damage, such as villous atrophy, mucosal swelling, submucosal hemorrhage and exfoliation, and led to an increase of intestinal permeability [7]. Several blood indictors have been used to evaluate the intestinal permeability. DAO, an enzyme found at high levels in the mammalian intestinal mucosa, is a marker of maturation and integrity in response to intestinal epithelium [49]. The levels of serum DAO are positively correlated with intestinal permeability. In this study, YH supplementation reduced DAO concentrations, indicating that YH had beneficial effects on attenuating the increase of intestinal permeability of piglets challenged with LPS. Meanwhile, intestinal permeability can usually be assessed with intestinal epithelial barrier function [6]. Therefore, it is possible that YH supplementation can mitigate systemic inflammatory associated with the improvement of intestinal barrier function.
Intestinal morphology is regarded as a visual reflection for the growth and development of gut and determined by villus height and crypt depth [50]. This study showed that LPS-injection induced intestinal damage including villous atrophy, mucosal detachment, and a decrease of villus height in jejunum of piglets. While dietary inclusion of YH could counteract the morphological changes of jejunum after LPS challenge, thereby maintaining the villus integrity and structure of intestinal mucosa. Considering the importance of intestinal epithelial barrier in intestinal function, the tight junctions between epithelial cells were analyzed in this study. Tight junctions are multi-protein complexes including claudins, occludin and ZOs, which defend against the passage of luminal antigens, pathogenic bacterium and their toxins [6]. Recently, the experimental results showed that YH inclusion markedly attenuated the down-regulation mRNA expressions of OCC, and the protein abundances of OCC and ZO-1 in jejunal mucosa of LPS-challenged piglets. Thus, dietary YH administration could maintain the integrity of intestinal barrier by inhibiting the decrease of tight junction protein expression under immunological stress, which might be the potential reasons that YH mitigated the impairments of intestinal permeability and morphology in LPS-injected piglets.
It is generally accepted that cytokines are the critical modulators of intestinal inflammation [47]. Over-production of pro-inflammatory cytokines (e.g., TNF-α and IL-1β) has been demonstrated to directly impair the tight junctional function of some epithelial and endocrine cells [51]. Previous studies have indicated that alleviating intestinal inflammation might effectively prevent pathogenic bacteria and their toxins from disrupting intestinal barrier function [2]. Consequently, the reduction of intestinal pro-inflammatory cytokine concentrations under stressful conditions is one of the major strategies to protect the intestinal barrier and mitigate intestinal inflammation. In the present study, YH supplementation significantly reduced the mRNA expressions of TNF-α and IL-1β, and the corresponding protein abundances in jejunal mucosa in LPS challenged-piglets. Similarly, Waititu et al. [ 52] reported that piglets receiving a YH riched in cell wall polysaccharides reduced TNF-α level in ileum challenged by LPS. Hence, the protected effect of YH on intestinal barrier appears to be achieved by suppressing the inflammatory response. This is also a possible reason that YH ameliorates the systemic inflammatory response.
With a further view to investigating the molecular mechanisms of YH in the alleviation of intestinal inflammation, we evaluated the activation of TLR4 signaling pathway. TLR4, a typical pattern recognition receptor in the TLR protein family, is widely distributed on the surface of various intestinal cells and plays an essential role in LPS-mediated signaling [11]. Mechanistically, TLR4 activation triggered by LPS can induce the increased expression of downstream molecules (such as NF-κB), and then enhance the expression of proinflammatory cytokines-related genes, resulting in intestinal barrier damage [7, 53]. Currently, we observed that YH supplementation significantly decreased the protein abundance and immunofluorescence intensity of TLR4 in jejunal mucosa of piglet challenged with LPS, as well as downregulated the protein levels of p-NF-κB. NF-κB, the master transcription factor for TLR4 signaling, is phosphorylated and translocated to the nucleus in the stimulated state, then promoting the release of pro-inflammatory cytokines [54]. Accordingly, YH supplementation inhibits the TLR4/NF-κB signaling pathway, which may represent a potential mechanism to decelerate the signaling of LPS, thereby enabling the inflammatory response to be dampened.
## Conclusion
Dietary YH supplementation improves the growth performance and attenuates LPS-induced intestinal inflammation and barrier injury. The underlying molecular mechanism is YH supplementation inhibits the activation of TLR4/NF-κB signaling pathway induced by LPS, to prevent the over-production of inflammatory cytokines, and thus improved the intestinal barrier function.
## Supplementary Information
Additional file 1: Fig. S1. SDS-PAGE analysis of yeast hydrolysate. Additional file 2: Table S1. Chemical component of yeast hydrolysate.
## References
1. Rana AS, Michel B, Thomas M. **Mechanism of cytokine modulation of epithelial tight junction barrier**. *Front Biosci* (2009.0) **14** 2765. DOI: 10.2741/3413
2. Croschwitz KP, Hogan SP. **Intestinal barrier function: molecular regulation and disease pathogenesis**. *J Allergy Clin Immun* (2009.0) **124** 3-20. DOI: 10.1016/j.jaci.2009.05.038
3. Weström B, Arévalo-Sureda E, Pierzynowska K, Pierzynowski SG, Pérez-Cano FJ. **The immature gut barrier and its importance in establishing immunity in newborn mammals**. *Front Immun* (2020.0) **11** 1153. DOI: 10.3389/fimmu.2020.01153
4. Ng QX, Soh AYS, Loke W, Lim DY, Yeo WS. **The role of inflammation in irritable bowel syndrome (IBS)**. *J Inflamm Res* (2018.0) **11** 345. DOI: 10.2147/JIR.S174982
5. Barani M, Rahdar A, Sargazi S, Amiri MS, Sharma PK, Bhalla N. **Nanotechnology for inflammatory bowel disease management: Detection, imaging and treatment**. *Sensing Bio-Sensing Res* (2021.0) **32** 100417. DOI: 10.1016/j.sbsr.2021.100417
6. 6.Anderson R, Dalziel J, Gopal P, Bassett S, Ellis A, Roy N. Colitis: The role of intestinal barrier function in early life in the development of colitis. In: Fukata M, editor. New Zealand: Institute; 2012;Chapter 1:1-30.
7. Williams JM, Duckworth CA, Watson AJ, Frey MR, Miguel JC, Burkitt MD. **A mouse model of pathological small intestinal epithelial cell apoptosis and shedding induced by systemic administration of lipopolysaccharide**. *Dis Model Mech* (2013.0) **6** 1388-1399. DOI: 10.1242/dmm.013284
8. Hiippala K, Jouhten H, Ronkainen A, Hartikainen A, Kainulainen V, Jalanka J. **The potential of gut commensals in reinforcing intestinal barrier function and alleviating inflammation**. *Nutrients* (2018.0) **10** 988. DOI: 10.3390/nu10080988
9. Moeser AJ, Pohl CS, Rajput M. **Weaning stress and gastrointestinal barrier development: Implications for lifelong gut health in pigs**. *Anim Nutr* (2017.0) **3** 313-321. DOI: 10.1016/j.aninu.2017.06.003
10. Mao J, Qi S, Cui Y, Dou X, Luo XM, Liu J. *J Nutr* (2020.0) **150** 1313-1323. DOI: 10.1093/jn/nxaa009
11. Ciesielska A, Matyjek M, Kwiatkowaska K. **TLR4 and CD14 trafficking and its influence on LPS-induced pro-inflammatory signaling**. *Cell Mol Life Sci* (2021.0) **78** 1233-1261. DOI: 10.1007/s00018-020-03656-y
12. Liu YL, Wang XY, Leng WB, Pi DG, Tu ZX, Zhu HL. **Aspartate inhibits LPS-induced MAFbx and MuRF1 expression in skeletal muscle in weaned pigs by regulating Akt, AMPKα and FOXO1**. *Innate Immun* (2017.0) **23** 34-43. DOI: 10.1177/1753425916673443
13. Touchette KJ, Carroll JA, Allee GL, Matteri RL, Zannelli ME. **Effect of spray-dried plasma and lipopolysaccharide exposure on weaned pigs: I. Effects on the immune axis of weaned pigs**. *J Anim Sci* (2002.0) **80** 494-501. DOI: 10.2527/2002.802494x
14. Parapouli M, Vasileiadis A, Afendra AS, Hatziloukas E. *AIMS Microbiol* (2020.0) **6** 1. DOI: 10.3934/microbiol.2020001
15. Jung EY, Hong YH, Kim JH, Park Y, Bae SH, Chang UJ. **Effects of yeast hydrolysate on hepatic lipid metabolism in high-fat-diet-induced obese mice: yeast hydrolysate suppresses body fat accumulation by attenuating fatty acid synthesis**. *Ann Nutr Metab* (2012.0) **61** 89-94. DOI: 10.1159/00033844
16. Mosser M, Chevalot I, Olmos E, Blanchard F, Kapel R, Oriol E. **Combination of yeast hydrolysates to improve CHO cell growth and IgG production**. *Cytotechnology* (2013.0) **65** 629-641. DOI: 10.1007/s10616-012-9519-1
17. Jung EY, Lee JW, Hong YH, Chang UJ, Suh HJ. **Low dose yeast hydrolysate in treatment of obesity and weight loss**. *Prev Nutr Food Sci* (2017.0) **22** 45. DOI: 10.3746/pnf.2017.22.1.45
18. Park Y, Kim JH, Lee HS, Jung EY, Lee H, Noh DO. **Thermal stability of yeast hydrolysate as a novel anti-obesity material**. *Food Chem* (2013.0) **136** 316-321. DOI: 10.1016/j.foodchem.2012.08.047
19. Hu J, Park JW, Kim IH. **Effect of dietary supplementation with brewer’s yeast hydrolysate on growth performance, faecal microbial counts, diarrhoea score, blood profile, rectal temperature in weanling pigs challenged with lipopolysaccharide**. *J Anim Physiol Anim Nutr* (2020.0) **104** 629-636. DOI: 10.1111/jpn.13301
20. Kim JM, Lee SW, Kim KM, Chang UJ, Song JC, Suh HJ. **Anti-stress effect and functionality of yeast hydrolysate SCP-20**. *Eur Food Res Technol* (2003.0) **217** 168-172. DOI: 10.1007/s00217-003-0723-2
21. Lee HS, Jung EY, Suh HJ. **Chemical composition and anti-stress effects of yeast hydrolysate**. *J Med Food* (2009.0) **12** 1281-1285. DOI: 10.1089/jmf.2009.0098
22. Fu RQ, Liang C, Chen DW, Yan H, Tian G, Zheng P. **Effects of dietary**. *J Anim Physiol Anim Nutr* (2021.0) **105** 898-907. DOI: 10.1111/jpn.13529
23. Gong Y, Yang F, Hu J, Liu C, Liu H, Han D. **Effects of dietary yeast hydrolysate on the growth, antioxidant response, immune response and disease resistance of largemouth bass (**. *Fish Shellfish Immun.* (2019.0) **94** 548-557. DOI: 10.1016/j.fsi.2019.09.044
24. Jung EY, Lee HS, Chang UJ, Bae SH, Kwon KH, Suh HJ. **Acute and subacute toxicity of yeast hydrolysate from**. *Food Chem Toxicol* (2010.0) **48** 1677-1681. DOI: 10.1016/j.fct.2010.03.044
25. Zhang JY, Park JW, Kim IH. **Effect of supplementation with brewer’s yeast hydrolysate on growth performance, nutrients digestibility, blood profiles and meat quality in growing to finishing pigs**. *Asian-Austral J Anim Sci* (2019.0) **32** 1565. DOI: 10.5713/ajas.18.0837
26. Fu RQ, Chen DW, Tian G, Zheng P, Mao XB, Yu J. **Effect of dietary supplementation of**. *Anim Nutr* (2019.0) **5** 366-372. DOI: 10.1016/j.aninu.2019.06.003
27. Andriamialinirina HJT, Irm M, Taj S, Lou JH, Jin M, Zhou Q. **The effects of dietary yeast hydrolysate on growth, hematology, antioxidant enzyme activities and non-specific immunity of juvenile Nile tilapia, Oreochromis niloticus**. *Fish Shellfish Immun* (2020.0) **101** 168-175. DOI: 10.1016/j.fsi.2020.03.037
28. Latimer Junior G. *Official methods of analysis of AOAC International* (2016.0)
29. Wang Y, Liu J, Wei F, Liu X, Yi C, Zhang Y. **Improvement of the nutritional value, sensory properties and bioavailability of rapeseed meal fermented with mixed microorganisms**. *LWT* (2019.0) **112** 108238. DOI: 10.1016/j.lwt.2019.06.005
30. Shi C, He J, Yu J, Yu B, Mao X, Zheng P. **Physicochemical properties analysis and secretome of Aspergillus niger in fermented rapeseed meal**. *PLoS One* (2016.0) **11** e0153230. DOI: 10.1371/journal.pone.0153230
31. 31.NRC (National Academy of Sciences-National Research Council)Nutrient requirements of swine201211Washington, DCNational Academies Press. *Nutrient requirements of swine* (2012.0)
32. Liu Y, Huang J, Hou Y, Zhu H, Zhao S, Ding B. **Dietary arginine supplementation alleviates intestinal mucosal disruption induced by Escherichia coli lipopolysaccharide in weaned pigs**. *Brit J Nutr* (2008.0) **100** 552-560. DOI: 10.1017/S0007114508911612
33. Arce R, Barros S, Wacker B, Peters B, Moss K, Offenbacher S. **Increased TLR4 expression in murine placentas after oral infection with periodontal pathogens**. *Placenta* (2009.0) **30** 156-162. DOI: 10.1016/j.placenta.2008.11.017
34. Pu JN, Chen DW, Tian G, He J, Huang ZQ, Zheng P. **All-trans retinoic acid attenuates transmissible gastroenteritis virus-induced inflammation in IPEC-J2 cells via suppressing the RLRs/NF-κB signaling pathway**. *Front Immun* (2022.0) **31** 13. DOI: 10.3389/fimmu.2022.734171
35. 35.SASStatistical Analysis System Version 9.42013NC, USASAS Institute Inc. *Statistical Analysis System Version 9.4* (2013.0)
36. Kim KW, Thomas R. **Antioxidative activity of chitosans with varying molecular weights**. *Food Chem* (2007.0) **101** 308-313. DOI: 10.1016/j.foodchem.2006.01.038
37. Jung EY, Kim SY, Bae SH, Chang UJ, Choi JW, Suh HJ. **Weight reduction effects of yeast hydrolysate below 10 kDa on obese young women**. *J Food Biochem* (2011.0) **35** 337-350. DOI: 10.1111/j.1745-4514.2010.00385.x
38. Jung EY, Cho MK, Hong YH, Kim JH, Park Y, Chang UJ. **Yeast hydrolysate can reduce body weight and abdominal fat accumulation in obese adults**. *Nutrition* (2014.0) **30** 25-32. DOI: 10.1016/j.nut.2013.02.009
39. Kim JM, Kim SK, Jung EY, Bae SH, Suh HJ. **Yeast hydrolysate induces longitudinal bone growth and growth hormone release in rats**. *Phytother Res* (2009.0) **23** 731-736. DOI: 10.1002/ptr.2720
40. Molist F, van Eerden E, Parmentier H, Vuorenmaa J. **Effects of inclusion of hydrolyzed yeast on the immune response and performance of piglets after weaning**. *Anim Feed Sci Tech* (2004.0) **195** 136-141. DOI: 10.1016/j.anifeedsci.2014.04.020
41. Wang T, Cheng K, Yu C, Tong Y, Yang Z, Wang T. **Effects of yeast hydrolysate on growth performance, serum parameters, carcass traits, meat quality and antioxidant status of broiler chickens**. *J Sci Food Agric* (2021.0) **102** 575-583. DOI: 10.1002/jsfa.11386
42. Jin M, Xiong J, Zhou QC, Yuan Y, Wang XX, Sun P. **Dietary yeast hydrolysate and brewer’s yeast supplementation could enhance growth performance, innate immunity capacity and ammonia nitrogen stress resistance ability of Pacific white shrimp (**. *Fish Shellfish Immun.* (2018.0) **82** 121-129. DOI: 10.1016/j.fsi.2018.08.020
43. Perez L. **Acute phase protein response to viral infection and vaccination**. *Arch Biochem Biophys* (2019.0) **671** 196-202. DOI: 10.1016/j.abb.2019.07.013
44. Chami R, Monteleone AM, Treasure J, Monteleone P. **Stress hormones and eating disorders**. *Mol Cell Endocrinol* (2019.0) **497** 110349. DOI: 10.1016/j.mce.2018.12.009
45. Sánchez-Lemus E, Benicky J, Pavel J, Saavedra JM. **In vivo Angiotensin II AT1 receptor blockade selectively inhibits LPS-induced innate immune response and ACTH release in rat pituitary gland**. *Brain Behav Immun* (2009.0) **23** 945-957. DOI: 10.1016/j.bbi.2009.04.012
46. Hou X, Wang T, Ahmad H, Xu Z. **Ameliorative effect of ampelopsin on LPS-induced acute phase response in piglets**. *J Funct Foods* (2017.0) **35** 489-498. DOI: 10.1016/j.jff.2017.05.044
47. Luan YY, Yao YM. **The clinical significance and potential role of C-reactive protein in chronic inflammatory and neurodegenerative diseases**. *Front Immun* (2018.0) **9** 1302. DOI: 10.3389/fimmu.2018.01302
48. Moya SL, Boyle L, Lynch PB, Arkins S. **Pro-inflammatory cytokine and acute phase protein responses to low-dose lipopolysaccharide (LPS) challenge in pigs**. *Anim Sci* (2006.0) **82** 527-534. DOI: 10.1079/ASC200665
49. Luk GD, Bayless TM, Baylin SB. **Diamine oxidase (histaminase). A circulating marker for rat intestinal mucosal maturation and integrity**. *J Clin Invest* (1980.0) **66** 66-70. DOI: 10.1172/JCI109836
50. Chelakkot C, Ghim J, Ryu SH. **Mechanisms regulating intestinal barrier integrity and its pathological implications**. *Exp Mol Med* (2018.0) **50** 1-9. DOI: 10.1038/s12276-018-0126-x
51. Capaldo CT, Nusrat A. **Cytokine regulation of tight junctions**. *Biochim Biophys Acta* (2009.0) **1788** 864-871. DOI: 10.1016/j.bbamem.2008.08.027
52. Waititu SM, Yin F, Patterson R, Rodriguez-Lecompte JC, Nyachoti CM. **Short-term effect of supplemental yeast extract without or with feed enzymes on growth performance, immune status and gut structure of weaned pigs challenged with Escherichia coli lipopolysaccharide**. *J Anim Sci Biotechn* (2016.0) **7** 1-13. DOI: 10.1186/s40104-016-0125-5
53. Guo S, Al-Sadi R, Said HM, Ma TY. **Lipopolysaccharide causes an increase in intestinal tight junction permeability in vitro and in vivo by inducing enterocyte membrane expression and localization of TLR-4 and CD14**. *Am J Pathol* (2013.0) **182** 375-387. DOI: 10.1016/j.ajpath.2012.10.014
54. Tang J, Xu L, Zeng Y, Gong F. **Effect of gut microbiota on LPS-induced acute lung injury by regulating the TLR4/NF-kB signaling pathway**. *Int Immunopharmacol* (2021.0) **91** 107272. DOI: 10.1016/j.intimp.2020.107272
|
---
title: 'Older persons experiences of healthcare in rural Burkina Faso: Results of
a cross sectional household survey'
authors:
- Ellen M. Goldberg
- Mamadou Bountogo
- Guy Harling
- Till Baernighausen
- Justine I. Davies
- Lisa R. Hirschhorn
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021992
doi: 10.1371/journal.pgph.0000193
license: CC BY 4.0
---
# Older persons experiences of healthcare in rural Burkina Faso: Results of a cross sectional household survey
## Abstract
Ensuring responsive healthcare which meets patient expectations and generates trust is important to increase rates of access and retention. This need is important for aging populations where non-communicable diseases (NCDs) are a growing cause of morbidity and mortality. We performed a cross-sectional household survey including socio-demographic; morbidities; and patient-reported health system utilization, responsiveness, and quality outcomes in individuals 40 and older in northwestern Burkina Faso. We describe results and use exploratory factor analysis to derive a contextually appropriate grouping of health system responsiveness (HSR) variables. We used linear or logistic regression to explore associations between socio-demographics, morbidities, and the grouped-variable, then between these variables and health system quality outcomes. Of 2,639 eligible respondents, $26.8\%$ had least one NCD, $56.3\%$ were frail or pre-frail and $23.9\%$ had a recent healthcare visit, including only $\frac{1}{3}$ of those with an NCD. Highest ratings of care experience (excellent/very good) included ease of following instructions ($86.1\%$) and trust in provider skills ($81.1\%$). The HSR grouping with the greatest factor loading included involvement in decision-making, clarity in communication, trust in the provider, and confidence in providers’ skills, labelled Shared Understanding and Decision Making (SUDM). In multivariable analysis, higher quality of life (OR 1.02,$95\%$CI 1.01–1.04), frailty (OR 1.47,$95\%$CI 1.00–2.16), and SUDM (OR 1.06,$95\%$CI 1.05–1.09) were associated with greater health system trust and confidence. SUDM was associated with overall positive assessment of the healthcare system (OR 1.02,$95\%$CI 1.01–1.03) and met healthcare needs (OR 1.09,$95\%$CI 1.08–1.11). Younger age and highest wealth quintile were also associated with higher met needs. Recent healthcare access was low for people with existing NCDs, and SUDM was the most consistent factor associated with higher health system quality outcomes. Results highlight the need to increase continuity of care for aging populations with NCDs and explore strengthening SUDM to achieve this goal.
## Introduction
As access to care has improved in low and middle income country (LMIC) settings, understanding and ensuring the quality of this care has emerged as a critical step to reach effective universal health coverage and health-related sustainable development goals [1]. The Institute of Medicine (IOM) has defined six domains of quality, including effectiveness (often measured by technical quality), safety, timeliness, equity, efficiency, and patient-centeredness [2]. Patient-centeredness has been further emphasized through the World Health Organization’s (WHO) initiative for Integrated People Centered Health Care, which puts the patient at the center of the health care system, and is a core outcome in the framework from the Lancet Global Health Commission on High Quality Health Systems [1, 3].
Poor quality in any of the IOM domains is now a leading cause of preventable mortality, overtaking access as a major cause; poor quality contributes to a persistent equity gap and results in costs to the individual, health care system, and society [1, 4]. Gaps in quality are particularly apparent in non-communicable diseases (NCDs), which represent a growing burden across all countries as populations age [5]. Multiple studies are now showing the magnitude of gaps in both overall quality of care and resulting clinical outcomes (corresponding to technical outcomes, e.g. having a condition recognized and adequately managed) in NCDs and among older individuals [1, 6–10].
Receipt of person-centered care, a key IOM domain, and focus of initiatives to improve health care more broadly have been associated with improved healthcare utilization, better health outcomes, and patient safety. In contrast, poor experiences and perceived quality due to non-responsive care is associated with delay in accessing or returning to care or bypassing the formal care system, whether because of personal experience or through word-of-mouth [11–13]. Confidence and trust in the health system and overall satisfaction with care are also important quality outcomes of the care system, critical for ensuring willingness to access and return to care and consequently for the management of chronic conditions which are more frequent in populations as they age [14–16].
Measurement around patient-centeredness builds on the WHO Health Systems Responsiveness Framework which identified seven components of responsive outpatient care: dignity, confidentiality, involvement in decision making (autonomy), communication, choice of provider, prompt attention, and quality of basic amenities [17]. Larson directly linked health system responsiveness to experiential quality of care and proposed two areas for measurement: [1] patient experience of care, a process measure; and [2] patient satisfaction, a health system quality outcome measure of how well provided care meets patient needs and expectations [18]. The relationship between components of responsiveness of care and the health system quality outcomes is not well described, although recent work from Ghana found that higher reported responsiveness was associated with improved measures of outcomes including reported met medical needs (a measure of satisfaction with care) and confidence in the health care system [19].
As health burden and care needs continue to shift to individuals as they age and risk of NCDs increases, there is a need to expand the measurement of quality beyond providing technically correct treatment to care which is also empowering and meets these patients’ needs through shared decision making (SDM) [20]. SDM has been demonstrated to be important for improving self-management and care outcomes, including among people with risk factors for existing chronic conditions [21]. SDM involves the patient and provider collaborating through better communication to identify preferences and make treatment choices that meet the patient’s goals. This approach addresses health system responsiveness domains including autonomy, communication, and trust between the patient and provider.
Burkina *Faso is* one of the poorest countries in the world, with $43.8\%$ of the population living in extreme poverty [22]. Health care expenditure as a percentage of GDP has increased since 2000, reaching $7\%$ by 2016, but out of pocket sources still contribute a large amount to healthcare funding [23]. The population is aging and NCDs are increasing, now estimated to account for $24\%$ deaths in Burkina Faso [24–30]. Gaps in both screening and care-seeking for NCDs and those at risk, including older individuals, is also noted to be of concern. For example, Cisse et al. reported rates of hypertension of close to $13\%$ with very low rates of diagnosis and treatment and Wagner et al. reported low rates of care seeking among individuals with high cardiovascular risk factors [29, 30].
The formal public health system within the district level includes primary care centers (known as a Center for Health and Social Promotion (CSPS)) and a district hospital (known as a medical center with surgical antennae) as well as private clinics and pharmacies. Health services in Burkina Faso have historically been tailored towards maternal and child health and infectious diseases. However, there is increasing attention being given to NCDs, including establishment of an NCD division in Ministry of Health (MOH) and a national integrated NCD policy [31], as well as a strategic plan (2016–2020) which included goals of strengthening healthcare quality coordination for the elderly (M Bountogo, personal communication).
We describe the causes of recent healthcare seeking and reported experiences of care in public sector primary and secondary level facilities among adults aged 40 and older in Nouna, a rural region in Burkina Faso. These results are important for providers and policy makers in Burkina Faso and similar settings to facilitate improved experiences of care to increase care seeking and retention of the aging population and begin to reverse the growing burden of NCD-related morbidity and mortality. While there have been different definitions of the age cut-off for older adults, in Burkina Faso, life expectancy is 62 years, hence people 40 years of age or older are considered “older” in this study relative to the ages of other people in the population. This is also the age above which WHO recognizes people to be at increased risk of cardiovascular disease risk factors (for example diabetes and hypertension)–prominent NCDs of aging [32]. Addressing gaps in all domains of healthcare quality is required to respond to the needs of this aging population and reach the goals of global Healthy *Aging agenda* [33].
## Study setting
The study was set in the Nouna Health and Demographic Surveillance System (HDSS) area, led by the Centre de Recherche en Santé de Nouna (CRSN) in the Boucle du Mouhoun region, north-western Burkina Faso. The demographic surveillance area of the Nouna HDSS consists of the market town of Nouna and 59 surrounding villages with a total population of 107 000 [34].
## Data collection
Data for this cross-sectional study were obtained during the baseline wave of the CRSN Heidelberg Aging Study (CHAS) and survey and data collection procedures have been described in detail elsewhere—the survey instrument is included as S1 Study Tool [28]. Briefly, we randomly sampled 4000 older adults (aged 40 years or older) from the 2015 CRSN census population. In villages with more than 90 adults aged 40 or older, a random sample of households with at least one age-eligible person was created, and one age-eligible adult in each selected household was randomly selected to complete the survey. In villages with fewer than 90 adults aged 40 or older, all households with one or more age-eligible individual were included. Data were collected using Open Data Kit (ODK) software on tablet computers at the participants’ houses between May and July 2018. Interviews were conducted either in French or translated into the local languages of Dioula or Mooré by the interviewers.
The household survey contained questions on sociodemographics; self-reported presence of diseases or other health conditions; visits to a healthcare provider for themselves in the last 3 months and facility-type last attended; reasons for last health facility visit; reasons for not attending a facility in the last three months; and selected measures of health system responsiveness and health system quality outcomes (Table 1). Other measures included Anxiety (measured using the Generalized Anxiety Disorder question (GAD-2) score) [35], depression (using Patient Health Questionnaire (PHQ- 9)) [36] and Quality of life (measured using the validated EuroHIS 8-item version of WHOQOL) [37]. Disability was measured using the 12 item WHO Disability Assessment Schedule, version 2 (WHODAS-II) disability score [38]. Both WHO DAS and QOL were scaled between 0–100 with 0 representing the lowest and 100 the highest values, as is standard for these scores [28]. Cognitive functioning was assessed using CSI-D [39]. The Fried frailty score was constructed as described previously [40].
**Table 1**
| Unnamed: 0 | Unnamed: 1 | Overall population | Attended facility in last 3 months | Did not attend facility in last 3 months | P-value |
| --- | --- | --- | --- | --- | --- |
| | Group | N (%) | N (%) | N (%) | |
| Total | | 2639 | 632 | 2007 | |
| Sex | Female | 1315 (49.8) | 338 (53.5) | 977 (48.7) | 0.035 |
| Age | 40–49 | 1141 (43.2) | 271 (42.9) | 870 (43.3) | 0.079 |
| Age | 50–59 | 755 (28.6) | 148 (23.4) | 607 (30.2) | |
| Age | 60–69 | 475 (18) | 145 (22.9) | 330 (16.4) | |
| Age | 70–79 | 217 (8.2) | 61 (9.7) | 156 (7.8) | |
| Age | 80+ | 53 (2) | 11 (1.7) | 42 (2.1) | |
| Educational attainment | No formal schooling | 2215 (83.9) | 515 (81.5) | 1700 (84.7) | 0.055 |
| Educational attainment | Some education (any primary school or higher) | 424 (16.1) | 117 (18.5) | 307 (15.3) | |
| Marital status | Widowed/divorced/single | 606 (23) | 164 (25.9) | 442 (22) | 0.041 |
| Marital status | Married or cohabitating | 2033 (77) | 468 (74.1) | 1565 (78) | |
| Wealth quintile | 1 (least wealthy) | 499 (18.9) | 103 (16.3) | 396 (19.7) | <0.0001* |
| Wealth quintile | 2 | 522 (19.8) | 103 (16.3) | 419 (20.9) | |
| Wealth quintile | 3 | 525 (19.9) | 124 (19.6) | 401 (20) | |
| Wealth quintile | 4 | 549 (20.8) | 132 (20.9) | 417 (20.8) | |
| Wealth quintile | 5 (most wealthy) | 544 (20.6) | 170 (26.9) | 374 (18.6) | |
| Self-reported non-communicable diseases (NCD)† | Hypertension | 463 (17.5) | 171 (27.1) | 292 (14.5) | |
| Self-reported non-communicable diseases (NCD)† | Diabetes | 62 (2.3) | 18 (2.8) | 44 (2.2) | |
| Self-reported non-communicable diseases (NCD)† | Hypercholesterolemia | 11 (.4) | 7 (1.1) | 4 (.2) | |
| Self-reported non-communicable diseases (NCD)† | Heart disease | 163 (6.2) | 61 (9.7) | 102 (5.1) | |
| Self-reported non-communicable diseases (NCD)† | Stroke | 36 (1.4) | 12 (1.9) | 24 (1.2) | |
| Self-reported non-communicable diseases (NCD)† | Chronic respiratory disease | 92 (3.5) | 33 (5.2) | 59 (2.9) | |
| Self-reported non-communicable diseases (NCD)† | Cancer | 14 (.5) | 9 (1.4) | 5 (.2) | |
| Self-reported non-communicable diseases (NCD)† | ≥1 NCD | 708 (26.8) | 250 (39.6) | 458 (22.8) | <0.0001* |
| Self-reported TB or HIV† | TB | 26 (1) | 12 (1.9) | 14 (.7) | |
| Self-reported TB or HIV† | HIV | 16 (.61) | 6 (.9) | 10 (.5) | |
| Self-reported TB or HIV† | HIV and/or TB | 41 (1.55) | 17 (2.69) | 24 (1.2) | 0.0001* |
| Self-reported other conditions for > 3 months† | Cough | 17 (.6) | 8 (1.3) | 9 (.4) | |
| Self-reported other conditions for > 3 months† | Headache or dizziness | 50 (1.9) | 16 (2.5) | 34 (1.7) | |
| Self-reported other conditions for > 3 months† | Musculoskeletal or back pain | 189 (7.2) | 60 (9.5) | 129 (6.4) | |
| Self-reported other conditions for > 3 months† | Dental | 17 (.6) | 5 (.8) | 12 (.6) | |
| Self-reported other conditions for > 3 months† | Gastrointestinal | 85 (3.2) | 40 (6.3) | 45 (2.2) | |
| Self-reported other conditions for > 3 months† | ≥1 other condition | 502 (19) | 181 (28.6) | 321 (16) | <0.0001* |
| Symptoms of mental health disorders (MHD)† | Cognitive impairment on testing | 163 (6.2) | 36 (5.7) | 127 (6.3) | |
| Symptoms of mental health disorders (MHD)† | Symptoms of anxiety on testing | 301 (11.4) | 89 (14.1) | 212 (10.6) | |
| Symptoms of mental health disorders (MHD)† | Depressive symptoms on testing | 202 (7.7) | 55 (8.7) | 147 (7.3) | |
| Symptoms of mental health disorders (MHD)† | >1 MHD | 518 (19.6) | 142 (22.5) | 376 (18.7) | 0.039 |
| Frailty | Not frail | 1163 (44.1) | 233 (36.9) | 930 (46.3) | <0.0001* |
| Frailty | Frail/pre-frail | 1476 (55.9) | 399 (63.1) | 1077 (53.7) | |
| Disability | WHO DAS s†† | 8.3 (0–20.8) | 14.6 (4.2–27.1) | 6.3 (0–18.8) | <0.0001* |
| Quality of life | WHO QoL †† | 59.4 (46.9–65.6) | 56.3 (43.8–65.6) | 59.4 (46.9–68.8) | <0.0001* |
| Facility type for last visit | Center for Health and Social Promotion | 2206 (83.6) | 496 (78.5) | 1710 (85.2) | <0.0001* |
| Facility type for last visit | District Hospital | 433 (16.4) | 136 (21.5) | 297 (14.8) | |
| Financial access | Did not borrow or sell anything | 2250 (85.3) | 539 (85.3) | 1711 (85.3) | 0.98 |
| Financial access | Borrowed or sold something to attend clinic | 389 (14.7) | 93 (14.7) | 296 (14.7) | |
| Health System Quality Outcomes | | | | | |
| Reported met need** | Excellent | 234 (8.9) | 56 (8.9) | 178 (8.9) | 0.0058* |
| Reported met need** | Very Good | 968 (36.7) | 262 (41.5) | 706 (35.2) | |
| Reported met need** | Good | 1293 (49) | 275 (43.5) | 1018 (50.7) | |
| Reported met need** | Fair | 116 (4.4) | 33 (5.2) | 83 (4.1) | |
| Reported met need** | Poor | 28 (1.1) | 6 (.9) | 22 (1.1) | |
| Trust and confidence in health care system*** | Very confident | 872 (33) | 246 (38.9) | 626 (31.2) | 0.0003* |
| Trust and confidence in health care system*** | Somewhat confident | 1610 (61) | 358 (56.6) | 1252 (62.4) | |
| Trust and confidence in health care system*** | Not very confident | 138 (5.2) | 26 (4.1) | 112 (5.6) | |
| Trust and confidence in health care system*** | Not at all confident | 19 (.7) | 2 (.3) | 17 (.8) | |
| Overall view of the health care system in this country**** | Positive | 1612 (61.1) | 408 (64.6) | 1204 (60) | |
| Overall view of the health care system in this country**** | Neutral | 956 (36.2) | 212 (33.5) | 744 (37.1) | |
| Overall view of the health care system in this country**** | Poor | 71 (2.7) | 12 (1.9) | 59 (2.9) | 0.040 |
## Health system responsiveness and health system quality outcomes
A subset of all possible health system responsiveness domains was included due to constraints of the survey length. Questions were selected based on discussion between investigators and their perceived relevance to the local context and focus on experiential quality. They were taken from published studies in sub-Saharan Africa (Table A in S1 Appendix) [19, 41, 42]. Health system quality outcome questions included trust and confidence in receiving effective treatment, patient satisfaction (how well the received care met health need), and the overall view of the health system.
## Demographic characteristics
Marital status was categorized as married/cohabiting versus single/widowed/divorced. Educational level was dichotomized as no education or any education (any primary school or higher). Participants were asked 37 questions on household assets and dwelling characteristics; from these, wealth quintiles were derived from the Filmer and Pritchett first principal component method [43]. Age was categorized in 10-year groups for the descriptive and univariate analysis and as a continuous variable in the multivariable analyses.
## Disease categories
We included several self-reported conditions including non-communicable conditions (hypertension, diabetes, hypercholesterolemia, heart disease, stroke, chronic respiratory disease, and epilepsy), and communicable diseases (HIV and tuberculosis (TB)). Self-reported chronic symptoms (lasting for more than 3 months) included cough, headache, musculoskeletal or back pain, dental, or gastrointestinal manifestations. Some health conditions were captured as free text; these were translated and categorized through discussions among authors where necessary.
Participants were defined as having symptoms of anxiety based on a GAD-2 score ≥3, depression based on PHQ-9 score ≥ 10, and cognitive functioning was defined as possible/probable cognitive impairment for CSI-D score <7. Participants with at least one symptom of anxiety, depression, or cognitive impairment on testing were categorized as having a neurological or mental health diagnosis. WHODAS-II and quality of life were normalized to 0–100. For frailty, participants were dichotomized as robust versus prefrail/frail/unable to complete assessment.
## Analytic sample
We limited our sample to those who sought care at their last visit from a public sector primary (Center for Health and Social Promotion) or secondary level (District Hospital) facility to reflect our focus on local care seeking and the most common sources of care ($93\%$ of individuals for the variables of interest (see CONSORT diagram S1 Fig)). Using unweighted data, we described demographic characteristics, disease state, visit characteristics, and health system outcomes both among the whole sample surveyed and separately for participants who recently sought care (within the last 3 months) and those who did not. We used a Bonferroni correction to adjust for multiple comparisons.
## Health system responsiveness and health system quality outcomes among recent care users
We conducted an exploratory factor analysis of the experiential quality questions (S2 Fig) to explore grouping of these variables based on our assumption that one or more common constructs related to engagement in care and health system quality outcomes underlay our observed variables. We first scaled all HSR variables from 0 to 1, 1 representing the highest possible rating with wait time capped at 4 hours and consultation time capped at 1 hour based on a histogram of responses. We then ran a factor analysis with the scaled HSR variables as a measure of construct validity and used an eigenvalue cutoff of ≥1.0 for retained factors. We used a factor loading cutoff of ≥0.40 for individual variables within the qualifying factors. We used the resultant composite variable in subsequent analyses by scaling each individual variable to 0–100 with 0 representing the lowest and 100 the highest possible rating and averaged them to arrive at a final variable between 0–100.
## Bivariate analyses
We described individual HSR process ratings among recent care seekers. We limited these analyses to individuals with a visit in the last 3 months to reduce recall bias as we were not able to determine if less recent visits had occurred more than a year ago, which is the maximum duration used when assessing HSR [17]. We then tested for bivariate associations between demographic characteristics, health status (one or more self-reported NCD, one or more self-reported “other” condition, one or more symptom of mental health disorder, quality of life, frailty, and disability), facility type, financial access, wait time, and the HSR-group variable. We conducted similar analyses including the HSR-group variable with each of the three health system quality outcomes as the dependent variable of interest. Finally, we compared HSR variables dependent on whether the health facility visit occurred more or less than three months ago.
## Multivariable analyses
We ran logistic regressions and a generalized linear model regression (with a gaussian distribution and log link) for health system quality outcomes and the HSR-group variable respectively. Variables that met an inclusion criterion of $P \leq 0.2$ in the bivariate analyses were included. We also included age, sex, educational attainment, and wealth quintile, given their associations with reported experiential quality and selected health systems quality outcomes in previous studies [19, 44–46].
All statistical analyses were performed using Stata software (version 15.1; StataCorp LLC, College Station, Texas).
IRB: Ethical approval was obtained from Ethics Commission of the medical faculty Heidelberg (S-$\frac{120}{2018}$), the Burkina Faso Comité d’Ethique pour la Recherche en Santé (CERS) in Ouagadougou [2018-4-045] and the Institutional Ethics Committee (CIE) of the CRSN (2018–04). CRSN colleagues approached village leadership identified through existing channels (e.g., from the census and past studies), informed them about the study aims and activities, and obtained approval to come into the village to conduct the work. Written informed consent was obtained from each participant and a literate witness assisted in cases of illiteracy. Participants with abnormal results were contacted and referred for clinical care based on specifications determined in collaboration with the health system authorities. Medical services were also alerted of the conduct of the study and that they may receive patients as a result of the study.
Patients or the public were not involved in the design, conduct, reporting, or dissemination plans of our research.
## Population
Overall, 3,028 individuals responded to the survey including questions about care seeking with 177 excluded for missing data and 212 for care at a private sector facility or tertiary care hospital (S1 Fig). Among the 2,639 who reported their last visit to a public sector primary or secondary level facility, 632 ($23.9\%$) sought care at one of these facilities in the 3 months prior to the survey (Table 1). Overall, one half ($50\%$) were women, with $42.8\%$ age 40–49 and $10.5\%$ age 70 or older. Education was low ($83.8\%$ reported no formal education), and three quarters ($76.4\%$) were married or cohabitating. One quarter reported at least one NCD ($26.8\%$), with lower rates of communicable diseases such as HIV or TB ($2.8\%$). The median WHO DAS score was 8.3 (interquartile range (IQR) 0–22.9) and for QoL was 59.4 (IQR 46.9–65.6), while $56.3\%$ were categorized as frail or pre-frail.
Individuals who attended care in the last 3 months were significantly wealthier than those who did not attend care in this timeframe, and were more likely to have at least one NCD, either HIV or TB or both, or other conditions lasting for more than 3 months. Despite individuals with chronic diseases having attended clinic more recently, $65\%$ of respondents with these conditions did not report attending care in this timeframe, including $62.7\%$ of patients reporting hypertension and $66.7\%$ of individuals reporting diabetes.
People who had attended in the last 3 months also had significantly higher disability measured by DAS scores (14.6 versus 6.3), lower QoL (56.3 versus 59.4), and were more likely to be frail or pre-frail ($63.1\%$ versus $53.7\%$) than those with no visits in the last 3 months; all $p \leq .0001.$
## Visits characteristics
The most common reasons overall for seeking care were for acute conditions ($79.1\%$) including fever or malaria ($51.6\%$), musculoskeletal pain ($9.6\%$), and diarrhea or stomach-ache ($8.4\%$). Chronic conditions accounted for care seeking in $12.9\%$ including hypertension ($6.2\%$), other cardiac conditions ($2.1\%$) and diabetes ($0.6\%$) (Table B in S1 Appendix). The most common reasons for care-seeking within the past 3 months were fever or malaria (37.8), high blood pressure ($12.8\%$), musculoskeletal pain ($12.0\%$), complaints related to the ear, nose, or throat ($7.4\%$), or diarrhea or stomach-ache ($7.0\%$). Not being sick was the most frequent reason for no recent care-seeking ($87.3\%$) (Table C in S1 Appendix). Among those who stated other reasons for not seeking care, cost was the most common reason ($50.4\%$), followed by preferring to see a traditional healer ($11.6\%$) and poor previous experiences with the health system ($6.0\%$).
## Health system quality outcomes
Overall, $32.7\%$ of respondents were very confident that if they got sick, the health system could meet their needs. Compared with individuals with a visit over 3 months ago, individuals with recent visits had higher trust and confidence in the health system to provide effective care if they were sick ($38.3\%$ versus $30.8\%$ very confident, $p \leq .0004$), although rates remained low. No differences were seen in their needs being met from their last visit or in overall opinion of the national health system (Table 1).
## Experiences of care at facilities (Health system responsiveness variables)
Among individuals with a visit to a public sector primary (CSPS) or secondary level (district hospital) public facility in the last 3 months, the median wait time was 20 minutes (IQR 10–30) while time spent with the provider was 15 minutes (IQR 10–25). Financial access was a challenge with $14.7\%$ borrowing money or selling something to pay for health care. The highest ratings of experience of care (defined as excellent or very good) were in ease of following instructions ($86.1\%$) and trust in the skills and abilities of the facility providers ($81.1\%$). Lower ratings were seen for provider medical knowledge and skills ($51.2\%$), clarity of communications ($48.2\%$), with the lowest ratings in involvement in decision making ($30.7\%$) (Table 2). Individuals without a recent facility visit reported lower ratings in clarity of communication, involvement in decision making, and trust in the skills and ability of the providers from their last visit (see S1 Appendix).
**Table 2**
| Measure | Rating | N (%) |
| --- | --- | --- |
| Clarity of provider communication | Excellent | 45 (7.1) |
| Clarity of provider communication | Very Good | 260 (41.1) |
| Clarity of provider communication | Good | 287 (45.4) |
| Clarity of provider communication | Fair | 39 (6.2) |
| Clarity of provider communication | Poor | 1 (.2) |
| Ease of following provider advice | Excellent | 81 (12.8) |
| Ease of following provider advice | Very Good | 463 (73.3) |
| Ease of following provider advice | Good | 76 (12) |
| Ease of following provider advice | Fair | 11 (1.7) |
| Ease of following provider advice | Poor | 1 (.2) |
| Provider medical knowledge and skills | Excellent | 64 (10.1) |
| Provider medical knowledge and skills | Very Good | 259 (41) |
| Provider medical knowledge and skills | Good | 280 (44.3) |
| Provider medical knowledge and skills | Fair | 27 (4.3) |
| Provider medical knowledge and skills | Poor | 2 (.3) |
| Trust in skills and abilities of health workers at the facility | Very much | 104 (16.5) |
| Trust in skills and abilities of health workers at the facility | Quite a bit | 408 (64.6) |
| Trust in skills and abilities of health workers at the facility | Some | 105 (16.6) |
| Trust in skills and abilities of health workers at the facility | Very little | 13 (2.1) |
| Trust in skills and abilities of health workers at the facility | Not at all | 2 (.3) |
| Involvement in decision making | Excellent | 35 (5.5) |
| Involvement in decision making | Very Good | 159 (25.2) |
| Involvement in decision making | Good | 268 (42.4) |
| Involvement in decision making | Fair | 101 (16) |
| Involvement in decision making | Poor | 69 (10.9) |
| Shared Understanding and Decision Making (SUDM) | Median (Interquartile range) | 62.5 (50–75) |
| Borrowed money or sold anything to pay for health care | Yes | 93 (14.7) |
| Borrowed money or sold anything to pay for health care | | |
| Borrowed money or sold anything to pay for health care | No | 539 (85.3) |
| Wait time (median, IQR) | | 20 (10–30) |
The variable grouping with the greatest factor loading (the HSR-group variable) combined the results for questions on involvement in decision-making (autonomy), clarity in communication, trust in the provider, and confidence in providers’ skills (factor loadings of 0.44, 0.73, 0.57, and 0.69, respectively) (S2 Fig). After discussion between authors, we agreed that these variables reflected components necessary for shared understanding and decision making and termed the resultant variable as such (SUDM). We used the scaled variable as described in the methods and chose to not weight variable components as all were assumed to be equally important for SUDM. The median score for SUDM was 58.3 (Interquartile range (IQR) 50–75). In a multivariable analysis, only being seen in a district hospital was associated with higher SUDM (regression coefficient (β) 5.91 ($95\%$ CI 2.87–8.96)) (Table 3).
**Table 3**
| Unnamed: 0 | Unnamed: 1 | Bivariate analysis coefficient (95% CI) | P value | Multivariable analysis Coefficient (95% CI) | P value.1 |
| --- | --- | --- | --- | --- | --- |
| Sex | Male | Reference | | Reference | |
| Sex | Female | -1.74 (-4.01–0.54) | 0.13 | -2.43 (-4.76 - -0.11) | 0.040 |
| Age (per year) | | 0.12 (0.016–0.22) | 0.023 | 0.072 (-0.046–0.19) | 0.23 |
| Educational attainment | No formal schooling | Reference | | Reference | |
| Educational attainment | Some education | 0.32 (-2.61–3.24) | 0.83 | -0.57 (-3.64–2.50) | 0.72 |
| Marital status | Widowed/divorced/single | Reference | | | |
| Marital status | Married/cohabitating | -0.44 (-3.03–2.15) | 0.74 | | |
| Wealth quintile† | 1 | Reference | | Reference | |
| Wealth quintile† | 2 | 0.24 (-3.87–4.36) | 0.91 | -0.33 (-4.28–3.62) | 0.87 |
| Wealth quintile† | 3 | 0.37 (-3.45–4.19) | 0.85 | 0.15 (-3.64–3.94) | 0.94 |
| Wealth quintile† | 4 | -0.76 (-4.64–3.12) | 0.70 | -1.12 (-4.88–2.64) | 0.56 |
| Wealth quintile† | 5 | 1.38 (-2.31–5.07) | 0.46 | -0.13 (-3.87–3.60) | 0.95 |
| Facility type | Center for Health and Social Promotion | Reference | | Reference | |
| Facility type | Medical Center with Surgical Antenna | 6.04 (3.32–8.76) | <0.001 | 5.48 (2.58–8.38) | < 0.001 |
| Financial Accessibility | Borrowed or sold anything to attend clinic | Reference | | | |
| Financial Accessibility | Did not borrow or sell anything | 1.06 (-2.14–4.27) | 0.56 | | |
| Non-communicable diseases (NCD) | No NCDs | Reference | | Reference | |
| Non-communicable diseases (NCD) | ≥1 NCD | 1.94 (-0.38–4.26) | 0.10 | 1.28 (-1.09–3.64) | 0.29 |
| TB or HIV | No TB or HIV | Reference | | | |
| TB or HIV | TB and/or HIV | -4.05 (-11.07–2.96) | 0.26 | | |
| Other conditions for > 3 months | No other conditions | Reference | | | |
| Other conditions for > 3 months | ≥1 other condition | -0.88 (-3.39–1.63) | 0.49 | | |
| Mental health disorders (MHD) | No MHDs | Reference | | Reference | |
| Mental health disorders (MHD) | ≥1 MHD | 2.51 (-0.20–5.23) | 0.070 | 1.53 (-1.39–4.45) | 0.30 |
| Frailty | Not frail | Reference | | | |
| Frailty | Pre-frail/Frail | 1.16 (-1.19–3.51) | 0.33 | | |
| Disability | WHO DAS score†† | 0.063 (0.00077–0.12) | 0.047 | 0.013 (-0.06–0.09) | 0.74 |
| Quality of life | WHO QoL score †† | -0.036 (-0.11–0.039) | 0.34 | | |
## Factors associated with health system quality outcomes
In the multivariable analysis, higher quality of life (OR 1.02, $95\%$ CI 1.01–1.04), frailty (OR 1.47, $95\%$ CI 1.00–2.16), and SUDM (OR 1.06, $95\%$ CI 1.05–1.09) were all associated with greater trust and confidence in the health system (Table 4). SUDM was associated with overall positive assessment of the health care system in Burkina Faso (OR 1.02, $95\%$ CI 1.01–1.03) and met healthcare needs in the last visit (OR 1.09, $95\%$ CI 1.08–1.11). Younger age and highest wealth quintile were also associated with higher scores for met needs, while having at least one mental health condition was associated with less positive ratings of the overall health system.
**Table 4**
| Unnamed: 0 | Unnamed: 1 | Trust and Confidence in healthcare system | Trust and Confidence in healthcare system.1 | Overall view of healthcare system | Overall view of healthcare system.1 | Health care needs met | Health care needs met.1 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| | | Bivariate analysis | Multivariable analysis | Bivariate analysis | Multivariable analysis | Bivariate analysis | Multivariable analysis |
| | | OR (95% CI) | OR (96% CI) | OR (95% CI) | | OR (95% CI) | |
| Sex | Male | Reference | Reference | Reference | Reference | Reference | Reference |
| Sex | Female | 0.73 (0.53–1.01) | 0.83 (0.58–1.19) | 0.89 (0.64–1.23) | 0.99 (0.70–1.40) | 0.86 (0.63–1.17) | 1.03 (0.71–1.49) |
| Age* | | 1.00 (0.98–1.01) | 1.00 (0.98–1.01) | 1.00 (0.99–1.01) | 1.01 (0.99–1.03) | 0.99 (0.98–1.01) | 0.98 (0.97–1.00) |
| Education | No formal schooling | Reference | Reference | Reference | Reference | Reference | Reference |
| Education | Some education | 1.02 (0.68–1.54) | 0.97 (0.60–1.56) | 0.93 (0.61–1.42) | 1.07 (0.67–1.70) | 1.09 (0.73–1.63) | 0.92 (0.56–1.50) |
| Wealth quintile | 1 | Reference | Reference | Reference | Reference | Reference | Reference |
| Wealth quintile | 2 | 1.58 (0.90–2.79) | 1.49 (0.79–2.81) | 1.00 (0.56–1.77) | 1.00 (0.55–1.80) | 1.17 (0.68–2.03) | 1.24 (0.65–2.35) |
| Wealth quintile | 3 | 1.86 (1.08–3.21) | 1.60 (0.87–2.94) | 1.05 (0.61–1.82) | 1.03 (0.58–1.82) | 1.53 (0.90–2.58) | 1.58 (0.85–2.94) |
| Wealth quintile | 4 | 1.13 (0.66–1.96) | 0.99 (0.53–1.83) | 1.24 (0.71–2.14) | 1.27 (0.72–2.25) | 1.19 (0.71–2.00) | 1.25 (0.68–2.30) |
| Wealth quintile | 5 | 1.31 (0.78–2.20) | 0.88 (0.48–1.63) | 0.77 (0.46–1.27) | 0.77 (0.44–1.34) | 1.83 (1.11–2.99) | 1.85 (1.00–3.42) |
| Facility | CSPS | Reference | | Reference | Reference | Reference | Reference |
| Facility | District Hospital | 1.13 (0.77–1.66) | | 0.68 (0.46–1.00) | 0.70 (0.46–1.08) | 1.43 (0.98–2.10) | 0.90 (0.56–1.45) |
| Financial access | Did not borrow/sell anything | Reference | | Reference | | Reference | |
| Financial access | Borrowed/sold something | 0.94 (0.60–1.48) | | 0.85 (0.54–1.33) | | 1.24 (0.80–1.92) | |
| NCDs | | Reference | | Reference | | Reference | |
| NCDs | ≥1 NCD | 1.05 (0.76–1.45) | | 0.86 (0.61–1.19) | | 1.21 (0.88–1.67) | |
| HIV or TB | | Reference | Reference | Reference | | Reference | |
| HIV or TB | HIV and/or TB | 0.47 (0.15–1.47) | 0.64 (0.19–2.18) | 0.77 (0.29–2.08) | | 0.53 (0.19–1.45) | |
| MHD | | Reference | | Reference | Reference | Reference | |
| MHD | ≥1 MHD | 1.07 (0.73–1.56) | | 0.51 (0.35–0.74) | 0.52 (0.34–0.80) | 1.18 (0.81–1.72) | |
| Frailty | Not frail | Reference | Reference | Reference | Reference | Reference | |
| Frailty | Pre-frail/Frail | 1.32 (0.95–1.85) | 1.47 (1.00–2.16) | 0.75 (0.53–1.06) | 0.83 (0.57–1.21) | 1.12 (0.81–1.55) | |
| Disability (DAS) | DAS score†† | 1.00 (0.99–1.01) | | 0.99 (0.98–1.00) | 0.99 (0.98–1.01) | 1.00 (0.99–1.01) | |
| QOL | QoL score †† | 1.02 (1.01–1.03) | 1.02 (1.01–1.04) | 1.00 (0.99–1.01) | | 1.01 (0.99–1.02) | |
| Wait time | | 0.64 (0.18–2.27) | | 2.18 (0.61–7.78) | | 1.19 (0.34–4.12) | |
| SUDM | | 1.06 (1.05–1.07) | 1.06 (1.05–1.09) | 1.02 (1.00–1.03) | 1.02 (1.01–1.03) | 1.09 (1.07–1.11) | 1.09 (1.08–1.11) |
## Discussion
Ensuring longitudinal preventive, promotive, and curative primary care among older adults in resource constrained settings is critical to reducing the burden of morbidity and mortality related to NCDs. In this household survey of individuals aged 40 or older in Nouna, Burkina Faso, we found that while about one-quarter of individuals sought care at a public primary or secondary care facility in the last three months, significant gaps existed in care seeking among individuals with NCDs or frailty. In addition to healthcare needs and wealth, we also found that higher ratings of health system quality outcomes were associated with care seeking behavior.
Acute conditions were the most common reason for care seeking among this older population overall, with just one-fifth of recent care seeking for more chronic conditions. However, care seeking overall was low—only one third of patients who self-reported an NCD and $41\%$ of those with TB or HIV had a visit in the last three months, despite recommendations from many institutions including the World Health Organization that individuals with NCDs be seen at least every three months [32]. The lack of recent visits for individuals with chronic conditions requiring longitudinal care is of concern given the importance of ongoing management even when symptoms are not present. A study of the hypertension care cascade in Burkina Faso found that only $17.5\%$ of patients with elevated blood pressure were aware of their diagnosis, and less than half were on treatment [29]. These gaps in both awareness and care are similar to other countries in the region including Sierra Leone where knowledge about cardiovascular disease risk factors and costs were identified barriers to accessing care [47]. The lower ratings of health system quality outcomes including trust and satisfaction (met need) among people not seeking recent care was consistent with work from the Lancet HQSS Commission highlighting their importance in achieving the quality needed for effective, people centered care [1]. More work to understand the scope and causes of this challenge in similar settings is needed to develop effective interventions to strengthen the quality and experience of primary and secondary care to ensure not just once-off access but continuity and comprehensiveness of care, core dimensions of effective primary care [48, 49].
Shared decision making is defined as "a process jointly shared by patients and their health care provider”. In our study, SUDM was found to be the most consistent factor associated with higher health system quality outcomes including satisfaction, confidence in the health systems, and health system quality outcomes. It aims at helping patients play an active role in decisions concerning their health, which is the ultimate goal of patient-centered care [50]. Shared decision making has been studied since the 1990’s and seen as increasingly important as the push for more people-centered primary care has emerged from the World Health Organizations and the Astana Declaration in 2018 [51]. The importance of shared decision making and effective communication for management of chronic conditions has been a focus of research in high income countries with lower rates of shared decision making being found among older individuals and those with poorer health, and associated with lower adherence to care and treatment [20, 21]. Achieving shared decision making requires engagement in decision making, effective communication, and good provider-patient relationships, factors which were captured in our SUDM measure. Similarly to our study, higher rates of shared decision making have been associated with better satisfaction, identifying an area for improving quality and outcomes of care for older individuals and people with NCDs [52].
Rating of care experience variables again pointed to areas where change is needed. Participants reported high ratings of some areas of visit experience (ability to follow advice and trust in provider skills), while other areas were lower, with one-half or fewer reporting high provider technical skills, clarity of communication, or involvement in decision making. Compared with other studies, clarity of communications was lower in our study ($48\%$ versus 66–$100\%$ in Tanzania and close to $60\%$ in Ghana), although variability in populations, survey questions, and scoring makes comparisons challenging [20, 53]. In contrast, in Ghana female patients gave lower ratings for involvement, although the population was younger overall than in our study [19].
Overall trust and confidence in the health system was high, but lower among those not seeking recent care, who also reported lower met needs during their last care encounter. In another study in Burkina Faso, perceived quality of care was a determinant for retention in care, which is important for the continuity needed for NCDs and effective primary care more broadly, and identifying an area where improvement is needed [54]. This evidence for the relationship between uptake, retention, process, and outcomes of care experience offer a potential opportunity for improving continuity for the aging population and growing numbers of people with chronic conditions.
While geographic access was only rarely given as a reason for no recent care seeking, $14.5\%$ had to borrow or sell something to attend a clinic, representing a significant burden among a population with high poverty. This measure also may underestimate cost burdens such as individuals who had to forgo consumption of other goods or services such as food to access their health care. While the lower wealth among non-users was similar to findings to another study, they also found higher rates of financial access as a barrier than in our study [55].
Our study had some key limitations. First, we were unable to collect all the dimensions of the traditional health systems responsiveness domains—aspects such as respect and confidentiality might have added to our understanding of care experience in this population. The self-reported nature of previous health information may have underestimated actual prevalence due to absent or forgotten diagnoses. We also limited our analyses to individuals visiting a public sector facility providing primary or secondary level care to focus on the local care system delivery, excluding the small proportion of participants using private or higher-level facilities. However, given the expanding role of the private sector in many countries, future work focusing on these facilities should be planned. We also did not collect the provider cadre who delivered the care, so can not comment on differences based on provider type. Finally, although statistically significant, the clinical significance was more limited for some variables where the odds ratio was close to one. One exception is the results for the SUDM variable which is measuring for every one-point increase in the variable, so the association increases when aggregated over multiple point changes.
In conclusion, we provide a comprehensive picture of public-sector health facility care seeking behaviors and user quality experiences among older individuals in rural Burkina Faso. A minority of individuals have sought recent care, most frequently for acute conditions, despite a burden of NCDs which need continuity of care. Among those with recent visits, the importance of shared understanding and engagement in decision making was seen across all measured health systems quality outcomes. Situating our findings was limited by the availability of comparable population-representative samples in rural, low-income settings–efforts to measure similar patient experiences should provide substantial benefit. Our findings provide insights into designing health system and care delivery interventions to improve the experience and involvement in care of the growing older population in rural LMICs. These interventions are particularly important for those with chronic conditions for whom ongoing care is critical to reducing preventable mortality and mortality.
## References
1. Kruk ME, Gage AD, Arsenault C, Jordan K, Leslie HH, Roder-dewan S. **The Lancet Global Health Commission High-quality health systems in the Sustainable Development Goals era: time for a revolution**. (2018.0) 1-57
2. 2Institute of Medicine. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: National Academies Press; 2001.. (2001.0)
3. **Framework on integrated, people-centred health services**. (2016.0) 1-12
4. Sharma J, Leslie HH, Kundu F, Kruk ME. **Poor quality for poor women? Inequities in the quality of antenatal and delivery care in Kenya**. *PLoS ONE* (2017.0) **12** 1-14. DOI: 10.1371/journal.pone.0171236
5. Piot P, Caldwell A, Lamptey P, Nyrirenda M, Mehra S, Cahill K. **Addressing the growing burden of non-communicable disease by leveraging lessons from infectious disease management**. *Journal of Global Health* (2016.0) **6** 5-7. DOI: 10.7189/jogh.06.010304
6. Manne-Goehler J, Geldsetzer P, Agoudavi K, Andall-Brereton G, Aryal KK, Bicaba BW. **Health system performance for people with diabetes in 28 low-and middle-income countries: A cross-sectional study of nationally representative surveys**. *PLoS Medicine* (2019.0) **16** 1-21. DOI: 10.1371/journal.pmed.1002751
7. Kruk ME, Chukwuma A, Leslie HH. **Variation in quality of primary-care services in Kenya, Malawi, Namibia, Rwanda, Senegal, Uganda and the United Republic of Tanzania**. *Bull World Health Organ* (2017.0) **95** 408-18. DOI: 10.2471/BLT.16.175869
8. Das J, Holla A, Das V, Mohanan M, Tabak D, Chan B. **In urban and rural India, a standardized patient study showed low levels of provider training and huge quality gaps**. *Health Affairs* (2012.0) **31** 2774-84. DOI: 10.1377/hlthaff.2011.1356
9. Geldsetzer P, Manne-Goehler J, Marcus ME, Ebert C, Zhumadilov Z, Wesseh CS. **The state of hypertension care in 44 low-income and middle-income countries: a cross-sectional study of nationally representative individual-level data from 1·1 million adults**. *The Lancet* (2019.0) **394** 652-62. DOI: 10.1016/S0140-6736(19)30955-9
10. Davies JI, Reddiar SK, Hirschhorn LR, Ebert C, Marcus ME, Seiglie JA. **Association between country preparedness indicators and quality clinical care for cardiovascular disease risk factors in 44 lower- And middle-income countries: A multicountry analysis of survey data**. *PLoS Medicine* (2020.0) **17** 1-25. DOI: 10.1371/journal.pmed.1003268
11. Odland ML, Whitaker J, Nepogodiev D, Aling CA, Bagahirwa I, Dushime T. **Identifying, Prioritizing and Visually Mapping Barriers to Injury Care in Rwanda: A Multi-disciplinary Stakeholder Exercise.**. *World Journal of Surgery* (2020.0) **44** 2903-18. DOI: 10.1007/s00268-020-05571-6
12. Kruk ME, Mbaruku G, McCord CW, Moran M, Rockers PC, Galea S. **Bypassing primary care facilities for childbirth: A population-based study in rural Tanzania.**. *Health Policy and Planning* (2009.0) **24** 279-88. DOI: 10.1093/heapol/czp011
13. Doyle C, Lennox L, Bell D. **A systematic review of evidence on the links between patient experience and clinical safety and effectiveness**. *BMJ Open* (2013.0) **3**. DOI: 10.1136/bmjopen-2012-001570
14. Woskie LR, Fallah MP. **Overcoming distrust to deliver universal health coverage: Lessons from Ebola**. *BMJ* (2019.0) 366. DOI: 10.1136/bmj.l5482
15. Mirzoev T, Kane S. **What is health systems responsiveness? Review of existing knowledge and proposed conceptual framework**. *BMJ Global Health* (2017.0) **2**. DOI: 10.1136/bmjgh-2017-000486
16. Tille F, Röttger J, Gibis B, Busse R, Kuhlmey A, Schnitzer S. **Patients’ perceptions of health system responsiveness in ambulatory care in Germany.**. *Patient Education and Counseling* (2019.0) **102** 162-71. DOI: 10.1016/j.pec.2018.08.020
17. de Silva A.. **A Framework of Measuring Responsiveness**. *WHO GPE Discussion Paper Series* (2000.0)
18. Larson E, Sharma J, Bohren MA, Tunçalp Ö. **When the patient is the expert: Measuring patient experience and satisfaction with care**. *Bull World Health Org* (2019.0) **97** 563-9. DOI: 10.2471/BLT.18.225201
19. Ratcliffe HL, Bell G, Awoonor-Williams K, Bitton A, Kim JH, Lipstiz S. **Towards patient-centred care in Ghana: health system responsiveness, self-rated health and experiential quality in a nationally representative survey**. *BMJ Open Qual* (2020.0) **9**
20. Spatz ES, Spertus JA. **Shared decision making: A path toward improved patient-centered outcomes. Circulation**. *Cardiovascular Quality and Outcomes* (2012.0) **5** 75-7
21. Joosten EAG, DeFuentes-Merillas L, de Weert GH, Sensky T, van der Staak CPF, de Jong CAJ. **Systematic review of the effects of shared decision-making on patient satisfaction, treatment adherence and health status**. *Psychotherapy and Psychosomatics* (2008.0) **77** 219-26. DOI: 10.1159/000126073
22. **Burkina Faso Poverty and Equity Brief [Internet].**. (2021.0)
23. 23World Bank. World Health Organization Global Health Expenditure database [Internet]. [cited 2021 May 15]. Available from: https://data.worldbank.org/indicator/SH.XPD.CHEX.GD.ZS?locations=BF
24. Ramsay M, Crowther N, Tambo E, Agongo G, Baloyi V, Dikotope S. **H3Africa AWI-Gen Collaborative Centre: A resource to study the interplay between genomic and environmental risk factors for cardiometabolic diseases in four sub-Saharan African countries.**. *Global Health, Epidemiology and Genomics* (2016.0) 1
25. 25World Health Organization (WHO).
Rapport de l’enquête nationale sur la prévalence des principaux risques communs aux maladies non transmissibles au Burkina Faso. 2014;78.. *Rapport de l’enquête nationale sur la prévalence des principaux risques communs aux maladies non transmissibles au Burkina Faso* (2014.0) 78
26. Millogo T, Bicaba BW, Soubeiga JK, Dabiré E, Médah I, Kouanda S. **Diabetes and abnormal glucose regulation in the adult population of Burkina Faso: Prevalence and predictors.**. *BMC Public Health* (2018.0) **18**. DOI: 10.1186/s12889-018-5257-4
27. Brinkmann B, Payne CF, Kohler I, Harling G, Davies J, Witham M. **Depressive symptoms and cardiovascular disease: A population-based study of older adults in rural Burkina Faso**. *BMJ Open* (2020.0) **10** 1-10
28. Odland ML, Payne C, Witham MD, Siedner MJ, Bärnighausen T, Bountogo M. **Epidemiology of multimorbidity in conditions of extreme poverty: A population-based study of older adults in rural Burkina Faso**. *BMJ Global Health* (2020.0) **5** 1-14. DOI: 10.1136/bmjgh-2019-002096
29. Cissé K, Kouanda S, Coppieters’T Wallant Y, Kirakoya-Samadoulougou F. **Awareness, Treatment, and Control of Hypertension among the Adult Population in Burkina Faso: Evidence from a Nationwide Population-Based Survey**. *International Journal of Hypertension* (2021.0) 2021
30. Wagner RG, Crowther NJ, Micklesfield LK, Boua PR, Nonterah EA, Mashinya F. **Estimating the burden of cardiovascular risk in community dwellers over 40 years old in South Africa, Kenya, Burkina Faso and Ghana**. *BMJ Global Health* (2021.0) **6**
31. **Noncommunicable diseases country profiles 2018**. (2016.0)
32. **WHO Package of Essential Noncommunicable (PEN) Disease Interventions for Primary Health Care [Internet].**. (2020.0)
33. 33United Nations. Decade of Healthy Aging 2020–2030 [Internet]. 2020 [cited 2021 Dec 23]. Available from: https://cdn.who.int/media/docs/default-source/decade-of-healthy-ageing/final-decade-proposal/decade-proposal-final-apr2020-en.pdf?sfvrsn=b4b75ebc_25&download=true
34. Sié A, Louis Valérie R, Gbangou A, Müller O, Niamba L, Stieglbauer G. **The Health and Demographic Surveillance System (HDSS) in Nouna, Burkina Faso, 1993–2007**. *Global Health Action* (2010.0) **3** 5284. DOI: 10.3402/gha.v3i0.5284
35. 35Anxiety and Depression Assoc of America. GAC-7 Anxiety [Internet]. [cited 2021 Jun 10]. Available from: https://adaa.org/sites/default/files/GAD-7_Anxiety-updated_0.pdfdisorder-gad
36. Kroenke K, Spitzer RL, Williams JBW. **The PHQ-9: Validity of a brief depression severity measure**. *Journal of General Internal Medicine* (2001.0) **16** 606-13. DOI: 10.1046/j.1525-1497.2001.016009606.x
37. 37World Health Organization. WHOQOL: Measuring Quality of Life [Internet]. [cited 2021 Jun 10]. Available from: https://www.who.int/tools/whoqol
38. 38World Health Organization. WHO Disability Assessment Schedule 2.0 (WHODAS 2.0) [Internet]. [cited 2021 Jun 10]. Available from: https://www.who.int/standards/classifications/international-classification-of-functioning-disability-and-health/who-disability-assessment-schedule
39. Hall K, Gao S, Emsley C, Ogunniyi A, Morgan O, Hendrie H. **Community screening interview for dementia (CSI ‘D’); performance in five disparate study sites.**. *Int J Geriatr Psychiatry* (2000.0) **15** 521-21. DOI: 10.1002/1099-1166(200006)15:6<521::aid-gps182>3.0.co;2-f
40. Harling G, Witham MD, Davies JI, Bärnighausen T, Bountogo M, Manne-Goehler J. **Frailty and physical performance in the context of extreme poverty: A population-based study of older adults in rural Burkina Faso**. *Wellcome Open Research* (2019.0) **4** 1-16. PMID: 31245630
41. Baltussen R, Ye Y. **Quality of care of modern health services as perceived by users and non-users in Burkina Faso.**. *International Journal for Quality in Health Care* (2006.0) **18** 30-4. DOI: 10.1093/intqhc/mzi079
42. Miller JS, Mhalu A, Chalamilla G, Siril H, Kaaya S, Tito J. **Patient satisfaction with HIV/AIDS care at private clinics in Dar es Salaam, Tanzania.**. *AIDS Care* (2014.0) **26** 1150-4. DOI: 10.1080/09540121.2014.882487
43. Filmer D, Pritchett L. **Estimating wealth effects without expenditure data-or tears: an application to educational enrollments in states of India**. *Demography* (2001.0) **38** 115-32. DOI: 10.1353/dem.2001.0003
44. Baltussen R.. **Perceived quality of care of primary health care services in Burkina Faso.**. *Health Policy and Planning* (2002.0) **17** 42-8. DOI: 10.1093/heapol/17.1.42
45. Geldsetzer P, Haakenstad A, James EK, Atun R. **Non-technical health care quality and health system responsiveness in middle-income countries: A cross-sectional study in China, Ghana, India, Mexico, Russia, and South Africa.**. *Journal of Global Health* (2018.0) **8**
46. Peltzer K.. **Patient experiences and health system responsiveness in South Africa.**. *BMC Health Services Research* (2009.0) **9** 117. DOI: 10.1186/1472-6963-9-117
47. Ignatowicz A, Odland ML, Bockarie T, Wurie H, Ansumana R, Kelly AH. **Knowledge and understanding of cardiovascular disease risk factors in Sierra Leone: A qualitative study of patients’ and community leaders’ perceptions**. *BMJ Open* (2020.0) **10** 7-9. DOI: 10.1136/bmjopen-2020-038523
48. Starfield B, Leiyu S, Macinko J. **Contribution of Primary Care to Health Systems and Health**. *The Milbank Quarterly* (2005.0) **83** 457-502. DOI: 10.1111/j.1468-0009.2005.00409.x
49. Bitton A, Ratcliffe HL, Veillard JH, Kress DH, Barkley S, Kimball M. **Primary Health Care as a Foundation for Strengthening Health Systems in Low- and Middle-Income Countries**. *J Gen Int Med* (2016.0) **32** 566-571. DOI: 10.1007/s11606-016-3898-5
50. Truglio-Londrigan M, Slyer JT, Singleton JK, Worral PS. **A qualitative systematic review of internal and external influences on shared decision-making in all health care settings**. *JBI Database of Systematic Reviews and Implementation Reports* (2014.0) **12** 121-94
51. 51WHO, UNICEF. Declaration of Astana.
WHO. 2018.. *Declaration of Astana.* (2018.0)
52. Milky G, Thomas J. **Shared decision making, satisfaction with care and medication adherence among patients with diabetes**. *Patient Education and Counseling* (2020.0) **103** 661-9. DOI: 10.1016/j.pec.2019.10.008
53. Kapologwe NA, Kibusi SM, Borghi J, Gwajima DO, Kalolo A. **Assessing health system responsiveness in primary health care facilities in Tanzania.**. *BMC Health Services Research* (2020.0) **20** 1-10. DOI: 10.1186/s12913-020-4961-9
54. Mugisha F, Bocar K, Dong H, Chepng’eno G, Sauerborn R. **The two faces of enhancing utilization of health-care services: Determinants of patient initiation and retention in rural Burkina Faso**. *Bull World Health Organ* (2004.0) **82** 572-9. PMID: 15375446
55. Dong H, Gbangou A, De Allegri M, Pokhrel S, Sauerborn R. **The differences in characteristics between health-care users and non-users: Implication for introducing community-based health insurance in Burkina Faso.**. *European Journal of Health Economics* (2008.0) **9** 41-50. DOI: 10.1007/s10198-006-0031-4
|
---
title: 'Pattern and correlates of physical activity and sedentary behaviours of pregnant
women in Ibadan, Nigeria: Findings from Ibadan pregnancy cohort study'
authors:
- Ikeola A. Adeoye
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10021993
doi: 10.1371/journal.pgph.0001153
license: CC BY 4.0
---
# Pattern and correlates of physical activity and sedentary behaviours of pregnant women in Ibadan, Nigeria: Findings from Ibadan pregnancy cohort study
## Abstract
Globally, physical inactivity is the fourth leading risk factor for premature death. Pregnancy is associated with reduced physical activity because of physiological and anatomical changes and socio-cultural barriers. Even though physical activity provides many benefits, such as improved insulin sensitivity and reduced cardiometabolic risk, it is not emphasized among pregnant women in Nigeria. This study described the pattern of physical activity and sedentary behaviours of pregnant women from the Ibadan Pregnancy Cohort Study in Ibadan, Nigeria. The Ibadan Pregnancy Cohort Study (IbPCS) is a prospective cohort study investigating the associations between maternal obesity, lifestyle factors on glycaemia control, gestational weight gain, pregnancy and postpartum outcomes among pregnant women in Ibadan. The Pregnancy Physical Activity Questionnaire (PPAQ) was used to assess physical activity and sedentary behaviour. Sedentary time was estimated from the time spent watching television, sitting at work and the computer. Bivariate and multivariate logistic regression analyses were done to investigate associations at a $5\%$ level of statistical significance. None of the pregnant women met the WHO recommendation of 150mins of moderate-intensity activity per week. The average time spent engaged in moderate-intensity activity was 26.3 ± 22.9 mins. The mean daily sedentary time was 6.5 ± 4.2 hours. High parity para ≥ 4: [AOR 0.57 $95\%$ CI: (0.36–0.89) $$p \leq 0.014$$] and being employed [AOR 0.23 $95\%$ CI: (0.15–0.33) $p \leq 0.001$] reduced the odds of having inadequate physical activity. Correlates of sedentary behavior after adjusting for confounders were high parity: para 1–3 AOR 0.73, $95\%$ CI: (0.58–0.91) $$p \leq 0.004$$], tertiary education: AOR 2.39 $95\%$ CI: (1.16–4.91) $$p \leq 0.018$$] and earning a higher income: AOR 1.40: $95\%$ CI: (1.11–1.78) $$p \leq 0.005$$]. Pregnant women’s physical activity and sedentary behaviours are emerging public health issues, especially in Nigeria. The level of physical activity was inadequate among pregnant women, while the sedentary time was high. There is a need to implement programmes that promote physical activity and discourage sedentary behaviour among pregnant women in Nigeria.
## Introduction
Physical inactivity (PA) is globally the number four leading risk factor for premature death [1,2]. The current obesity epidemic has drawn attention to the importance of physical activity among the general population. Regular physical activity is beneficial throughout the life course because it improves cardiorespiratory fitness, reduces the risk of obesity, diabetes, cardiovascular diseases, and certain cancers, and prolongs life [3]. Additionally, during pregnancy, physical activity improves insulin sensitivity and glucose uptake; hence can prevent cardio-metabolic outcomes such as gestational diabetes mellitus and excessive gestational weight gain [4–6]. The other benefits of physical activity during pregnancy include improved cardiovascular fitness, sleep quality, and psychological well-being [7]. Generally, pregnancy is associated with reduced physical activity because of physiological and anatomical changes. For example, increased lumbar lordosis, weight gain and fatigue which affects the ease, willingness and safety involved in physical activity, particularly as the pregnancy progresses [8]. The other reasons for reduced physical activity include socio-cultural beliefs, misinformation, and the fear of complications [9]. Also, pregnant women are often encouraged to rest rather than exercise because of the presumed maternal and foetal complications, such as miscarriage, preterm delivery and intrauterine growth restriction that results from reduced placenta circulation. Exercise is thought to shunt blood away from the placenta and other vital organs to the skeletal muscles during exercise [10].
Physical activity is defined "as any bodily movement produced by skeletal muscles that result in energy expenditure" [11]. The degree of energy expenditure is influenced by the intensity, duration, and frequency of muscular contractions and muscle mass. Physical activity is classified according to types, which have been described in four distinct domains: occupational, household task, leisure-time, e.g. sports, and transport [12]. It could also be categorized by intensity (light, moderate or heavy intensity) or time of the week (weekday and weekend). In contrast, exercise is a planned, structured, repetitive physical activity to improve physical fitness. The types of exercise are aerobic exercises (e.g. walking, which increases strength and cardiovascular fitness), resistance exercises (which increase muscle mass and strength), and stretching exercises (which improve flexibility by increasing muscle fibre size) [11].
The World Health Organization and some allied professional organizations, such as the American Congress of Obstetricians and Gynaecologists (ACOG), and Royal College of Obstetricians and Gynaecologists (RCOG), among others, have formulated physical activity guidelines during pregnancy [1,2,13]. These guidelines are evidence-based recommendations for practising physical activity during pregnancy to ensure the health and safety of the mother, foetus and neonate. The recommendation of WHO is that pregnant women should engage in at least 150 minutes of moderate-intensity physical activity during the week. Similarly, the ACOG advocated that "in the absence of either medical or obstetric contraindications, 30 min or more of moderate exercise a day on most, if not all, days is recommended for pregnant women" [13]. In 2015, ACOG revised the earlier recommendation based on the current scientific data [1], which still affirmed the previous guidance of 30 minutes of moderate-intensity exercise on most days of the week. Unfortunately, physical activity among pregnant women has received much less attention in sub-Saharan Africa, including Nigeria [14]. Physical activity is not yet an essential component of maternal health services in Nigeria regarding available policy, guidelines, counselling and recommendation.
Sedentary behaviour (SB) has been defined as "any waking behaviour characterized by an energy expenditure ≤ 1.5 metabolic equivalents (METs) while sitting or reclining position" [15,16]. At the same time, sedentary time (ST) is the time spent for any duration in sedentary behaviour [15]. Too much time spent in sedentary behaviour compromises metabolic health and is a significant risk factor for obesity, type 2 diabetes, cardiovascular disease and all-cause mortality [17–19]. The Canadian Society for Exercise Physiology recommends that sedentary time should be less than 8 hours (with less than 3 hours of recreational screen time [20]. Sedentary behaviour differs from physical inactivity in that individuals who meet physical activity recommendations are at higher risk of adverse metabolic complications if much time is spent on sedentary behaviour [19,21]. Sedentary behaviour during pregnancy is associated with adverse outcomes, for example, gestational weight gain [22], gestational diabetes mellitus [23], hypertensive disorders [24] and macrosomia [25]. Pregnant women’s physical activity and sedentary behaviours are emerging public health issues, especially in Sub-Saharan Africa. This results from epidemiologic and nutritional transitions characterized by changes in dietary patterns, reduced physical activity and increased sedentary behaviours [26–29] particularly in the cities. This is a departure from the traditional African culture in which women were physically active during pregnancy, particularly the rural dwelling women of Africa belonging to the lower socioeconomic status [30]. While there is a paucity of studies assessing physical activity among pregnant women in Nigeria, information on sedentary behaviour among pregnant is lacking in Nigeria. Therefore, this study evaluated the pattern and the correlates of physical activity and sedentary behaviour among pregnant women in Ibadan, Nigeria, using the Ibadan Pregnancy Cohort Study.
## Study design and participants
The Ibadan Pregnancy Cohort Study (IbPCS) is a prospective cohort study investigating the associations between maternal obesity, lifestyle factors on glycaemic control, gestational weight gain, pregnancy and postpartum outcomes among pregnant women in Ibadan. The lifestyle factors examined included dietary patterns, sugar-sweetened beverages, physical activity, sedentary behaviour, tobacco use, alcohol consumption, and sleep patterns. The study was conducted at four medical facilities within the Ibadan metropolis: University College Hospital, Adeoyo Maternity Teaching Hospital, Jericho Specialist Hospital, Saint Mary Catholic Hospital, Oluyoro, Ibadan, Oyo State, Nigeria. These hospitals offer comprehensive obstetric services to pregnant women and are the major referral centres in the city. IbPCS recruited 1745 pregnant women in early pregnancy (≤ 20 weeks’ gestation) during their antenatal booking visit and followed them until delivery. Data were collected using pretested, interviewer-administered questionnaires and desktop review of medical records at three points during the study–booking, third trimester, and delivery. The level of physical activity and sedentary behaviours were assessed at baseline and other parameters described in the protocol [31]. The details of the methodology have been published elsewhere. Physical activity (PA) and sedentary behaviour were assessed using the Pregnancy Physical Activity Questionnaire (PPAQ). Sedentary time was estimated from the number of hours spent sitting in a day: i.e. the number of hours spent on average watching television per day and on weekends, sitting at the computer, and the number of hours spent sitting at work daily.
## Assessment of physical activity
The Pregnancy Physical Activity Questionnaire (PPAQ) assessed physical activity patterns and levels. The PPAQ was developed and validated by Chasan-Taber in 2004 to determine the physical activity levels in pregnant women [32]. It is a 32-item questionnaire that measures the levels of physical activity across five domains–Household and caregiving activities (13 items), occupational (5 items), sports and exercise (12 items), and transport (3 items). The analysis of physical activity was done according to the PPAQ instruction guide [32]. In summary, the activity assessed was classified by intensity (sedentary, light, moderate and vigorous) and by the pattern of activity (household/care, occupation, transport, sports). Responses to each question were reckoned in time: none, less than ½ hours per day, ½ to almost 1 hour per day, 1 to nearly 2 hours per day, 2 to almost 3 hours per day, and three or more hours per day. Energy expenditure (MET-hours per week) was estimated by multiplying the mid-point of the time option by the metabolic equivalent allotted to the activity. The metabolic equivalent (MET) is the ratio of a person’s working metabolic rate relative to the resting metabolic rate. The total energy expenditure was obtained by summing all activities’ energy expenditure. Total physical activity was categorized as inadequate if the total score was within the 50th percentile and high if it was > the 50th percentile. Factors associated with inadequate physical activity were assessed.
## Assessment of sedentary behaviour
Sedentary behaviour was also assessed using the PPAQ according to the PPAQ instruction guide. Sedentary behaviour was calculated as the sum of the product of intensity and duration (MET-h/day) of sedentary activities (questions #11, 12, 13, 22, 32). Duration of moderate-intensity exercise (minutes per week) is the summation of time spent on moderate-intensity activity per week. Sedentary time (ST) was estimated by the time reportedly spent (in hours) watching television, sitting at the computer, and sitting at work. An eight-hour threshold is the recommended limit for ST within 24 hours [33].
## Data analysis
Statistical analysis was performed using STATA 13. Univariate analysis was used to describe the intensity and types of physical activity during pregnancy. Boxplots displayed the duration of moderate-intensity activity per week, and the pie chart shows the proportion that exceeds the recommended threshold for ST. A Chi-square test was performed to assess the relationship between the women’s characteristics and total physical activity in tertiles. The outcome variables: total physical activity and sedentary behaviour, were categorized into binary variables (Inadequate = 1 and high = 0) using the cut-off of the 50th percentile. The explanatory variables included demographic variables, BMI and having motorized transport. Factors associated with inadequate physical activity were investigated using bivariate logistic analysis. Variables significant at a $5\%$ level of statistical significance at bivariate logistic analysis (parity, employment status, religion, ownership of a motorized transport) were subjected to multiple logistic regression analysis. Unadjusted odds ratios (UOR), adjusted odds ratios (AOR), $95\%$ confidence intervals and p-values ($p \leq 0.05$) were reported.
## Ethical consideration
The ethical approval for this study was obtained from the University of Ibadan/University College Hospital (UI/UCH) Institutional Review Board (UI/EC/$\frac{15}{0060}$) and Oyo State Ministry of Health Ethical Committee (AD/$\frac{13}{479}$/710). Both verbal and written informed consent was obtained from respondents before recruitment into the study. The study protocol and conduct adhered to the principles in the Declaration of Helsinki.
## Characteristics of pregnant women according to level of physical activity (Table 1)
A total of 1745 pregnant women were recruited for this study. The characteristics of pregnant women according to tertiles of total physical activity in the Ibadan Pregnancy Cohort Study are shown in Table 1. Maternal age ($$p \leq 0.007$$), parity ($p \leq 0.001$), marital status ($$p \leq 0.005$$), maternal education ($$p \leq 0.019$$), occupation ($p \leq 0.001$), ownership of motorized transport ($$p \leq 0.009$$) were associated with the total physical activity. Specifically, older women 40 years and above ($47.1\%$) compared with younger women less than 20 years ($33.3\%$), as well as unemployed women ($68.3\%$) compared with women with gainful employment ($29.1\%$) reported lower levels of physical activity. However, physical activity levels increased significantly with parity; (nulliparous: $26.3\%$), (1–3: $38.1\%$) (≥ 4: $44.2\%$) in a dose-response fashion. Also, married women ($34.0\%$) had higher physical activity levels than single women ($21.6\%$). Women with motorized transport ($36.6\%$) also reported significantly higher physical activity levels than those without one ($30.1\%$).
**Table 1**
| Unnamed: 0 | Unnamed: 1 | Tertiles of Total Physical Activity | Tertiles of Total Physical Activity.1 | Tertiles of Total Physical Activity.2 | Unnamed: 5 | Unnamed: 6 |
| --- | --- | --- | --- | --- | --- | --- |
| Characteristics | Total | Low | Middle | High | Chi-square | p-value |
| Age group | | | | | | |
| < 20 | 33 (1.9) | 11 (33.3) | 15(45.5) | 7(21.2) | 17.66 | 0.007 |
| 20–29 | 832 (47.7) | 299(36.0) | 271(33.0) | 262(32.0) | | |
| 30–39 | 812 (46.5) | 240(30.0) | 276(34.0) | 296(36.5) | | |
| ≥ 40 years | 68 (3.9) | 32(47.1) | 20(29.4) | 16(24.0) | | |
| Parity | | | | | | |
| Nulliparous | 760 (43.7) | 304 (40.0) | 256 (33.7) | 200 (26.3) | 40.83 | <0.001 |
| 2–4 | 882 (50.8) | 246 (27.9) | 300 (34.0) | 336 (38.1) | | |
| ≥ 5 | 95 (5.5) | 29 (30.5) | 24 (25.6) | 42 (44.2) | | |
| Marital Status | | | | | | |
| Single | 102 (5.8) | 48 (47.1) | 32 (31.4) | 20 (21.6) | 10.70 | 0.005 |
| Married | 1643 (94.2) | 534 (32.5) | 550 (33.5) | 559 (34.0) | | |
| Maternal Education | | | | | | |
| Primary or less | 49 (2.8) | 13 (26.5) | 19 (38.8) | 17 (34.7) | 11.82 | 0.019 |
| Secondary | 504 (28.9) | 180 (35.7) | 185 (36.7) | 139 (27.6) | | |
| Tertiary | 1188 (68.2) | 387 (32.6) | 377 (31.7) | 424 (35.7) | | |
| Occupation | | | | | | |
| Employed | 1556 (89.2) | 453(29.1) | 542(34.8) | 561 (36.1) | 119.63 | <0.001 |
| Unemployed | 186 (10.8) | 129 (68.3) | 40 (21.2) | 20 (10.6) | | |
| Religion | | | | | | |
| Christianity | 1010 (58.2) | 321 (31.8) | 332 (32.9) | 357 (35.4) | 5.20 | 0.074 |
| Islam | 726 (41.8) | 260 (35.8) | 245 (33.8) | 221 (30.4) | | |
| Ethnicity | | | | | | |
| Yorubas | 1564 (89.8) | 521 (33.3) | 534 (34.1) | 509 (32.5) | 5.94 | 0.051 |
| Non-Yorubas | 178 (10.2) | 61 (34.3) | 46 (25.8) | 71 (39.9) | | |
| Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) |
| <20,000 | 583 (38.0) | 189 (32.4) | 200 (34.3) | 194 (33.3) | 4.45 | 0.348 |
| 20,000–99,999 | 843 (55.0) | 234 (27.7) | 293 (34.8) | 316 (37.5) | | |
| ≥ 100,000 | 108 (7.0) | 30 (27.8) | 39 (36.1) | 39 (36.1) | | |
| Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) |
| Underweight | 50 (3.0) | 21 (42.0) | 20 (40.0) | 9 (18.0) | 11.65 | 0.070 |
| Normal weight | 845 (49.8) | 294 (34.8) | 267 (31.6) | 284 (33.6) | | |
| Overweight | 473 (27.9) | 151 (31.9) | 154 (32.6) | 168 (35.5) | | |
| Obese | 328 (19.3) | 98 (29.9) | 126 (38.4) | 104 (31.7) | | |
| Motorized Transport | | | | | | |
| Yes | 885 (50.7) | 264 (30.7) | 218 (32.7) | 315 (36.6) | 9.47 | 0.009 |
| No | 860 (49.3) | 318 (35.9) | 301 (34.0) | 266 (30.1) | | |
## Factors associated with inadequate physical activity among respondents (Table 2)
Table 2 shows the factors associated with inadequate physical activity among pregnant women in the Ibadan Pregnancy Cohort Study. By crude and adjusted logistic models, positive associations were found between parity, religion and occupational status and inadequate total physical activity. The odds of being less physically active decreased with increasing parity: para 1–3: AOR 0.67 $95\%$ CI: (0.55–0.82) $p \leq 0.001$ and para ≥ 4: AOR 0.57 $95\%$ CI: (0.37–0.89) $$p \leq 0.014$$ compared with nulliparous women after adjusting for other factors. Additionally, employed women with AOR 0.23 $95\%$ CI: (0.16–0.34) $p \leq 0.001$ had a lower likelihood of being less physically active than women that are unemployed.
**Table 2**
| Characteristics | Unadjusted OR (95% CI) | p-value | Adjusted OR (95% CI) | p-value.1 |
| --- | --- | --- | --- | --- |
| Age group | | | | |
| < 20 | 1.00 | | | |
| 20–29 | 0.85 (0.42–1.72) | 0.654 | | |
| 30–39 | 0.61 (0.30–1.23) | 0.170 | | |
| ≥ 40 years | 1.05 (0.45–2.44) | 0.905 | | |
| Parity | | | | |
| Nulliparous | 1.00 | | 1.00 | |
| 1–3 | 0.59 (0.48–0.71) | <0.001 | 0.67 (0.55–0.82) | <0.001 |
| ≥ 4 | 0.51 (0.33–0.79) | 0.002 | 0.57 (0.36–0.89) | 0.014 |
| Marital Status | | | | |
| Single | 1.00 | | | |
| Married | 0.69 (0.46–1.03) | 0.068 | | |
| Education | | | | |
| Primary or less | 1.00 | | | |
| Secondary | 1.11 (0.62–2.01) | 0.713 | | |
| Tertiary | 0.90 (0.51–1.59) | 0.719 | | |
| Occupation | | | | |
| Employed | 0.20 (0.14–0.30) | <0.001 | 0.23 (0.16–0.34) | <0.001 |
| Unemployed | 1.00 | | 1.00 | |
| Religion | | | | |
| Christianity | 1.00 | | 1.00 | |
| Islam | 1.27 (1.06–1.55) | 0.011 | 1.30 (1.07–1.59) | 0.009 |
| Ethnicity | | | | |
| Yorubas | 1.19 (0.88–1.63) | 0.262 | | |
| Non-Yorubas | 1.00 | | | |
| Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) |
| <20,000 | 1.00 | | | |
| 20,000–99,999 | 0.83 (0.67–1.02) | 0.087 | | |
| ≥ 100,000 | 0.68 (0.45–1.04) | 0.081 | | |
| Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) | Body Mass Index (kg/m 2 ) |
| Underweight | 1.00 | | | |
| Normal weight | 0.96(0.55–1.72) | 0.917 | | |
| Overweight | 0.80 (0.45–1.44) | 0.460 | | |
| Obese | 1.02 (0.56–1.85) | 0.954 | | |
| Motorized transport | 0.79 (0.66–0.96) | 0.016 | 0.88 (0.72–1.07) | 0.188 |
## Factors associated with sedentary behaviour among pregnant women (Table 3)
Table 3 shows the factors associated with sedentary behaviour among pregnant women in Ibadan with the unadjusted and adjusted odds ratios and $95\%$ confidence intervals. On univariate analysis, high parity, tertiary education, employment status, religion and income were significantly associated with sedentary behaviour. The adjusted analysis showed that women with higher parity had lower odds for sedentary behaviour: para 1–3 AOR 0.73, $95\%$ CI: (0.58–0.91) $$p \leq 0.004$$ compared with nulliparous women after adjusting for confounders. Women with tertiary education also had a higher likelihood of sedentary behaviour AOR 2.39 $95\%$ CI: (1.16–4.91) $$p \leq 0.018$$ compared with women with primary education. Earning a higher income also increased the odds of sedentary behaviour: "20,000–99,999" naira: 1.40: $95\%$ CI: (1.11–1.78) $$p \leq 0.005$$ compared with lower-income earning women (<20,000.00).
**Table 3**
| Characteristics | Unadjusted odds ratio(95% CI) | p-value | Adjusted odds ratio(95% CI) | p-value.1 |
| --- | --- | --- | --- | --- |
| Age in years | | | | |
| < 20 | 1 | | | |
| 20–29 | 1.11 (0.51–2.43) | 0.790 | | |
| 30–39 | 0.90 (0.42–1.99) | 0.809 | | |
| ≥ 40 years | 0.76 (0.30–1.92) | 0.569 | | |
| Parity | | | | |
| Nulliparous | 1 | | 1 | |
| 2–4 | 0.70 (0.56–0.86) | 0.001 | 0.73 (0.58–0.91) | 0.006 |
| ≥ 5 | 0.57 (0.35–0.88) | 0.011 | 0.81 (0.49–1.33) | 0.398 |
| Marital Status | | | | |
| Single | 1 | | | |
| Married | 0.73 (0.45–1.16) | 0.182 | | |
| Education | | | | |
| Primary or less | 1 | | 1 | |
| Secondary | 1.31 (0.69–2.52) | 0.421 | 1.26 (0.61–2.58 | 0.536 |
| Tertiary | 2.84 (1.49–5.40) | 0.001 | 2.39 (1.16–4.90) | 0.018 |
| Occupation | | | | |
| Employed | 0.63 (0.36–1.09) | 0.099 | | |
| Unemployed | 1 | | | |
| Religion | | | | |
| Christianity | 1 | | 1 | |
| Islam | 0.70 (0.57–0.86) | 0.001 | 0.87 (0.69–1.0) | 0.221 |
| Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) | Income per month (Naira) |
| <20,000 | 1 | | 1 | |
| 20,000–99,999 | 1.72 (1.38–2.16) | <0.001 | 1.41 (1.11–1.79) | 0.005 |
| ≥ 100,000 | 1.78 (1.17–2.75) | 0.008 | 1.31 (0.69–2.52) | 0.177 |
| Body mass index | | | | |
| Underweight | 1 | | | |
| Normal weight | 0.79 (0.42–1.46) | 0.444 | | |
| Overweight | 0.84 (0.45–1.59) | 0.597 | | |
| Obese | 0.77 (0.40–1.47) | 0.427 | | |
## Pattern, types and duration of physical activity and sedentary behaviour among pregnant women (Figs 1–3)
The intensity and types of physical activity engaged in by the pregnant women are shown in Fig 1. The mean total energy expenditure was 290.5 ± 124.8 MET-h/week. Almost seventy per cent ($67.8\%$) of total activity was low in intensity: sedentary (90.1 ± 40.9 MET-h/week) and light intensity activity 106.9 ± 50.6 MET-h/week ($36.8\%$), followed by moderate intensity 91.9 ± 79.8 MET-h/week, and vigorous activity was rare ($0.6\%$). The pattern of energy expenditure was: occupation-related activity ($45.4\%$), household/caregiving activities ($41.4\%$), and transport-related ($10.9\%$), while sports or exercise were infrequent ($2.2\%$). Fig 2 displays a boxplot that shows the duration of moderate-intensity activity per week in minutes by pregnant women. The mean duration was 26.3 ± 22.9 minutes. None of the individuals in the study met the WHO recommendation for physical activity. Fig 3 shows the proportion of pregnant women that exceeded the recommended ST. About a third ($29.1\%$) of the women exceeded the recommended limit pattern and duration (in hours) of sedentary behaviours. The mean daily sedentary time was 6.5 ± 4.2 hours.
**Fig 1:** *Intensity and types of physical activity during pregnancy.* **Fig 2:** *Duration of moderate intensity activity among the pregnant women in Ibadan, Nigeria.* **Fig 3:** *Proportion of women that exceed the recommended sedentary time limit by pregnant women in Ibadan, Nigeria.*
## Discussion
Physical activity and sedentary behaviour have emerged as modifiable risk factors for adverse health outcomes in the general population and pregnant women. However, they have received very scant attention especially among pregnant women in Nigeria. Therefore, we assessed the pattern and factors associated with physical activity and sedentary behaviour of pregnant women using the PPAQ. The most crucial finding was the low level of physical activity among the study participants, as none of the pregnant women in this study met the WHO recommendation that pregnant women should engage in at least 150 minutes of moderate-intensity physical activity per week. Adeniyi et al. [ 2014] reported that in the same study setting, none of their study participants met the physical activity recommendation [34]. Other studies from different parts of the world have also documented low physical activity among pregnant women [35–37]. Very few pregnant women ever meet the recommended allowance for physical activity. This study’s average duration of moderate-intensity physical activity was 26.3 ± 22.9 minutes per week, implying that Nigerian women do not obtain sufficient benefits from physical activity during pregnancy. These benefits include improved cardiovascular fitness, reduced cardio-metabolic outcomes such as gestational diabetes mellitus, excessive gestational weight gain, hypertension, and enhanced mood [4–6]. The reason for the deficient level of physical activity may include the following: a lack of awareness of the benefits of physical activity by pregnant women and the health care providers, failure to incorporate and promote physical activity in routine maternal care services, the fear of complication especially miscarriage, lack of time, cultural inhibitions that encourage rest rather than physical activity during pregnancy [8,9]. It is necessary to explore the reasons for low physical activity levels in our environment in future studies using quantitative and qualitative methods. Zhang et al. [ 2014] reported that the fear of miscarriage was the most crucial reason for physical inactivity among urban Chinese pregnant women [35]. It is recommended that physical activity be encouraged among pregnant women during antenatal care.
Low-intensity physical activity was the study’s most common form of physical activity, a finding corroborated by several researchers [34–38] as physical activity usually declines during pregnancy [36,39]. Hence, a light-intensity activity like walking (< 3.0 METs) is better than no activity. Walking has several maternal and fetal benefits and is the preferred mode of physical activity among pregnant women [40]. Besides, its intensity can be increased to moderate (3.0–6.0 METs) and even vigorous (> 6.0 METs) by increasing the pace. The maternal-fetal benefits of walking during pregnancy have been documented to include reduced risk of gestational diabetes mellitus, pre-eclampsia, unhealthy weight gain, and weight-related neonatal outcomes such as macrosomia, shoulder dystocia, and congenital anomaly [40]. Our maternal health care services should incorporate exercise routines that encourage walking. On the other hand, the prominence of light intensity activity in our study may be due to the characteristics of our study population, which include urban residence and high-level education. Several African countries’ ongoing epidemiological and nutritional transitions, especially cities, are associated with physical inactivity and sedentary occupations [29]. Hence this finding is not generalizable to the rural population. For example, a study that assessed physical activity levels during pregnancy in Northern Ethiopia [30] found that most women were involved in moderate-intensity physical activity almost every day, and $77.1\%$ met the recommendation for physical activity. Notably, these women were of low socioeconomic status, performing household chores, and spent the highest energy on household activities (69.4 MET-h/weeks).
The correlates of inadequate physical activity were parity, occupation and religion. Other researchers have reported that parity, previous maternal history of miscarriage, mothers’ education, maternal age, and employment were significantly associated with physical activity [35–37]. Women with higher parity had lower odds for inadequate physical activity in a dose-response fashion, i.e. "para 1–3" (AOR = 0.67) and para ≥ 4 (AOR = 0.57) compared with nulliparous women after controlling for confounders. A low level of physical activity among nulliparous women may result from inexperience, lack of information or misinformation, the notion that pregnancy is a time of rest rather than physical exercise, lower levels of household and caring duties compared to women of higher parities. Conversely, women with higher parity are likely to be engaged in more household and care-related activities. Apart from walking, household chores also allow pregnant women to take part in physical activity. Researchers have reported that household activities were the most accustomed activity among pregnant women [30,34]. Furthermore, women employed had a much lower likelihood (AOR = 0.23) of have inadequate physical activity compared with unemployed women. In this study, occupational-related physical activity was the most conversant physical activity reported by our study population. Probably due to the high-level education and the women mostly being in the working-class. The workplace also provides the opportunity for physical activity and energy expenditure during pregnancy. However, several occupations are now sedentary, requiring prolonged hours of sitting at work, associated with minimal energy expenditure. In this study, the average time spent sitting at work was 4.1 ± 2.7 hours, and about $60\%$ of the women reported that their work required prolonged sitting.
Studies evaluating sedentary behaviour among the Nigerian pregnant population are lacking. In this study, sedentary behaviour was assessed using the PPAQ; sedentary time was also estimated to be spent sitting at work, using the computer, and watching television during the week and weekends. Earlier studies, mostly from high-income countries, had used various indicators to assess sedentary behaviour and time, for example, the time spent watching television [41,42], time sitting at work and other sedentary behaviours [43,44] or used objective measures [45,46]. Recently, the Sedentary Behaviour Research Network–a global collaboration of researchers and professionals on sedentary behaviour, developed consensus definitions and conceptual frameworks to be used in sedentary behaviour-related studies [15]. Additionally, the Canadian movement guideline for adults recommended that sedentary time be limited to eight hours or less with 24 hours in addition to adequate physical activity and sleep [33]. In this study, the estimated sedentary time was (6.5± 4.2) hours, and over a quarter of our study participants exceeded the eight hours limit for sedentary behaviour. Sedentary behaviour is a health challenge because it increases cardio-metabolic risk by reducing metabolic rate, contractility of muscles, glucose uptake, blood flow and so on [43,47]. It also increases the risk of adverse perinatal outcomes as gestational diabetes [48], excessive gestational weight gain [49], hypertension and deep vein thrombosis [21]. Hence, the need for targeted public health messages and antenatal education about the benefits of physical activity and the need to reduce sedentary behaviour among pregnant women. However, some have documented not smoking [46], non-compliance with physical activity recommendations [50], maternal age and level of education [21] as factors associated with sedentary behaviour. For example, a US study which reported pregnant women spent $50\%$ of their daily activity in sedentary behaviour also found that the odds of passive behaviour were lower among pregnant women who met the physical activity recommendation in their study population [46].
The factors associated with sedentary behaviour in the univariate analysis included parity, education, religion, and income. However, the multivariate analysis showed that high parity was protective of sedentary behaviour after controlling for confounding variables. The association between increasing parity and lower odds of sedentary behaviour is likely linked to expanded household and child caring activities. Tertiary level education and rising income, which are measures of socioeconomic status, had a direct association with sedentary behaviour. Tertiary education often affords high-paying, sedentary jobs with minimal energy expenditure. In this study, sedentary occupation in which participants majorly carried their duties in the sitting position was reported by $60\%$ of our pregnant population. Jones et al. [ 2021], in a recent study, noted that primarily sitting occupation was associated with a higher odds of sedentary behaviour among pregnant women in their study [51]. Additionally, high income affords the means for sedentary entertainment and automated devices [52]. Few studies have explored the factors associated with sedentary behaviour [53], among pregnant women. However, some have documented not smoking [46], non-compliance with physical activity recommendations [50], maternal age and level of education [21] as factors associated with sedentary behaviour. For example, a US study which reported pregnant women spent $50\%$ of their daily activity in sedentary behaviour also found that the odds of sedentary behaviour were lower among pregnant women who met the physical activity recommendation in their study population [46].
Our study fills a crucial but neglected gap in maternal health in Nigeria by examining pregnant women’s patterns and correlates of physical activity and sedentary behaviours in Ibadan, Nigeria, using a large sample from the Ibadan Pregnancy Cohort Study in Nigeria. However, it has some limitations, which include a limited external validity to rural-dwelling women. Using a self-reported questionnaire may be associated with misclassification bias from under or over-reporting. Also, the study was conducted during early gestation, so the findings may not apply to late gestation. Further studies should explore the relationship between physical activity and pregnancy outcomes among Nigerian women.
## Conclusion and implication of the study
Pregnant women’s physical activity and sedentary behaviours are emerging public health issues, especially in Nigeria. This study examined patterns and correlates of physical activity and sedentary behaviours of pregnant women in Ibadan, Nigeria, using the Pregnancy Physical Activity Questionnaire (PPAQ). The physical activity level was low as none of the pregnant women met the physical activity recommendation, and sedentary behaviour was prevalent. Parity and being gainfully employed were factors associated with physical activity, while parity, level of education and income were associated with sedentary behaviour. There is a need to emphasize the benefits of physical activity and limit sedentary behaviour among pregnant women in Nigeria.
## References
1. **Exercise in pregnancy and the postpartum period. Number 267, January 2002 ACOG.**. *American College of Obstetricians and Gynecologists Int J Gynecol Obstet* (2003.0) **37** 6-12. DOI: 10.1016/S0020-7292(02)80004-2
2. 2WHO. Physical activity. World Health Organization. 2016.. *WHO. Physical activity. World Health Organization* (2016.0)
3. Dipietro L, Evenson KR, Bloodgood B, Sprow K, Troiano RP, Piercy KL. **Benefits of Physical Activity during Pregnancy and Postpartum: An Umbrella Review.**. *Medicine and science in sports and exercise* (2019.0) **51** 1292-302. DOI: 10.1249/MSS.0000000000001941
4. Baker JH, Rothenberger SD, Kline CE, Okun ML. **Exercise during early pregnancy is associated with greater sleep continuity.**. *Behavioral sleep medicine.* (2018.0) **16** 482-93. DOI: 10.1080/15402002.2016.1228649
5. da Silva SG, Ricardo LI, Evenson KR, Hallal PC. **Leisure-time physical activity in pregnancy and maternal-child health: a systematic review and meta-analysis of randomized controlled trials and cohort studies.**. *Sports medicine.* (2017.0) **47** 295-317. DOI: 10.1007/s40279-016-0565-2
6. Russo LM, Nobles C, Ertel KA, Chasan-Taber L, Whitcomb BW. **Physical activity interventions in pregnancy and risk of gestational diabetes mellitus: a systematic review and meta-analysis.**. *Obstetrics & Gynecology.* (2015.0) **125** 576-82. DOI: 10.1097/AOG.0000000000000691
7. 7U.S. Department of Health and Human Services.
Physical Activity Guidelines for Americans. 2nd ed. Washington, DC: US Department of Health and Human Services; 2018. 2018.. *Physical Activity Guidelines for Americans* (2018.0) **2018**
8. Davenport MH, Marchand A-A, Mottola MF, Poitras VJ, Gray CE, Garcia AJ. **Exercise for the prevention and treatment of low back, pelvic girdle and lumbopelvic pain during pregnancy: a systematic review and meta-analysis**. *British journal of sports medicine* (2019.0) **53** 90-8. DOI: 10.1136/bjsports-2018-099400
9. Harrison AL, Taylor NF, Shields N, Frawley HC. **Attitudes, barriers and enablers to physical activity in pregnant women: a systematic review**. *Journal of physiotherapy* (2018.0) **64** 24-32. DOI: 10.1016/j.jphys.2017.11.012
10. Davenport M, Kathol AJ, Mottola MF, Skow RJ, Meah VL, Poitras VJ. **Prenatal exercise is not associated with fetal mortality: a systematic review and meta-analysis**. *British Journal of Sports Medicine* (2019.0) **53** 108. DOI: 10.1136/bjsports-2018-099773
11. Caspersen CJ, Powell KE, Christenson GM. **Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research**. *Public health reports (Washington, DC: 1974).* (1985.0) **100** 126-31
12. Sallis JF, Cervero RB, Ascher W, Henderson KA, Kraft MK, Kerr J. **An ecological approach to creating active living communities.**. *Annu Rev Public Health* (2006.0) **27** 297-322. DOI: 10.1146/annurev.publhealth.27.021405.102100
13. **ACOG Committee Opinion No. 650: Physical Activity and Exercise During Pregnancy and the Postpartum Period.**. *Obstetrics and gynecology.* (2015.0) **126** e135-e42. DOI: 10.1097/AOG.0000000000001214
14. Mukona D, Munjanja SP, Zvinavashe M, Stray-Pederson B. **Physical activity in pregnant women in Africa: A systematic review**. *International Journal of Nursing and Midwifery* (2016.0) **8** 28-34
15. Tremblay MS, Aubert S, Barnes JD, Saunders TJ, Carson V, Latimer-Cheung AE. **Sedentary Behavior Research Network (SBRN)—Terminology Consensus Project process and outcome.**. *The international journal of behavioral nutrition and physical activity* (2017.0) **14** 75. DOI: 10.1186/s12966-017-0525-8
16. Bames J, Behrens TK, Benden ME, Biddle S, Bond D, Brassard P. **Standardized use of the terms" sedentary" and" sedentary behaviours".**. *Applied Physiology Nutrition and Metabolism-Physiologie Appliquee Nutrition Et Metabolisme* (2012.0) **37** 540-2. PMID: 22540258
17. Thorp AA, Healy GN, Owen N, Salmon J, Ball K, Shaw JE. **Deleterious associations of sitting time and television viewing time with cardiometabolic risk biomarkers: Australian Diabetes, Obesity and Lifestyle (AusDiab) study 2004–2005.**. *Diabetes care* (2010.0) **33** 327-34. DOI: 10.2337/dc09-0493
18. Wijndaele K, Healy GN, Dunstan DW, Barnett AG, Salmon J, Shaw JE. **Increased cardiometabolic risk is associated with increased TV viewing time**. *Medicine and science in sports and exercise* (2010.0) **42** 1511-8. DOI: 10.1249/MSS.0b013e3181d322ac
19. de Rezende LF, Rodrigues Lopes M, Rey-López JP, Matsudo VK, Luiz Odo C. **Sedentary behavior and health outcomes: an overview of systematic reviews.**. *PloS one* (2014.0) **9** e105620. DOI: 10.1371/journal.pone.0105620
20. 20Physiology CSfE. CSEP-PATH: Movement Counselling Tool for Adults Aged 18–64 years. 2021.
21. Fazzi C, Saunders DH, Linton K, Norman JE, Reynolds RM. **Sedentary behaviours during pregnancy: a systematic review.**. *The international journal of behavioral nutrition and physical activity* (2017.0) **14** 32. DOI: 10.1186/s12966-017-0485-z
22. Chasan-Taber L, Silveira M, Lynch KE, Pekow P, Solomon CG, Markenson G. **Physical activity and gestational weight gain in Hispanic women.**. *Obesity (Silver Spring, Md).* (2014.0) **22** 909-18. PMID: 23804434
23. Hayes L, Bell R, Robson S, Poston L. **Association between physical activity in obese pregnant women and pregnancy outcomes: the UPBEAT pilot study.**. *Annals of nutrition & metabolism* (2014.0) **64** 239-46. DOI: 10.1159/000365027
24. Chasan-Taber L, Silveira M, Pekow P, Braun B, Manson JE, Solomon CG. **Physical activity, sedentary behavior and risk of hypertensive disorders of pregnancy in Hispanic women**. *Hypertension in pregnancy* (2015.0) **34** 1-16. DOI: 10.3109/10641955.2014.946616
25. Reid EW, McNeill JA, Alderdice FA, Tully MA, Holmes VA. **Physical activity, sedentary behaviour and fetal macrosomia in uncomplicated pregnancies: a prospective cohort study**. *Midwifery* (2014.0) **30** 1202-9. DOI: 10.1016/j.midw.2014.04.010
26. Popkin BM. **The nutrition transition and obesity in the developing world**. *The Journal of nutrition* (2001.0) **131** 871s-3s. DOI: 10.1093/jn/131.3.871S
27. Prentice AM. **The emerging epidemic of obesity in developing countries**. *International journal of epidemiology* (2006.0) **35** 93-9. DOI: 10.1093/ije/dyi272
28. Abubakari AR, Lauder W, Agyemang C, Jones M, Kirk A, Bhopal RS. **Prevalence and time trends in obesity among adult West African populations: a meta-analysis.**. *Obesity reviews: an official journal of the International Association for the Study of Obesity* (2008.0) **9** 297-311. DOI: 10.1111/j.1467-789X.2007.00462.x
29. Bosu WK. **An overview of the nutrition transition in West Africa: implications for non-communicable diseases.**. *The Proceedings of the Nutrition Society* (2015.0) **74** 466-77. DOI: 10.1017/S0029665114001669
30. Gebregziabher D, Berhe H, Kassa M, Berhanie E. **Level of physical activity and associated factors during pregnancy among women who gave birth in Public Zonal Hospitals of Tigray.**. *BMC research notes.* (2019.0) **12** 454. DOI: 10.1186/s13104-019-4496-5
31. 31Adeoye IA. Effect of Maternal Obesity, lifestyle characteristics on glycaemic control, gestational weight gain and pregnancy outcomes in Ibadan, Nigeria. PhD Dissertation, University of Ibadan, Nigeria. 2021.
32. Chasan-Taber L, Schmidt MD, Roberts DE, Hosmer D, Markenson G, Freedson PS. **Development and validation of a Pregnancy Physical Activity Questionnaire.**. *Medicine and science in sports and exercise* (2004.0) **36** 1750-60. DOI: 10.1249/01.mss.0000142303.49306.0d
33. 33Physiology CSfE. CSEP-PATH: Movement Counselling Tool for adults aged 18–64 years. 2021.
34. Adeniyi AF, Ogwumike OO. **Physical activity and energy expenditure: findings from the Ibadan Pregnant Women’s Survey.**. *African journal of reproductive health* (2014.0) **18** 117-26. PMID: 25022148
35. Zhang Y, Dong S, Zuo J, Hu X, Zhang H, Zhao Y. **Physical activity level of urban pregnant women in Tianjin, China: a cross-sectional study.**. *PloS one* (2014.0) **9** e109624. DOI: 10.1371/journal.pone.0109624
36. Santo EC, Forbes PW, Oken E, Belfort MB. **Determinants of physical activity frequency and provider advice during pregnancy.**. *BMC pregnancy and childbirth.* (2017.0) **17** 286. DOI: 10.1186/s12884-017-1460-z
37. Liu J, Blair SN, Teng Y, Ness AR, Lawlor DA, Riddoch C. **Physical activity during pregnancy in a prospective cohort of British women: results from the Avon longitudinal study of parents and children**. *European journal of epidemiology* (2011.0) **26** 237-47. DOI: 10.1007/s10654-010-9538-1
38. Wojtyła C, Ciebiera M, Wojtyła-Buciora P, Janaszczyk A, Brzęcka P, Wojtyła A. **Physical activity patterns in third trimester of pregnancy–use of pregnancy physical activity questionnaire in Poland.**. *Annals of Agricultural and Environmental Medicine.* (2019.0) **27**. DOI: 10.26444/aaem/110480
39. Evenson KR, Wen F. **National trends in self-reported physical activity and sedentary behaviors among pregnant women: NHANES 1999–2006.**. *Prev Med* (2010.0) **50** 123-8. DOI: 10.1016/j.ypmed.2009.12.015
40. Connolly CP, Conger SA, Montoye AHK, Marshall MR, Schlaff RA, Badon SE. **Walking for health during pregnancy: A literature review and considerations for future research.**. *J Sport Health Sci* (2019.0) **8** 401-11. DOI: 10.1016/j.jshs.2018.11.004
41. Dunstan DW, Barr EL, Healy GN, Salmon J, Shaw JE, Balkau B. **Television viewing time and mortality: the Australian Diabetes, Obesity and Lifestyle Study (AusDiab).**. *Circulation* (2010.0) **121** 384-91. DOI: 10.1161/CIRCULATIONAHA.109.894824
42. Dunstan D, Salmon J, Healy G, Shaw J, Jolley D, Zimmet P. **Dunstan DW, Salmon J, Healy GN, Shaw JE, Jolley D, Zimmet PZ, Owen NAssociation of television viewing with fasting and 2-h postchallenge plasma glucose levels in adults without diagnosed diabetes**. *Diabetes Care* (2007.0) **30** 516-22. PMID: 17327314
43. Hu FB, Li TY, Colditz GA, Willett WC, Manson JE. **Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women**. *Jama* (2003.0) **289** 1785-91. DOI: 10.1001/jama.289.14.1785
44. Gollenberg AL, Pekow P, Bertone-Johnson ER, Freedson PS, Markenson G, Chasan-Taber L. **Sedentary behaviors and abnormal glucose tolerance among pregnant Latina women**. *Medicine and science in sports and exercise* (2010.0) **42** 1079-85. DOI: 10.1249/MSS.0b013e3181c6dec8
45. Hjorth MF, Kloster S, Girma T, Faurholt-Jepsen D, Andersen G, Kæstel P. **Level and intensity of objectively assessed physical activity among pregnant women from urban Ethiopia.**. *BMC pregnancy and childbirth.* (2012.0) **12** 154. DOI: 10.1186/1471-2393-12-154
46. Evenson KR, Wen F. **Prevalence and correlates of objectively measured physical activity and sedentary behavior among US pregnant women.**. *Prev Med.* (2011.0) **53** 39-43. DOI: 10.1016/j.ypmed.2011.04.014
47. Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ. **Compendium of physical activities: an update of activity codes and MET intensities.**. *Medicine and science in sports and exercise* (2000.0) **32** S498-S504. DOI: 10.1097/00005768-200009001-00009
48. Oken E, Ning Y, Rifas-Shiman SL, Radesky JS, Rich-Edwards JW, Gillman MW. **Associations of physical activity and inactivity before and during pregnancy with glucose tolerance.**. *Obstetrics and gynecology.* (2006.0) **108** 1200. DOI: 10.1097/01.AOG.0000241088.60745.70
49. Jiang H, Qian X, Li M, Lynn H, Fan Y, Jiang H. **Can physical activity reduce excessive gestational weight gain? Findings from a Chinese urban pregnant women cohort study**. *International Journal of Behavioral Nutrition and Physical Activity* (2012.0) **9** 1-7. DOI: 10.1186/1479-5868-9-12
50. Di Fabio DR, Blomme CK, Smith KM, Welk GJ, Campbell CG. **Adherence to physical activity guidelines in mid-pregnancy does not reduce sedentary time: an observational study.**. *The international journal of behavioral nutrition and physical activity* (2015.0) **12** 27. DOI: 10.1186/s12966-015-0191-7
51. Jones MA, Whitaker K, Wallace M, Barone Gibbs B. **Demographic, Socioeconomic, and Health-Related Predictors of Objectively Measured Sedentary Time and Physical Activity During Pregnancy**. *Journal of physical activity & health* (2021.0) **18** 957-64. PMID: 34140419
52. Arigbede O, Adeoye I, Jarrett O, Yusuf O. **Prediabetes among Nigerian adolescents: A School-based study of the prevalence, risk factors and pattern of fasting blood glucose in Ibadan, Nigeria.**. *International Journal of Diabetes in Developing Countries.* (2017.0) **37** 437-45
53. Martins LCG, Lopes MVdO, Diniz CM, Guedes NG. **The factors related to a sedentary lifestyle: A meta-analysis review**. *Journal of Advanced Nursing* (2021.0) **77** 1188-205. DOI: 10.1111/jan.14669
|
---
title: 'Assessing the capacity of primary health care facilities in Nigeria to deliver
eye health promotion: Results of a mixed-methods feasibility study'
authors:
- Ada Aghaji
- Helen E. D. Burchett
- Shaffa Hameed
- Clare Gilbert
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10022001
doi: 10.1371/journal.pgph.0000645
license: CC BY 4.0
---
# Assessing the capacity of primary health care facilities in Nigeria to deliver eye health promotion: Results of a mixed-methods feasibility study
## Abstract
Over 25 million people in sub-Saharan Africa are blind or visually impaired, the majority from avoidable causes. Health promotion and disease prevention are important strategies for eye health, through good governance, health literacy and increasing access to eye care services. To increase equity in access for eyecare services, the World Health Organization Africa Region developed a package of interventions for primary eye care, which includes health promotion. The aim of this study was to assess the capacity of the primary healthcare system to deliver health promotion for eye care in Nigeria. Mixed methods were used during a survey of 48 government-owned primary health care facilities in Anambra state, Nigeria: interviews with district health supervisors, facility staff and village health workers, and a desk review of policy documents for primary health care and eye care in Nigeria. Findings were benchmarked against the capacities needed to deliver health promotion agreed through a Delphi exercise and were analysed using the World Health Organization’s health system building blocks. Eye health promotion policies exist but are fragmented across different national health policies. Health promotion activities focussed on “mobilising” community members to access care provided in facilities, particularly for women of childbearing age and young children, and health education was limited. Only one in ten facilities engaged the elderly and a fifth delivered health promotion for eye care. Health promotion activities were supervised in $43.2\%$ of facilities and transport to remote areas was limited. A robust eye health promotion strategy needs to be included in the National Eye Health Policy. The scope of existing health promotion will need to expand to include eye conditions and different age groups. Increasing eye health literacy should be emphasized. Governance, training health workers in eye health promotion, educational materials, and transport to visit communities will also be needed.
## Introduction
It is estimated that in sub-Saharan Africa (SSA), 25 million adults have vision loss and a further 51.6 million have uncorrected presbyopia [1]. In addition, over 400,000 children in SSA are blind [2] and twice that number are visually impaired. Over $90\%$ of the causes of vision loss are avoidable [3] and are amenable to health promotion or can be prevented. Primary health care (PHC) activities in facilities and communities are the vehicles through which health promotion and disease prevention are typically delivered. For example, many of the causes of blindness in children are preventable, through measles immunization, ocular prophylaxis of the new-born and vitamin A supplementation, with health education to promote exclusive breast feeding and a vitamin A-rich diet. Children with treatable causes such as cataract and glaucoma, need to be detected early and referred to specialist services [4, 5]. Health promotion also has a key role to play through policies which ensure that these services are accessible, and by health education to empower mothers to make the right choices about their child’s eye health. In adults, refractive error, cataract and glaucoma are the most common causes of distance visual impairment while presbyopia causes near visual impairment. Although these conditions may not be treated at PHC level, they can be detected, and individuals counselled and referred for eye care. Creating awareness about these conditions and where to seek appropriate management may encourage community members to seek treatment. Trachoma, which is endemic in rural, northern Nigeria, is the most common cause of preventable vision loss in adults and has effective strategies for control. These entail facial and environmental hygiene, mass drug administration of azithromycin, and surgery for the sight-threatening lid complications. Diabetes and diabetic retinopathy are emerging epidemics in SSA particularly in urban settings and will require significant health promotion for their control. Many of the interventions for the control of visual impairment are highly cost effective, but in SSA many people become or remain blind as they do not access services even when they are available. Commonly reported barriers are cost [6–9], cultural and social factors, lack of trust in the services [6], and lack of knowledge and awareness that many eye conditions can be prevented or treated and where treatment is available [6, 7, 9]. Health promotion activities in eye health are key to overcoming these barriers.
The fundamental role of health promotion is to “enable people to increase control over, and to improve their own health” [10], as stated in the Ottawa Charter for Health Promotion in 1986. Recently, the objectives of health promotion have been realigned with the United Nation’s Sustainable Development Goals, in response to globalisation and climate change [11]. Although strategies for health promotion have changed over time, the basic principle remains the same. In 2016, at the Shanghai Declaration on health promotion, the World Health Organization (WHO) identified three pillars for the delivery of health promotion; good governance, health literacy and healthy cities [11]. Good governance entails implementing clear policies, developing regulations and legislation to make healthy choices accessible and affordable to all, and creating sustainable systems for society. Health literacy entails empowering individuals to make the healthiest choices and decisions for themselves and their families by increasing their knowledge and social skills [12]. Healthy cities involves prioritising policies that create synergies between health and other city policies, supporting cities to promote equity and social inclusion of their diverse populations and re-orienting health and social services to make them more equitable [11]. The concepts articulated in the three pillars of health promotion apply directly to primary health care (PHC), and consequently can be applied to primary eye health care (PEC), the eye health component of PHC. Governance and policies (which increase access to services for eye care), and health literacy could address the high prevalence of visual impairment in SSA.
Concerning health literacy, PHC workers can promote relevant health services and in addition, provide appropriate health education messages to empower community members to make appropriate health choices and ultimately, have better eye health “Fig 1”.
**Fig 1:** *Conceptual model linking primary eye care to the World Health Organization’s three pillars of health promotion.*
In terms of healthy cities, re orienting health facilities to optimize access to health services and fostering social inclusion is crucial in reducing barriers to accessing healthcare [11].
Some of these health promotion strategies have been implemented in SSA. Data from 43 countries in SSA have shown that improving the quality of governance improves health outcomes in terms of reducing under five mortality rates, for example [13]. In SSA, mobile phone technology has been used as an intervention to improve health literacy for sexual reproductive health, maternal and child health, Ebola, tuberculosis and malaria [14]. In terms of healthy cities, health faciities located close to communities have shown higher odds of facility delivery for pregnant women and lower odds of neonatal mortality [15].
The WHO have developed a framework for the social determinants of health, which includes (i) governance and policies, (ii) social status and education, and (iii) material circumstances, i.e., living and working conditions [16]. The pillars of health promotion—governance, health literacy and healthy cities (where people live and work) bear a strong relationship with the social structures that determine equity and the health of populations. Effective eye health promotion could, therefore, play an important role in addressing inequity in eye health.
To increase access to eye care, the WHO Africa Region recently launched a package of interventions for PEC (WHO AFRO PEC package) which has two broad components: facility based management and eye health promotion [17]. WHO AFRO recommend that all aspects of their PEC package, including health promotion, be undertaken by suitably trained PHC staff. The health promotion component comprises specific eye health promotion messages, and the training curriculum includes how to give a health talk and how to counsel patients. The messages target mothers and caregivers of young children, people of all ages particularly the elderly, people with diabetes and relatives of adults with glaucoma. It is recommended that the health promotion messages be delivered using posters, health talks and in one-to-one counselling, and could include the use of mass media, such as radio, to reach communities [17]. However, before eye health promotion can be effectively delivered, it is important to determine the capacities needed to deliver health promotion, and the extent to which they are available in PHC facilities.
In PHC facilities in Nigeria (health centres and health posts), health promotion is mainly delivered by junior community health extension workers (JCHEWs) and village health workers (VHWs). JCHEWs undergo two years of training in approved schools of health technology after completing secondary education. They are employed by the government, are attached to health centres or health posts, and spend $90\%$ of their time in the community. VHWs are volunteers nominated and remunerated by the community; they are trained from six days to up to three weeks by the JCEHWs [18, 19]. Sometimes, health promotion may be conducted by community health extension workers (CHEWS) who undergo four years post-secondary school training in approved schools of health technology “Fig 2”. However, health promotion will require more than the human resource component to be effective. Key capacities required include relevant health promotion materials, supervision, transport, referral mechanisms, intersectoral communication and partnerships [20].
**Fig 2:** *Roles of the different cadres of PHC staff [21].*
The overall purpose of this study was to assess capacity gaps in PHC facilities in Nigeria which would need to be addressed to effectively deliver the WHO AFRO PEC package. PHC capacity for facility case management has already been published [22, 23]. This paper reports the capacities available to deliver the health promotion component of the package.
## Ethics statement
Ethical clearance was obtained from the National Research Ethics Committee of the Federal Ministry of Health, Nigeria, and the ethics review boards of the University of Nigeria Teaching Hospital and the London School of Hygiene & Tropical Medicine. Written informed consent was obtained from all participants and permission was taken to audio-record the interviews and to use anonymous quotes. All data collected were anonymised and every effort was made to ensure confidentiality.
## Overview
The study had several stages, which have been described in detail [21]. In brief, the methods included a literature review of PEC in SSA and of technical feasibility frameworks to identify a suitable framework to assess the feasibility of delivering PEC [24], In addition, a two-round Delphi exercise was conducted to provide consensus on the technical complexity of each component of the WHO AFRO PEC package using Gericke’s framework of technical complexity [24], and the capacities needed to implement each of them. These were then mapped onto the WHO health system building blocks. The capacities needed, which were derived from the agreed technical complexities, formed the basis of the data collection methods and tools, and guided the selection of relevant participant groups. We conducted a survey in PHC facilities, interviews with district PHC supervisors, facility staff and VHWs, and a desk review of policy documents for PHC and eye care in Nigeria to identify statements of relevance to health promotion for eye health. All the findings were mapped onto WHO’s health systems building blocks.
## Policy review
A desk review of national health policy documents that could support eye health promotion was undertaken. Relevant statements from specific policy documents were extracted and tabulated in a Microsoft Excel spreadsheet for analysis.
## Selection of PHC facilities and staff
The survey of PHC facilities (health centres and health posts) was conducted in Anambra state, Nigeria, between October 2018 and May 2019. Anambra is located in south-eastern Nigeria and has a population of 5.5 million [25]. The adult literacy rate is $78.2\%$ compared with the national average of $62\%$ [26]. There are 21 local government areas, or districts, which can be stratified into urban, semiurban, and rural. The main occupations are farming, manufacturing, and commerce as in many other states in Nigeria.
A representative sample of facilities was selected using two-stage stratified random sampling from a total of 235 PHC centres and 112 health posts. Six districts were selected to reflect their urban, rural or semi-urban location (ratio of 1:2:3, respectively). Within districts, facilities were selected to reflect the 2:1 ratio of health centres to health posts. Hence a representative sample of 33 health centres and 15 health posts was selected. Structured questionnaires and observational checklists were administered to one (J)CHEW in each facility by a trained research assistant. If a facility had more than one (J)CHEW, one was randomly selected. Nine facilities (three health posts and six health centres) were purposively selected based on an interim analysis to identify high and low performing facilities in relation to the health workforce, patient load and supervisory practices. In these facilities, facility heads (nurse-midwives, community health officers, or (J)CHEWs) were interviewed, and 18 VHWs (two VHWs attached to the selected facility) were administered a structured questionnaire to determine their health promotion practices. Finally, the supervisors for health in the six districts were interviewed using structured topic guides. Interview findings were triangulated with data from the observational checklists and structured questionnaires.
## Data management
Data from the questionnaires were entered into a custom-made database in Microsoft Access and transferred to STATA V.15.1 (Statcorp, Texas) using STATransfer for analysis. Frequency tables were generated from the data and simple descriptive analyses were performed. Interview recordings, which were all in English, were transcribed verbatim by AA and checked for accuracy. Analysis entailed re-reading the transcripts for familiarisation with the data which were then indexed, charted, mapped and interpreted by AA. Open Code Software V. 4.02 was used to assist analysis.
## Review of policy documents
The National Health Act [2014] and the National Health Policy [2016] support health promotion (Table 1). In addition, eye health promotion is specifically mentioned in the Nigeria Eye Health Strategic Plan (2014–2019), the National Eye Health Policy [2019] and the National Health Policy [2016]. Furthermore, the National Strategic Health Development Plan II (2018–2022) specifies that an element of the eye health package at community and PHC level should include health education to promote eye health and disease prevention. In addition, the NSHDP 2018–2022 recommends promoting the establishment of a multi-sectoral coordination platform for eye health. However, eye health is not included in the most recent National Health Promotion Policy [2019].
**Table 1**
| Policy document | Policy for Health Promotion | Policy for eye health promotion |
| --- | --- | --- |
| National Health Promotion Policy 2006 [27] | Health promotion priorities should reflect the health needs in Nigeria, including both communicable and non-communicable diseases like injury prevention, mental health and oral healthHealth promotion should empower individuals and communities to make informed decisions about their health | Not included |
| National Primary Health Care Development Agency; Guidelines for the development of PHC systems in Nigeria. 2012 [19] | The Minimum Health Package for Non-Communicable Diseases will include health promotion/education materials on non-communicable diseases displayed in all facilities. The health education programme will provide information on health promotion, disease prevention at community and individual levels. This requires creating awareness, demand and utilization/patronage of health services and programmesFor the minimum package for health education and community mobilization, every facility should have relevant health promotion/ education materials conspicuously displayed in culturally acceptable language with graphics | Not included |
| National Health Act 2014 [28] | The National Council on Health, is the highest policy making body in Nigeria on matters relating to health, to promote health and healthy lifestyles | Not included |
| National Primary Health Care Development Agency Management Guideline For Primary Health Care Under One Roof 2016 [29] | The ward minimum health care package should include health promotion and community mobilisation | Not included |
| National Eye Health Policy 2016 | Strengthen the capacity for health promotion at all levels | One of the activities is to improve public awareness of eye health |
| National Strategic Health Development Plan II. 2018–2022 [30] | | An element of the eye health package at community and PHC level is health education for eye health promotion and disease preventionStrategic interventions to deliver eye care include: to strengthen advocacy, social mobilization and behaviour change communication on eye health and expand access (financial, geographical, social etc.) to comprehensive (promotive, preventive, curative and rehabilitative), appropriate and quality eye health services at all levels |
| Implementation guidelines for primary health care under one roof (PHCUOR) 2018 [31] | The Local Government Health Authority Management Team will include a programme officer for health promotion | Not included |
| Nigeria Eye Health Strategic Plan. 2014–2019 [32] | | States to provide eye health education at community and PHC levels targeting those affected and those at risk of avoidable blindness. This should also focus on harmful eye practices, including *couchingThe National Eye Health Programme Secretariat will design and produce the eye health education materials |
| National Eye health Plan 2019 | | Healthcare facilities should deliver eye health promotionPHC workers should be trained to deliver eye health promotionGovernment will facilitate and regulate community participation in eye health promotion.Intersectoral collaboration will be done to improve eye health promotion |
## Facility survey
In the 48 facilities surveyed, all the heads of facility interviewed were female and representation was evenly distributed between urban, semi urban and rural facilities. However, four (J)CHEWs were not available; two were on study or maternity leave and two facilities did not have a (J)CHEW). All the (J)CHEWs were female, their mean age was 41.4 (SD± 8) years and $93.2\%$ had completed training in schools of health technology.
The majority of health promotion, whether delivered in communities or facilities, focussed on young children and pregnant women, with very little time spent on health promotion for the elderly, general health for people of all ages or for people with diabetes “Fig 3A and 3B”.
**Fig 3:** *A and B amount of time staff spent with different groups on health promotion A in the community and B in facilities.*
## Health service data collection
There was no documented communication between the community and facilities as there were no referral slips, referral registers or evidence of two-way communication.
As there were some differences in health promotion activities undertaken in the community and in facilities they are described separately.
## Human resources for health
Sixteen VHWs were interviewed as one facility had no VHW attached to it. Their mean age was 47.4 (SD±10) years and $25\%$ were male. All the VHWs lived in the community, knew it very well and spoke the local language. The mean number of years they had lived in the community was 29.8 (SD±12) years.
Less than half of the 48 facilities had a VHW who conducted health promotion in the community, while over $80\%$ of facilities reported that community health promotion was conducted by (J)CHEWs (Table 2). $25\%$ of the VHWs interviewed said they rarely conducted health promotion activities.
**Table 2**
| Unnamed: 0 | Health centre N = 32 | Health centre N = 32.1 | Health post N = 12 | Health post N = 12.1 | Total N = 44 | Total N = 44.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Health workforce | N | % | N | % | N | % |
| Health promotion conducted by VHWs | 14 | 43.8 | 5 | 41.7 | 19 | 43.2 |
| Health promotion conducted by (J)CHEWs | 28 | 87.5 | 9 | 75.0 | 37 | 84.1 |
| (J)CHEW fluent in local language | 31 | 96.9 | 12 | 100 | 43 | 97.7 |
| (J)CHEWs knowledge of community: good/moderately good | 27 | 84.4 | 12 | 100 | 39 | 88.6 |
| (J)CHEWs confidence in delivering eye health promotion | | | | | | |
| A little or more confident | 14 | 43.8 | 7 | 58.3 | 21 | 47.7 |
| Not confident | 18 | 56.2 | 5 | 41.7 | 23 | 52.3 |
| Willing to be trained on eye health promotion | 30 | 93.8 | 12 | 100 | 42 | 95.5 |
Health promotion in the community should, ideally, be conducted by VHWs, but this was mainly undertaken by facility staff, including facility heads. The term frequently used by staff to describe health promotion was “mobilization”, by which they meant that they inform communities about the services available in the facility and encourage them to attend. The following comments from facility heads illustrate the work they do in communities: During the interview with one facility head it became clear that UNICEF had engaged their own VHWs to undertake health promotion on maternal and child health topics: It appears that a shortage of staff was hampering health promotion activities, as one head of facility explained: Half ($48\%$) of (J)CHEWs were confident about delivering eye health promotion and $96\%$ were willing to undergo training (Table 2).
Health promotion activities. Health promotion activities by (J)CHEWS also mainly focused on mobilizing the community. For example, mothers of young children were informed that PHC facilities provide immunization and vitamin A supplements, or that medication was available for onchocerciasis, as explained by many heads of facility.
Most of the health promotion activities focussed on mothers and children (Table 3), which included health education about exclusive breast feeding. A key role of PHC staff was also to promote safe water and sanitation, and advise on household waste disposal, activities which were reported by a third to almost half of those interviewed.
**Table 3**
| Unnamed: 0 | Health centre N = 32 | Health centre N = 32.1 | Health post N = 12 | Health post N = 12.1 | Total N = 44 | Total N = 44.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Frequency of health promotion by VHWs | N | % | N | % | N | % |
| At least weekly | 4 | 12.5 | 3 | 25.0 | 7 | 15.9 |
| At least monthly | 3 | 9.4 | 3 | 25.0 | 6 | 13.6 |
| Rarely/never | 25 | 78.1 | 6 | 50.0 | 31 | 70.5 |
| Community health promotion by (J)CHEWs: | Community health promotion by (J)CHEWs: | Community health promotion by (J)CHEWs: | Community health promotion by (J)CHEWs: | Community health promotion by (J)CHEWs: | Community health promotion by (J)CHEWs: | Community health promotion by (J)CHEWs: |
| Measles immunisation | 30 | 93.8 | 12 | 100 | 42 | 95.5 |
| Vitamin A supplementation of young children | 29 | 90.6 | 10 | 83.3 | 39 | 88.6 |
| Exclusive infant breastfeeding | 21 | 65.6 | 8 | 66.7 | 29 | 65.9 |
| Diabetes prevention | 8 | 25.0 | 2 | 16.7 | 10 | 22.7 |
| Eye health | 6 | 18.8 | 3 | 25.0 | 9 | 20.5 |
| Care of the elderly | 3 | 9.4 | 2 | 16.7 | 5 | 11.4 |
| Safe water | 12 | 37.5 | 3 | 25.0 | 15 | 34.1 |
| Safe sanitation | 15 | 46.9 | 5 | 41.7 | 20 | 45.5 |
| Household waste disposal | 16 | 50.0 | 2 | 16.7 | 18 | 40.9 |
| How messages are communicated in the community | How messages are communicated in the community | How messages are communicated in the community | How messages are communicated in the community | How messages are communicated in the community | How messages are communicated in the community | How messages are communicated in the community |
| Meetings with the target audience | 29 | 90.6 | 12 | 100 | 41 | 93.2 |
| Town criers/announcers | 22 | 68.8 | 8 | 66.7 | 30 | 68.2 |
| Home visits | 8 | 25.0 | 3 | 25.0 | 11 | 25.0 |
| Community leaders | 7 | 21.9 | 1 | 8.3 | 8 | 18.2 |
A range of avenues were available for health workers to meet with target audiences in the community for health promotion, such as village meetings, and using town criers/announcers (Table 3), as two facility heads explained:
## Health promotion materials and transport
Over two thirds of facilities had health promotion materials for use in the community, but only a fifth were in the local language (Table 4). The commonest audiences for the posters were mothers of young children ($63.6\%$) and pregnant women ($54.5\%$) with none for people with diabetes or the elderly. The only education materials which mentioned eyes were posters on vitamin A supplementation.
**Table 4**
| Unnamed: 0 | Health centre N = 32 | Health centre N = 32.1 | Health post N = 12 | Health post N = 12.1 | Total N = 44 | Total N = 44.1 |
| --- | --- | --- | --- | --- | --- | --- |
| | N | % | N | % | N | % |
| Ease of acquisition materials for all health promotion | | | | | | |
| Fairly easy or easy | 30 | 93.8 | 12 | 100 | 42 | 95.5 |
| Difficult to acquire / not available | 2 | 6.2 | 0 | 0 | 2 | 4.5 |
| Materials for health promotion in the community | | | | | | |
| Available | 20 | 62.5 | 10 | 83.3 | 30 | 68.2 |
| With explanatory graphics | 19 | 59.4 | 10 | 83.3 | 29 | 65.9 |
| In local languages | 5 | 15.6 | 4 | 33.3 | 9 | 20.5 |
| Available for the following target populations: | | | | | | |
| Mothers with children 0–5 years | 19 | 59.4 | 9 | 75.0 | 28 | 63.6 |
| Pregnant women | 13 | 40.6 | 9 | 75.0 | 24 | 54.5 |
| All ages | 3 | 9.4 | 3 | 25.0 | 6 | 13.6 |
| People with diabetes | 0 | 0 | 0 | 0 | 0 | 0 |
| The elderly | 0 | 0 | 0 | 0 | 0 | 0 |
Less than $10\%$ of facilities had transport for PHC staff to visit communities. A head of facility commented on the challenges faced when visiting some hard-to-reach communities:
## Governance
A higher proportion of health promotion activities conducted in the community by (J)CHEWs were supervised if they worked in health posts ($25\%$) than in health centres ($9.4\%$), and supervision of VHWs was only reported by two heads of health centres ($6.3\%$). It became clear that (J)CHEWs were largely not aware of what the VHWs were doing, and they had limited control over their activities, as explained by two facility heads: One of the reasons given for the lack of VHW supervision was that they no longer dispense medication: All the health promotion materials observed for use in the community had the official logos.
## Service delivery
Staff in all the facilities reported that they deliver health talks for measles immunisation and vitamin A supplementation but less than $10\%$ of facilities gave health talks for the elderly or for the prevention of diabetes (Table 5). Only one facility, a health centre, had a list of topics to be covered in health talks, but this was not a weekly or monthly schedule.
**Table 5**
| Unnamed: 0 | Health centre N = 32 | Health centre N = 32.1 | Health post N = 12 | Health post N = 12.1 | Total N = 44 | Total N = 44.1 |
| --- | --- | --- | --- | --- | --- | --- |
| Leadership and Governance | n | % | n | % | n | % |
| List of topics for health education in the facility | 1 | 3.1 | 0 | 0 | 1 | 2.3 |
| Supervision of health promotion in the facility | 15 | 46.9 | 4 | 33.3 | 19 | 43.2 |
| Facilities that conduct health talks | 32 | 100 | 12 | 100 | 44 | 100 |
| Topics covered in health talks | | | | | | |
| Vitamin A supplementation of young children | 32 | 100 | 12 | 100 | 44 | 100 |
| Measles immunisation | 32 | 100 | 12 | 100 | 44 | 100 |
| Exclusive infant breastfeeding | 30 | 93.8 | 6 | 50.0 | 36 | 81.8 |
| Diabetes prevention | 4 | 12.5 | 3 | 25.0 | 7 | 15.9 |
| Care of the elderly | 1 | 3.1 | 2 | 16.7 | 3 | 6.8 |
| Materials for health promotion in facilities | | | | | | |
| Available | 32 | 100.0 | 12 | 100.0 | 44 | 100. |
| With explanatory graphics | 30 | 93.8 | 11 | 91.7 | 41 | 93.2 |
| In local languages | 24 | 75.0 | 9 | 75.0 | 33 | 75.0 |
| Target populations for materials | | | | | | |
| Mothers of children 0–5 years* | 28 | 87.5 | 12 | 100 | 40 | 90.9 |
| Pregnant women | 29 | 90.6 | 9 | 75.0 | 38 | 86.4 |
| All ages | 8 | 25.0 | 4 | 33.3 | 12 | 27.3 |
| People with diabetes | 0 | 0.0 | 0 | 0.0 | 0 | 0.0 |
| The elderly | 0 | 0.0 | 0 | 0.0 | 0 | 0.0 |
Staff also engaged in health promotion and health education talks in the facility, as explained by a facility head: Another facility head thought that delivering eye care in the facility would be likely to lead to a reduction in the use of traditional eye remedies: Infrastructure, equipment and consumables. Posters to support health promotion in facilities were similar to those for use in the community, but posters in facilities were more likely to be in the local language ($75\%$), and a high proportion had illustrative graphics suitable for audiences with low levels of literacy. There were no health promotion posters for the elderly or for people with diabetes (Table 5).
Governance. A total of 19 JCHEWs ($43.2\%$) reported that their health promotion activities in the facility were supervised, which was slightly higher in health centres ($46.9\%$) than in health posts ($33.3\%$) (Table 5).
To be effective, health promotion should be delivered to the appropriate target audience. Health promotion activities were mainly targeted at women of child-bearing age and their young children. Similar findings were reported in a study of community based health workers in Ethiopia where health promotion for measles immunisation, vitamin A supplementation and Crede’s prophylaxis was conducted by over $70\%$ of PHC workers, while eye health promotion was conducted by less than a third [35]. Maternal and child health has been and remains a major and important focus of health promotion activities as a component of PHC. However, over the last decade or so there has been a global shift towards promoting the health of everyone in the community [41]. For health promotion to contribute towards a reduction in visual impairment and blindness, there needs to be an awareness of all the appropriate target groups. The elderly, who have the highest prevalence of blinding eye conditions, and people with diabetes who are at risk of diabetic retinopathy are very important target audiences. In addition, women, especially widows who have lower access to eye care in low- and middle-income countries (LMICs), should also be targeted [42]. Delivering targeted eye health promotion at community level may be one step on the ladder to reduce health inequities in accessing appropriate eye care. Indeed, health promotion has been identified as a key strategy to reduce health inequalities and it has been suggested that targeted health promotion should be prioritised [43]. In this study, the (J)CHEWs used a range of mechanisms to reach different target audiences, including talking to village, town criers and community leaders. One of the PHC workers in our study indicated that the topics of the health talks were spontaneous, depending on what seems to be relevant at the time. While this is a form of “people-centredness,” a prepared list of topics, including eye conditions, would ensure relevant health topics are covered. Other mechanisms for reaching the elderly and those with visual impairment could be informal providers of health care, such as Propriety Patent Medicine Vendors who are ubiquitous in Nigeria [44] and traditional healers, and religious leaders who played a key role in promoting the polio immunization campaign in Nigeria, for example [45]. One head of facility reported that oral herbal remedies were used to treat eye conditions in her community and suggested that introducing PEC may reduce their use. The use of traditional eye remedies is widespread in SSA, some of which are harmful [46]. In addition, the use of such remedies may delay seeking treatment from orthodox sources. This emphasizes the need for accessible eye care services and targeted eye health promotion.
## Policy gaps
Eye health promotion is key to delivering WHO’s Integrated People Centre Eye Care [33] and should be a crucial pillar of any eye health policy. However, our study shows that some although enabling national policies for eye health promotion are in place, these are limited, particularly at the PHC level. In addition, these policies are not being implemented possibly because they are fragmented across several other policies, and their relevance and importance may not be appreciated by non-eye care professionals. To address this, as recommended in the Nigeria National Strategic Health Development Plan II, 2018–2022 [30], as part of a multisectoral coordination platform for eye health, stakeholders from other health disciplines and non-health sectors need to work together to strengthen eye health promotion policies for the Nigeria National Eye Health Policy, addressing the current limitations particularly at the primary level. The resulting cohesive and comprehensive strategy would need to be ratified by the National Council on Health, the highest health policy making body in Nigeria, and relevant components integrated into other policies, such as child health, non-communicable diseases, care of the elderly and water and sanitation. The National Eye Health *Policy is* the main policy and therefore advocacy instrument for eye care. We recommend bringing together all eye health promotion strategies in the National Eye Health Policy because in South Africa, failure to implement eye health promotion at the primary level was attributed to the lack of a specific eye health promotion policy [34].
## Human resource development for eye health promotion
Appropriately trained health workers are crucial in the delivery of a health intervention, and workers skilled in health promotion should be available to deliver appropriate health education messages. However, imparting health education to the community, which would improve heath literacy, was limited in our study. Instead, the thrust of health promotion activities in the community was to increase uptake of the services provided in health facilities. The reason for the apparent lack of health education could be that staff lacked training in health education and behaviour change communication.
In our study, less than half of the (J)CHEWs were confident about delivering eye health promotion but almost all were willing to be trained in this area. Similar findings were reported in a study of eye health promotion in Ethiopia, where $47\%$ of PHC workers were confident in delivering PEC and $75\%$ felt that more training was needed [35]. Addressing this training capacity gap is crucial because health workers delivering health promotion need to be purposefully trained and have the necessary skills to deliver clear and relevant messages to the appropriate audiences [36, 37].
## Appropriate health promotion materials
Health promotion needs to be supported by appropriate educational materials. In our study, although health promotion materials were readily available and easy to acquire, over a third of facilities did not use them in the community and most were not in the local language. While this may have contributed to the lack of health education in communities, the majority of the posters had self-explanatory graphics. However, the focus of the health promotion posters was on maternal and child health. One of the posters, on vitamin A supplementation, included a message that supplements keep children’s eyes healthy, and there is scope for a similar message to be included in materials for measles immunization. Advocacy will be needed to ensure that messages for people with diabetes include that diabetes can causes serious eye complications, with messages which would increase their health literacy on what they could do to reduce the risk [38]. In addition, posters on eye conditions in older adults e.g. on presbyopia and cataract, could create awareness and stimulate conversations around accessing appropriate care for these conditions. More recently there have been calls for innovative methods of delivering health promotion messages [39]. A recent study in India demonstrated the effectiveness of using mobile phones to improve eye health literacy [40]. With increasing mobile phone penetration and internet connectivity in SSA, mobile phones could become an important tool for disseminating health promotion messages, including for eye health.
## Supervision
Supervision is an important aspect of leadership and governance in health service delivery. However, supervision of health promotion in facilities and in the community was very limited. In the facilities, this may explain why none of the facilities had a schedule for health education talks and topics, with only one having a list of topics which should be covered in health talks.
Concerning community health promotion, VHWs were hardly supervised at all and the (J)CHEWs interviewed had very limited awareness of what the VHWs were doing. An explanation given for this was that VHWs do not require supervision as they are no longer permitted to treat patients. Lack of supervision may also explain why materials were not available in the local languages, particularly for activities in the community where it could be argued that they are more important. There was anecdotal evidence that a United Nations agency had appointed VHWs for health promotion, so by-passing the supervisory function of staff in health facilities. The fragmented structure of health promotion and its supervision may lead to the dissemination of disparate messages, and so there is need to develop a unified managerial structure for health promotion. Although there is conflicting evidence about the impact of supervision on health outcomes [47], it has been suggested that health system challenges, particularly a shortage of human resources in poor countries, make supportive supervision even more important, and that it is the quality of the supervision which is more important than the frequency [48].
## Infrastructure
Health promotion and its supervision in communities will require appropriate transport arrangements, which was a challenge in this study, particularly in hard-to-reach areas. Local transport will be needed by staff to visit communities [49] and lack of transport for field work in eye care has been a recurring problem [50]. Transport promotes access to health centres by community members as well as access to communities by health workers. Policy makers need to recognise the impact of lack of transport on the delivery of health services, and inter-sectoral collaborations which align health and transport policies, may help to address this. Creating synergies between health and transport policies aligns with the “healthy cities” policy of the Shanghai declaration on health promotion [11]. This is in line with the good governance pillar of the Shanghai Health Promotion Policy and the Nigerian health policy which encourage the promotion of inter-sectoral action for health and effective partnerships among all relevant stakeholders for health development by mainstreaming ‘Health-in-All’ policies [51].
## Generalizability to other settings
Eye health promotion, which encompasses good governance, health literacy and healthy cities with equitable access to health care, cannot exist in a vacuum and can only be embedded in existing health promotional structures. Although there are many critical gaps for health promotion, one of the strengths is the focus on maternal and childhood conditions, and eye health promotion for young children can be integrated into this. This can be an entry point for eye health promotion. Indeed, there are efforts underway to embed eye health and eye health prevention activities into WHO’s Integrated Management of Childhood Illnesses (IMCI) [52]. In many countries, eye health is not included in national health plans [33]. To implement and deliver eye health promotion effectively, eye health policies need to be intentional about including specific policies for PEC and eye health promotion which align with existing national policies. For SSA countries that will be implementing the WHO AFRO PEC, including the eye health promotion component, it is important to assess the capacity of the health system at the community and primary health care level to implement it.
## Strengths and limitations
To the best of our knowledge, this is the first study to assess the capacity of the health system to deliver eye health promotion. Assessment tools were developed based on a rigorous process that involved a review of the literature, appropriate technical capacity frameworks and consensus from experts in eye health in SSA using a Delphi exercise. The findings may be generalizable to other countries in SSA. A limitation of this study is that it did not assess community participation, which is central to the success of health promotion.
Further studies will be needed on the impact of health promotion on the uptake of eye care services in health facilities.
## Conclusions
Our study identified several capacity gaps in health promotion which will need to be addressed to implement eye health promotion effectively. A robust eye health promotion strategy needs to be included in the National Eye Health Policy. The scope of existing health promotion will need to expand to include those with eye conditions, the elderly and people with diabetes. This will require trained health workers, targeted health promotion supported by relevant health promotion materials in the local languages which are suitable for populations with low levels of literacy, transport to visit communities and supportive supervision.
## References
1. Bourne R, Steinmetz JD, Flaxman S, Briant PS, Taylor HR, Resnikoff S. **Trends in prevalence of blindness and distance and near vision impairment over 30 years: an analysis for the Global Burden of Disease Study**. *The Lancet Global Health* (2021.0) **9** e130-e43. DOI: 10.1016/S2214-109X(20)30425-3
2. Chandna A, Gilbert C. **When your eye patient is a child**. *Community Eye Health* (2010.0) **23** 1-3. PMID: 20523854
3. Burton MJ, Ramke J, Marques AP, Bourne RR, Congdon N, Jones I. **The lancet global health commission on global eye health: vision beyond 2020**. *The Lancet Global Health* (2021.0) **9** e489-e551. DOI: 10.1016/S2214-109X(20)30488-5
4. Gilbert C, Foster A. **Childhood blindness in the context of VISION 2020—the right to sight**. *Bulletin of the World Health Organization* (2001.0) **79** 227-32. PMID: 11285667
5. Aghaji A, Okoye O, Bowman R. **Causes and emerging trends of childhood blindness: findings from schools for the blind in Southeast Nigeria**. *The British journal of ophthalmology* (2015.0) **99** 727-31. DOI: 10.1136/bjophthalmol-2014-305490
6. Aboobaker S, Courtright P. **Barriers to Cataract Surgery in Africa: A Systematic Review**. *Middle East African journal of ophthalmology* (2016.0) **23** 145-9. DOI: 10.4103/0974-9233.164615
7. Ajibode H, Jagun O, Bodunde O, Fakolujo V. **Assessment of barriers to surgical ophthalmic care in South-Western Nigeria**. *J West Afr Coll Surg* (2012.0) **2** 38-50. PMID: 25453003
8. Muhammad N, Mansur RM, Dantani AM, Elhassan E, Isiyaku S. **Prevalence and causes of blindness and visual impairment in sokoto state, Nigeria: baseline data for vision 2020: the right to sight eye care programme**. *Middle East African journal of ophthalmology* (2011.0) **18** 123-8. DOI: 10.4103/0974-9233.80700
9. Odugbo OP, Mpyet CD, Chiroma MR, Aboje AO. **Cataract blindness, surgical coverage, outcome, and barriers to uptake of cataract services in Plateau State, Nigeria**. *Middle East African journal of ophthalmology* (2012.0) **19** 282-8. DOI: 10.4103/0974-9233.97925
10. 10World Health Organisation. Ottawa Charter for Health Promotion. In: WHO E, editor. 1986.
11. **Shanghai declaration on promoting health in the 2030 Agenda for Sustainable Development**. *Health promotion international* (2017.0) **32** 7. DOI: 10.1093/heapro/daw103
12. Nutbeam D.. **The evolving concept of health literacy**. *Social science & medicine* (2008.0) **67** 2072-8. DOI: 10.1016/j.socscimed.2008.09.050
13. Makuta I, O’Hare B. **Quality of governance, public spending on health and health status in Sub Saharan Africa: a panel data regression analysis**. *BMC Public Health* (2015.0) **15** 932. PMID: 26390867
14. 14Ouedraogo I, Some BMJ, Benedikter R, Diallo G. Mobile technology as a health literacy enabler in African rural areas: a literature review. 2021.
15. Karra M, Fink G, Canning D. **Facility distance and child mortality: a multi-country study of health facility access, service utilization, and child health outcomes**. *International journal of epidemiology* (2017.0) **46** 817-26. DOI: 10.1093/ije/dyw062
16. 16World Health Organization. A conceptual framework for action on the social determinants of health. World Health Organization. 2010; https://apps.who.int/iris/handle/10665/44489.
17. 17World Health Organisation Africa Region. Primary Eye Care Training Manual-A course to strengthen the capacity of health personnel to manage eye patients at primary-level health facilities in the African Region. Brazzaville: World Health Organization. Regional Office for Africa2018. p. https://www.afro.who.int/publications/primary-eye-care-training-manual.
18. Bamisaiye A, Olukoya A, Ekunwe EO, Abosede O. **A village health worker programme in Nigeria**. *World Health Forum (World Health Organisation)* (1989.0) **10** 386-92. PMID: 2637712
19. 19National Primary Health Care Development Agency. National Guidelines for the Development of Primary Health Care System in Nigeria. Abuja Nigeria: Federal Governmet of Nigeria; 2012.
20. Aghaji A, Burchett HE, Mathenge W, Faal HB, Umeh R, Ezepue F. **Technical capacities needed to implement the WHO’s primary eye care package for Africa: results of a Delphi process**. *BMJ open* (2021.0) **11** e042979. DOI: 10.1136/bmjopen-2020-042979
21. Aghaji A, Burchett H, Hameed S, Webster J, Gilbert C. **The Technical Feasibility of Integrating Primary Eye Care Into Primary Health Care Systems in Nigeria: Protocol for a Mixed Methods Cross-Sectional Study**. *JMIR research protocols* (2020.0) **9** e17263. DOI: 10.2196/17263
22. Aghaji A, Burchett HED, Oguego N, Hameed S, Gilbert C. **Human resource and governance challenges in the delivery of primary eye care: a mixed methods feasibility study in Nigeria**. *BMC Health Serv Res* (2021.0) **21** 1321. DOI: 10.1186/s12913-021-07362-8
23. Aghaji A, Burchett HED, Oguego N, Hameed S, Gilbert C. **Primary health care facility readiness to implement primary eye care in Nigeria: equipment, infrastructure, service delivery and health management information systems**. *BMC Health Serv Res* (2021.0) **21** 1360. DOI: 10.1186/s12913-021-07359-3
24. Gericke CA, Kurowski C, Ranson MK, Mills A. **Intervention complexity: a conceptual framework to inform priority-setting in health**. *Bulletin of the World Health Organization* (2005.0) **83** 285-93. PMID: 15868020
25. 25National Bureau of Statistics. 2017 Demographics Statistics Bulletin. https://nigerianstatgovng/download/775. 2018.
26. 26National Bureau of Statistics. Socioeconomic statistics, Literacy- Nigeria Data Portal. 2014. https://nigeria.opendataforafrica.org/qwfntxd/profile?states=Anambra
27. 27Federal Government of Nigeria. National Health Promotion Policy. https://wwwafrowhoint/sites/default/files/2017-06/Nigeria_national_health_promotion_policy_feb2006pdf. 2006;Accessed 5th February 2020.
28. 28Federal Republic of Nigeria. The National Health Act. Federal Republic of Nigeria Official Gazette2014. p. A139-72.
29. 29National Primary Health Care Development Agency. Integrating Primary Health Care Governance in Nigeria: PHC under one roof (Management Guidelines). Abuja, Nigeria: Federal Ministry of Health; 2016.
30. 30Federal Ministry of Health. National Strategic Health Development Plan II. In: https://www.health.gov.ng/doc/NSHDP%20II%20Final.pdf, editor. Abuja2019. p. https://www.health.gov.ng/doc/NSHDP%20II%Final.pdf.
31. 31National Primary Health Care Development Agency. Implementation Guidelines for Primary Health Care under one roof (PHCUOR). Abuja, Nigeria: Federal Ministry of Health; 2018.
32. 32Federal Ministry of Health. Nigeria National Eye Health Strategic Plan. 2014. 2014.
33. 33World Health Organisation. World Report on Vision. https://wwwwhoint/publications-detail/world-report-on-vision. 2019;Accessed 23rd February 2020.
34. Sithole H, Oduntan O. **Eye health promotion in the South African primary health care system**. *African Vision and Eye Health* (2010.0) **69** 200-6
35. Aemero A, Berhan S, Yeshigeta G. **Role of health extension workers in eye health promotion and blindness prevention in Ethiopia**. *JOECSA* (2014.0) **18** 68-74
36. Hubley J, Gilbert C. **Eye health promotion and the prevention of blindness in developing countries: critical issues**. *British journal of ophthalmology* (2006.0) **90** 279-84. DOI: 10.1136/bjo.2005.078451
37. 37World Health Organisation. A Vision for primary health care in the 21st century. https://wwwwhoint/docs/default-source/primary-health/visionpdf. 2018;Accessed 6th February 2020.
38. Gurmu Y, Gela D, Aga F. **Factors associated with self-care practice among adult diabetes patients in West Shoa Zone, Oromia Regional State, Ethiopia**. *BMC health services research* (2018.0) **18** 732. DOI: 10.1186/s12913-018-3448-4
39. Macnab AJ, Mukisa R. **Celebrity endorsed music videos: innovation to foster youth health promotion**. *Health Promot Int* (2019.0) **34** 716-25. DOI: 10.1093/heapro/day042
40. Sharma IP, Chaudhry M, Sharma D, Kaiti R. **Mobile health intervention for promotion of eye health literacy**. *PLOS Global Public Health* (2021.0) **1** e0000025
41. 41World Health Organization. The world health report 2008: primary health care now more than ever. 2008. Geneva, WHO. 2008.
42. Ramke J, Kyari F, Mwangi N, Piyasena M, Murthy G, Gilbert CE. **Cataract Services are Leaving Widows Behind: Examples from National Cross-Sectional Surveys in Nigeria and Sri Lanka**. *International journal of environmental research and public health* (2019.0) **16**. DOI: 10.3390/ijerph16203854
43. Barsanti S, Salmi LR, Bourgueil Y, Daponte A, Pinzal E, Ménival S. **Strategies and governance to reduce health inequalities: evidences from a cross-European survey**. *Global health research and policy* (2017.0) **2** 18. DOI: 10.1186/s41256-017-0038-7
44. Beyeler N, Liu J, Sieverding M. **A systematic review of the role of proprietary and patent medicine vendors in healthcare provision in Nigeria**. *PLoS One* (2015.0) **10**. DOI: 10.1371/journal.pone.0117165
45. Nasir S-G, Aliyu G, Ya’u I, Gadanya M, Mohammad M, Zubair M. **From intense rejection to advocacy: How Muslim clerics were engaged in a polio eradication initiative in Northern Nigeria**. *PLoS medicine* (2014.0) **11**
46. Aghaji A, Ezeome I, Ezeome E. **Evaluation of content and cost of traditional eye medication in a resource-poor country–Implications for eye care practice and policy**. *Nigerian journal of clinical practice* (2018.0) **21** 1514-9. DOI: 10.4103/njcp.njcp_201_18
47. Bosch-Capblanch X, Liaqat S, Garner P. **Managerial supervision to improve primary health care in low-and middle-income countries**. *Cochrane Database Syst Rev* (2011.0) **9**. DOI: 10.1002/14651858.CD006413.pub2
48. Avortri GS, Nabukalu JB, Nabyonga-Orem J. **Supportive supervision to improve service delivery in low-income countries: is there a conceptual problem or a strategy problem?**. *BMJ global health* (2019.0) **4** e001151. DOI: 10.1136/bmjgh-2018-001151
49. Müller A, Murenzi J, Mathenge W, Munana J, Courtright P. **Primary eye care in Rwanda: gender of service providers and other factors associated with effective service delivery**. *Tropical Medicine & International Health* (2010.0) **15** 529-33. DOI: 10.1111/j.1365-3156.2010.02498.x
50. Sutter EE. **Training of eye care workers and their integration in Gazankulu’s comprehensive health services**. *Social science & medicine (1982)* (1983.0) **17** 1809-12. DOI: 10.1016/0277-9536(83)90396-9
51. 51Federal Ministry of Health Abuja Nigeria. National Health Policy 2016: Promoting the Health of Nigerians to Accelerate Socio-economic Development. 2016.
52. Malik ANJ, Mafwiri M, Gilbert C, Kim MJ, Schellenberg J. **Integrating eye health training into the primary child healthcare programme in Tanzania: a pre-training and post-training study**. *BMJ paediatrics open* (2020.0) **4** e000629. DOI: 10.1136/bmjpo-2019-000629
|
---
title: '“I see salt everywhere”: A qualitative examination of the utility of arts-based
participatory workshops to study noncommunicable diseases in Tanzania and Malawi'
authors:
- Maria Bissett
- Cindy M. Gray
- Sharifa Abdulla
- Christopher Bunn
- Amelia C. Crampin
- Angel Dillip
- Jason M. R. Gill
- Heri C. Kaare
- Sharon Kalima
- Elson Kambalu
- John Lwanda
- Herbert F. Makoye
- Otiyela Mtema
- Mia Perry
- Zoë Strachan
- Helen Todd
- Sally M. Mtenga
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10022006
doi: 10.1371/journal.pgph.0000927
license: CC BY 4.0
---
# “I see salt everywhere”: A qualitative examination of the utility of arts-based participatory workshops to study noncommunicable diseases in Tanzania and Malawi
## Abstract
The burden of noncommunicable diseases (NCDs) including hypertension, diabetes, and cancer, is rising in Sub-Saharan African countries like Tanzania and Malawi. This increase reflects complex interactions between diverse social, environmental, biological, and political factors. To intervene successfully, new approaches are therefore needed to understand how local knowledges and attitudes towards common NCDs influence health behaviours. This study compares the utility of using a novel arts-based participatory method and more traditional focus groups to generate new understandings of local knowledges, attitudes, and behaviours towards NCDs and their risk factors. Single-gender arts-based participatory workshops and focus group discussions were conducted with local communities in Tanzania and Malawi. Thematic analysis compared workshop and focus group transcripts for depth of content and researcher-participant hierarchies. In addition, semiotic analysis examined the contribution of photographs of workshop activities to understanding participants’ experiences and beliefs about NCD risk factors. The arts-based participatory workshops produced in-depth, vivid, emotive narratives of participants’ beliefs about NCDs and their impact (e.g., “… it spreads all over your body and kills you—snake’s poison is similar to diabetes poison”), while the focus groups provided more basic accounts (e.g., “diabetes is a fast killer”). The workshops also empowered participants to navigate activities with autonomy, revealing their almost overwhelmingly negative beliefs about NCDs. However, enabling participants to direct the focus of workshop activities led to challenges, including the perpetuation of stigma (e.g., comparing smells associated with diabetes symptoms with sewage). Semiotic analysis of workshop photographs provided little additional insight beyond that gained from the transcripts. Arts-based participatory workshops are promising as a novel method to inform development of culturally relevant approaches to NCD prevention in Tanzania and Malawi. Future research should incorporate more structured opportunities for participant reflection during the workshops to minimise harm from any emerging stigma.
## Introduction
Communicable diseases such as malaria and HIV/AIDS have traditionally been the primary focus of public health initiatives in Sub-Saharan Africa [1]. However, the additional burden of noncommunicable diseases (NCDs) is now of growing concern [2, 3]. Between 1990 and 2017, NCDs accounted for a $67\%$ increase in disability-adjusted life-years in Sub-Saharan Africa [4] and current estimates predict that NCD mortality will outstrip that from communicable, maternal and perinatal diseases by 2030 [5].
Tanzania and Malawi, like other Sub-Saharan Africa countries, are experiencing major challenges in the face of this additional burden of disease [6, 7]. In 2019, NCDs were estimated to account for $34\%$ of deaths in Tanzania and $40\%$ in Malawi [8]. Recent studies suggest that prevalence of hypertension is high ($28\%$) in parts of Tanzania and uncontrolled among $95\%$ of patients [9], and up to $6.8\%$ of adults have diabetes [10]. Similarly, both hypertension ($33\%$) and diabetes ($5.6\%$) are highly prevalent and underdiagnosed in Malawi [7, 11].
The drivers of many NCDs are complex and caused by interactions between social, environmental, biological, and political factors [2]. Evidence indicates there may be a predisposition to poor metabolic health associated with historic and persistent undernutrition and communicable disease in low-middle income countries like Tanzania and Malawi [2, 3]. In addition, there is an increasingly obesogenic environment due to the commercial influx of calorific, low nutrient foods, alcohol, and tobacco, while structural barriers associated with rising urbanisation in the region may further restrict access to healthy food and spaces to exercise [2]. Finally, different socio-cultural norms, such as a large body size being associated with wealth and health [3] and traditional food preferences [12, 13] mean that the lifestyle interventions used to address NCDs in high-income countries may not be translatable to the Sub-Saharan African context [1]. Culturally relevant approaches are needed to understand how local beliefs and culture influence health behaviours in Tanzania and Malawi to inform effective NCD prevention initiatives.
Historically, health research in Sub-Saharan Africa has prioritised a Western scientific approach to understand the lived realities of local communities [14, 15]. This approach has generated mistrust of Western researchers among local people [14, 16]. New methodologies should therefore seek to foreground the cultural significance of local people’s understandings of health and, where possible, repair any damaged trust between communities and researchers [17, 18].
Community-based participatory research (CBPR) has emerged to promote equitable collaborations between health researchers and study participants [18]. Genuine partnerships between researchers and community members, where each partner benefits from mutual knowledge generation, are key to CBPR, which has been used successfully with marginalised groups in high-income contexts [18]. The use of CBPR in Sub-Saharan Africa can be traced back to earlier participatory approaches in the 1970s, including action research and Theatre for Development [19–25]. Although CBPR approaches can bring new learning opportunities and rich perspectives for those involved, there are still many practical challenges with ensuring that participation is truly equitable [18, 23, 26].
Arts-based methods are also becoming increasingly used within health research. They incorporate a range of artistic practices including performance, visual arts and written narratives, and can be used at any stage of the research process [27] to enable the communication of rich, affective narratives that are difficult to express through words alone [28]. Arts-based methods have been widely used in communicable disease research in Sub-Saharan Africa, particularly HIV/AIDS, but remain largely underutilised in NCD research [17]. However, some researchers have criticised the use of arts-based research in Sub-Saharan Africa for unintentionally reinforcing researcher-participant hierarchies. Examples include failing to actively engage performers in creating the narratives [29] and communicating health messages that prioritise western biomedical knowledge over the lived experiences of local communities [17].
Integrating arts-based methods and CBPR provides a promising way of overcoming these challenges. Some researchers have demonstrated the utility of participant-centred arts-based approaches in narrowing the distance between the researcher and the researched [30] and liberating ‘performers’ to express themselves about the issues they find most meaningful and important [27, 28, 31]. Building on this evidence, the current study sought to examine the utility of arts-based participatory workshops in exploring local knowledges, attitudes, and behaviours towards common NCDs and their risk factors in Tanzania and Malawi. The aim was to compare the workshops to traditional focus group discussions in relation to the depth of content generated and researcher-participant hierarchies. A secondary aim was to examine the extent to which visual data (photographs) from the workshops contributed to understanding local experiences and beliefs around NCD risk factors.
## Ethics statement
Ethical approvals were obtained from the Ethical Committee at the National Institute for Medical Research (NIMR) in Tanzania (NIMR/HQ/R.8a/Vol. IX/2959), the National Committee on Research in the Social Sciences and Humanities in Malawi (Ref. NCST/RTT/$\frac{2}{6}$), and the University of Glasgow College of Social Science Ethics Committee (Ref. 400170232). Participants were provided with information sheets containing full details of the project in their local language. The facilitator explained that participation was fully voluntary and summarised the information sheet before prompting participants to ask questions if they required further clarification or information. All participants provided written informed consent. The individuals pictured in Fig 2 have also provided written informed consent (as outlined in the media consent form) to publish their image alongside the manuscript.
## Study design
The current study is part of a larger MRC/AHRC-funded project exploring interdisciplinary approaches to NCD prevention in Tanzania and Malawi (Culture and Bodies, MC_PC_MR_R024448, 21). Culture and Bodies was a collaboration between researchers from Ifakara Health Institute in Tanzania, the Malawi Epidemiology and Intervention Research Unit (MEIRU), the University of Malawi, the University of Glasgow; and local participatory arts practitioners from Bagamoyo Arts and Cultural Institute (*Taasisi ya* *Sanaa na* Utamaduni Bagamoyo, TaSUBa) in Tanzania and Art and Global Health Centre Africa (ArtGlo) and Art House Africa in Malawi.
The current study involved three main components. First, community engagement activities were undertaken to allow the researchers to develop relationships with the local communities and generate initial insight into local NCD knowledge and community arts. Then focus group discussions and arts-based participatory workshops were used to explore local attitudes to NCDs and risk factors. This paper reports the qualitative findings from the focus group discussions and arts-based participatory workshops.
## Setting and participants
The study was conducted in Bagamoyo, Tanzania and Area 25, Lilongwe, Malawi. Bagamoyo is a coastal town in Tanzania with a population of 89,000 in 2008 [32]. Area 25 is a high-density area of the Malawian capital, Lilongwe, with a population of 64,650 in 2008 [33]. Both communities have a history of involvement in previous health research, and therefore may be more familiar with health issues than other similar communities.
Four focus group discussions ($$n = 7$$–10) and four arts-based participatory workshops ($$n = 7$$–10) were conducted in single-gender groups to minimise the impact of cultural gender norms on participants’ contributions. Local residents were recruited using a convenience sampling approach with the help of community leaders. All were aged over 18 years and had lived in the community for at least 12 months. Participants were initially invited to take part in the focus group discussions and then invited to also attend the arts-based participatory workshops. Additional participants were recruited as required to make up numbers for the workshops. Some participants already knew each other; however, most met for the first time during the research.
## Data collection and analysis
The arts-based participatory workshops and focus group discussions aimed to explore people’s knowledges, attitudes, and behaviours towards NCDs and their risk factors including symptoms, and prevention. In Tanzania, the workshops (November/December 2018) were facilitated by SMM with support from AD, CMG and HCK, and the focus groups (February/March 2019) were conducted by SMM and AD. In Malawi, the workshops (February 2019) were co-facilitated by SK and OM, and the focus groups (February 2019) were co-conducted by OM and HN. A topic guide was used to steer the focus group discussions, which were designed to last around 1.5–2 hours, around the issues of interest, including a range of common NCDs.
In the workshops, participants were asked to select one NCD to focus on. In Tanzania, both groups chose to explore diabetes; in Malawi, the men selected hypertension and the women chose cancer. Half a day was allocated to the workshops, as the enjoyable nature of the creative activities lessened the burden on participants. As shown in Fig 1, each workshop consisted of three activities (Verbalised senses, Embodied images, and Performance), each followed by reflective discussion. The activities were designed to build on each other, with the first two providing a scaffold to support the development of the performance activity. Embodied images and performances were created in small groups and interpreted by other participants. For the final performance, participants could use an arts-based approach of their choice. Theatre was selected in three of four workshops, while men in Tanzania chose poetry.
**Fig 1:** *Arts-based participatory workshop activities.Description of the arts-based participatory workshop verbalised senses, embodied images, and performance activities.*
The focus groups and workshops were audio-recorded, transcribed, and translated from the local languages (Kiswahili and Chichewa) to English by local fieldworkers. Fieldnotes were also taken during each method. Photographs of the embodied images were taken in the Tanzanian women’s workshops and both Malawi workshops.
The transcripts were analysed using a thematic framework approach [34] and NVivo12 software. One focus group and one workshop transcript were independently coded by MB, CMG and SMM, who then met to agree nine broad themes for analysis, six of which are reported here: Beliefs about NCDs and risk factors; Impact of NCDs; Experiences of NCDs and risk factors; Causes of NCDs and risk factors; Barriers to NCD prevention; and NCD symptoms. A full description of all nine themes is provided in S1 Table. The broad themes were then applied to all transcripts by MB and checked by CMG. MB generated relevant subthemes and applied them to Excel matrices. Data were then sorted according to method and each sub-theme summarised with focus on the similarities and differences between the arts-based participatory workshops and focus group discussions. The comparisons were aligned with the main aims of the research; specifically, depth of content and researcher-participant hierarchy.
Semiotic analysis was used to analyse eleven photographs of embodied images from the workshops. Semiotics conceptualise images as a series of signs (e.g., body language) to understand how meaning is created [35, 36]. MB applied semiotic analysis to explore the characterisation and the emotions portrayed within the images as follows: first, literal descriptions of what the characters were doing in the images were written without assigning any further meaning (e.g., the character holds a bottle to their lips vs. the character is drinking alcohol). The images were then broken down further into visual sign categories (i.e., physical appearances of actors and body language/facial expressions) summarised in an Excel matrix. As cultural knowledge is key to assigning higher levels of meaning, the matrix also referenced the relevant participant discussions within transcripts [35].
The first author (MB) is a Western researcher who was not present at data collection. Analysis was, therefore, approached critically with care taken to recognise and correct any biases and misinterpretations. It was closely supported by Culture and Bodies investigators, including SMM, a Tanzanian national.
## Results
The analysis compared arts-based participatory workshops to traditional focus group discussions in relation to i) depth of content generated; and ii) researcher-participant hierarchies during each method. Relevant findings to i) and ii) are therefore summarised and presented in turn below.
## i) Depth of content
The workshops stimulated participants to use vivid and emotive imagery, which often generated more insight into their own deeply held beliefs about the chosen NCDs in comparison to the focus groups. This was especially apparent in the verbalised senses activities: for example, participants equated the chosen NCD to a “bomb”, “bullet” and “plane crash”. This language suggests sudden, catastrophic impacts of the diseases as depicted in the following workshop extract from Tanzania where men used powerful similes to portray amputation as a complication of diabetes: By comparing diabetes to a bomb, the participant conveys a sense of destruction. This analogy strongly evokes feelings of fear, worry and pain, and therefore carries more weight than simply saying that limb amputation is a negative impact of diabetes. Others communicated beliefs of the chosen NCDs being like predatory animals, including a “hyena”, “lion” and “snake”. Animals play important roles in African cultures, often described in symbolism, folklore, and spirituality [37–39]. Although the associated traits may vary between cultures, the hyena is usually depicted as an antagonistic symbol, described as sly and unintelligent [37, 39]. The lion is frequently considered a mark of power and strength [39], and, in some circumstances, danger [40]. When reflecting on these ideas, one Tanzanian man explained that their belief stemmed from the perceived impact of the disease: This extract evokes a sense of powerlessness to diabetes, in which fatal consequences are unavoidable. In Sub-Saharan Africa, communities face several barriers to controlling diabetes, including regular access to medications [13]. Therefore, the powerful association of diabetes with a snake’s poison could reveal both beliefs about the severity of the disease and the lived experience of barriers to controlling it. In contrast, the focus groups provided more basic ‘bare-bones’ accounts of the impacts of NCDs: In some cases, while the workshops provided rich analogies, participants could have been supported to reflect on emerging beliefs more fully. During the verbalised senses activity in Tanzania, one male participant remarked; “When I see diabetes, I see salt everywhere”. This statement evokes a vivid mental image that could be interpreted as salt being perceived as causing diabetes. Salt is a well-documented risk factor for hypertension, but not for diabetes [41]. However, hypertension and diabetes present together in a high proportion of patients [42–44]. Therefore, the analogy could also signify familiarity with the healthy eating advice provided to patients with diabetes in Sub-Saharan Africa, which typically includes salt-reduction to prevent or control comorbid hypertension [45, 46]. However, on this occasion, the participant was not asked to elaborate, and his intended meaning remained unclear.
Where opportunity was provided for reflection, the workshop activities helped participants to consider alternative causes of NCD risk behaviours. They produced emotive and empathetic insight compared to the focus groups. For example, following a performance about alcohol leading to high blood pressure in Malawi, participants had different opinions on the “drunkard” character. Some argued that the character’s life choices made him a poor and unhelpful husband. However, when prompted to consider different scenarios about why somebody would drink, others imagined a backstory for the character and appeared more compassionate towards him: While the workshops provided more emotive insight into local experiences of NCDs, the focus groups provided important narratives on topics not discussed in the workshops. These included reflections on how circumstances outwith personal control could act as barriers to prevention. For example, when asked what type of cooking oil was used locally, one Tanzanian man complained about the prohibitive cost of healthier options: This extract demonstrates that knowledge about NCD risk factors alone is insufficient to change behaviour, and the use of a rhetorical question gives a sense of the frustration and lack of empowerment experienced by the community to do the ‘right thing’. Thus, whilst the workshop activities appeared to facilitate participants’ exploration of personal and wider social factors that contribute to NCD risk, the focus groups tended to focus more on macro level structural factors, such as unaffordability of healthy lifestyle practices.
## ii) Researcher-participant hierarchies
Focus group participants covered a broad range of topics. For example, when considering the causes of NCDs, some traditional widely held beliefs around their supernatural and fatalistic nature emerged. However, as the following extract indicates, these beliefs were often elicited only after direct prompts from the researcher: In contrast, the workshop’s verbalised senses activity was particularly valuable when allowing participants to describe their chosen NCD with autonomy. Most participants used this activity to highlight negative beliefs, indicating that they themselves recognise the increasing challenge of NCDs in their communities. For example, Malawian men associated hypertension with things that were dangerous (“axe”, “fire”) and startling (“thunder”, “fireworks”), leading one participant to express his surprise at the extent of the negativity that had emerged from the group: Encouraging autonomy during the workshops also brought unexpected challenges, particularly in relation to the focus of the workshop activities. For example, the Malawian women selected cancer as their NCD of choice, but then explored a range of cancers (skin, cervical and prostate) during the embodied images and performance activities, this restricted the opportunity for the research team to gain depth of insight into local knowledges, attitudes, and behaviours in relation to a single type of cancer. Another unintended consequence of allowing participants to direct the focus of the workshop activities was that their imaginative reflections appeared to perpetuate stigma to a greater degree than the focus group discussions. In Tanzania, diabetic symptoms, such as frequent urination and slow-healing wounds were raised in both men’s and women’s focus groups. For example, in the women’s focus group in Tanzania it was simply stated that wounds “smell a lot if not treated”. However, in the workshops, the verbalised senses activity stimulated participants to provide a complex narrative around the negative social norms associated with people with diabetes: The comparison to a sewage tank is a rich analogy that fully portrays local perceptions of the unpleasant smells associated with people with diabetes. They do not merely smell ‘bad’ but appear to be intolerable to the point of disrupting social interactions. Similarly, the Tanzanian men’s reflections on the verbalised senses activity drew parallels between diabetes and sewage (“a pit with a strong smell”) which was later reinforced within their final performance (poem):
Therefore, although the workshop activities provided the researchers with a deeper understanding of local experiences of diabetes, they could also perpetuate stigma by providing validation of participants’ negative views.
## iii) Using visual data to explore local understandings of NCD risk factors
A secondary aim of the study was to examine the extent to which photographs of the workshop embodied images activity could contribute to developing researchers’ understanding of local experiences and beliefs around NCD risk factors. The semiotic analysis found that characterisation (physical appearances) had some utility. For example, as Fig 2 shows, the Tanzania women’s workshop chose a participant with a larger body size to portray a physically inactive individual with diabetes, thus also suggesting the recognition of overweight/obesity as a risk factor for diabetes locally. However, the analysis of emotion (through body language and facial expression) was often inconclusive or did not contribute any further understanding beyond what was available from workshop transcripts. Indeed, interpretation of the visual data was often only possible with the addition of important cultural insights from the transcripts.
**Fig 2:** *Embodied image from the women’s arts-based participatory workshop in Tanzania.One woman supporting the arms of another woman to depict a lack of exercise as a cause of diabetes. The individuals pictured have provided written informed consent (as outlined in the media consent form) to publish their image alongside the manuscript.*
## Discussion
In comparison to focus groups, arts-based participatory workshops generated more in-depth insight into local knowledges, attitudes, and behaviours around NCDs and their risk factors. In relation to depth of content, the workshop activities evoked strong emotive portrayals of participants’ fear, worry and powerlessness in relation to NCDs. Reflective discussions encouraged workshop participants to further explore and examine their own and other group members’ perspectives of NCDs and NCD risk factors. However, on occasion, reflective discussions needed further development to clarify ambiguity around the intended meanings of participants’ symbolic representations. Focus group discussions, on the other hand, provided insight into more tangible concepts and the structural causes of NCDs. In addition, workshops clearly reduced researcher-participant hierarchies and supported participants’ autonomy around their chosen NCD, particularly in relation to their negative beliefs. However, this autonomy may have led to greater stigmatisation of the NCD than was apparent in the focus group discussions by providing a platform for participants to gain validation of their negative views. Finally, although photographs of workshop activities provided some additional understanding of local experiences and beliefs of NCD risk factors, the contribution of these visual data was enhanced by the cultural insights present in the workshop transcripts.
It has been argued that arts-based methods enable participants to delve into the imaginary, go beyond literal depictions of illness and provide new in-depth perspectives compared to traditional forms of research [28, 47, 48]. This was particularly true in the current study in relation to the emergence of fears and negative beliefs about NCDs, and the description of symptoms (e.g., diabetic wounds) during the verbalised senses activities. Our findings align with previous arts-based research: for example, Horne [49] noted that metaphors generated through body mapping and storytelling enabled HIV-positive women in South Africa to communicate deep pain about the lived realities of their condition, as well as confront their personal experience of the disease. Similarly, Abdulla [40] emphasised how ‘play’ and ‘make believe’ during community arts workshops in rural Malawi contributed to creating safe spaces between reality and the imaginary where participants could speak openly about their perspectives and experiences of HIV/AIDs without fear of real-life consequences.
It has been argued that where community members are directly involved in the creation of arts-based activities, they may be better able to explore and navigate the new knowledge produced [24]. In the current study, using verbalised senses and embodied images as a scaffold for participants to create their performances appeared to empower them to actively engage in the workshop in a way that was consistent with their personal interests and concerns. In addition, opportunities for reflection throughout the workshops provided participants with space to consider the new knowledges emerging from the activities (e.g., the overwhelmingly negative response to hypertension in the Malawian men’s workshop). Our approach echoes a previous study where a creative, reflective participatory workshop conducted as part of an exploration of community resilience in the Netherlands encouraged participants to understand and share each other’s feelings and discover common concerns. These findings led the authors to conclude that the creative empathy which emerged during the workshop supported in-depth communication of participants’ attitudes and feelings in a way that moved beyond “cognitive ways of knowing” (pg.6; [28]).
Some researchers have suggested that narratives stimulated by arts-based methods can further marginalise vulnerable individuals [27, 50]. During the workshops, the combination of the arts-based activities providing participants with a ‘safe space’ to demonstrate negative beliefs about NCDs and the reflective discussion encouraging creative empathy may have unintentionally reinforced NCD-related stigma. However, arts-based methods have been shown to be helpful in challenging health-related stigma in Sub-Saharan Africa [17, 51]. For example, one photovoice study in South Africa encouraged secondary school students to use their own portrayals of HIV/AIDS-related stigmatisation to reflect on misunderstandings and negative attitudes towards people affected by HIV/AIDS [50]. Future workshops should include additional reflective activities to encourage participants to consider how their narratives could contribute to stigma and explore how they might support people who experience it.
Another challenge associated with participant autonomy during one of the workshops emerged in relation to cancer being the chosen NCD. This led to somewhat superficial considerations of different types of cancer rather than in-depth consideration of one. Howard [52] reported similar difficulties when using interactive theatre as a vehicle to generate information about eating habits and body image among US women. Spectators redirected the intended focus of the performance from broader social issues to individual situations and events that reflected their own personal experiences rather the collective viewpoint.
The current study was facilitated by close collaboration between researchers, arts-practitioners and community members from Tanzania, Malawi, and Scotland. Such cross-cultural interdisciplinarity means grappling with diverse epistemologies on what constitutes truth and different definitions of the topic being researched [53, 54]. Navigating what and how knowledge is shared also requires critical awareness of where participant contributions and researcher goals conflict [54]. Some of the challenges outlined above, regarding the workshop focus and need for different forms of reflection, could also reflect the complexity of navigating the interests of multiple disciplines and partners and expectations of what the collected data should look like. These examples serve as a reminder of the importance of transparency, reflexivity and dialogue between different partners to explore these challenges [54, 55].
Finally, the contribution of the photographs of the embodied images activity to our understandings of local experiences and beliefs about NCD risk factors was limited. When exploring emotions, it was sometimes difficult to ascertain to what extent the emotions conveyed in the photographs were intended as part of the image or were spontaneous facial expressions that emerged from taking part in the activity. This limitation could be overcome by using the principles of semiotic analysis (e.g., physical appearances of actors and body language/facial expressions) during future workshops to guide participants’ reflective interpretations of images and performances. This approach would be beneficial in further empowering participants by supporting their involvement in the interpretation of the data generated and thus the production of culturally relevant insight [27, 56].
## Strengths and limitations
A major strength of this study is the involvement of different teams of local researchers, arts practitioners, and community members in two Sub-Saharan Africa countries. This not only provides important insights into how arts-based participatory approaches could be applied in NCD and wider health research in Sub-Saharan Africa but also the strengths of co-creation in art-based participatory research as a new way to generate new rich, in-depth understandings.
However, there are some limitations. First, as arts-based research can encompass a diverse range of activities, the findings may not be generalisable of the findings beyond the activities employed in the current study. Second, as both communities have had long involvement in health research, participants may have had increased exposure to health information compared to the general populations of Malawi and Tanzania. Although people’s knowledges, attitudes and behaviours will undoubtably be influenced by prior experience of health research, there is no reason to believe that the arts-based participatory approach described here should not be translatable to other communities with different levels of pre-existing health knowledge. Third, there were some differences in the questions explored in the workshops and focus group discussions. However, it is unlikely that these differences affected the conclusions about the emergence of more in-depth content and the reduction of participant-researcher hierarchies within the workshops. Finally, the lead author (MB) was not present during the workshops or focus groups and thus may have missed important contextual cues, particularly during the arts-based participatory activities. However, the analysis was closely supported by the lead Tanzanian researcher (SMM) who was present at both Tanzanian workshops and informed by fieldnotes describing researcher experiences and perceptions of data collection. The lead Malawian researcher (OM) was also consulted closely to ensure adequate representation of local context in the findings presented.
## Conclusions
Compared to traditional focus group discussions, the arts-based participatory workshops revealed more emotive and personal narratives of these NCD experiences. They appeared to liberate participants from the trappings of cognitive ways of knowing and generate new in-depth understandings. The workshops also succeeded in empowering participants to voice their own perspectives, including fears and negative beliefs, and in supporting the development of a performance that reflected participants’ personal interests and concerns about NCDs. As such, the arts-based participatory workshops show promise as a new approach to inform the development of culturally relevant NCD prevention initiatives among at-risk populations in Sub-Saharan Africa and potentially elsewhere.
## References
1. Mendenhall E, Kohrt BA, Norris SA, Ndetei D, Prabhakaran D. **Non-communicable disease syndemics: poverty, depression, and diabetes among low-income populations**. *Lancet* (2017) **389** 951-63. DOI: 10.1016/S0140-6736(17)30402-6
2. Miranda JJ, Barrientos-Gutiérrez T, Corvalan C, Hyder AA, Lazo-Porras M, Oni T. **Understanding the rise of cardiometabolic diseases in low- and middle-income countries**. *Nature Medicine* (2019) **25** 1667-79. DOI: 10.1038/s41591-019-0644-7
3. Nyirenda MJ. **Non-communicable diseases in sub-Saharan Africa: understanding the drivers of the epidemic to inform intervention strategies.**. *International Health* (2016) **8** 157-8. DOI: 10.1093/inthealth/ihw021
4. Gouda HN, Charlson F, Sorsdahl K, Ahmadzada S, Ferrari AJ, Erskine H. **Burden of non-communicable diseases in sub-Saharan Africa, 1990–2017: results from the Global Burden of Disease Study 2017**. *The Lancet Global Health* (2019) **7** e1375-e87. DOI: 10.1016/S2214-109X(19)30374-2
5. Mathers CD, Loncar D. **Projections of Global Mortality and Burden of Disease from 2002 to 2030.**. *PLOS Medicine.* (2006) **3** e442. DOI: 10.1371/journal.pmed.0030442
6. Peck R, Mghamba J, Vanobberghen F, Kavishe B, Rugarabamu V, Smeeth L. **Preparedness of Tanzanian health facilities for outpatient primary care of hypertension and diabetes: a cross-sectional survey**. *Lancet Glob Health* (2014) **2** e285-92. DOI: 10.1016/S2214-109X(14)70033-6
7. Price AJ, Crampin AC, Amberbir A, Kayuni-Chihana N, Musicha C, Tafatatha T. **Prevalence of obesity, hypertension, and diabetes, and cascade of care in sub-Saharan Africa: a cross-sectional, population-based study in rural and urban Malawi.**. *Lancet Diabetes Endocrinol.* (2018) **6** 208-22. DOI: 10.1016/S2213-8587(17)30432-1
8. 8World Health Organization. Noncommunicable diseases progress monitor 2022. Geneva: World Health Organization, 2022. Available from: https://www.who.int/publications/i/item/9789240047761. *Noncommunicable diseases progress monitor 2022* (2022)
9. Galson SW, Staton CA, Karia F, Kilonzo K, Lunyera J, Patel UD. **Epidemiology of hypertension in Northern Tanzania: a community-based mixed-methods study**. *BMJ Open* (2017) **7** e018829. DOI: 10.1136/bmjopen-2017-018829
10. Hodel NC, Hamad A, Reither K, Mwangoka G, Kasella I, Praehauser C. **Assessment of diabetes and prediabetes prevalence and predictors by HbA1c in a population from sub-Saharan Africa with a high proportion of anemia: a prospective cross-sectional study.**. *BMJ Open Diabetes Research & Care.* (2020) **8** e000939. DOI: 10.1136/bmjdrc-2019-000939
11. Amberbir A, Lin SH, Berman J, Muula A, Jacoby D, Wroe E. **Systematic Review of Hypertension and Diabetes Burden, Risk Factors, and Interventions for Prevention and Control in Malawi: The NCD BRITE Consortium.**. *Glob Heart.* (2019) **14** 109-18. DOI: 10.1016/j.gheart.2019.05.001
12. Thakwalakwa C, Flax VL, Phuka JC, Garcia H, Jaacks LM. **Drivers of food consumption among overweight mother-child dyads in Malawi.**. *PLOS ONE.* (2020) **15** e0243721. DOI: 10.1371/journal.pone.0243721
13. Zimmermann M, Bunn C, Namadingo H, Gray CM, Lwanda J. **Experiences of type 2 diabetes in sub-Saharan Africa: a scoping review**. *Global Health Research and Policy* (2018) **3** 25. DOI: 10.1186/s41256-018-0082-y
14. Myers RA. **Review of Curing Their Ills: Colonial Power and African Illness, by M. Vaughan**. *African Studies Review.* (1993) **36** 147-9. DOI: 10.2307/524758
15. Smith LT. *Decolonizing methodologies: research and indigenous peoples* (2021)
16. Leininger M.. **Becoming Aware of Types of Health Practitioners and Cultural Imposition**. *Journal of Transcultural Nursing* (1991) **2** 32-9. DOI: 10.1177/104365969100200205
17. Bunn C, Kalinga C, Mtema O, Abdulla S, Dillip A, Lwanda J. **Arts-based approaches to promoting health in sub-Saharan Africa: a scoping review**. *BMJ Global Health* (2020) **5** e001987. DOI: 10.1136/bmjgh-2019-001987
18. Wallerstein NB, Duran B. **Using Community-Based Participatory Research to Address Health Disparities.**. *Health Promotion Practice.* (2006) **7** 312-23. DOI: 10.1177/1524839906289376
19. Walker M. **Context, Critique and Change: doing action research in South Africa.**. *Educational Action Research* (1995) **3** 9-27. DOI: 10.1080/0965079950030102
20. Visser M.. **Development of structured support groups for HIV-positive women in South Africa**. *SAHARA-J:Journal of Social Aspects of HIV/AIDS* (2005) **2** 333-43. DOI: 10.1080/17290376.2005.9724858
21. Kamanda A, Embleton L, Ayuku D, Atwoli L, Gisore P, Ayaya S. **Harnessing the power of the grassroots to conduct public health research in sub-Saharan Africa: a case study from western Kenya in the adaptation of community-based participatory research (CBPR) approaches.**. *BMC Public Health* (2013) **13** 91. DOI: 10.1186/1471-2458-13-91
22. Kamlongera C.. **Theatre for Development in Africa**. *In: CLACSO, editor. Media and Global Change: Rethinking Communication for Development. Buenos Aires* (2005)
23. Kidd R, Byram M. **Demystifying Pseudo-Freirian Development: The Case of Laedza Batanani.**. *Community Development Journal* (1982) **17** 91-105. DOI: 10.1093/cdj/17.2.91-a
24. Abdulla S.. **The Art of Inclusion: Contradictions Affecting Theatre for Development Interventions in Malawi.**. *Handbook on Promoting Social Justice in Education* (2020) 999-1020. DOI: 10.1007/978-3-030-14625-2_15
25. Chimberengwa PT, Naidoo M. **A description of community-based participatory research of hypertension awareness, prevention and treatment in a district of Matabeleland South Province, Zimbabwe.**. *Afr J Prim Health Care Fam Med.* (2019) **11** e1-e9. DOI: 10.4102/phcfm.v11i1.1839
26. Weber S, Hardiman M, Kanja W, Thomas S, Robinson-Edwards N, Bradbury-Jones C. **Towards Ethical International Research Partnerships in Gender-Based Violence Research: Insights From Research Partners in Kenya.**. *Violence Against Women.* (2021) 10778012211035798. DOI: 10.1177/10778012211035798
27. Coemans S, Hannes K. **Researchers under the spell of the arts: Two decades of using arts-based methods in community-based inquiry with vulnerable populations.**. *Educational Research Review* (2017) **22** 34-49. DOI: 10.1016/j.edurev.2017.08.003
28. Van der Vaart G, van Hoven B, Huigen PP. **Creative and arts-based research methods in academic research. Lessons from a participatory research project in the Netherlands**. *Forum Qualitative Sozialforschung/Forum: Qualitative Social Research* (2018) **19**. DOI: 10.17169/fqs-19.2.2961
29. Chinyowa KC. **Participation as ‘repressive myth’: a case study of the Interactive Themba Theatre Organisation in South Africa**. *Research in Drama Education: The Journal of Applied Theatre and Performance* (2015) **20** 12-23. DOI: 10.1080/13569783.2014.975109
30. Daniels D.. **Learning about community leadership: Fusing methodology and pedagogy to learn about the lives of settlement women**. *Adult Education Quarterly* (2003) **53** 189-206. DOI: 10.1177/0741713603053003004
31. Francis DA. **‘Sex is not something we talk about, it’s something we do’: using drama to engage youth in sexuality, relationship and HIV education.**. *Critical Arts: A Journal of South-North Cultural Studies.* (2010) **24** 228-44. DOI: 10.1080/02560041003786508
32. Mwangoka GW, Burgess B, Aebi T, Sasi P, Abdulla S. **The Ifakara Health Institute’s Bagamoyo Research and Training Centre: a well-established clinical trials site in Tanzania.**. *International Health.* (2009) **1** 85-90. DOI: 10.1016/j.inhe.2009.06.009
33. 33Malawi Epidemiology and Intervention Research Unit. Lilongwe: Lilongwe City and Area 25: Malawi Epidemiology and Intervention Research Unit; 2022. Available from: https://www.meiru.info/lilongwe/.. *Lilongwe: Lilongwe City and Area 25: Malawi Epidemiology and Intervention Research Unit* (2022)
34. Spencer L, Ritchie J, O’Connor W, Morrell G, Ormston J, Metzler K. *Qualitative Research Practice: A Guide for Social Science Students and Researchers* (2014) 297-309
35. Penn G., Bauer M. W., Gaskell G. *Qualitative Researching with Text, Image and Sound* (2000)
36. Röttger K.. **The Mystery of the In-Between. A Methodological Approach to Intermedial Performance Analysis**. *Forum Modernes Theater* (2013) **28** 105-116. DOI: 10.1353/fmt.2013.0014
37. Banda D, Morgan WJ. **Folklore as an instrument of education among the Chewa people of Zambia**. *International review of education* (2013) **59** 197-216. DOI: 10.1007/s11159-013-9353-5
38. Curran D.. **Nyau Masks and Ritual**. *African Arts.* (1999) **32** 68-77. DOI: 10.2307/3337711
39. Olupona JK. **Some Notes on Animal Symbolism in African Religion and Culture.**. *Anthropology and Humanism.* (1993) **18** 3-12. DOI: 10.1525/ahu.1993.18.1.3
40. Culture Abdulla S.. *play and health: A folk media approach to HIV and AIDS research in rural Malawi* (2021)
41. Grillo A, Salvi L, Coruzzi P, Salvi P, Parati G. **Sodium Intake and Hypertension.**. *Nutrients* (2019) **11**. DOI: 10.3390/nu11091970
42. Munyogwa MJ, William R, Kibusi SM, Gibore NS. **Clinical characteristics and health care received among patients with type 2 diabetes attending secondary and tertiary healthcare facilities in Mwanza Region, Tanzania: a cross-sectional study**. *BMC Health Services Research* (2020) **20** 527. DOI: 10.1186/s12913-020-05407-y
43. Stanifer JW, Cleland CR, Makuka GJ, Egger JR, Maro V, Maro H. **Prevalence, Risk Factors, and Complications of Diabetes in the Kilimanjaro Region: A Population-Based Study from Tanzania**. *PLOS ONE* (2016) **11** e0164428. DOI: 10.1371/journal.pone.0164428
44. Mohamed SF, Uthman OA, Mutua MK, Asiki G, Abba MS, Gill P. **Prevalence of uncontrolled hypertension in people with comorbidities in sub-Saharan Africa: a systematic review and meta-analysis**. *BMJ Open* (2021) **11** e045880. DOI: 10.1136/bmjopen-2020-045880
45. Stephani V, Opoku D, Beran D. **Self-management of diabetes in Sub-Saharan Africa: a systematic review**. *BMC Public Health* (2018) **18** 1148. DOI: 10.1186/s12889-018-6050-0
46. Birkinshaw A, Nel R, Walsh C. **Adherence of patients with type 2 diabetes mellitus with the SEMDSA lifestyle guidelines**. *Journal of Endocrinology, Metabolism and Diabetes of South Africa.* (2018) **23** 39-45. DOI: 10.1080/16089677.2018.1433110
47. Woodgate RL, Zurba M, Tennent P. **Worth a Thousand Words? Advantages, Challenges and Opportunities in Working with Photovoice as a Qualitative Research Method with Youth and their Families.**. *Forum Qualitative Sozialforschung / Forum: Qualitative Social Research.* (2016) **18**. DOI: 10.17169/fqs-18.1.2659
48. Dunn VJ, Mellor T. **Creative, participatory projects with young people: Reflections over five years**. *Research for All* (2017) **1** 284-299. DOI: 10.18546/RFA.01.2.05
49. Horne F.. **Conquering AIDS through narrative: LONGLIFE positive HIV stories**. *English Studies in Africa* (2011) **54** 71-87. DOI: 10.1080/00138398.2011.626186
50. Moletsane R, de Lange N, Mitchell C, Stuart J, Buthelezi T, Taylor M. **Photo-voice as a tool for analysis and activism in response to HIV and AIDS stigmatisation in a rural KwaZulu-Natal school**. *Journal of child and adolescent mental health* (2007) **19** 19-28. DOI: 10.2989/17280580709486632
51. Boneh G, Jaganath D. **Performance as a Component of HIV/AIDS Education: Process and Collaboration for Empowerment and Discussion.**. *American Journal of Public Health* (2011) **101** 455-64. DOI: 10.2105/AJPH.2009.171991
52. Howard LA. **Speaking theatre/doing pedagogy: re‐visiting theatre of the oppressed.**. *Communication Education* (2004) **53** 217-33. DOI: 10.1080/0363452042000265161
53. Barnett T, Pfeiffer DU, Hoque MA, Giasuddin M, Flora MS, Biswas PK. **Practising co-production and interdisciplinarity: Challenges and implications for one health research.**. *Preventive veterinary medicine.* (2020) **177** 104949. DOI: 10.1016/j.prevetmed.2020.104949
54. Janes JE. **Democratic encounters? Epistemic privilege, power, and community-based participatory action research**. *Action Research* (2016) **14** 72-87. DOI: 10.1177/1476750315579129
55. Boydell KM, Hodgins M, Gladstone BM, Stasiulis E, Belliveau G, Cheu H. **Arts-based health research and academic legitimacy: transcending hegemonic conventions.**. *Qualitative Research* (2016) **16** 681-700. DOI: 10.1177/1468794116630040
56. Lee J-A, Finney SD. **Using popular theatre for engaging racialized minority girls in exploring questions of identity and belonging.**. *Child & youth services.* (2005) **26** 95-118. DOI: 10.1300/J024v26n02_06
|
---
title: 'Health literacy strengths and challenges among residents of a resource-poor
village in rural India: Epidemiological and cluster analyses'
authors:
- Reetu Passi
- Manmeet Kaur
- P. V. M. Lakshmi
- Christina Cheng
- Melanie Hawkins
- Richard H. Osborne
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10022012
doi: 10.1371/journal.pgph.0001595
license: CC BY 4.0
---
# Health literacy strengths and challenges among residents of a resource-poor village in rural India: Epidemiological and cluster analyses
## Abstract
Cluster analysis can complement and extend the information learned through epidemiological analysis. The aim of this study was to determine the relative merits of these two data analysis methods for describing the multidimensional health literacy strengths and challenges in a resource poor rural community in northern India. A cross-sectional survey ($$n = 510$$) using the Health Literacy Questionnaire (HLQ) was undertaken. Descriptive epidemiology included mean scores and effect sizes among sociodemographic characteristics. Cluster analysis was based on the nine HLQ scales to determine different health literacy profiles within the population. Participants reported highest mean scores for Scale 4. Social support for health (2.88) and Scale 6. Ability to actively engage with healthcare professionals (3.66). Lower scores were reported for Scale 3. Actively managing my health (1.81) and Scale 8. Ability to find good health information (2.65). Younger people (<35 years) had much higher scores than older people (ES >1.0) for social support. Eight clusters were identified. In Cluster A, educated younger men (mean age 27 years) reported higher scores on all scales except one (Scale 1. Feeling understood and supported by a healthcare professional) and were the cluster with the highest number ($43\%$) of new hypertension diagnoses. In contrast, Cluster H also had young participants (mean age 30 years) but with low education ($72\%$ illiterate) who scored lowest across all nine scales. While epidemiological analysis provided overall health literacy scores and associations between health literacy and other characteristics, cluster analysis provided nuanced health literacy profiles with the potential to inform development of solutions tailored to the needs of specific population subgroups.
## Introduction
Common to health literacy research is descriptive epidemiological analysis where average population scores are used to estimate associations between health literacy and a range of sociodemographic variables, such as education, age, gender, and cultural background [1]. However, average population scores do not reveal the diversity of individuals or different groups within a population [2]. When the focus of data interpretation is on average population scores, a one-size-fits-all approach will be applied to public health, education, and communication initiatives, which leads to limited or no attention given to the needs of groups of people outside the norm. One-size-fits-all initiatives tend not to cater for diversity and may increase the health equity divide [3]. Complementary to epidemiological analyses are methods that can reveal patterns in health literacy differences and the sociodemographic characteristics of groups of people within populations. Identification of these patterns enables public health and health service initiatives to be developed in response to the specific needs of these groups, especially when responding to the various and often complex needs of people who experience disadvantage and marginalisation. One method to identify patterns of characteristics within groups is hierarchical cluster analysis [4,5], and this method has been used previously applied to data collected using the Health Literacy Questionnaire which is informed by a strong and comprehensive a priori model of health literacy [6–10].
*In* general, epidemiological analysis provides information about prevalence and also associations between target variables and independent sociodemographic variables [11–17]. This approach is problematic in health literacy research because health literacy is a multidimensional concept with several independent components where one or more may play a dominant role in outcomes in some contexts but not others [18]. An analytic method such as cluster analysis can complement epidemiological information by analysing health literacy patterns across a range of sociodemographic variables and investigating interactions between these variables [2].
Data from instruments that capture the multidimensional elements of health literacy are now being used for making decisions about public health and policy [19–21]. Importantly, the multiple elements of health literacy have the potential to provide practical guidance for actions to improve public health programs and policies around the world [20]. A globally relevant perspective of health literacy is one that seeks to recognize the diverse ways in which knowledge is produced, transferred, exchanged and used in different countries, cultures and settings around the world, and especially how health knowledge accumulates in families, communities and societies through daily, often communal, activities and social interactions [18]. This global perspective of health literacy recognizes that one-size-fits-all strategies and initiatives to develop health literacy will not be effective in all settings, nor for all people within a setting. Measurement of health literacy in global settings must not only allow for but also look for differences across the multiple dimensions of health literacy [18]. To do otherwise would be to potentially impose dominant culture notions of health literacy where these are not relevant and increase health inequities [3]. The majority of health literacy research has been undertaken in countries such as the USA and Australia [22].
India is a highly diverse country with a rich range of languages, cultures, and geography, and with wide socioeconomic gradients. To date, only a few studies describing health literacy in India have been published (for example, [23–26]) despite wide recognition that it is a useful concept for public health development [19,27]. The study reported in this paper is the first stage of a larger study in a rural area of India. The larger study investigates the use of the community-based participatory Ophelia (Optimising Health Literacy and Access) process [5,6], which is being used to co-design and implement public health initiatives to support medication adherence among people with hypertension.
The focus of this paper is to comprehensively describe the health literacy of the study population using a widely used health literacy questionnaire that was developed in a Western context but has been increasingly tested and applied in non-Western settings. The specific aim of this study is to investigate and compare epidemiological analysis and cluster analysis and explore the utility of each method for describing health literacy strengths and challenges. The study outcomes will inform the next stages of the larger study and are expected to lead to the development of meaningful and context appropriate public health initiatives and policies.
## Study design and setting
A cross-sectional survey was conducted in Chandigarh in the north of India, which is a union territory and joint capital for two neighbouring states. The region comprises 22 villages [28]. One village, Faidan Nizampur, in the south of the region, was selected for this study. This rural village was selected purposively because it has a diverse population and is one of the least developed villages in terms of infrastructure. Most of the houses are small and situated in narrow streets with no sealed roads. It has poor sanitation and water supply. Illiteracy in urban areas of *Chandigarh is* high ($14\%$) and higher in the rural areas ($19\%$) [29]. Faidan Nizampur has 1,072 households and a population of approximately 9,728 people, of which 3,475 are adults.
## Participants
A sample size of 500 was estimated for the main study that will seek to develop community-based interventions to improve medication adherence among people with hypertension. It was expected that this sample size will allow the detection of small differences in health literacy scores between demographic groups, assuming group mean differences between demographic factors of 0.3 and a standard deviation of 0.5 across HLQ scales, with alpha = 0.05 and beta = 0.80. There are no established standards for a minimum sample size for cluster analysis, but previous studies have demonstrated stable cluster analysis with fewer than 500 HLQ respondents [9,30].
There was no existing map of Faidan Nizampur village with marked locations of its households, and this study could not proceed with the systematic random selection of households without such a map. With the help of a geographic information systems (GIS) expert and a field worker, an estimated map of the village with its boundaries, households, and prominent land masses was prepared. First, the field worker surveyed the village and marked village boundaries, including specific places such as temples, clinics, schools, roads, drains, and houses. These data were then provided to the GIS expert who used the information to make the map of the village. Of the 1,072 households identified, 255 households were selected using computer-generated random numbers. It was assumed that two adults would be residing in each household. Adults from each household who were 18 years or over, and who intended to reside in the village for a minimum of one year, were invited to take part in the study. If, after two visits, there was no answer, or the occupants of the selected household chose not to participate, the researchers approached the occupants of the household to the right of the household initially selected.
## Health literacy measurement
The Health Literacy Questionnaire (HLQ) was developed through a grounded validity-driven approach [31], where the constructs were derived directly from the lived experiences of patients and frontline health workers [32]. The HLQ generates information about nine separate health literacy dimensions to provide a profile of the strengths, challenges, and preferences of populations. HLQ data can be used to inform public health planning and evaluation for diverse health literacy needs [5–7]. Validity testing in a wide range of cultural and linguistic contexts provides consistent evidence of good to excellent psychometric properties. The HLQ has been used and tested in Africa [33], Asia [34,35], South America [36], Europe [37–41]; and the Middle East [42].
A key attribute of the measurement theory of the HLQ is the recognition that health literacy is a complex multidimensional concept where different individuals can have different sets of health literacy strengths and challenges (i.e., different health literacy profiles). Health literacy profiles may be similar for people with similar backgrounds, who live in similar contexts, and who have similar lived experiences [43]. Given that different groups of people can have different patterns of health literacy strengths and challenges across the multiple dimensions, a single total score is a poor representation of health literacy diversity. It is for this reason that data from the HLQ are presented as nine separate scale scores to preserve the multiple dimensions and the diverse patterns of health literacy strengths and challenges within a population.
A licence from Swinburne University of Technology in Australia was obtained to use the Hindi version of the HLQ in this study. The HLQ consists of 44 questions within nine conceptually and psychometrically distinct scales: HLQ Scales 1 to 5 have response options of ’strongly disagree,’ ’disagree,’ ’agree’ and ’strongly agree’ and a score range of 1 to 4. Response options for Scales 6 to 9 are ’cannot do or always difficult,’ ’very difficult,’ ’quite difficult,’ quite easy’ and ’very easy’ and the score range is from 1 to 5. Scale scores are calculated as per the scoring instructions [32] by averaging the scores of each item within scales. All items have equal weighting.
## Data collection
Community health workers invited eligible people from the randomly selected households to participate. This included providing written and verbal information about the research, obtaining informed consent, and informing them they may stop or withdraw from the study at any time. All adults in a household who consented to participate were administered the questionnaire orally in Hindi. The interview also collected demographic data (see Table 1), including blood pressure (BP, twice at a gap of a minimum of five minutes) using a digital sphygmomanometer. A range of other clinical and anthropometric data were also collected and will be reported in future publications.
**Table 1**
| Socio- demographic variables | Number (%)510 (100) |
| --- | --- |
| Sex | |
| Women | 302 (59.2) |
| Men | 208 (40.8) |
| Age | |
| 18-35 years (young adults) | 317 (62.1) |
| 36-55 (middle-aged adults) | 160 (31.4) |
| 56-65 (older adults) | 25 (4.9) |
| Above 65 years (elderly) | 8 (1.6) |
| Religion | |
| Hindu | 435 (85.3) |
| Sikh | 37 (7.2) |
| Muslim | 33 (6.5) |
| Christian | 5 (1) |
| Caste | |
| General | 202 (39.6) |
| Other backward class | 61 (12) |
| Scheduled caste | 241 (47.2) |
| Scheduled tribes | 6 (1.2) |
| Education | |
| Illiterate | 179 (35.1) |
| Literate and below class 5 | 17 (3.3) |
| Primary school | 87 (17.1) |
| Middle school | 96 (18.7) |
| High or Secondary school | 111 (21.8) |
| Graduate or Post-graduate | 19 (3.7) |
| Professional or Honours | 1 (0.2) |
| Occupation | |
| Unemployed/ home maker | 303 (59.4) |
| Unskilled worker | 87 (17) |
| Semi-skilled worker | 84 (16.5) |
| Skilled worker | 8 (1.6) |
| Clerical, shop owner, farmer | 24 (4.7) |
| Semi- professional | 3 (0.6) |
| Professional | 1 (0.2) |
| Socio economic class | |
| Lower | 9 (1.8) |
| Upper lower | 411 (80.5) |
| Lower middle | 85 (16.7) |
| Upper middle | 5 (1) |
| Upper | 0 (0) |
| State of origin | |
| Uttar Pradesh | 235 (46.1) |
| Bihar | 101 (19.8) |
| Punjab | 57 (11.1) |
| Haryana | 48 (9.4) |
| Himachal Pradesh | 9 (1.8) |
| Chandigarh | 43 (8.4) |
| Other | 17 (3.4) |
## Epidemiological analysis
HLQ scale scores for all participants were compared across sociodemographic characteristics, as previously described by Beauchamp et al [43]. Discrete data were presented as proportions while continuous variables were expressed as means, standard deviations (SD) and $95\%$ confidence intervals (CI). With the nine scale scores of the HLQ as dependent variables and sociodemographic characteristics and history of chronic conditions as independent variables, we conducted robust analysis of variance (ANOVA) and a post-hoc test using Games-Howell method where required. Independent variables included male/female; age (18–35 years, 36–55 years, 56–65 years and above 65 years); education (illiterate, literate and below grade 5, primary school, middle school, high or secondary school, graduate or post-graduate, and professional); socio-economic status (SES) using modified Kuppuswamy scale 2019 (lower, upper-lower, lower-middle, upper-middle and upper) [44]; hypertension diagnosed by a physician at least two weeks ago; new case of hypertension (average systolic BP at or above 140 and/or average diastolic BP at or above 90 mm Hg at the time of survey); internal migrant (people coming from other states and residing in the village for livelihood or other reason); having a health card permitting free medical services at a hospital; and self-reported chronic illnesses.
Effect size (ES) using Cohen’s d was used to estimate the strength of the associations between health literacy mean scores and other variables. Cohen’s d is determined by calculating the mean difference between sub-groups and dividing by the pooled standard deviation. ES was assumed to be small if it was from 0.20–0.49, medium if it was from 0.50–0.79, and large if it was >0.80 [45]. Statistical significance was set at $p \leq 0.05.$
Table 2 displays the mean scores and Fig 1 displays the distribution of mean scores for the nine HLQ scales. Overall, for scales 1 to 5 (score range from 1 to 4), participants reported the highest scores for Scale 4. Social support for health (mean 2.88; SD 0.25), indicating that on average they had good social support. The lowest scores (mean 1.81; SD 0.28) were for Scale 3. Actively managing my health, indicating that many people do not proactively manage their own health care.
**Fig 1:** *Distribution of scores across the nine health literacy questionnaire scales.* TABLE_PLACEHOLDER:Table 2 For Scales 6 to 9 (score range from 1 to 5), the highest reported scores were for Scale 6. Ability to actively engage with healthcare providers (mean 3.66; SD 0.26) and the lowest scores were for Scale 8. Ability to find good health information (mean 2.65; SD 0.46) These scores indicate that, although participants proactively communicated with their healthcare providers, they were experiencing difficulties in finding good health information.
Table 3 provides details about the associations between sociodemographic characteristics and the nine HLQ scales. The largest effect sizes (greater than 1) between demographic groups were observed among age groups for Scale 4. Social support for health, where the youngest group (18–35 years) scored lowest. Also, people with middle SES scored significantly higher than those with lower SES for Scales 7. Navigating the healthcare system, 8. Finding good health information, and 9. Understanding health information well enough to know what to do. Moderate to large differences were observed between men and women in Scales 7. ( ES = 0.67), 8. ( ES = 0.44) and 9. ( ES = 0.53), but lower than women (ES = 0.24) in Scale 6. Ability to actively engage with healthcare provider.
**Table 3**
| Factors(n) | Factors(n).1 | Factors(n).2 | Factors(n).3 | Part 1 | Part 1.1 | Part 1.2 | Part 2 | Part 2.1 | Part 2.2 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| | 1 (HPS) Mean (SD) | 2 (HSI) Mean (SD) | 3 (AMH) Mean (SD) | 4 (SS) Mean (SD) | 5 (CA) Mean (SD) | 6 (AE) Mean (SD) | 7 (NHS) Mean (SD) | 8 (FHI) Mean (SD) | 9 (UHI) Mean (SD) |
| Gender | | | | | | | | | |
| Women (302) | 2.75 (0.25) | 1.81 (0.25) | 1.78 (0.24) | 2.87 (0.28) | 1.92 (0.26) | 3.68 (0.19) | 2.84 (0.28) | 2.57 (0.38) | 3.02 (0.35) |
| Men (208) | 2.63 (0.42) | 1.79 (0.33) | 1.84 (0.33) | 2.90 (0.31) | 1.90 (0.34) | 3.61 (0.34) | 3.11 (0.48) | 2.78 (0.54) | 3.26 (0.51) |
| Difference | 0.01 | 0.02 | 0.06 | 0.03 | 0.02 | 0.07 | 0.27 | 0.21 | 0.24 |
| ES | 0.02 | 0.04 | 0.19 | 0.11 | 0.03 | 0.24 | 0.67 | 0.44 | 0.53 |
| Age (in years) | | | | | | | | | |
| 18-35 (317) | 2.57 (0.32) | 1.82 (0.28) | 1.80 (0.29) | 2.66 (0.25) | 1.93 (0.29) | 3.64 (0.27) | 2.91 (0.30) | 2.67 (0.46) | 3.14 (0.44) |
| 36-55 (160) | 2.71 (0.33) | 1.79 (0.27) | 1.83 (0.26) | 2.93 (0.25) | 1.90 (0.28) | 3.07 (0.38) | 3.07 (0.38) | 2.69 (0.42) | 3.14 (0.43) |
| 56-65 (25) | 2.70 (0.40) | 1.65 (0.40) | 1.74 (0.37) | 2.94 (0.25) | 1.90 (0.28) | 2.85 (0.42) | 2.85 (0.42) | 2.34 (0.67) | 2.91 (0.39) |
| >65 (8) | 2.70 (0.15) | 1.82 (0.16) | 1.75 (0.18) | 2.90 (0.17) | 1.74 (0.40) | 2.94 (0.31) | 2.94 (0.31) | 2.37 (0.36) | 2.62 (0.51) |
| Difference (1&2) | 0.14 | 0.03 | 0.03 | 0.27 | 0.03 | 0.57 | 0.16 | 0.02 | 0.00 |
| ES (1&2) | 0.43 | 0.10 | 0.10 | 1.08 | 0.10 | 0.11 | 0.46 | 0.04 | 0.00 |
| Difference (1&3) | 0.13 | 0.17 | 0.06 | 0.28 | 0.03 | 0.21 | 0.06 | 0.33 | 0.23 |
| ES (1&3) | 0.35 | 0.49 | 0.18 | 1.12 | 0.54 | 0.16 | 0.16 | 0.57 | 0.55 |
| Difference (1&4) | 0.13 | 0.00 | 0.05 | 0.24 | 0.19 | 0.70 | 0.70 | 0.30 | 0.52 |
| ES (1&4) | 0.52 | 0.00 | 0.20 | 1.12 | 0.07 | 0.13 | 0.13 | 0.72 | 1.90 |
| Education | | | | | | | | | |
| 1-Illiterate (179) | 2.62 (0.32) | 1.73 (0.31) | 1.71 (0.29) | 2.83 (0.26) | 1.84 (0.31) | 3.62 (0.26) | 2.77 (0.38) | 2.40 (0.46) | 2.87 (0.37) |
| 2-Below Secondary (104) | 2.58 (0.32) | 1.73 (0.31) | 1.80 (0.26) | 2.88 (0.26) | 1.83 (0.30) | 3.62 (0.27) | 2.92 (0.35) | 2.59 (0.36) | 3.06 (0.29) |
| 3-Secondary or above (227) | 2.65 (0.34) | 1.89 (0.22) | 1.88 (0.27) | 2.93 (0.22) | 2.01 (0.25) | 3.70 (0.25) | 3.11 (0.37) | 2.88 (0.39) | 3.34 (0.43) |
| Difference (1&2) | 0.4 | 0 | 0.09 | 0.05 | 0.01 | 0 | 0.15 | 0.19 | 0.29 |
| ES (1&2) | 0.12 | 0 | 0.32 | 0.19 | 0.03 | 0 | 0.41 | 0.46 | 0.57 |
| Difference (1&3) | 0.03 | 0.16 | 0.17 | 0.10 | 1.77 | 0.08 | 0.34 | 0.48 | 0.47 |
| ES (1&3) | 0.09 | 0.62 | 0.60 | 0.19 | 0.60 | 0.31 | 1.21 | 1.12 | 1.17 |
| Difference (2&3) | 0.07 | 0.16 | 0.08 | 0.05 | 1.78 | 0.08 | 0.19 | 0.29 | 0.28 |
| ES (2&3) | 0.21 | 0.33 | 0.29 | 0.20 | 0.65 | 0.30 | 0.90 | 0.79 | 0.76 |
| Socio-economic class | | | | | | | | | |
| Lower (420) | 2.61 (0.31) | 1.78 (0.28) | 1.78 (0.28) | 2.86 (0.24) | 1.88 (0.29) | 3.64 (0.26) | 2.87 (0.35) | 2.57 (0.41) | 3.02 (0.38) |
| Middle (90) | 2.68 (0.41) | 1.92 (0.25) | 1.93 (0.27) | 2.97 (0.26) | 2.07 (0.27) | 3.70 (0.27) | 3.34 (0.38) | 3.07 (0.46) | 3.57 (0.42) |
| Difference | 0.07 | 0.14 | 0.15 | 0.11 | 0.19 | 0.06 | 0.47 | 0.50 | 0.55 |
| ES | 0.19 | 0.52 | 0.52 | 0.43 | 0.67 | 0.22 | 1.28 | 1.14 | 1.37 |
| Internal migration | | | | | | | | | |
| Yes (467) | 2.63 (0.34) | 1.79 (0.29) | 1.80 (0.29) | 2.85 (0.26) | 1.90 (0.29) | 3.65 (0.29) | 2.95 (0.40) | 2.64 (0.46) | 3.11 (0.43) |
| No (43) | 2.57 (0.28) | 1.91 (0.29) | 1.84 (0.26) | 2.92 (0.15) | 2.01 (0.27) | 3.69 (0.27) | 2.98 (0.33) | 2.78 (0.45) | 3.18 (0.53) |
| Difference | 0.06 | 0.12 | 0.04 | 0.07 | 0.11 | 0.04 | 0.03 | 0.16 | 0.07 |
| ES | 0.19 | 0.45 | 0.14 | 0.18 | 0.39 | 0.14 | 0.08 | 0.30 | 0.14 |
| Chronic Illness | | | | | | | | | |
| Yes | 2.71 (0.28) | 1.84 (0.25) | 1.85 (0.35) | 2.88 (0.24) | 1.94 (0.26) | 3.72 (0.24) | 2.95 (0.39) | 2.63 (0.46) | 3.09 (0.44) |
| No | 2.60 (0.34) | 1.79 (0.29) | 1.79 (0.26) | 2.88 (0.27) | 1.90 (0.30) | 3.64 (0.26) | 2.95 (0.40) | 2.66 (0.46) | 3.13 (0.44) |
| Difference | 0.11 | 0.05 | 0.06 | 0 | 0.04 | 0.08 | 0 | 0.03 | 0.04 |
| ES | 0.35 | 0.01 | 0.18 | 0.01 | 0.11 | 0.33 | 0.002 | 0.06 | 0.08 |
| Health-card | | | | | | | | | |
| Yes (45) | 2.89 (0.31) | 1.79 (0.33) | 1.90 (0.30) | 3.01 (0.25) | 1.88 (0.32) | 3.67 (0.28) | 3.25 (0.40) | 2.79 (0.52) | 3.27 (0.44) |
| No (465) | 2.60 (0.32) | 1.80 (0.28) | 1.80 ((0.28) | 2.87 (0.25) | 1.91 (0.29) | 3.65 (0.26) | 2.93 (0.38) | 2.64 (0.46) | 3.10 (0.43) |
| Difference | 0.29 | 0.01 | 0.10 | 0.14 | 0.03 | 0.02 | 0.32 | 0.15 | 0.17 |
| ES | 0.92 | 0.03 | 0.34 | 0.56 | 0.09 | 0.07 | 0.82 | 0.30 | 0.39 |
Education was clearly associated with health literacy in all scales, where people with secondary education scored higher than those who did not complete secondary education. ES ranged from 0.20 for Scale 4. to 0.90 for Scale 7. SES was most strongly associated with ability to navigate the healthcare system (Scale 7), as well as finding and understanding health information (Scales 8 and 9), but not for Scale 1. Feeling understood and supported by healthcare providers. Participants who had immigrated from other states scored lower than non-migrants in Scale 2. Having sufficient information to manage health (ES = 0.45), and Scales 5. Appraisal of health information (ES = 0.39) and 8. Ability to find good information (ES = 0.30).
People with a history of a chronic condition reported better healthcare provider support (Scale 1) and better ability to actively engage with their healthcare provider (Scale 6) than those who were not sick (ES 0.35 and 0.33, respectively). People with a health-card had substantially higher health literacy across several scales. They scored much higher (ES = 0.93) in Scale 1. Feeling understood and supported by healthcare providers and had better healthcare navigation skills (Scale 7) (ES = 0.81) than those who did not have a health-card. Also, they had better social support (Scale 4) (ES = 0.56) and skills to find and understand health information to manage health (ES = 0.30 and ES = 0.39 in Scales 8 and 9, respectively).
## Cluster analysis
Cluster analysis is an exploratory multivariate method used to identify groups of people with similar patterns of the predefined variables. Cluster analysis does not distinguish between dependent and independent variables and so allows for grouping (or clustering) of all the variables to reveal different patterns within populations. In this study, cluster analysis was used to reveal groups of people who reported similar patterns of health literacy scores across all nine scales, and then to explore related patterns across sociodemographic characteristics and chronic conditions.
The nine mean scale scores from respondents were standardized and converted into Z-scores. Hierarchical cluster analysis was then undertaken using Ward’s method for linkage as recommended by the developers of the Ophelia process [5,7]. Three criteria were used to determine the number of clusters: demographic and clinical variables; distance coefficient in agglomeration schedule; and standard deviation (<0.6 for each scale’s mean score). The sociodemographic variables examined were gender, age, education, occupation, socio-economic status, history of internal migration, clinical variables (chronic illness, previous diagnosis of hypertension, hypertension at time of survey), and the number of people who had free health-cards. Working from 3 through to 16 cluster solutions, the pattern generated by each consecutive cluster split was examined and compared with the parent cluster to identify differences between the patterns until further splits no longer provided new or meaningful patterns in the data.
Cluster analysis identified eight health literacy profiles in the study population (Table 4). Clusters were given designations from A to H. The number of people in each cluster varied from 9 (cluster G) to 276 (cluster C), with characteristics described below.
**Table 4**
| Cluster | A | B | C | D | E | F | G | H |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Number of participantsn (%) | 14(2.7) | 54(10.6) | 276(54.1) | 40(7.8) | 50(9.8) | 49(9.6) | 9(1.8) | 18(3.5) |
| Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) | Within cluster HLQ mean score (SD) |
| 1. Feeling understood and supported by healthcare providers | 2.07(0.15) | 3.00(0.25) | 2.64(0.09) | 3.03(0.17) | 2.21(0.36) | 2.54(0.23) | 3.11(0.45) | 2.07(0.52) |
| 2. Having sufficient information to manage my health | 2.09(0.29) | 2.01(0.13) | 1.87(0.07) | 1.33(0.24) | 1.78(0.28) | 1.95(0.33) | 1.17(0.28) | 1.14(0.20) |
| 3. Actively managing my health | 2.03(0.25) | 2.13(0.31) | 1.83(0.07) | 1.63(0.30) | 1.96(0.37) | 1.65(0.26) | 1.18(0.27) | 1.17(0.18) |
| 4. Social support for health | 3.14(0.17) | 3.09(0.19) | 2.89(0.05) | 3.10(0.23) | 2.66(0.37) | 2.85(0.15) | 3.16(0.17) | 2.20(0.39) |
| 5. Appraisal of health information | 2.31(0.37) | 2.10(0.21) | 1.97(0.06) | 1.46(0.30) | 1.84(0.27) | 2.12(0.22) | 1.24(0.17) | 1.16(0.15) |
| 6. Ability to actively engage with healthcare providers | 3.90(0.15) | 3.87(0.18) | 3.66(0.11) | 3.69(0.25) | 3.37(0.37) | 3.92(0.16) | 3.27(0.37) | 3.13(0.28) |
| 7. Navigating the healthcare system | 3.60(0.29) | 3.55(0.28) | 2.98(0.09) | 2.95(0.38) | 2.86(0.37) | 2.57(0.28) | 2.26(0.30) | 2.02(0.49) |
| 8. Ability to find good health information | 3.76(0.32) | 3.17(0.30) | 2.68(0.11) | 2.41(0.39) | 2.76(0.48) | 2.38(0.36) | 1.33(0.22) | 1.59(0.46) |
| 9. Understand health information well enough to know what to do | 4.03(0.37) | 3.67(0.33) | 3.11(0.14) | 3.15(0.48) | 3.22(0.53) | 2.62(0.36) | 2.31(0.15) | 2.48(0.44) |
| Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics | Sociodemographic characteristics |
| Mean age (years) | 27.7 | 40.7 | 33.6 | 38.2 | 32.1 | 35.3 | 44.6 | 30.3 |
| Women (%) | 7.1 | 14.8 | 76.1 | 40.0 | 26.0 | 83.7 | 55.6 | 44.4 |
| Illiteracy (%) | 0.0 | 1.9 | 40.9 | 27.5 | 18.0 | 46.9 | 100.0 | 72.2 |
| Secondary education or higher (%) | 85.7 | 85.2 | 41.3 | 35.0 | 48.0 | 32.7 | 0.0 | 5.6 |
| Occupation (%) | | | | | | | | |
| Unemployed | 14.2 | 16.6 | 67.7 | 55 | 38 | 89.7 | 77.8 | 72.3 |
| Clerical/ shop owner/ farmer | 0 | 7.4 | 1.4 | 0 | 0 | 0 | 0 | 0 |
| Average Income | 4.3 | 3.6 | 3.2 | 3.5 | 3.3 | 3.1 | 3.9 | 3.2 |
| Average Socio-economic status | 2.9 | 2.6 | 2.1 | 2.2 | 2.2 | 2.0 | 2.0 | 1.9 |
| Migration (%) | 85.7 | 96.3 | 88.0 | 97.5 | 94.0 | 98.0 | 100.0 | 94.4 |
| Health card (%) | 7.1 | 27.8 | 5.1 | 22.5 | 6.0 | 2.0 | 22.2 | 0.0 |
| Medical conditions | Medical conditions | Medical conditions | Medical conditions | Medical conditions | Medical conditions | Medical conditions | Medical conditions | Medical conditions |
| Any chronic illness (%) | 7.1 | 31.5 | 23.6 | 15.0 | 18.0 | 30.6 | 22.2 | 16.7 |
| Known hypertension (%) | 7.1 | 18.5 | 15.6 | 10.0 | 8.0 | 26.5 | 11.1 | 11.1 |
| New hypertension case (%) | 42.9 | 31.5 | 17.8 | 20 | 30 | 10 | 11 | 22 |
This was the second smallest cluster of younger adults, mainly men, who were educated and had a good income. They scored highest among all clusters in Scales 8. Ability to find good information (mean 3.76) and 9. Understand health information well enough to know what to do (4.03), and lowest in Scale 1. Feeling understood and supported by a healthcare professional (2.07). This cluster had the highest rate of new cases of high blood pressure ($42.9\%$) and most were unaware they had this health condition.
Cluster B: Have knowledge of healthcare resources and support from healthcare providers but lack good health information ($$n = 54$$) This cluster constituted mainly middle-aged males (mean 40.7 years). Almost one-third ($31.5\%$) had a chronic illness and equal proportion were found to have previously unknown high blood pressure. More of these participants had health-cards than in any other cluster. They were educated but did not think they had enough information, as indicated by the mean score for Scale 2. Having sufficient information to manage own health (2.01).
Cluster C: Have the ability to understand health information well but have support from a healthcare provider and do not actively manage own health ($$n = 276$$) This was the largest cluster, mainly comprised of young adult women (mean age 33.6 years). Many of them were illiterate ($40\%$) with about two-third ($67.7\%$) unemployed. Scores in Scale 1 (2.61) indicated the people in this cluster don’t always feel understood and supported by healthcare providers and the low mean score (1.87) in Scale 3 suggests they were not spending much time actively managing their health.
Cluster D: Have good support from healthcare providers but limited ability to find or appraise health information ($$n = 40$$) The people in this cluster (men $60\%$; women $40\%$) had low education (only $35\%$ had completed secondary education or higher) and had middle range incomes. They felt they didn’t have adequate information, as shown by a very low score (1.33) for Scale 2. Having sufficient information to manage health, despite having good support from their healthcare providers (Scale 1 mean score of 3.03).
Cluster E: Low support from healthcare providers and limited ability to engage with them ($$n = 50$$) People in this cluster were mostly young adult men ($74\%$), with almost half having completed secondary education and who could generally understand health information (Scale 9 score of 3.22). However, they felt they did not have adequate health information to manage their health (score of 1.78 for Scale 2. Having sufficient information to manage health). They also felt they had little support from a healthcare provider (Scale 1 score of 2.21) and were not confident in actively engaging with them (Scale 6 score of 3.37).
Cluster F: Highest ability to engage with healthcare providers but do not have sufficient information and are not managing their own health ($$n = 49$$) The people in this cluster were mostly women ($83.7\%$), of whom almost half were illiterate and with low SES. This cluster had the highest number of participants with chronic conditions ($30.6\%$) and a low score (1.65) for Scale 3. Actively managing my health. They could, however, communicate well with doctors and scored the highest among all clusters (3.92) for Scale 6. Ability to actively engage with healthcare providers.
Cluster G: Good social and healthcare provider support but limited access to and understanding of health information ($$n = 9$$) This was a small group of people, but this cluster had the highest average age (44 years) among all clusters. All participants were illiterate migrants belonging to low SES, and $22\%$ had health-cards. These participants reported having good healthcare support (Scale 1 score of 3.11) and social support (Scale 4 score of 3.16). However, they were experiencing challenges in all other HLQ dimensions.
Cluster H: Limited healthcare and social support with inadequate understanding of health information ($$n = 18$$) This was a cluster with younger people (mean age 30.3 years) and almost three quarters ($72.2\%$) were illiterate. No participant reported having a health-card. Unlike other clusters, which generally had a higher score for social support (Scale 4), this cluster reported the lowest social support among all clusters (2.20). In fact, this cluster also had the lowest scores among all clusters for all scales except Scale 1 (2.07), which had the same scores as Cluster A (2.07).
## Ethical considerations
The study plan was approved by Institute Ethics Committee, Post-Graduate Institute of Medical Education and Research Chandigarh (INT/IEC/$\frac{2019}{000414}$; Date: $\frac{01}{03}$/2019). Informed consent was obtained from all participants and confidentiality was maintained. The study was registered under the Clinical Trial Registry of India (CTRI/$\frac{2019}{10}$/021827; $\frac{31}{10}$/2019).
## Sociodemographic characteristics of participants
Data from 199 households (510 participants) were included. The household response rate was $76.3\%$ and on 62 ($23.7\%$) of occasions, the adjacent house was approached where a response was collected on $100\%$ of approaches. The overall response rate was $76.2\%$. Among the 510 participants, $59.2\%$ were women, most ($62.1\%$) were young adults aged 18–35 years and only $1.6\%$ were older than 65 years. Most of the participants were internal migrants ($91.6\%$) and were Hindus ($85.3\%$). Almost half ($47.2\%$) belonged to scheduled caste, a socially disadvantaged class in India. More than one-third of participants were illiterate ($35.1\%$) and only $3.7\%$ were college or university graduates or had obtained further higher education. One-third of participants ($33.3\%$) were unskilled or semi-skilled workers. Most participants belonged to upper-lower class ($80.5\%$) as per Kuppuswami classification of socio-economic status (SES) 2019 [44].
## Discussion
In this study we conducted standard descriptive analysis and cluster analysis to explore the health literacy, sociodemographic, and other health-related characteristics of people living in a resource poor context in northern India. We sought to generate a nuanced understanding of what the data mean to inform the development of an intervention strategy to address the health literacy needs of this rural population. The study also allowed for comparison of methods by which data can be interpreted and how those interpretations may influence decisions about public health initiatives and policies.
The epidemiological analysis examined overall population health literacy mean scores and indicated that, on average, most participants in this study had multiple health literacy challenges, except in the areas of having social support for health and their ability to engage with and feel supported by their healthcare providers. Using this method of data analysis, decisions might be made about population-level initiatives to support people to navigate the healthcare system, to actively manage their health, to understand and appraise health information, and to help people to find and have sufficient health information.
Cluster analysis examined the health literacy patterns among participants and related these patterns to demographic and health-related data, which resulted in eight groups (clusters) of people with distinct patterns of demographic, health condition, and health literacy characteristics. Importantly, this analysis uncovered a group of young men with relatively few health literacy challenges but with high rates of undiagnosed high blood pressure. While the descriptive analysis showed that this population, overall, has health literacy challenges, the cluster analysis revealed sub-groups of people with similar demographics and health conditions who have specific patterns of health literacy strengths, challenges and preferences. Cluster analysis, although not commonly used in public health, has the potential to inform the development of public health initiatives that are tailored to specific population subgroups, especially groups that frequently experience disadvantage or marginalisation [6,8,46–48].
## Descriptive epidemiological analysis and health literacy needs
In public health research, descriptive epidemiological analysis is useful for studying disease distribution, exploring risk-factors of a disease, and identifying public health hazards and other factors [49]. Score distributions were examined in this study in India and indicated that there was a wide distribution of health literacy across the population (Fig 1), and most people appeared to have several health literacy challenges. The epidemiological analysis provided the average health literacy scores and associations with other characteristics of the population, and this helped to draw attention to some key characteristics, such as people generally having good social support for health, which suggests these are potential strengths that could be built on. Importantly, although most people in this study were migrants, and some were living alone, the majority reported having good social support for health in their community. On average, the data indicated that people also had good support from at least one healthcare provider and were able to actively engage with them (Scales 1 and 6).
The epidemiological analysis provided information on some health literacy challenges experienced by the participants. For example, most participants who did not have secondary education reported markedly lower scores in Scales 5, 7, 8, and 9, which concurs with the results of other studies where education levels were found to be associated with health literacy [6,50]. Similarly, low SES was associated with lower scores for all health literacy scales except for Scale 1. Feeling understood and supported by healthcare providers [51]. Being male seems to have a role in enabling access to health services because men tended to have higher scores than women in scales relating to navigating the healthcare system and finding and understanding health information. This is expected because many women have reduced access to resources in this population [52,53].
Having a health-card permitting free treatment was associated with people who had a relatively good ability to navigate the healthcare system, felt understood and supported by healthcare providers, and had the ability to get good health information and understand information to manage their health. Studies have shown that patients who had opportunities to avail low-cost or free treatments had better chronic disease management, therefore, such facilities are necessary for supporting health as well as promoting health literacy, especially for low SES populations [54].
## Clusters and health literacy needs
The descriptive epidemiological analysis investigated average associations between the overall population health literacy scores and individual characteristics. Cluster analysis extended this investigation to uncover eight groups of people with different patterns of specific health literacy strengths and needs and related sociodemographic and clinical characteristics. Specific needs, especially for groups that experience disadvantage or marginalisation, are often masked when investigations only go as far as using average population scores to describe a population’s characteristics. For example, Cluster A had people who were generally well educated, male, with a higher SES, and with HLQ scores that were higher on most scales than other clusters, but scores were surprisingly low in Scale 1. Feeling understood and supported by a healthcare provider. It could be that they were young, considered themselves healthy, and did not see the need to have a regular healthcare provider. Therefore, a striking finding was that almost half of the men in this group were not aware, at the time of data collection, that they had high blood pressure and were in urgent need of active care. This finding demonstrates that efforts are required to uncover the differences among groups within populations so that initiatives can be tailored to meet the different needs of people in different contexts.
Cluster B, which consisted mainly of educated men with a chronic illness and a health-card, had good navigation skills and support and understanding from their healthcare providers. Conversely, although Cluster C had a somewhat similar cluster pattern (but with lower scores), it was by far the largest group and primarily comprised lower educated women. They reported low support from healthcare providers and struggled to navigate the healthcare system. The different education levels of people in the clusters might play a role in this inequity, but gender-based differences could also be a reason for women having lower health literacy. Inequitable access to health care for women is evident in India [55]. This finding calls for interventions to change systems and cultural norms to support women and promote equal access.
People in Cluster D had good support from healthcare providers and some education, but they reported low scores in dimensions linked to health information and they were not able to manage their health well. This finding suggests that people in this cluster, while having some good skills and support, struggle to find health information and understand how to use it to manage their health. The role of service providers is crucial in such circumstances to provide the necessary information in accessible and understandable formats [56].
People in Cluster E (primarily younger men with some education) tended not to engage and had limited ability to engage with healthcare providers and did not report having health conditions. However, $30\%$ of the men in this cluster were found to have high blood pressure at the time of the study. Health check-ups, including blood pressure monitoring, are necessary to detect disease early and would be an essential health promotion program for this group [57,58].
Cluster G was a small, somewhat older, and illiterate group of people who reported very low health literacy on most scales except social support for health and feeling understood by healthcare providers. Only a quarter of this group had a health-card. However, the positive engagements with healthcare providers may be a key strength that could be leveraged to support this group of people to better understand and manage their health.
## Using epidemiological or cluster analysis to inform intervention studies
The epidemiological analysis provided confirmation of associations that are already known: i.e., low education and low SES are associated with health literacy. Effect sizes revealed new information about the relative magnitude of these associations and indicated priority areas in which population health literacy could be improved. In this study context, access to health services could be improved by better understanding and generating system and policy initiatives to change how being a woman, having limited education, and having low SES result in health inequities. However, it is known that these sociodemographic factors are embedded structural inequities in many societies, that they are related to health literacy challenges, and that one-size-fits-all solutions are often employed but rarely lead to systemic and long-lasting changes in health equity. What is needed is to get beyond the known generalisations and to reveal the nuances of needs in different groups within populations.
The cluster analysis results reveal a diverse range of health literacy strengths and needs of groups of people in populations. This means that fit-for-purpose interventions can be developed for each group rather than a generalised whole population intervention. For example, programs, including using digital initiatives, could be developed for and with young people to motivate and support them to go for regular health check-ups, even if they feel well (Clusters A, C, E and H). Also, local healthcare services and community health workers could join with women’s groups to develop ways in which to make healthcare more accessible for women in Cluster C, and other women like them. Provision of door-to-door health information in a range of communication formats to inform and motivate people, such as those in Clusters F and H, would assist them in their self-care. People in Clusters A and B may also benefit from provision of tailored health information and self-care motivation skills training to enable them to use healthcare services to help manage and control their hypertension. Some groups of people may respond well to healthcare providers using communication methods, such as teachback [59,60], when delivering health information to patients, so health literacy responsiveness training for healthcare providers may support better interactions with patients and community members. These interventions mostly include actions that are required from service providers, so they are better able to reach out and respond to the needs of the people they serve, especially the people who experience disadvantage and marginalisation. Government funded health programs should be designed to not leave anyone behind and should therefore be able to reach the unreached, irrespective of societal, physical, and environmental barriers.
This study found that cluster analysis is useful for going beyond generalisations about populations to identify specific strengths and needs of the diverse groups of people within a population. People with different patterns of strengths and needs may need support to overcome barriers to accessing health services in ways that are different to people with other strengths and needs. Cluster analysis extends descriptive epidemiological analysis to reveal the different needs of different groups so that interventions can be tailored to promote health and equity to people in all their diversity.
## Strengths and limitations
A strength of the study is that it was population-based and reached many people who are usually not included in research studies. The study was conducted in a rural setting with a large and young migrant population. Compared with the general population, there was a slight underrepresentation of people older than 60 years ($5.6\%$), which is somewhat lower than the general population (>$8\%$) [61]. We administered the HLQ using one-on-one interviews to promote a high response rate among people who were illiterate. While this increased the generalizability of the findings to this particular village setting, a possible limitation is that interviewer administration may result in bias related to people providing socially desirable responses.
## Conclusion
Descriptive epidemiological and cluster analysis methods were used to explore health needs among residents of a rural north Indian village. The utilisation of population averages, without careful consideration of disparate demographic groups, might lead to the neglect of population subgroups who need tailored solutions to address their needs. This study demonstrates how cluster analysis is useful to expand the detail of epidemiological analyses to inform intervention development by generating nuanced evidence-based profiles across a whole community such that public health planners have actionable data on health literacy strengths and needs so that no group is left behind.
## References
1. Mantwill S, Monestel-Umaña S, Schulz PJ. **The Relationship between Health Literacy and Health Disparities: A Systematic Review.**. *PLoS One.* (2015.0) **10** e0145455. DOI: 10.1371/journal.pone.0145455
2. Nguena Nguefack HL, Pagé MG, Katz J, Choinière M, Vanasse A, Dorais M. **Trajectory Modelling Techniques Useful to Epidemiological Research: A Comparative Narrative Review of Approaches.**. *Clin Epidemiol.* (2020.0) **12** 1205-1222. DOI: 10.2147/CLEP.S265287
3. Osborne RH, Cheng CC, Nolte S, Elmer S, Besançon S, Budhathoki SS. **Health literacy measurement: embracing diversity in a strengths-based approach to promote health and equity, and avoid epistemic injustice**. *BMJ Global Health* (2022.0) **7** e009623
4. Ward J. H., Hook M. E.. **Application of an Hierarchical Grouping Procedure to a Problem of Grouping Profiles.**. *Educ Psychol Meas* (1963.0) **23** 69-81
5. Batterham RW, Buchbinder R, Beauchamp A, Dodson S, Elsworth GR, Osborne RH. **The OPtimising HEalth LIterAcy (Ophelia) process: study protocol for using health literacy profiling and community engagement to create and implement health reform.**. *BMC Public Health.* (2014.0) **14** 694. DOI: 10.1186/1471-2458-14-694
6. Beauchamp A, Batterham RW, Dodson S, Astbury B, Elsworth GR, McPhee C. **Systematic development and implementation of interventions to OPtimise Health Literacy and Access (Ophelia).**. *BMC Public Health* (2017.0) **17** 230. DOI: 10.1186/s12889-017-4147-5
7. Goeman D, Conway S, Norman R, Morley J, Weerasuriya R, Osborne RH. **Optimising Health Literacy and Access of Service Provision to Community Dwelling Older People with Diabetes Receiving Home Nursing Support.**. *J Diabetes.* (2016.0) **2016** 2483263. DOI: 10.1155/2016/2483263
8. Aaby A, Beauchamp A, O’Hara J, Maindal HT. **Large diversity in Danish health literacy profiles: perspectives for care of long-term illness and multimorbidity.**. *Eur J Public Health.* (2020.0) **30** 75-80. DOI: 10.1093/eurpub/ckz134
9. Anwar WA, Mostafa NS, Hakim SA, Sos DG, Cheng C, Osborne RH. **Health Literacy Co-Design in a Low Resource Setting: Harnessing Local Wisdom to Inform Interventions across Fishing Villages in Egypt to Improve Health and Equity.**. *Int J Environ Res Public Health* (2021.0) **18** 4518. DOI: 10.3390/ijerph18094518
10. Cheng C, Elsworth GR, Osborne RH. **Co-designing eHealth and Equity Solutions: Application of the Ophelia (Optimizing Health Literacy and Access) Process.**. *Front Public Health.* (2020.0) **8** 604401. DOI: 10.3389/fpubh.2020.604401
11. Elisabeth Stømer U, Klopstad Wahl A, Gunnar Gøransson L, Hjorthaug Urstad K. **Health Literacy in Kidney Disease: Associations with Quality of Life and Adherence.**. *J Ren Care.* (2020.0) **46** 85-94. DOI: 10.1111/jorc.12314
12. Jessup RL, Osborne RH, Beauchamp A, Bourne A, Buchbinder R. **Health literacy of recently hospitalised patients: a cross-sectional survey using the Health Literacy Questionnaire (HLQ).**. *BMC Health Serv Res.* (2017.0) **17** 52. DOI: 10.1186/s12913-016-1973-6
13. Storey A, Hanna L, Missen K, Hakman N, Osborne RH, Beauchamp A. **The Association between Health Literacy and Self-Rated Health Amongst Australian University Students.**. *J Health Commun.* (2020.0) **25** 333-343. DOI: 10.1080/10810730.2020.1761913
14. Sukys S, Cesnaitiene VJ, Ossowsky ZM. **Is Health Education at University Associated with Students’ Health Literacy? Evidence from Cross-Sectional Study Applying HLS-EU-Q**. *Biomed Res Int* (2017.0) **2017** 8516843. DOI: 10.1155/2017/8516843
15. Levin-Zamir D, Baron-Epel OB, Cohen V, Elhayany A. **The Association of Health Literacy with Health Behavior, Socioeconomic Indicators, and Self-Assessed Health From a National Adult Survey in Israel.**. *J Health Commun.* (2016.0) **21** 61-68. DOI: 10.1080/10810730.2016.1207115
16. Marrie RA, Salter A, Tyry T, Fox RJ, Cutter GR. **Health literacy association with health behaviors and health care utilization in multiple sclerosis: a cross-sectional study.**. *Interact J Med Res.* (2014.0) **3** e3. DOI: 10.2196/ijmr.2993
17. Jacobson AF, Sumodi V, Albert NM, Butler RS, DeJohn L, Walker D. **Patient activation, knowledge, and health literacy association with self-management behaviors in persons with heart failure**. *Heart Lung* (2018.0) **47** 447-451. DOI: 10.1016/j.hrtlng.2018.05.021
18. 18World Health Organization. Health literacy development for the prevention and control of noncommunicable diseases: Volume 2. A globally relevant perspective. Licence: CC BY-NC-SA 3.0 IGO. Geneva; 2022. https://www.who.int/publications/i/item/9789240055339.
19. Boot GR, Lowell A. **Acknowledging and promoting Indigenous knowledges, paradigms, and practices within health literacy-related policy and practice documents across Australia, Canada, and New Zealand.**. *Int. Indig. Policy J* (2019.0) **10**
20. Bakker MM, Putrik P, Aaby A, Debussche X, Morrissey J, Råheim Borge C. **Acting together–WHO National health literacy demonstration projects (NHLDPs) address health literacy needs in the European region.**. *Public Health Panorama.* (2019.0) **5**
21. 21Scottish Government, NHS Scotland. Making it Easy: A Health Literacy Action Plan for Scotland [Internet]. The Scottish Government, NHS Scotland: Edinburgh; 2014 [cited 2022 Aug 27]. Available from: http://www.gov.scot/Publications/2014/06/9850.
22. Wang J, Shahzad F. **A Visualized and Scientometric Analysis of Health Literacy Research.**. *Front Public Health.* (2022.0) **9** 811707. DOI: 10.3389/fpubh.2021.811707
23. Asharani PV, Lau JH, Roystonn K, Devi F, Peizhi W, Shafie S. **Health Literacy and Diabetes Knowledge: A Nationwide Survey in a Multi-Ethnic Population**. *Int J Environ Res Public Health* (2021.0) **18** 9316. DOI: 10.3390/ijerph18179316
24. Ahmad D, Mohanty I, Hazra A, Niyonsenga T. **The knowledge of danger signs of obstetric complications among women in rural India: evaluating an integrated microfinance and health literacy program.**. *BMC Pregnancy Childbirth.* (2021.0) **21** 79. DOI: 10.1186/s12884-021-03563-5
25. Garner SL, George CE, Young P, Hitchcock J, Koch H, Green G. **Effectiveness of an mHealth application to improve hypertension health literacy in India**. *Int Nurs Rev* (2020.0) **67** 476-483. DOI: 10.1111/inr.12616
26. Johri M, Subramanian SV, Koné GK, Dudeja S, Chandra D, Minoyan N. **Maternal Health Literacy Is Associated with Early Childhood Nutritional Status in India**. *J Nutr* (2016.0) **146** 1402-10. DOI: 10.3945/jn.115.226290
27. Dodson S., Good S., Osborne R.H.. *Health literacy toolkit for low and middle-income countries: a series of information sheets to empower communities and strengthen health systems.* (2015.0)
28. 28Municipal Corporation Chandigarh. Wards [Internet]. Chandigarh, India: The official website of Municipal Corporation Chandigarh, 2021 [cited on 2021 Dec 3]. Available from: http://mcchandigarh.gov.in/?q=wards.. *Wards [Internet]* (2021.0)
29. 29Registrar General and Census Commissioner of India. Chandigarh population 2011–2022 [Internet]. New Delhi: Office of Registrar General and Census Commissioner of India; 2011 [cited 2022 May 29]. Available from: https://www.census2011.co.in/census/state/chandigarh.html.. *Chandigarh population 2011–2022 [Internet]* (2011.0)
30. Bakker MM, Putrik P, Rademakers J, van de Laar M, Vonkeman H, Kok MR. **Addressing Health Literacy Needs in Rheumatology: Which Patient Health Literacy Profiles Need the Attention of Health Professionals?**. *Arthritis Care Res (Hoboken).* (2021.0) **73** 100-109. DOI: 10.1002/acr.24480
31. Buchbinder R, Batterham R, Elsworth G, Dionne CE, Irvin E, Osborne RH. **A validity-driven approach to the understanding of the personal and societal burden of low back pain: development of a conceptual and measurement model**. *Arthritis Res Ther* (2011.0) **13** R152. DOI: 10.1186/ar3468
32. Osborne RH, Batterham RW, Elsworth GR, Hawkins M, Buchbinder R. **The grounded psychometric development and initial validation of the Health Literacy Questionnaire (HLQ).**. *BMC Public Health* (2013.0) **13** 658. DOI: 10.1186/1471-2458-13-658
33. Boateng MA, Agyei-Baffour P, Angel S, Enemark U. **Translation, cultural adaptation and psychometric properties of the Ghanaian language (Akan; Asante Twi) version of the Health Literacy Questionnaire.**. *BMC Health Serv Res.* (2020.0) **20** 1064. DOI: 10.1186/s12913-020-05932-w
34. Budhathoki SS, Pokharel PK, Jha N, Moselen E, Dixon R, Bhattachan M. **Health literacy of future healthcare professionals: a cross-sectional study among health sciences students in Nepal.**. *Int Health.* (2019.0) **11** 15-23. DOI: 10.1093/inthealth/ihy090
35. Park JH, Osborne RH, Kim HJ, Bae SH. **Cultural and linguistic adaption and testing of the Health Literacy Questionnaire (HLQ) among healthy people in Korea.**. *PLoS One.* (2022.0) **17** e0271549. DOI: 10.1371/journal.pone.0271549
36. Moraes KL, Brasil VV, Mialhe FL, Sampaio HA, Sousa AL, Canhestro MR. **Validation of the Health Literacy Questionnaire (HLQ) to brazilian portuguese.**. *Acta Paulista de Enfermagem* (2021.0) **34**
37. Wahl AK, Hermansen Å, Osborne RH, Larsen MH. **A validation study of the Norwegian version of the Health Literacy Questionnaire: A robust nine-dimension factor model.**. *Scand J Public Health.* (2021.0) **49** 471-478. DOI: 10.1177/1403494820926428
38. Debussche X, Lenclume V, Balcou-Debussche M, Alakian D, Sokolowsky C, Ballet D. **Characterisation of health literacy strengths and weaknesses among people at metabolic and cardiovascular risk: Validity testing of the Health Literacy Questionnaire.**. *SAGE Open Med* (2018.0) **6**. DOI: 10.1177/2050312118801250
39. Nolte S, Osborne RH, Dwinger S, Elsworth GR, Conrad ML, Rose M. **German translation, cultural adaptation, and validation of the Health Literacy Questionnaire (HLQ).**. *PLoS One* (2017.0) **12** e0172340. DOI: 10.1371/journal.pone.0172340
40. Maindal HT, Kayser L, Norgaard O, Bo A, Elsworth GR, Osborne RH. **Cultural adaptation and validation of the Health Literacy Questionnaire (HLQ): robust nine-dimension Danish language confirmatory factor model.**. *Springerplus* (2016.0) **5** 1232. DOI: 10.1186/s40064-016-2887-9
41. Kolarcik P, Cepova E, Madarasova Geckova A, Elsworth GR, Batterham RW, Osborne RH. **Structural properties and psychometric improvements of the Health Literacy Questionnaire in a Slovak population.**. *Int J Public Health.* (2017.0) **62** 591-604. DOI: 10.1007/s00038-017-0945-x
42. Anwar WA, Mostafa NS, Hakim SA, Sos DG, Abozaid DA, Osborne RH. **Health literacy strengths and limitations among rural fishing communities in Egypt using the Health Literacy Questionnaire (HLQ).**. *PLoS One.* (2020.0) **15** e0235550. DOI: 10.1371/journal.pone.0235550
43. Beauchamp A, Buchbinder R, Dodson S, Batterham RW, Elsworth GR, McPhee C. **Distribution of health literacy strengths and weaknesses across socio-demographic groups: a cross-sectional survey using the Health Literacy Questionnaire (HLQ).**. *BMC Public Health.* (2015.0) **15** 678. DOI: 10.1186/s12889-015-2056-z
44. Saleem SM, Jan SS. **Modified Kuppuswamy socioeconomic scale updated for the year 2019.**. *Indian J Forensic Community Med.* (2019.0) **6** 1-3
45. Cohen J.. *Statistical power analysis for the behavioral sciences [e book].* (2013.0)
46. Jessup RL, Osborne RH, Buchbinder R, Beauchamp A. **Using co-design to develop interventions to address health literacy needs in a hospitalised population.**. *BMC Health Serv Res.* (2018.0) **18** 989. DOI: 10.1186/s12913-018-3801-7
47. Cheng C, Elsworth GR, Osborne RH. **Co-designing eHealth and Equity Solutions: Application of the Ophelia (Optimizing Health Literacy and Access) Process.**. *Front Public Health.* (2020.0) **8** 604401. DOI: 10.3389/fpubh.2020.604401
48. Debussche X.. **Addressing health literacy responsiveness in diabetes**. *Diabetes Epidemiol. Manag* (2021.0) **4** 100033
49. Rothman K.. *Epidemiology: An Introduction* (2004.0)
50. Van der Heide I, Wang J, Droomers M, Spreeuwenberg P, Rademakers J, Uiters E. **The relationship between health, education, and health literacy: results from the Dutch Adult Literacy and Life Skills Survey.**. *J Health Commun.* (2013.0) **18** 172-84. DOI: 10.1080/10810730.2013.825668
51. Stormacq C, Van den Broucke S, Wosinski J. **Does health literacy mediate the relationship between socioeconomic status and health disparities? Integrative review.**. *Health Promot Int.* (2019.0) **34** e1-e17. DOI: 10.1093/heapro/day062
52. Roy K, Chaudhuri A. **Influence of socioeconomic status, wealth and financial empowerment on gender differences in health and healthcare utilization in later life: evidence from India.**. *Soc Sci Med.* (2008.0) **66** 1951-62. DOI: 10.1016/j.socscimed.2008.01.015
53. Patel R, Chauhan S. **Gender differential in health care utilisation in India.**. *Clin. Epidemiology Glob. Health* (2020.0) **8** 526-30
54. Selvaraj S, Subramanian SV. **Financial risk protection & chronic disease care**. *Indian J Med Res* (2012.0) **136** 544-6. PMID: 23168693
55. Zodpey S, Negandhi P. **Inequality in health and social status for women in India—A long-standing bane.**. *Indian J Public Health.* (2020.0) **64** 325-327. DOI: 10.4103/ijph.IJPH_1312_20
56. Kapadia-Kundu N, Sullivan TM, Safi B, Trivedi G, Velu S. **Understanding health information needs and gaps in the health care system in Uttar Pradesh, India.**. *J Health Commun.* (2012.0) **17** 30-45. DOI: 10.1080/10810730.2012.666625
57. Chow CK, Teo KK, Rangarajan S, Islam S, Gupta R, Avezum A, Bahonar A. **Prevalence, awareness, treatment, and control of hypertension in rural and urban communities in high-, middle-, and low-income countries**. *JAMA* (2013.0) **310** 959-68. DOI: 10.1001/jama.2013.184182
58. Mirzaei M, Mirzaei M, Bagheri B, Dehghani A. **Awareness, treatment, and control of hypertension and related factors in adult Iranian population.**. *BMC Public Health* (2020.0) **20** 667. DOI: 10.1186/s12889-020-08831-1
59. Talevski J, Wong Shee A, Rasmussen B, Kemp G, Beauchamp A. **Teach-back: A systematic review of implementation and impacts.**. *PLoS One.* (2020.0) **15** e0231350. DOI: 10.1371/journal.pone.0231350
60. Talevski J, Beauchamp A, Wong Shee A, Rasmussen B, Hilbers J. *The Teach-Back Toolkit* (2021.0)
61. 61Registrar General and Census Commissioner of India. Census of India: Primary Census Abstract [Internet]. New Delhi: Office of Registrar General and Census Commissioner of India; 2011 [cited 2021 Aug 12]. Available from: https://censusindia.gov.in/pca/default.aspx.. *Census of India: Primary Census Abstract [Internet]* (2011.0)
|
---
title: 'Racial and neighborhood disparities in mortality among hospitalized COVID-19
patients in the United States: An analysis of the CDC case surveillance database'
authors:
- Atarere Joseph
- Tarsicio Uribe-Leitz
- Tanujit Dey
- Joaquim Havens
- Zara Cooper
- Nakul Raykar
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10022015
doi: 10.1371/journal.pgph.0000701
license: CC BY 4.0
---
# Racial and neighborhood disparities in mortality among hospitalized COVID-19 patients in the United States: An analysis of the CDC case surveillance database
## Abstract
### Background
Black and Hispanic populations have higher overall COVID-19 infection and mortality odds compared to Whites. Some state-wide studies conducted in the early months of the pandemic found no in-hospital racial disparities in mortality.
### Methods
We performed chi-square and logistic regression analyses on the CDC COVID-19 Case Surveillance Restricted Database. The primary outcome of the study was all-cause in-hospital mortality. The primary exposures were racial group (White, Black, Hispanic and Others) and neighborhood type (low vulnerability, moderate vulnerability, high vulnerability, very high vulnerability).
### Findings
The overall unadjusted mortality rate was $33\%$ and was lowest among Hispanics. In the fully adjusted models, Blacks and Hispanics had higher overall odds of dying [OR of 1.20 ($95\%$ CI 1.15, 1.25) and 1.23 ($95\%$ CI 1.17, 1.28) respectively] compared with White patients, and patients from neighborhoods with very high vulnerability had the highest mortality odds in the Northeast, Midwest and overall [Adjusted OR 2.08 ($95\%$ CI 1.91, 2.26)]. In the Midwest, Blacks and Hispanics had higher odds of mortality compared with Whites, but this was not observed in other regions.
### Interpretation
Among hospitalized COVID-19 patients, Blacks and Hispanics were more likely to die compared to Whites in the Midwest. Patients from highly vulnerable neighborhoods also had the highest likelihood of death in the Northeast and Midwest. These results raise important questions on our efforts to curb healthcare disparities and structural racism in the healthcare setting.
## Introduction
The Coronavirus disease (COVID-19) pandemic, caused by SARS-COV-2 has resulted in millions of deaths worldwide, with more than 1 million of these deaths occurring in the U.S. [1, 2]. Race and neighborhood of residence have been identified as important risk factors for infection and mortality from COVID-19 [3, 4]. Several studies have also shown that Blacks and Hispanics are more likely to be infected with COVID-19 and have a higher burden of mortality than Whites [5–9]. During the early months of the pandemic, residents of neighborhoods in New York with large proportions of Blacks/African Americans were found to be at a significantly higher risk of COVID-19 infection compared to residents of predominantly White neighborhoods [6]. Similarly, the Bronx which has the highest proportion of racial/ethnic minorities, poverty levels and the lowest educational levels had the highest rates of hospitalization and death from COVID-19 in New York [10].
However, some studies conducted among hospitalized COVID-19 patients have found no difference in in-hospital mortality by race [11–13]. A retrospective cohort study conducted on members of an integrated-delivery health system in Louisiana found no racial difference in the hazard of death among hospitalized COVID-19 patients despite a higher overall COVID-19 mortality rate among Blacks [11]. A similar study conducted in New York found that among hospitalized patients, Blacks had lower odds of severe illness and death when compared with Whites [12]. A study conducted among patients admitted to 92 hospitals in 12 U.S. states by Yehia et al. also found no difference in all-cause in-hospital mortality by race [13]. These findings, though, are in contradiction to published reports in other inpatient populations that consistently demonstrate a link between in-hospital mortality and race [14, 15].
The studies in New York and Louisiana, however, were state-level analyses using state and hospital-level datasets [11, 12]. Of the 92 hospitals included in the analysis by Yehia et al., only 2 were from Northeastern states which were the most affected during the early months of the pandemic [13]. Hence, the findings from these studies may not represent the state of disparities in mortality among hospitalized COVID-19 patients nationwide. In addition, although studies in New York have established a relationship between neighborhood racial composition and in-hospital mortality, there are no larger studies evaluating this link on the national level [10]. We sought to characterize in-hospital mortality rates nationwide, determine whether in-hospital mortality for COVID-19 varied based on race and neighborhood type, and evaluate for differences across census regions using a large nationwide database (The CDC COVID-19 Case Surveillance Restricted Access Detailed Data).
## Study design
We conducted a cross-sectional study of all hospitalized cases of COVID-19 in the 4 major U.S. census regions between January 1st and November 19th, 2020. This study was classified as exempt by the Institutional Review Board (IRB) of Mass General Brigham (MGB) in Boston, Massachusetts. The CDC Case Surveillance database is a de-identified dataset and does not directly involve human subjects as defined by federal regulations and guidance [16].
## Primary data
The analyses in this study were carried out on the Center for Disease control and prevention (CDC) COVID-19 Case Surveillance Restricted Access Detailed Data which is a 32-element dataset provided by the Case Surveillance Task Force and Surveillance Review and Response Group [17]. The CDC COVID-19 Case Surveillance Restricted Access database was created on April 4, 2020, is updated monthly and the version used in this analysis was last updated on December 4, 2020. It includes de-identified individual level data for 8,405,079 individuals collected across all U.S. states and territories from 1st January to 19th November 2020. The information on each individual in the database was collected using a standard questionnaire (the CDC case report form) [18]. The case definitions for COVID-19 used in this paper are based on the current Council of State and Territorial Epidemiologists case definitions for COVID-19 [19].
## Additional sources
The National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme (URCS): a classification system for U.S. counties based on population size. It was developed for use in studying and monitoring health disparities across the urban-rural continuum [20]. The most recent [2013] NCHS scheme which we used in this study is based on the 2010 census and the February 2013 office of Management and Budget delineation of metropolitan and micropolitan statistical areas [20].
The CDC Social Vulnerability Index (SVI): created by the Agency for Toxic Substances and Disease Registry Geospatial Research, Analysis & Services Program to “help public health officials and emergency response planners identify and map the communities that will most likely need support before, during, and after a hazardous event” [21]. The CDC SVI categorizes the relative vulnerability of every U.S. census tract or county in four summary themes (Socioeconomic, Minority Status and Language, Household Composition and Disability, Housing Type and Transportation) which rank counties based on their overall vulnerability with higher percentile ranks indicating higher vulnerability [21].
## Patient population
Our inclusion criteria were based on: [1] positive result for SARS-COV-2 infection by a molecular amplification detection test [2] hospitalization for COVID-19 in any of the 4 major census regions (Northeast, West, Midwest and South) designated by the United States Census Bureau [22].
Patients of all ages were included in the analyses.
## Outcomes and variables
The primary outcome of this study was all cause in-hospital mortality among patients diagnosed with COVID-19, defined by the categorical variable ‘death’ (yes/no) from the COVID-19 Case Surveillance database.
The primary predictors of interest are ‘race’ and ‘neighborhood type’.
The CDC case report form provides multiple options for reporting race/ethnicity. For our analyses, we categorized the variable into 4 different groups as done in previous studies investigating the relationship between race and COVID-19 mortality: Non-Hispanic White, Non-Hispanic Black, Hispanic and Other races (which includes Asian, American Indian/Alaska Native, Native Hawaiian/Other Pacific Islander and multiple ethnicities) [8, 23]. For simplicity, in this paper, we will refer to these racial groups as White, Black, Hispanic and Others respectively.
The Minority Status and Language summary theme of the CDC SVI provides a composite rank for counties based on the proportion of their residents who are non-white and have limited English proficiency (LEP). *To* generate a variable for ‘neighborhood type’, we grouped the percentiles into quartiles (Q1- low vulnerability, Q2- moderate vulnerability, Q3- high vulnerability, Q4- very high vulnerability) with higher quartiles indicating neighborhoods with a higher proportion of non-white residents and individuals with limited English proficiency.
The socioeconomic summary theme of the CDC SVI provides a composite rank for counties based on the income of residents, proportion who live below poverty, are unemployed or have no high school diploma. To adjust for socioeconomic status, we grouped the percentiles of the socioeconomic summary themes into quartiles (Q1- higher SES, Q2- upper middle SES, Q3- lower middle SES, Q4- lower SES) for our analyses with higher quartiles indicating lower socioeconomic status.
The 2013 NCHS URCS classifies counties into 6 main categories by population size: large central metropolitan, large fringe metropolitan, medium metropolitan, small metropolitan, micropolitan and rural/noncore [20]. For our analyses, county size was defined as metropolitan, micropolitan or rural/noncore [24].
Other variables from the CDC COVID-19 Case Surveillance database which were identified a priori from previous studies as independently associated with mortality from COVID-19 were also included in the analysis. These include sex (male or female) [25], age group (<40, 40–59, 60–79 and 80+) [25, 26], disease severity (non-critical or critical) [27], and presence of comorbidities [5, 28, 29]. We classified disease severity (using the COVID-19 WHO severity classification system) as follows: [1] Critical: Patients with acute respiratory distress syndrome (ARDS), those in the ICU, and those who were mechanically ventilated. [ 2] Non-critical: Symptomatic patients who do not meet the criteria for critical illness [27]. A patient was considered to have a comorbidity if they had any of the following conditions: Diabetes Mellitus, Hypertension, Severe Obesity (BMI> 40mg/kg2) Cardiovascular disease, Chronic Liver disease, Chronic Kidney Disease, Immunosuppressive or autoimmune diseases [18]. A current or previous history of smoking was also classified as a comorbidity [18].
## Statistical analysis
Descriptive statistics for the categorical variables have been presented as frequencies and percentages and were compared using chi-squared tests to evaluate the associations between the categorical variables.
Due to the significant confounding between race and neighborhood type, both variables were not included in the same model (see S4 Table). Multivariable logistic regression models were used to evaluate for disparities in COVID-19 mortality by race and neighborhood type among hospitalized COVID-19 patients on the national level. The models included the primary predictors (model one- race, model two- neighborhood type) and other sociodemographic and health-related variables including age group, sex, socioeconomic status, presence of comorbidity, disease severity, and county size. For model 1, we included an interaction term between race and age group (see S1 Table). We also conducted a subgroup analysis evaluating for racial disparities among patients from neighborhoods with very high vulnerability. In addition, we conducted stratified analyses of racial disparities by (i) disease severity and (ii) time-period (January to June vs July to November).
When interaction terms between [1] race and census region and [2] neighborhood type and census region, were added to the respective multivariable logistic regression models, the race-by-region and neighborhood-by-region interactions were statistically significant (see S2 and S3 Tables respectively).
Therefore, to evaluate, for differences in the pattern of racial and neighborhood disparities across the 4 major census regions, we used different multivariable logistic regression models (one for each census region) which included the outcome of interest (death), the primary predictors (race/ neighborhood type) and other sociodemographic and health-related variables including age group, sex, socioeconomic status, presence of comorbidity, disease severity, and county size. Additionally, based on the results from (ii) above, we conducted a subgroup analysis evaluating patterns of racial disparities across the 4 major census regions in the July to November time-period.
All analyses were performed on the complete cases of the patients by removing all patients with missing information. Missingness for race, neighborhood status and death in the dataset were $24.63\%$, $1.23\%$ and $29.44\%$ respectively. Detailed information on missing data is provided in the S6 Table. The statistical software R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria [2020]) was used to merge the variables in CDC SVI and the NCHS URCS databases to the CDC COVID-19 Case Surveillance database by matching with the county federal information processing codes. All other statistical analyses were conducted using Stata 16.1 (Stata Corp, College Station, TX USA). All tests were two-sided and p values < 0.05 were considered statistically significant.
## Results
A total of 106,962 hospitalized COVID-19 patients were included in this analysis. Of these, 55,468 ($51.9\%$) were Whites, 22,589 ($21.1\%$) were Blacks, 20,846 ($19.5\%$) were Hispanics and 8,059 ($7.5\%$) were of other races. Figs 1 and 2 show the mortality rate by racial group in each census region stratified by age (<60 years vs ≥60 years), all of which were highest in the Northeast. In the Northeast, mortality rates among Blacks <60 years was about twice the mortality rates in Whites; among those ≥ 60 years, the mortality rates were about even. Fig 3 shows the mortality rate by neighborhood type. In the Northeast and Midwest, patients from neighborhoods with very high vulnerability had the highest mortality rates. National mortality rates during the study were highest in April and the mortality rates trended downwards towards November (Fig 4).
**Fig 1:** *Unadjusted mortality rates (%) per 100 hospitalized COVID-19 patients by racial group among patients <60 years.* **Fig 2:** *Unadjusted mortality rates (%) per 100 hospitalized COVID-19 patients by racial group among patients 60+ years.* **Fig 3:** *Unadjusted mortality rates (%) per 100 hospitalized COVID-19 patients by neighborhood type.* **Fig 4:** *Monthly mortality rates among hospitalized COVID-19 patients by census region.*
Over three-fifths of the hospitalized cases were ≥60 years old and about one-fifth were over 80 years. The sex distribution of the hospitalized population was about even. Most patients were from metropolitan counties (93,673, $87.6\%$), had at least one comorbidity (92,809, $86.8\%$), and about one-third (33,845, $31.6\%$) were in critical condition. The mortality rate among all hospitalized COVID-19 patients during the study period was about $33\%$ (35,$\frac{569}{106}$,962). Detailed characteristics of the study population by racial category is shown in Table 1.
**Table 1**
| Demographic and Clinical variables | Total | White | Black | Hispanic | Others | p-value |
| --- | --- | --- | --- | --- | --- | --- |
| Demographic and Clinical variables | 106,962 (%) | 55,468 (%) | 22,589 (%) | 20,846 (%) | 8,059 (%) | p-value |
| Sex | | | | | | |
| Female | 50,510 (47.22) | 25,967 (46.81) | 11,716 (51.87) | 9,068 (43.50) | 3,759 (46.64) | <0.0001 |
| Male | 56,452 (52.78) | 29,501 (53.19) | 10,873 (48.13) | 11,778 (56.50) | 4,300 (53.36) | |
| Age Group | | | | | | |
| <40 years | 13,278 (12.41) | 4,094 (7.38) | 3,048 (13.49) | 4,887 (23.44) | 1,249 (15.50) | <0.0001 |
| 40–59 years | 28,164 (26.33) | 10,625 (19.16) | 6,877 (30.44) | 8,272 (39.68) | 2,390 (29.66) | |
| 60–79 years | 44,154 (41.28) | 25,157 (45.35) | 9,638 (42.67) | 6,088 (29.20) | 3,271 (40.59) | |
| 80+ years | 21,366 (19.98) | 15,592 (28.11) | 3,026 (13.40) | 1,599 (7.67) | 1,149 (14.26) | |
| Comorbidities | | | | | | |
| Absent | 14,153 (13.23) | 6,717 (12.11) | 2,192 (9.70) | 3,938 (18.89) | 1,306 (16.21) | <0.0001 |
| Present | 92,809 (86.77) | 48,751 (87.89) | 20,397 (90.30) | 16,908 (81.11) | 6,753 (83.79) | |
| Disease Severity | | | | | | |
| Non-critical | 73,117 (68.36) | 39,290 (70.83) | 15,138 (67.01) | 13,624 (65.36) | 5,065 (62.85) | <0.0001 |
| Critical | 33,845 (31.64) | 16,178 (29.17) | 7,451 (32.99) | 7,222 (34.64) | 2,994 (37.15) | |
| Socioeconomic Status (SES) | | | | | | |
| Quartile 1 (Higher SES) | 33,993 (31.78) | 21,461 (38.69) | 4,455 (19.72) | 5,385 (25.83) | 2,692 (33.40) | <0.0001 |
| Quartile 2 (Upper middle SES) | 29,986 (28.03) | 15,509 (27.96) | 5,951 (26.34) | 6,320 (30.32) | 2,206 (27.37) | |
| Quartile 3 (Lower middle SES) | 32,648 (30.52) | 14,265 (25.72) | 8,697 (38.50) | 7,083 (33.98) | 2,603 (32.30) | |
| Quartile 4 (Lower SES) | 10,335 (9.66) | 4,233 (7.63) | 3,486 (15.43) | 2,058 (9.87) | 558 (6.92) | |
| Neighborhood Type | | | | | | |
| Quartile 1 (low vulnerability) | 5,351 (5.00) | 5,210 (9.39) | 40 (0.18) | 23 (0.11) | 78 (0.97) | <0.0001 |
| Quartile 2 (moderate vulnerability) | 8,986(8.40) | 7,641 (13.78) | 812 (3.59) | 331 (1.59) | 202 (2.51) | |
| Quartile 3 (high vulnerability) | 24,971 (23.35) | 16,550 (29.84) | 4,696 (20.79) | 2,681 (12.86) | 1,044 (12.95) | |
| Quartile 4 (very high vulnerability) | 67,654 (63.25) | 26,067 (46.99) | 17,041 (75.44) | 17,811 (85.44) | 6,735 (83.57) | |
| Census Region | | | | | | |
| Northeast | 40,326 (37.70) | 17,672 (31.86) | 7,856 (34.78) | 11,054 (53.03) | 3,744 (46.46) | <0.0001 |
| Midwest | 36,889 (34.49) | 23,120 (41.68) | 7,413 (32.82) | 4,019 (19.28) | 2,337 (29.00) | |
| South | 20,111 (18.80) | 10,663 (19.22) | 6,400 (28.33) | 2,436 (11.69) | 612 (7.59) | |
| West | 9,636 (9.01) | 4,013 (7.23) | 920 (4.07) | 3,337 (16.01) | 1,366 (16.95) | |
| County Size | | | | | | |
| Metropolitan | 93,673 (87.58) | 45,454 (81.95) | 20,901 (92.53) | 19,558 (93.82) | 7,760 (96.29) | <0.0001 |
| Micropolitan | 8,419 (7.87) | 6,163 (11.11) | 1,103 (4.88) | 916 (4.39) | 237 (2.94) | |
| Rural/Noncore | 4,870 (4.55) | 3,851 (6.94) | 585 (2.59) | 372 (1.78) | 62 (0.77) | |
| Death | | | | | | |
| No | 71,393 (66.75) | 36,295 (65.43) | 14,851 (65.74) | 14,998 (71.95) | 5,249 (65.13) | <0.0001 |
| Yes | 35,569 (33.25) | 19,173 (34.57) | 7,738 (34.26) | 5,848 (28.05) | 2,810 (34.87) | |
In the unadjusted model, Hispanics had lower odds of dying [OR 0.74, $95\%$ CI (0.71, 0.76)] while there was no difference in the mortality odds between Blacks and White patients [OR 0.99, $95\%$ CI (0.95, 1.02)]. However, after adjusting for age group, sex, presence of comorbidity, disease severity, county size, and socioeconomic status, Blacks and Hispanics had higher odds of dying [OR of 1.20, $95\%$ CI (1.15, 1.25) and 1.23, $95\%$ CI (1.17, 1.28) respectively] compared with White patients. In the subgroup analysis among residents from neighborhoods with very high vulnerability, Blacks had the highest mortality odds of all the racial groups. In both unadjusted and adjusted analyses, the mortality odds among patients from neighborhoods with moderate and high vulnerability did not differ significantly from mortality odds of those from neighborhoods with low vulnerability. Patients from neighborhoods with very high vulnerability however had the highest mortality odds [Adjusted OR 2.08, $95\%$ CI (1.91, 2.26)]. These results are shown in Table 2A and 2B. The adjusted mortality odds for the other covariates are almost identical between Model 1 (race as primary predictor) and Model 2 (neighborhood type as primary predictor). Results of the fully adjusted model are shown in Table 3. The model assessing for effect modification of the relationship between race and mortality by age group showed that among patients <40 years, Blacks and Hispanics had higher mortality odds than Whites and this trend is consistent in the 40–79-year age group. However, in the 80+ year age group, Whites had the highest mortality odds (see S1 Table).
In the analysis stratified by disease severity, among patients in critical condition, Black and Hispanic patients had higher odds of dying [OR of 1.40, $95\%$ CI (1.32, 1.50) and 1.56, $95\%$ CI (1.46, 1.67) respectively] compared with White patients (Table 4A). In the analysis stratified by time-period, there was no difference in mortality between Black, Hispanic, and White patients between January and June. Between July and November, compared to White patients, Black patients had higher odds of mortality [OR of 1.10, $95\%$ CI (1.02, 1.19)] and these findings were driven by disparities in the Midwest. There was however no difference in the overall mortality odds between Hispanic and White patients [OR of 1.04, $95\%$ CI (0.94, 1.14)] during this period. These results are shown in Table 4B and 4C.
**Table 4**
| (a) | (a).1 | (a).2 | (a).3 | (a).4 |
| --- | --- | --- | --- | --- |
| Reference is White | Non-Critical Unadjusted OR | Non-Critical *Adjusted OR, | Critical Unadjusted OR | Critical *Adjusted OR, |
| Reference is White | (95% CI) | (95% CI * ) | (95% CI) | (95% CI * ) |
| Black | 0.73 (0.69, 0.76) | 1.09 (1.03, 1.15) | 1.35 (1.28, 1.43) | 1.40 (1.32, 1.50) |
| Hispanic/Latino | 0.41 (0.39, 0.43) | 0.97 (0.91, 1.04) | 1.10 (1.04, 1.16) | 1.56 (1.46, 1.67) |
| Other races | 0.71 (0.66, 0.76) | 1.09 (1.01, 1.19) | 1.22 (1.13, 1.32) | 1.44 (1.31, 1.57) |
| (b) | (b) | (b) | (b) | (b) |
| Reference is White | January to June Unadjusted OR | January to June *Adjusted OR, | July to November Unadjusted OR | July to November *Adjusted OR, |
| Reference is White | (95% CI) | (95% CI * ) | (95% CI) | (95% CI * ) |
| Black | 0.86 (0.82, 0.89) | 0.98 (0.93, 1.03) | 0.80 (0.75, 0.85) | 1.10 (1.02, 1.19) |
| Hispanic/Latino | 0.58 (0.56, 0.61) | 0.97 (0.92, 1.03) | 0.57 (0.53, 0.62) | 1.04 (0.94, 1.14) |
| Other races | 0.92 (0.87, 0.98) | 1.10 (1.02, 1.18) | 0.68 (0.61, 0.75) | 0.92 (0.81, 1.04) |
| (c) | (c) | (c) | (c) | (c) |
| Racial Group (White is reference category) | Northeast OR * | Midwest OR * | South OR*, | West, OR * |
| Racial Group (White is reference category) | (95% CI) | (95% CI) | (95% CI) | (95% CI) |
| Black | 1.26 (0.93, 1.70) | 1.19 (1.05, 1.35) | 1.05 (0.95, 1.18) | 0.98 (0.74, 1.30) |
| Hispanic/Latino | 0.74 (0.55, 0.99) | 1.22 (1.02, 1.45) | 0.95 (0.77, 1.17) | 0.95 (0.79, 1.13) |
| Other races | 1.28 (0.93, 1.76) | 0.74 (0.59, 0.91) | 0.97 (0.70, 1.34) | 0.90 (0.72, 1.13) |
In the Midwest, Blacks [OR 1.27, $95\%$ CI (1.18, 1.37)] and Hispanics [OR 1.99, $95\%$ CI (1.80, 2.20)] had higher odds of mortality compared with Whites. In the Northeast, South and West, Hispanics had lower mortality odds compared to Whites while there was no difference in mortality odds between Blacks and Whites. In the Northeast and Midwest, patients from neighborhoods with very high vulnerability had the highest mortality odds. In the South, patients from neighborhoods in quartile 3 (high vulnerability) had lower mortality odds than those in quartile 1 (low vulnerability). In the West, most patients, $\frac{8995}{9636}$ ($93.4\%$) are from neighborhoods with very high vulnerability (quartile 4) and as a result, there are very wide confidence intervals around the estimates. These results are summarized in Table 5. The neighborhood distribution of patients in the *West is* shown in the S5 Table.
**Table 5**
| Racial Group (White is reference category) | Northeast OR | Midwest OR | South OR, | West, OR |
| --- | --- | --- | --- | --- |
| Racial Group (White is reference category) | (95% CI) | (95% CI) | (95% CI) | (95% CI) |
| Black | 1.06 (0.99, 1.14) | 1.27 (1.18, 1.37) | 1.07 (0.97, 1.16) | 0.97 (0.79, 1.18) |
| Hispanic/Latino | 0.80 (0.74, 0.85) | 1.99 (1.80, 2.20) | 0.76 (0.66, 0.89) | 0.82 (0.72, 0.94) |
| Other races | 1.15 (1.05, 1.26) | 0.98 (0.87, 1.11) | 0.94 (0.75, 1.20) | 0.83 (0.70, 0.97) |
| Neighborhood type (reference is low vulnerability) | | | | |
| Quartile 2 (moderate vulnerability) | 2.37 (1.66, 3.38) | 0.89 (0.79, 0.99) | 0.96 (0.78, 1.17) | 2.30 (0.17, 31.82) |
| Quartile 3 (high vulnerability) | 2.81 (2.03, 3.89) | 0.85 (0.76, 0.95) | 0.82 (0.69, 0.98) | 1.25 (0.10, 15.45) |
| Quartile 4 (very high vulnerability) | 4.79 (3.48, 6.61) | 2.85 (2.55, 3.19) | 0.96 (0.81, 1.14) | 0.96 (0.08, 11.88) |
## Discussion
The association between race and neighborhood with COVID-19 mortality is complex and nuanced. Although there are differences overall, they are driven by disparities in time-period and specific geographical regions. When adjusted for age group, sex, presence of comorbidity, disease severity, socioeconomic status, and county size, hospitalized Blacks and Hispanics were more likely to die compared with Whites, but these differences are driven by disparities in the Midwest and the July to November time-period. We also found that patients from high vulnerability neighborhoods (those with the highest proportion of racial minorities and individuals with limited English proficiency) had the highest odds of mortality among hospitalized COVID-19 patients, though these differences were limited to the Northeast and Midwest.
Overall, disparities in COVID-19 in-hospital mortality by race were driven by differences in the Midwest–there were no differences in the West, South, Northeast. Additionally, between January and June, there were no disparities in COVID-19 mortality between Whites, Blacks and Hispanics. This is consistent with findings from retrospective cohort studies conducted in Louisiana, Georgia (both Southern states) and California (western state) during the January to June time-period which found no difference in mortality rates between Blacks and Whites among hospitalized COVID-19 patients [11, 30, 31]. It contrasts, however, with the findings of the study conducted in New York by Ogedegbe et al. which found that Blacks were less likely than Whites to die from COVID-19 among hospitalized patients [12]. Their study was conducted using information from patients in a single health system which mainly included patients from Manhattan, Brooklyn, Queens, and Long Island and may not completely represent the characteristics of the entire New York population [12].
Several studies and reports in the Mid-West during periods of ICU bed shortage documented a higher overall mortality from COVID-19 among Blacks and Hispanics compared to Whites [32–34]. The finding of a higher overall mortality rate in Blacks [9, 32] but comparable in-hospital mortality rates between Blacks and Whites across multiple studies [11–13], suggest that factors related to care access contribute to the racial disparities seen in COVID-19 mortality. Reduced access to care is a multifaceted problem and could be due to underinsurance, geographic disparities in hospital location leading long transportation and emergency room wait times, all of which affect healthcare seeking behavior [35]. In addition, distrust of healthcare professionals and perceived racial discrimination within the healthcare system has been shown to significantly affect the healthcare seeking behavior of racial minorities [36].
Our study goes further to show that racial disparities in mortality exist even among hospitalized COVID-19 patients in the Midwest. This suggests that other in-hospital factors may contribute to the observed racial disparities in mortality. Previous studies conducted in other emergency settings have identified factors including physician race/ethnic case mix and implicit bias among healthcare workers as responsible for racial disparities in health outcomes [37, 38]. In multiple emergency settings physician treatment recommendations have been found to vary by race. A study conducted among patients with acute coronary syndrome found that Black patients were less likely than Whites to be referred for cardiac catheterization [39]. These disparities are worsened during periods of hospital overcrowding and bed shortage (as occurred in the Midwest during the early months of the pandemic), resulting in racial minorities receiving poorer quality care [40]. The importance of race as a predictor of in-hospital COVID-19 mortality is made more compelling by the finding that Blacks had a higher likelihood of death compared to Whites even within the very high vulnerability neighborhoods, and after stratification by disease severity. Although Hispanics in non-critical condition were just as likely as to die as Whites, Hispanics in critical condition were much more likely to die compared to Whites. As patients in critical condition are more likely to require a higher level of care, these findings further highlight the racial inequities in care that occur within the hospital.
Race is a complex concept in public health, and its categories may be too broad for appropriate nuance. Although there have been arguments that higher comorbidity levels are responsible for the higher level of mortality seen among Hispanics and Blacks, Qeadan et al. found in their study, which was stratified by comorbidity index, that mortality in Blacks was consistently higher than that of Whites. Hispanics however had a lower risk of mortality [41]. This is demonstrated again in our study, where compared to Whites, Hispanics had significantly lower mortality odds in the Northeast, South and West. As early as the 1980’s, researchers found that the health status of Hispanics in the southwestern states of the U.S. was closer to the health status of Whites than that of Blacks [42]. Studies on the ‘Hispanic paradox’ have even found better health outcomes among Hispanics compared to their White counterparts in the U.S. [43–45]. The yet unmeasured factors accounting for these better outcomes may explain why Hispanics in the South and West have lower odds of mortality than the Blacks in our study. The Hispanic population is also not monolithic and health outcomes vary between different Hispanic subpopulations [46]. The different compositions Hispanic populations across the U.S. may in part explain the difference in COVID-19 mortality observed in our study between Hispanics in the Midwest and those in the other census regions.
Disparities in COVID-19 in-hospital mortality by neighborhood vulnerability were driven by differences in the Midwest and Northeast–there were no differences in the West and South. These disparities may be due to inequities in healthcare resource allocation between hospitals in different neighborhoods; inequities which were worsened by the hospital overcrowding and ICU bed shortage that occurred disproportionately in the Northeast and Midwest during the peak of the pandemic [3, 6, 10]. The optimum hospital bed occupancy rate is estimated to be between $80\%$-$85\%$ with discernible mortality risks above these rates [47, 48]. Data from the American Hospital Association showed hospital bed occupancy rates of over $90\%$ in the Northeast and Midwest during our study period, but not for the South or West [49]. In New York City, Manhattan which has the highest proportion of White residents and the most equipped medical centers recorded the lowest number of COVID-19 deaths [10, 50]. These hospitals were properly staffed and had access to experimental drugs like Remdesivir and life-saving devices like heart-lung bypass machines [50]. Conversely, the Bronx which has the highest proportion of minority residents and less equipped hospitals recorded the highest number of COVID-19 deaths per 100,000 population [10]. The understaffing and inadequate access to high-technology diagnostic and therapeutic procedures in these hospitals were exacerbated by the increased demands of the pandemic resulting in disproportionately higher mortality [51]. A recent report suggests that increased access to novel therapeutics for patients with limited English proficiency can help close the language-based disparity gaps in patient outcomes within the acute care setting [52].
We found an overall mortality rate of $33\%$ among hospitalized COVID-19 patients. The mortality rate in our study is higher than the $20.3\%$ mortality rate reported in the multi-state retrospective study conducted by Yehia et al. across 92 hospitals in the U.S. [13]. Their study however collected data from only 2 hospitals in New York where mortality from COVID-19 was the highest during the February to May study period and thus likely underestimated the overall mortality rate. The authors also reported no difference in mortality by race which contrasts with our study’s findings. The 92 hospitals in their study were however located in only 12 states and over a third of them were from the South alone (where we found no difference in mortality between Blacks and Whites).
Our study has limitations. [ 1] Like all studies conducted using secondary data, our ability to adjust for confounding depends on the number of variables in our dataset and the results of our analysis are subject to the accuracy of the information provided by the different reporting authorities. For example, we were unable to adjust for hospital and provider characteristics in our analyses. As such, residual confounding cannot be eliminated. [ 2] *Socioeconomic status* was measured at zip-code level and may not reflect individual factors. [ 3] The CDC COVID-19 Case Surveillance database contained varying amounts of missing data, some of which could not be assumed to be missing completely at random. Missing at random is an assumption (MAR) and missing not at random (MNAR) cannot be ruled out empirically [53]. Analysis of data that are MNAR do not however guarantee that the study estimates will be biased; it only implies that we cannot correct for bias if present [53]. Studies have also shown that even when the exposure and/or confounders are MNAR, complete case analysis is a valid approach [54]. We however used the multiple imputation with chained equations (MICE) algorithm on our data and conducted analyses on the imputed datasets to test the robustness of the results from our complete case analyses (see S7 Table). For our primary objective, the results from the imputation model are similar (albeit attenuated) to those from the complete case analyses.
The biggest strength of our study is its large sample size which reduces the variability in our effect estimates. The data used in this study was collected from all states and territories in the United States. Hence, despite its limitations, our study provides important epidemiological data on Blacks and Hispanics in the context of the COVID-19 pandemic.
## Conclusion
Our study findings show that among hospitalized COVID-19 patients in the United States, Blacks and Hispanics have an overall higher odd of mortality in the Midwest, and residents of neighborhoods with the highest proportion of racial minorities and individuals with limited English proficiency have higher odds of mortality in the Northeast and Midwest. These results suggest that efforts to curb healthcare disparities, eliminate structural racism and reduce inequity in resource allocation in the healthcare setting have largely been unsuccessful. The timing of these findings in the middle of a global pandemic presents a unique opportunity to address these issues.
## Source of the data
Centers for Disease Control and Prevention (CDC), COVID-19 Response. COVID-19 Case Surveillance Data Access, Summary, and Limitations (version date: December 4, 2020).
The National Center for Health Statistics (NCHS) Urban-Rural Classification Scheme (URCS)- downloaded March 23, 2021.
The CDC Social Vulnerability Index (SVI)- downloaded April 23, 2021
## References
1. Rothan H, Byrareddy S. **The epidemiology and pathogenesis of coronavirus disease (COVID-19) outbreak.**. (2020.0) **109** 102433. DOI: 10.1016/j.jaut.2020.102433
2. Allen J, Almukhtar S, Aufrichtig A. **Coronavirus in the U.S.: Latest Map and Case Count**. *The New York Times* (2021.0)
3. Mahajan U., Larkins-Pettigrew M.. **Racial demographics and COVID-19 confirmed cases and deaths: A correlational analysis of 2886 US counties**. (2020.0) **42** 445-447. DOI: 10.1093/pubmed/fdaa070
4. Hanson A, Hains D, Schwaderer A, Starr MC. **Variation in COVID-19 Diagnosis by Zip Code and Race and Ethnicity in Indiana.**. (2020.0) **8**. DOI: 10.3389/fpubh.2020.593861
5. Lighter J, Philips M, Hochman S, Sterling S, Johnson D, Francois F. **Obesity in Patients Younger Than 60 Years Is a Risk Factor for COVID-19 Hospital Admission**. (2020.0) **71** 895-896. DOI: 10.1093/cid/ciaa409
6. DiMaggio C, Klein M, Berry C, Frangos S. **Black/African American Communities are at highest risk of COVID-19: spatial modeling of New York City ZIP Code–level testing results.**. (2020.0) **51** 7-13. DOI: 10.1016/j.annepidem.2020.08.012
7. Millett G, Jones A, Benkeser D, Baral S, Mercer L, Beyrer C. **Assessing differential impacts of COVID-19 on black communities.**. (2020.0) **47** 37-44. DOI: 10.1016/j.annepidem.2020.05.003
8. Rentsch C, Kidwai-Khan F, Tate J, Park L, King J, Skanderson M. **Patterns of COVID-19 testing and mortality by race and ethnicity among United States veterans: A nationwide cohort study.**. (2020.0) **17** 1-17. DOI: 10.1371/journal.pmed.1003379
9. Yancy CW. **COVID-19 and African Americans.**. (2020.0) **323** 1891-1892. DOI: 10.1001/jama.2020.6548
10. Wadhera R, Wadhera P, Gaba P, Figueroa J, Maddox K, Yeh R. **Variation in COVID-19 Hospitalizations and Deaths Across New York City Boroughs**. (2020.0) **323** 2192-2195. DOI: 10.1001/jama.2020.7197
11. Price-Haywood E, Burton J, Fort D, Seoane L. **Hospitalization and Mortality among Black Patients and White Patients with Covid-19**. (2020.0) **382** 2534-2543. DOI: 10.1056/NEJMsa2011686
12. Ogedegbe G, Ravenell J, Adhikari S, Butler M, Cook T, Francois F. **Assessment of Racial / Ethnic Disparities in Hospitalization and Mortality in Patients With COVID-19 in New York City.**. (2020.0) **3** 1-14. DOI: 10.1001/jamanetworkopen.2020.26881
13. Yehia B, Winegar A, Fogel R, Fakih M, Ottenbacher A, Jesser C. **Association of Race With Mortality Among Patients Hospitalized With Coronavirus Disease 2019 (COVID-19) at 92 US Hospitals.**. (2020.0) **3** e2018039. DOI: 10.1001/jamanetworkopen.2020.18039
14. Salihu H, Dongarwar D, Ikedionwu C, Shelton A, Jenkins C, Onyenaka C. **Racial/Ethnic Disparities in the Burden of HIV/Cervical Cancer Comorbidity and Related In-hospital Mortality in the USA.**. (2021.0) **8**. DOI: 10.1007/s40615-020-00751-5
15. Nguyen G, Patel A. **Racial Disparities in Mortality in Patients Undergoing Bariatric Surgery in the USA.**. (2013.0) **23**. DOI: 10.1007/s11695-013-0957-4
16. 16Office for Human Research Protections. Federal Policy for the Protection of Human Subjects (’Common Rule’). United States Department of Health and Human Services. Published 2016. Accessed July 4, 2021. https://www.hhs.gov/ohrp/regulations-and-policy/regulations/common-rule/index.html
17. 17Centers for Disease Control and Prevention. COVID-19 Case Surveillance Restricted Use Data. Published 2021. Accessed March 8, 2021. https://data.cdc.gov/Case-Surveillance/COVID-19-Case-Surveillance-Restricted-Access-Detai/mbd7-r32t/
18. 18Centers for Disease Control and Prevention. Human Infection with 2019 Novel Coronavirus Case Report Form. Published 2021. Accessed June 8, 2021. https://www.cdc.gov/coronavirus/2019-ncov/downloads/pui-form.pdf
19. 19Centers for Disease Control and Prevention. Coronavirus Disease 2019 (COVID-19) 2020 Interim Case Definition, Approved August 5, 2020. Published 2021. Accessed January 26, 2021. https://wwwn.cdc.gov/nndss/conditions/coronavirus-disease-2019-covid-19/case-definition/2020/08/05/
20. 20Centers for Disease Control and Prevention. NCHS Urban-Rural Classification Scheme for Counties. Published 2017. Accessed April 6, 2021. https://www.cdc.gov/nchs/data_access/urban_rural.htm#Data_Files_and_Documentation
21. 21Centers for Disease Control and Prevention, Agency for Toxic Substances and Disease Registry, Geospatial Research Analysis and Services Program. CDC SVI 2018 Documentation: 1/31/2020. Published online 2020:1–29. https://www.atsdr.cdc.gov/placeandhealth/svi/documentation/pdf/SVI2018Documentation-H.pdf
22. 22U.S. Census Bureau. Census Bureau Regions and Divisions with State FIPS Codes. Accessed March 17, 2021. https://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf
23. Golestaneh L, Neugarten J, Fisher M, Billett H, Gil M, Johns T. **The association of race and COVID-19 mortality.**. (2020.0) **25** 100455. DOI: 10.1016/j.eclinm.2020.100455
24. Karim S, Chen H. **Deaths From COVID-19 in Rural, Micropolitan, and Metropolitan Areas: A County-Level Comparison.**. (2021.0) **37** 124-132. DOI: 10.1111/jrh.12533
25. Mohamed M, Gale C, Kontopantelis E, Doran T, Belder M, Asaria M. **Sex Differences in Mortality Rates and Underlying Conditions for COVID-19 Deaths in England and Wales**. (2020.0) **95** 2110-2124. DOI: 10.1016/j.mayocp.2020.07.009
26. Imam Z, Odish F, Gill I, O’Connor D, Armstrong J. **Older age and comorbidity are independent mortality predictors in a large cohort of 1305 COVID-19 patients in Michigan, United States.**. DOI: 10.1111/joim.13119
27. 27WHO. Clinical Management of COVID-19: Interim Guidance. 2016;4(1):64–75.. *Clinical Management of COVID-19: Interim Guidance* (2016.0) **4** 64-75
28. Wu Z, McGoogan J. **Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China.**. (2020.0) **323**. DOI: 10.1001/jama.2020.2648
29. Harrison S, Fazio-Eynullayeva E, Lane D, Underhill P, Lip G. **Comorbidities associated with mortality in 31,461 adults with COVID-19 in the United States: A federated electronic medical record analysis.**. (2020.0) **17** 1-11. DOI: 10.1371/journal.pmed.1003321
30. Gold J, Wong K, Szablewski C, Patel P, Rossow J, da Silva J. **Characteristics and Clinical Outcomes of Adult Patients Hospitalized with COVID-19—Georgia, March 2020**. (2020.0) **69** 545-550. DOI: 10.15585/mmwr.mm6918e1
31. Azar K, Shen Z, Romanelli R, Lockhart S, Smits K, Robinson S. **Disparities in outcomes among COVID-19 patients in a large health care system in California.**. (2020.0) **39** 1253-1262. DOI: 10.1377/hlthaff.2020.00598
32. Reyes C, Husain N, Gutowski C, st. Clair S, Pratt G. *Chicago’s coronavirus disparity: Black Chicagoans are dying at nearly six times the rate of white residents, data show.* (2020.0)
33. Schencker L, Heinzmann D. **As COVID-19 cases increase, intensive care beds are filling up at Chicago-area hospitals.**. (2020.0)
34. Bryan MS, Sun J, Jagai J, Horton D, Montgomery A, Sargis R. **Coronavirus disease 2019 (COVID-19) mortality and neighborhood characteristics in Chicago.**. (2021.0) **56** 47-54. DOI: 10.1016/j.annepidem.2020.10.011
35. 35National Academies of Sciences Engineering and Medicine; Health and Medicine Division; Board on Healthcare services; Committee on Health Care Utilization and Adults with Disabilities. Health-Care Utilization as a Proxy in Disability Determination.
National Academies Press; 2018. doi: 10.17226/24969. (2018.0). DOI: 10.17226/24969
36. Eley N, Namey E, McKenna K, Johnson A, Guest G. **Beyond the Individual: Social and Cultural Influences on the Health-Seeking Behaviors of African American Men.**. (2019.0) **13**. DOI: 10.1177/1557988319829953
37. Udyavar N, Salim A, Cornwell E, Cooper Z, Haider A. **Do outcomes in emergency general surgery vary for minority patients based on surgeons’ racial/ethnic case mix?**. (2019.0) **218** 42-46. DOI: 10.1016/j.amjsurg.2019.01.001
38. Chapman E, Kaatz A, Carnes M. **Physicians and implicit bias: How doctors may unwittingly perpetuate health care disparities**. (2013.0) **28** 1504-1510. DOI: 10.1007/s11606-013-2441-1
39. Schulman KA, Berlin JA, Harless W, Kerner J, Sistrunk S, Gersh B. **The Effect of Race and Sex on Physicians’ Recommendations for Cardiac Catheterization.**. (1999.0) **340**. DOI: 10.1056/NEJM199902253400806
40. Hsia R, Sarkar N, Shen Y. **Impact Of Ambulance Diversion: Black Patients With Acute Myocardial Infarction Had Higher Mortality Than Whites.**. (2017.0) **36**. DOI: 10.1377/hlthaff.2016.0925
41. Qeadan F, VanSant-Webb E, Tingey B, Rogers T, Brooks E, Mensah N. **Racial disparities in COVID-19 outcomes exist despite comparable Elixhauser comorbidity indices between Blacks, Hispanics, Native Americans, and Whites.**. (2021.0) **11** 8738. DOI: 10.1038/s41598-021-88308-2
42. Markides K, Coreil J. **The health of Hispanics in the southwestern United States: an epidemiologic paradox**. (1986.0) **101** 253-265. PMID: 3086917
43. Lancet The. **The Hispanic paradox.**. (2015.0) **385**. DOI: 10.1016/S0140-6736(15)60945-X
44. Lariscy J, Nau C, Firebaugh G, Hummer R. **Hispanic-White Differences in Lifespan Variability in the United States.**. (2016.0) **53**. DOI: 10.1007/s13524-015-0450-x
45. Abraído-Lanza A, Dohrenwend B, Ng-Mak D, Turner J. **The Latino mortality paradox: a test of the “salmon bias” and healthy migrant hypotheses.**. (1999.0) **89**. DOI: 10.2105/ajph.89.10.1543
46. Maurer L, Rahman S, Perez N, Allar B, Witt E, Moya J. **Differences in outcomes after emergency general surgery between Hispanic subgroups in the New Jersey State Inpatient Database (2009–2014): The Hispanic population is not monolithic.**. (2021.0). DOI: 10.1016/j.amjsurg.2021.03.057
47. Bagust A, Place M, Posnett J. **Dynamics of bed use in accommodating emergency admissions: stochastic simulation model**. (1999.0) **319**. DOI: 10.1136/bmj.319.7203.155
48. Ravaghi H, Alidoost S, Mannion R, Bélorgeot V. **Models and methods for determining the optimal number of beds in hospitals and regions: A systematic scoping review.**. (2020.0) **20** 1-13. DOI: 10.1186/s12913-020-5023-z
49. 49American Hospital Association. COVID-19 Bed Occupancy Projection Tool. Published 2021. Accessed April 7, 2021. https://metricvu.aha.org/dashboard/covid-bed-shortage-detection-tool?Mean%2FLower%2FUpper=Standard
50. Rosenthal B, Goldstein J, Otterman S, Fink S. **Why Surviving the Virus Might Come Down to Which Hospital Admits You**. *The New York Times* (2020.0)
51. Madsen F, Ladelund S, Linneberg A. **High levels of bed occupancy associated with increased inpatient and thirty-day hospital mortality in Denmark.**. (2014.0) **33** 1236-1244. DOI: 10.1377/hlthaff.2013.1303
52. 52Mass General Brigham.
Non-English–Speaking Patients Hospitalized with COVID-19 Are at Risk of Worse Health Outcomes.; 2022. https://www.massgeneralbrigham.org/newsroom/press-releases/non-english-speaking-patients-hospitalized-covid-19-are-risk-worse-health. (2022.0)
53. Perkins N, Cole S, Harel O, Tchetgen E, Sun B, Mitchell E. **Principled Approaches to Missing Data in Epidemiologic Studies**. (2018.0) **187**. DOI: 10.1093/aje/kwx348
54. Hughes R, Heron J, Sterne J, Tilling K. **Accounting for missing data in statistical analyses: multiple imputation is not always the answer**. (2019.0) **48**. DOI: 10.1093/ije/dyz032
|
---
title: Qualitative assessment of family caregiver-centered neonatal education program
in Karnataka, India
authors:
- Shirley D. Yan
- Sahana S.D.
- Meghna Desai
- Megan Marx Delaney
- Lauren Bobanski
- N. Rajkumar
- Seema Murthy
- Natalie Henrich
journal: PLOS Global Public Health
year: 2023
pmcid: PMC10022017
doi: 10.1371/journal.pgph.0000524
license: CC BY 4.0
---
# Qualitative assessment of family caregiver-centered neonatal education program in Karnataka, India
## Abstract
Globally 2.5 million newborns die every year before they reach the age of one month; the majority of these deaths occur in low- and middle-income countries. Among other factors, inadequate knowledge and skills to take care of newborns contribute to these deaths. To fill this gap, training patients and family members on the behaviors needed to improve essential newborn care practices at home is a promising opportunity. One program that aims to do this is the Care Companion Program (CCP) which provides in-hospital, skills-based training on care of mothers and newborns to families. This study uses semi-structured interviews to understand how and why knowledge and behaviors of maternal and newborn care behaviors change (or don’t change) as a result of CCP sessions and participants’ perception of the impact of CCP on change. Interviews focused on knowledge and behaviors around key neonatal and newborn topics and health seeking behaviors for health complications. Forty-two in-depth interviews were conducted among families with recently-delivered babies at their homes from four districts in Karnataka, India. Respondents have a positive perception about CCP, found training useful and appreciated other family members presence during the training. CCP increased knowledge and awareness and provided critical details to key behaviors like breastfeeding. Respondents were more likely to be receptive toward details on already known topics, like hand washing before touching the baby. Awareness increased on newly learned behaviors, like skin-to-skin care, which don’t conflict with cultural norms. The CCP did not influence nonrestrictive maternal diet as much, which cultural norms heavily influence. In-hospital family caregiver education programs, like CCP, can positively influence key neonatal behaviors by imparting knowledge and key skills. However, the effect is not universal across health behaviors.
## Background
Currently 2.5 million deaths occur in the first month of neonatal life worldwide [1]. The majority ($90\%$) of these deaths occur in LMICs due to infections, complications of preterm birth, and intrapartum related complications [2, 3]. In India, about four-fifths of these deaths occur within the first week of delivery of the more than 750,000 neonates die annually in India [4–6]. Causes of death in the late newborn period include bacterial infections, pneumonia and infection of the umbilical cord stump, which hand washing, exclusive breastfeeding and dry umbilical cord care can help prevent, respectively [7]. Several key behaviors, such as exclusive breastfeeding, skin-to-skin contact, dry umbilical cord care, unrestrictive maternal diet, and hand washing with soap, can reduce health complications and adverse health outcomes [8–12]. Initiation of early and exclusive breastfeeding in the first six months of neonatal life can prevent $20\%$ of newborn deaths especially due to sepsis, pneumonia, tetanus, and diarrhea [8]. Additionally, maintaining dry cord care and maintaining hand hygiene can prevent neonatal deaths by reducing infection [9, 10]. During the neonatal period (first 28 days of life) and after, practice of key health behaviors are critical.
Globally, there is value to address knowledge gaps around postnatal behaviors. Though it is well documented in the literature that the previously mentioned key behaviors contribute to the prevention of poor neonatal and maternal health outcomes post-delivery, knowledge and practice of these behaviors vary widely. A Nepal based study found that while postnatal mothers knew about early breastfeeding, many of them had less knowledge on hand washing and recognition of danger signs [11]. In a cross sectional study in Nepal among recently delivered mothers, $48.7\%$ had inadequate knowledge and $33.8\%$ inadequate behavior for newborn care [13]. In a qualitative study in Uganda among mothers of low-birthweight babies, danger sign recognition, initiation and exclusivity of breastfeeding, and maintenance of these behaviors could be improved [14]. Maternal and neonatal care practices have been shown to vary across India but are largely influenced by religion, cultural practices beliefs, socioeconomic status, and decision of family members [15, 16]. In Karnataka, a state located in Southern India, persisting cultural beliefs and practices have influenced families to withhold nutritious foods for mothers post-delivery, apply various substances to the umbilical cord stump and eyes of the newborn, and give prelacteal feeds instead of exclusive breastfeeding [17]. Pre-discharge education on post-discharge behaviors must include the whole family to practice newborn care to reduce the risks of neonatal mortality [18, 19].
Behavior change communication has shown demonstrable impact on health behavior change and improved health outcomes. In an intervention in which frontline health workers provided counseling, posters, pamphlets, and demonstration on postnatal care, the intervention group reported reduction of potentially harmful behaviors, including withholding nutritious foods considered "hot" or "cold"; application of various substances to the umbilical stump and eyes of the newborn and giving prelacteal feeds; significant improvement in breastfeeding; and increase in knowledge on skin to skin care, identification, and management of danger signs in babies [20]. Evidence from a behavior change communication delivered through rural Indian self-help groups showed a 5–11 percentage point increase for positive antenatal, natal, and postnatal behaviors (e.g. breastfeeding, skin-to-skin care, and cord care) [21]. In addition to community based programs, hospital based programs focused on delivering neonatal information to parents have positively changed health behaviors and resulted in improved maternal outcomes (e.g. knowledge, self-efficacy, behavior, feeling supported, and attachment) and newborn health (e.g. practice of preventive behaviors, immunization, morbidity, and mortality) outcomes primarily due to uptake of knowledge for exclusive breastfeeding [19]. Based on a cross sectional analysis of 13,730 families who recently delivered in three Indian states, it is clear that newborn care topics are taught within facilities, but consistency on coverage across topics differ; for example breastfeeding is mentioned the most ($26.2\%$ of participants) and dry umbilical cord care the least ($0.3\%$) [18].
While parent and family caregiver education is promoted throughout the literature, it remains unclear the impact of this intervention on the knowledge and behaviors of trained families [19, 22]. This qualitative study aims to understand how trained families experience one such neonatal education program focused on mothers and family caregivers: the neonatal Care Companion Program (CCP). Additionally, this study aims to assess how and why knowledge and behaviors of maternal and newborn care behaviors changes as a result of the CCP sessions, and participants’ perception of the impact of CCP on their change.
## Program description: Noora health’s Care Companion Program
Noora Health along with its associate organizations supports implementation of Neonatal Care Companion Program (CCP), a family caregiver education program that teaches families knowledge and skills pre-discharge, with the aim to reduce complications and hospital readmissions. With technical support from Noora Health and its affiliate partners, district hospital nurses and counselors teach neonatal topics pre-discharge to new mothers and their family members in wards and waiting rooms postpartum before hospital discharge about caring for themselves and their newborns. The postnatal Care Companion Program (CCP) sessions teach families key preventive behaviors (non-restrictive diet for mothers, exclusive breastfeeding, skin-to-skin care, dry umbilical cord care, hand hygiene, danger sign recognition for the mother and baby, and danger sign recognition for the baby and recently delivered mother) which is implemented at district level hospitals in six states of India (Punjab, Karnataka, Madhya Pradesh, Maharashtra, Telangana, Andhra Pradesh) (Table 1). As of April 2022, CCP programs have been implemented in 321 hospitals and health facilities across India in maternal and child health, oncology, cardiology, and inpatient condition areas.
**Table 1**
| Behavior | Definition | CCP recommended behavior and how they are taught |
| --- | --- | --- |
| Breastfeeding | Exclusive breastfeeding for the first six months of life, unless prescribed by doctor | • Exclusive breastfeeding newborns up to six months and only feeding formula if recommended by a doctor• No complementary feeding with cow’s milk, honey, gripe water (typically given to infants who are colicky or have other gastrointestinal issues), sugar water and any other things• Complete mother’s diet to help with breast milk production• In CCP session, trainer teaches behavior importance, how to troubleshoot, and taught with doll to demonstrate baby positioning |
| Dry Cord Care | No application of any substance on the baby’s umbilical cord care until it falls off, unless prescribed by a doctor | • No touching the cord unnecessarily, no pulling off the cord, and no applying anything (powder/oil/cream) to the cord before it falls off or it fully heals• The cord area should be kept dry and clean at all times and the cord will fall off by itself in 1–2 weeks• In CCP session, trainer teaches warning sign recognition and health-seeking behavior |
| Postpartum maternal Diet | Non-restrictive diet for food or liquids for postpartum mothers | • Maternal diet should include vegetables, fruits, and dairy, and meats if the mother is non-vegetarian• No restrictions of any food nor the quantity of food. In particular, "cold" foods do not have to be avoided and drink at least 8–10 glasses (2–3 liters) of water every day• In CCP session, trainer teaches what all mothers can eat postpartum |
| Hand Washing | Hand washing with water and soap before touching or feeding baby, after going to the bathroom, and before eating | • Wash hands with soap for 30 seconds including the front and back of their hands, between fingers, and under nails.• Wash hands before touching or feeding the baby and after going to the bathroom.• In CCP session, trainer teaches handwashing importance and demonstration |
| Skin-to-skin contact | Performs skin-to-skin contact with baby when baby is cold, for low-birthweight babies, and for general bonding | • Skin-to-skin care can be performed by putting the bare baby (except the diaper, cap and socks) on the bare chest of the father/mother to provide steady warmth. CCP promotes skin-to-skin care for all babies and not just underweight or preterm babies.• It should be practiced regularly, not only if you think the baby is cold• In CCP session, trainer teaches importance and demonstrates how to perform skin-to-skin contact |
| Warning sign identification | Identify signs of diarrheal disease, jaundice, signs of infection, hypothermia | • Recognition of danger signs in newborns (such as fever and hypothermia, lethargy, poor feeding, and blood in stools) and how to manage symptoms• In CCP session, trainer teaches warning signs to look out for |
| Care-seeking behavior | Upon recognition of warning signs, families seek medical care at hospitals. | • Seek medical attention upon recognition of danger signs in newborns• In CCP session, trainer teaches when to see care at hospital |
Master trainers (nurses or counsellors) generally facilitate sessions in hospital hallways or wards. Noora Health supports these master trainers through training of trainers (two-three days), booster trainings (one-two days) and ongoing coaching which emphasize soft skills and tools master trainers can use to facilitate a session. An ideal CCP session involves use of flipcharts (S1 Text), visuals, and supporting text for the trainers to talk through the topics over 20–30 minutes (see flipchart example in the Annexure). Props such as a doll are used to demonstrate key skills, such as positioning for breastfeeding, how to perform skin to skin contact, and warning sign identification for jaundice. If there are televisions in the hospital, videos are used as part of the session to teach key preventive behaviors (i.e. dry cord care, skin-to-skin contact, exclusive breastfeeding, hand washing, and postpartum maternal diet; the same content that trainers cover). During a session, trainers should cover all topics, introduce the Whatsapp-based service, and encourage participants to ask questions as necessary. The Whatsapp service delivers information and behavior reinforcement messages and videos, covering the same content delivered in the CCP session. To enroll, the trainer asks the families to give a missed call to the postnatal Whatsapp number, which will then register the phone number into the service. At the end of the session, trainers distribute paper handouts to families that highlight the main topics taught in CCP. Full list of tools are shown in the annexure. Classes are conducted in groups, which has advantages over one-on-one counseling sessions that health professionals typically conduct with patients at their bedside [23–25]. These tools and overall approach for the sessions were created using a human-centered process and adult learning principles. Medical content is written and reviewed by an internal team of medically trained doctors, following from WHO and Ministry of Health, India literature and guidelines on what behaviors are linked to preventable complications. State Departments of Health and District Surgeons and Medical Officers provide final sign off to implement neonatal CCP sessions on a regular basis. The neonatal CCP program is an integrated component of the public health system in Punjab, Karnataka, Madhya Pradesh, Maharashtra, Andhra Pradesh, and Telangana implemented by the public health infrastructure. On average, sessions are run one-three times in a week to maximize the number of participants who receive the training, yet limit the burden on health care staff in delivering the program. Whether CCP sessions are implemented exactly as idealized depends on availability of infrastructure (e.g. televisions, printing challenges for handout distribution), hospital prioritization to run sessions (e.g. allocate time for nurses, hospital leadership’s interest in sessions), and individual nurse factors (e.g. nurse motivation).
## Study setting
This study was set in Karnataka, which had an infant mortality rate of 27 per 1000 live births in 2015–2016 (slightly below the national average of 32.3), down from 43 since 2005–6 [26]. Mothers and family members were recruited to participate in this study if they attended CCP sessions in one of the four district hospitals (large tertiary hospitals) in Chikkaballapur, Ramanagara, Dharwad and Bagalkot, Karnataka and who had completed a short survey 28-day post-discharge survey as part of a separate on-going study covered under a separate IRB [18]. The 15-minute post-discharge survey covered questions on knowledge and behavior of key behaviors; and self-report of complications or readmissions, which will be reported separately. This qualitative research provides insight into participants’ perception of the CCP sessions, how and why families performed behaviors, and explores facilitators and barriers to change. At the time of planning for this qualitative study (July 2019), the implementation and research team looked at the eight study sites from the larger neonatal study and selected facilities based on delivery load in the last month, CCP attendance in the last month, geographic location in Karnataka, number of sessions run each month, and quality of sessions. Recently delivered mothers participated in the survey if they were 18 years or older, delivered within the hospital, and neither the mother nor newborn died before discharge. These participants would have attended sessions within selected district hospitals within January-February 2020.
## Sampling strategy and inclusion criteria
Based on the post-discharge survey responses, study team members purposively recruited interview participants based on whether newborn complications were reported, given this is a main health outcome CCP aims to impact. The goal was to recruit equal numbers of mothers and families with and without reported newborn complications. Only individuals who spoke Kannada and/or Hindi were recruited. Mothers and family members who did not complete the 28-day post discharge survey were not included. The sampling targets were set at 40 participants total (20 family caregivers and 20 recently delivered mothers with half the participants reporting baby complications) due to expected saturation in responses.
## Data collection instruments
The research team developed separate semi-structured interview guides for mothers and family members with questions aimed at understanding how the CCP influenced behaviors, knowledge, and confidence among women and family members. Interview questions covered the following themes: attitudes, beliefs, and norms around newborn and maternal care practices (including perceptions, benefits, risks, challenges, facilitators), and acceptability and feasibility of the CCP. Particular focus was on specific health behaviors (i.e. exclusive breastfeeding, dry umbilical cord care until it falls off, handwashing before feeding and touching the baby, skin-to-skin care on a routine basis and if the baby is cold, unrestricted maternal diet for women, and danger sign recognition during the neonatal period), and experience for women and families who experienced baby complications (see Table 1). Among participants who reported a complication with their newborn, follow up questions around health seeking behavior and follow up actions were explored.
## Researcher team composition
A third-party external agency conducted the semi-structured interviews; four of the interviewers were female and one male. This external agency was trained on the CCP program, interview guide, and interview techniques. Staff had Kannada, Hindi, and English language capacities and came from social science and public health training. After an initial recruitment call to confirm participant interest, the interviewers would call to schedule the interview. Interviewers conducted the interviews within the household where other household members would often observe the interview. Logistically, it was challenging to enforce privacy with the new mother or family caregiver given the newborn needs and the local cultural context. An interviewer and note-taker pair conducted the interview at the participants’ home from February-March 2020 (before the COVID-19 pandemic), audio-recorded, and transcribed and translated into English. The research team read transcripts for quality and de-identified transcripts.
The research tools and protocol were first piloted by members of the research team (who were not involved in eventual data collection), and the interview guide was refined and simplified. After the third-party agency was trained, they piloted the interview guides as well and received interviewing feedback. The research team and external agency iterated on the interview guide during this period, refining language, probing questions, and question flow, based off of initial interview responses (S2, S3 Texts). Results of this qualitative research are reported against Standards Reporting for Qualitative Research guidelines [27]. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the S1 Checklist.
## Ethical review
This study received IRB clearances from Harvard University in Boston, USA and ACE Ethics Committee in Karnataka, India. All participants underwent a written consent process for the interview and audio recording and signed (or provided thumbprints if they couldn’t read) in their local language.
## Analysis process
Research team members, who did not conduct interviews, created a codebook after reviewing five de-identified interviews through a mix of inductive and deductive coding. A doer/non-doer analysis framework informed codebook development to understand why people performed or did not perform a specific behavior [28]. The finalized codebook included themes around barriers and facilitators for key behaviors, health seeking behavior in the event of neonatal complications, and impact from CCP sessions (S4 Text). The same researchers then coded the interviews using Dedoose online software and reviewed each other’s codes [29]. They kept in touch with the interview team as questions arose. The analysis is based on the key behaviors according to the doer/non-doer framework, to understand barriers, facilitators, and context for the key behaviors and associated impact of the CCP program.
## Results
The total sample size for this analysis is 42 respondents (23 mothers and 19 family caregivers), all from different families. Interviews lasted between 20–80 minutes. Table 2 describes the demographics of respondents and the profile of the family’s newborn, collected through the 28-day post discharge survey. Of note, the majority of the family caregivers interviewed were grandmothers to the newborn.
**Table 2**
| Unnamed: 0 | N = 42 |
| --- | --- |
| Respondent Type | Respondent Type |
| Mother | 23 (54.8%) |
| Family Caregiver (e.g. Maternal and paternal grandparents, paternal aunt) | 19 (45.2%) |
| Maternal Characteristics | Maternal Characteristics |
| Maternal Age (mean age and range) | 23.4 (18, 31) |
| Family owns smart phone | 35 (83.3%) |
| Birth Characteristics | Birth Characteristics |
| Singleton birth | 42 (100%) |
| Primiparous | 26 (61.9%) |
| Caesarean section | 26 (61.9%) |
| Baby sex: female | 20 (47.6%) |
| Special Newborn Care Unit admission | 8 (19.0%) |
| Reported Baby Complications post discharge | 23 (52.8%) |
## Overall perceptions of the Neonatal Care Companion Program (CCP)
Our findings show that respondents have a positive perception of the neonatal CCP sessions. More than half of respondents ($$n = 30$$) said they found the training useful (18 caretakers, 12 family members) and 32 respondents said they would recommend the CCP to a friend (18 caretakers, 14 family members). Respondents appreciated the knowledge and skills taught in the class: Mothers ($$n = 19$$) said it was helpful to have family members present during CCP and it made them feel supported.
Twelve respondents that expressed worry or concerns before the birth of their baby (or from the CCP session itself) said that knowledge and skills learned in the CCP helped alleviate some of their worry and concern.
## Influence on knowledge, preventive behaviors, and confidence of newborns’ family caregivers
We analyzed five main, preventive behaviors taught in the CCP: exclusive breastfeeding, handwashing, skin to skin care, dry cord care, and non-restrictive mother postpartum diet. The data indicates that respondents come into the CCP with prior knowledge on newborn care behaviors and receive knowledge from multiple different sources including healthcare providers, family, friends, and media channels. Cultural norms and social beliefs also have an influence on individual newborn care behavior.
## Exclusive breastfeeding
Breastfeeding is not a new behavior to most respondents. They have prior knowledge through their own personal experience (previous children, observing other women in family breastfeed), advice from health care workers (doctors and nurses at the hospital, Anganwadi workers) and advice from family and community members. Several ($$n = 7$$) respondents said that the CCP offered them new knowledge on breastfeeding including positioning, breastfeeding technique, using a pillow to help support the baby and the mother’s back, and burping.
Respondents felt most confident in their ability to breastfeed when they are producing enough milk for their baby and when their baby is consuming enough breastmilk. *In* general, the most commonly mentioned barrier to breastfeeding is a lack of milk production.
All respondents said their babies were fed breastmilk ($$n = 42$$) but several respondents said that they also fed their newborn other foods, indicating suboptimal or nonexclusive breastfeeding. *In* general, respondents that experienced problems with milk production supplemented breast milk with other items (prior to six months of age) including cow’s milk and glucose water. One respondent that experienced difficulty producing breast milk was prescribed powder milk and provided it to her newborn. Respondents said that they breastfed with complementary feeding of animal milk (cow, sheep) ($$n = 5$$) or glucose water ($$n = 1$$) because the mother was not producing enough breast milk. Other items fed to newborns included almond paste ($$n = 1$$), ground dates and almonds mixed in with breast milk “for strength” ($$n = 1$$), Ghutti (herbal medicine for infants) and honey paste ($$n = 1$$), and honey ($$n = 2$$). Several respondents also fed their baby gripe water (not prescribed) ($$n = 7$$) and prescribed medicine ($$n = 2$$) to help with digestion and gas, and growth.
## Handwashing
Most respondents have prior existing knowledge about good hand washing practice. Based on the data, handwashing seems to be a generally understood and accepted behavior (even prior to CCP attendance); it is a common behavior that is not unique to newborn care. The majority of respondents ($$n = 35$$) practiced regular hand washing before attending the CCP. Overall, mothers and family caregivers are aware of the importance of washing hands with soap. Respondents recognized the importance of handwashing ($$n = 39$$) and understood the benefits including being able to make the link between good handwashing practice and preventing infection.
Respondents attributed to CCP that they now regularly wash their hands with soap before touching their baby ($$n = 14$$), five respondents said that their family members also wash their hands now with soap before touching their baby and they are now more careful to wash hands more frequently with soap ($$n = 7$$). The CCP helped this group of respondents understand the benefits of using soap in addition to water when hand washing and helped respondents understand the importance of washing their hands prior to touching their newborn in order to help minimize spread of infection to their baby. One respondent stated that they do not follow the exact method of handwashing taught in the CCP but still wash their hands. Overall, respondents reported having regular access to water (in-house pipes, borewell, stored water) and soap which facilitates handwashing behavior. A few respondents also have access to hand sanitizer ($$n = 3$$). *In* general, the CCP influenced when respondents wash hands, including after using the bathroom, before eating and/or serving meals to family members, after doing housework, after coming home from doing outside work, and before touching their baby (including before breastfeeding and after handling a dirty diaper):
## Skin-to-skin care
Overall, responses indicate that there is limited existing knowledge about skin-to-skin care for newborns. The data show that there aren’t any strong existing cultural and social beliefs around the practice, and family and community members are not offering advice around this behavior. Respondents’ knowledge of skin-to-skin care seems to be coming from the CCP and other healthcare workers.
More than half of respondents said that they do not practice skin to skin care ($$n = 25$$) although 29 respondents believe that there are benefits to skin-to-skin care including newborn weight gain, increased bonding between mother and newborn, and transfer of warmth. Many respondents ($$n = 33$$) said they learned about skin to skin from the CCP and some respondents ($$n = 6$$) could recall the use of a doll to demonstrate skin-to-skin care (“hugging”, “wrapping of doll”). However, there is variation among respondents in their understanding of how and when to perform skin-to-skin care. Responses regarding how skin-to-skin care should be performed varied and were sometimes contradictory to what was taught in the CCP. The following are examples of responses when respondents were asked how and when to perform skin-to-skin care: A few respondents ($$n = 5$$) said that skin-to-skin care should only be done after the mother or family member bathes. Nineteen respondents said that skin-to-skin care should be practiced when the baby is cold or has a problem (when the newborn is “underweight” or “weak”, when the newborn has a “fever”).
A few respondents ($$n = 7$$) said they did not feel the need to practice skin-to-skin because their baby is healthy and others ($$n = 4$$) did not want to practice skin-to-skin because they felt their newborn was “too weak” (premature) and they were scared to carry their newborn because it could cause harm to the baby in some way.
## Cord care
Over half of respondents ($$n = 27$$) had prior knowledge around dry cord care practice. Sources of knowledge included advice around cord care from family members, influence of “elders”, prior personal experience, and information from healthcare workers (physicians, nurses, ASHA workers). Over half of respondents ($$n = 22$$) said they learned new information from the CCP or the CCP reinforced dry cord care knowledge.
The majority of respondents ($$n = 37$$) did not put anything on the cord or cord area and allowed the cord to naturally dry and fall off. Several of these respondents recalled that putting something on the cord could lead to infection ($$n = 7$$) and a few respondents recalled that dry cord care prevents infection and pus formation ($$n = 2$$). A few respondents ($$n = 5$$) put something onto the cord before it fell off (hot coconut oil, powder, castor oil, “boric powder”). Respondents that put something onto the cord before it fell off explained this was due to direct advice from a family member (“elders”, mother-in-law, sister). Thirteen respondents applied something (oil, powder) to the cord area after the cord fell off. Common reasons given for applying items to newborn cord or cord area include: to reduce pain, keep cord area dry, reduce redness, heal a cut, stop bleeding, or stop pus.
## Mother postpartum diet restriction
Almost all respondents ($$n = 38$$) said that the mother restricted specific foods and/or liquids post-pregnancy, either for a specified (e.g. days or months) or unspecified amount of time. Thirty-three of these respondents could recall that the CCP recommended that the mother not restrict any food or liquid. However, six respondents recalled being told during the CCP to avoid certain foods (cold food, spicy food, salty food, oily food). *In* general, respondents offered contradictory answers to interview questions about maternal postpartum diet. When initially asked if they restrict any foods, respondents typically answer “No” and say that they follow what was taught in the CCP, but when probed about specific food items (fruits, spicy foods etc.) then the respondent said that the mother does restrict certain foods and liquids.
The majority of respondents restrict despite what they recall being instructed to do in the CCP. The data shows that a complex set of factors influence maternal postpartum diet restriction including individual beliefs, advice from family, advice from extended community, and cultural practices. Mothers seem to receive mixed messages and contradictory advice around diet from various sources. Key sources of advice include elders within their family such as their mother, grandmother, and mother-in-law. The elder women of the household often prepare food for the mother and therefore have control over what and when she eats; she is dependent on what her family provides her and has little choice.
There are strong held social beliefs about risks to the newborn and mother from consuming certain foods or drinking too much liquids. Consequently, respondents avoid eating certain foods and consuming too much liquid. The most common restricted foods and liquids include: *In* general, participants seem to have regular access to meals with grain, fruits, vegetables, and water. A few respondents ($$n = 4$$) said that they do not restrict any foods or liquids post pregnancy. Three of these respondents directly attributed this to what they learned in the CCP.
Out of the three respondents, two said that food/liquid restriction was present for their first child but for the second baby there was no restriction due to what they learned in the CCP. Several respondents ($$n = 11$$) made the connection between a good maternal postpartum diet and breast milk production.
## Identifying danger or warning signs
We interviewed 23 respondents (13 mothers and 10 family members) that self-reported a post-discharge newborn complication in a follow-up survey conducted by Noora Health. Out of the 23 respondents identified using the survey, the most commonly reported complications were fever ($$n = 6$$) and cold/cough ($$n = 5$$), followed by cord infection ($$n = 4$$), jaundice ($$n = 2$$), rash ($$n = 2$$), infection from drinking “womb water” ($$n = 1$$), problems with urine production ($$n = 1$$), vomiting ($$n = 1$$), and gas ($$n = 1$$). From the analysis, we identified nine additional respondents from the non-complication stratum who mentioned that their newborn experienced some sort of problem (not originally mentioned in the 28-day post discharge survey). These respondents reported cord infection ($$n = 3$$), cold ($$n = 2$$), breathing problems ($$n = 1$$), rash ($$n = 1$$), stomach pain ($$n = 1$$), and discomfort when passing urine ($$n = 1$$). Two respondents (both family members) believed that the reason for their baby’s problem (fever, constipation, stomach pain) was due to maternal diet.
A few respondents ($$n = 4$$) said that the CCP influenced their knowledge and understanding of danger signs related to their newborn’s complications (3 fever, 1 cord infection). Four respondents said that the CCP influenced their symptom management and three respondents said that the CCP influenced their health seeking behavior. Several respondents ($$n = 4$$) said that the CCP had no influence on their ability to identify danger signs, symptom management, or health seeking behavior. *In* general, respondents either tried to manage symptoms at home with medicine from the pharmacy and/or they went to a clinic or hospital for treatment. There was no mention among respondents of utilizing a traditional healer.
Overall, respondents reported that they were able to recognize when their baby had a problem and understood the importance of seeking medical advice and treatment. The data indicate that health seeking behavior (taking baby to the clinic or hospital) is common among parents of newborns and that respondents believe they are able to recognize when something is wrong with their baby. *In* general, responses indicate that prior knowledge contributed to health seeking behavior. It is possible that the CCP reinforced certain knowledge around danger signs and behaviors (symptom management, health seeking behavior) but we cannot determine that with our current data.
At the start of the interview, all respondents ($$n = 42$$) were asked “In the group meeting, the Nurses talked about a lot of things that could go wrong for a baby. How did hearing about these things make you feel? Why?” Most respondents described danger signs taught. Fifteen respondents recalled being told about jaundice during the CCP and eight of these respondents recalled the symptoms of jaundice including yellowish discoloration of the eyes or body. Two respondents recalled that in the CCP the nurses said to take the baby to the hospital if they noticed breathing problems including if the baby started to “breathe heavy” and two respondents recalled that that in the CCP the nurses said to take the baby to the hospital if the “feet turn blue” and if the body turns blue. It is possible that respondents are able to recall the various danger signs discussed in the CCP but in this study the interview guide did not specifically explore each danger sign discussed in the CCP with each respondent.
## Discussion
Overall, participants perceive that CCP sessions have a positive impact on health behaviors though the extent varies across behaviors and type of participant. Mothers and family caregivers did report learning new information on new or already known behaviors, but the practice of these behaviors was not consistent. The practice of key behaviors varied, possibly due to nurse level factors, how new the information was to mothers and family caregivers, and whether behaviors stood in contrast with cultural and family advice. For example, steps and triggers for when to do skin-to-skin contact varied across participants, possibly because how some nurses promote skin-to-skin care for premature and low birthweight babies, compared to the CCP session which promotes skin-to-skin care regardless of premature or birthweight status. For other more common behaviors, such as exclusive breastfeeding or handwashing, which are already saturated with public health messaging, the value of the CCP session emerged when specific, new details were provided (e.g. how to troubleshoot breastfeeding challenges with hand positioning or when to wash one’s hands). A final potential explanation of whether the CCP changed skills depends on whether they align with cultural norms. For example, nonrestrictive maternal diet as taught in CCP conflicts with cultural norms, which may explain why the majority of mothers restricted food or liquids postpartum.
As for whether the CCP helped families identify warning signs, overall the data suggest that CCP information did help a few of the respondents who reported health complications and reinforced existing health seeking practices to go to a facility. Families either tried to manage symptoms with pharmacy-bought medicine or went to a health facility for treatment.
Based on the interviews, it was clear there were variations in the CCP session, like skin-to-skin care for postpartum maternal dietary habits. Based on the information from the interviews, we cannot determine why there is variation among respondents but it is possible that these topics may be inconsistently taught across CCP sites and/or respondents are receiving conflicting advice from healthcare workers (such as nurses and doctors at the hospital and frontline health workers).
These qualitative results completement a previous pilot evaluation of CCP in Punjab and Karnataka, which reported statistically significant improvements on dry cord care practice by $4\%$, skin-to-skin care by $78\%$, newborn complications reduced by $16\%$, mother complications by $12\%$, and newborn readmissions by $56\%$ in the postintervention group as compared with preintervention group [30]. However, the practice of exclusive breastfeeding, unrestricted maternal diet, hand-hygiene and being instructed on warning signs were not statistically different. The health behaviors for which no statistically significant changes were found between the pre and post intervention group are those that either have a high ceiling effect or have strong pre-existing norms, as found by this study.
Given the findings from this qualitative study, there are several opportunities for improvement in the CCP program, most notably to include family members (e.g. what can mothers, fathers, grandmothers, or other family caregivers do) with specific messaging and address beliefs and practices directly (e.g. target grandmothers who prepare maternal foods and emphasize that all foods can be eaten). CCP sessions can be improved through greater program consistency across trainers, speaking directly to socio-cultural norms that may conflict with medical advice, and allowing for more interactive question-answer sessions. CCP sessions are similar to other parent focused education programs delivered before hospital discharge, in that it uses a multitude of tools and techniques and covers multiple topics. Like CCP, in a scoping review of parent focused education programs before hospital discharge, $61\%$ focused on a single topic, almost all interventions delivered information verbally ($93.5\%$), $20\%$ through modeling, $17\%$ through images, $11.7\%$ using videos, and $11.7\%$ in groups [19].
Findings demonstrated that other family members, especially grandmothers to the baby, play a significant role in postnatal care. This is consistent with Lunkenheimer’s research which identifies that families, friends, and neighbors are more likely to influence maternal diet than ASHA workers and contribute to restricting postpartum maternal diet in Bihar [31]. These family caregivers have particular influence and control over the mother’s diet, especially when they prepare the mother’s diet. Globally across Africa, Asia, and Latin America, Aubel argues the importance grandmothers have in maternal and child health given their experience and access to social networks for help beyond the household, although grandmothers offer advice, sometimes with outdated, unscientific information that is not in line with current medical recommendations [32, 33]. Interventions that include grandmothers especially as part of the target audience are key to changing behaviors and improving outcomes, given their influence on maternal and newborn practices. Given the mother’s unique position as both the “patient” and a “caregiver,” there is value in training the whole family so the burden doesn’t only fall onto the mother. It is estimated that girls and women contribute more than $70\%$ of all caregiving hours globally [34]. Family caregiver training can help offset gendered care expectations for newborns, which often falls to the women of the family and involve male family members more.
The limitations of this study include the fact that respondents were not interviewed privately; limited location of the study; only inquiring about complications for the families who reported it; and the reliance on self-reported data. Though the intention was to interview mothers and family caregivers separately, practically this was not always possible due to the size of the house and the presence of others in the household. This invariably could have influenced the answers that participants gave. Second, the research team only recruited families from within four sites in Karnataka due to time and financial constraints. A wider sample across a more diverse population could have revealed how CCP programs interact with different cultural contexts and norms around neonatal health. Additionally, the majority of the newborn complications group reported non-serious complications (e.g. fever or cold), making it challenging to ascertain impact on health complications or hospital readmissions. Additionally, we chose only to ask complication related questions to families who reported neonatal complications during the 28-day post discharge survey. Finally, given the reliance on self-reported data, there could be social desirability bias or limited recall.
## Conclusion
The CCP, a health education program delivered within hospitals for postnatal caregivers, delivers valuable information for mothers and family caregivers to practice preventive behaviors. The CCP session not only provided details on already known behaviors like breastfeeding or handwashing, but taught new information such as skin-to-skin care. Throughout the interviews, it was clear that there are persisting cultural norms that conflict with health information provided during the CCP, so additional consideration of these norms are needed. Finally, CCP reinforced warning sign recognition.
## References
1. 1UNICEF. India (IND)—Demographics, Health & Infant Mortality [Internet].
UNICEF DATA. 2020 [cited 2021 Nov 19]. Available from: https://data.unicef.org/country/ind/. *UNICEF DATA* (2020.0)
2. Jain K, Sankar MJ, Nangia S, Ballambattu VB, Sundaram V, Ramji S. **Causes of death in preterm neonates (<33 weeks) born in tertiary care hospitals in India: analysis of three large prospective multicentric cohorts**. *Journal of Perinatology* (2019.0) **39** 13-9. PMID: 31485016
3. Berhan D, Gulema H, Khubchandani J. **Level of Knowledge and Associated Factors of Postnatal Mothers’ towards Essential Newborn Care Practices at Governmental Health Centers in Addis Ababa, Ethiopia**. *Advances in Public Health* (2018.0) **2018** 8921818
4. Fadel SA, Ram U, Morris SK, Begum R, Shet A, Jotkar R. **Facility Delivery, Postnatal Care and Neonatal Deaths in India: Nationally-Representative Case-Control Studies.**. *PLoS One [Internet]* (2015.0) **10**. DOI: 10.1371/journal.pone.0140448
5. **Causes of neonatal and child mortality in India: nationally representative mortality survey**. *Lancet* (2010.0) **376** 1853-60. PMID: 21075444
6. Ram U, Jha P, Ram F, Kumar K, Awasthi S, Shet A. **Neonatal, 1–59 month, and under-5 mortality in 597 Indian districts, 2001 to 2012: estimates from national demographic and mortality surveys**. *Lancet Glob Health* (2013.0) **1** e219-226. DOI: 10.1016/S2214-109X(13)70073-1
7. Herbert HK, Lee ACC, Chandran A, Rudan I, Baqui AH. **Care seeking for neonatal illness in low- and middle-income countries: a systematic review**. *PLoS Med* (2012.0) **9** e1001183. DOI: 10.1371/journal.pmed.1001183
8. Phukan D, Ranjan M, Dwivedi LK. **Impact of timing of breastfeeding initiation on neonatal mortality in India.**. *Int Breastfeed J* (2018.0) **13** 27. DOI: 10.1186/s13006-018-0162-0
9. Kinanu L, Odhiambo E, Mwaura J, Habtu M. **Cord Care Practices and Omphalitis among Neonates Aged 3–28 Days at Pumwani Maternity Hospital, Kenya**. *Journal of Biosciences and Medicines* (2016.0) **04** 27
10. Kuti BP, Ogunlesi TA, Oduwole O, Oringanje C, Udoh EE, Meremikwu MM. **Hand hygiene for the prevention of infections in neonates**. *Cochrane Database Syst Rev* (2021.0) **1** CD013326. DOI: 10.1002/14651858.CD013326.pub2
11. Shrestha T, Bhattarai SG, Silwal K. **Knowledge and practice of postnatal mother in newborn care**. *JNMA J Nepal Med Assoc* (2013.0) **52** 372-7. PMID: 24362663
12. Srivastava S, Gupta A, Bhatnagar A, Dutta S. **Effect of very early skin to skin contact on success at breastfeeding and preventing early hypothermia in neonates**. *Indian J Public Health* (2014.0) **58** 22-6. DOI: 10.4103/0019-557X.128160
13. Singh DR, Harvey CM, Bohara P, Nath D, Singh S, Szabo S. **Factors associated with newborn care knowledge and practices in the upper Himalayas.**. *PLoS One* (2019.0) **14** e0222582. DOI: 10.1371/journal.pone.0222582
14. Nabiwemba EL, Atuyambe L, Criel B, Kolsteren P, Orach CG. **Recognition and home care of low birth weight neonates: a qualitative study of knowledge, beliefs and practices of mothers in Iganga-Mayuge Health and Demographic Surveillance Site, Uganda.**. *BMC Public Health* (2014.0) **14** 546. DOI: 10.1186/1471-2458-14-546
15. Withers M, Kharazmi N, Lim E. **Traditional beliefs and practices in pregnancy, childbirth and postpartum: A review of the evidence from Asian countries**. *Midwifery* (2018.0) **56** 158-70. DOI: 10.1016/j.midw.2017.10.019
16. Pati S, Chauhan AS, Panda M, Swain S, Hussain MA. **Neonatal care practices in a tribal community of Odisha, India: A cultural perspective**. *Journal of Tropical Pediatrics* (2014.0) **60** 238-44. DOI: 10.1093/tropej/fmu005
17. George M, Johnson A. **Postpartum and Newborn Care-A Qualitative study.**. *Indian Journal of Community Health* (2018.0) **30** 163-5
18. Subramanian L, Murthy S, Bogam P, Yan SD, Marx Delaney M, Goodwin CDG. **Just-in-time postnatal education programees to improve newborn care practices: needs and opportunities in low-resource settings**. *BMJ Glob Health* (2020.0) **5**. DOI: 10.1136/bmjgh-2020-002660
19. Dol J, Campbell-Yeo M, Tomblin Murphy G, Aston M, McMillan D, Gahagan J. **Parent-targeted postnatal educational interventions in low and middle-income countries: A scoping review and critical analysis**. *Int J Nurs Stud* (2019.0) **94** 60-73. DOI: 10.1016/j.ijnurstu.2019.03.011
20. Parashar M, Singh SV, Kishore J, Kumar A, Bhardwaj M. **Effect of community-based behavior change communication on delivery and newborn health care practices in a resettlement colony of Delhi**. *Indian Journal of Community Medicine* (2013.0) **38** 42. DOI: 10.4103/0970-0218.106627
21. Hazra A, Atmavilas Y, Hay K, Saggurti N, Verma RK, Ahmad J. **Effects of health behaviour change intervention through women’s self-help groups on maternal and newborn health practices and related inequalities in rural india: A quasi-experimental study.**. *EClinicalMedicine* (2020.0) **18** 100198. DOI: 10.1016/j.eclinm.2019.10.011
22. Pillay T.. **Parent-Carer Education: Reducing the Risks for Neonatal and Infant Mortality.**. *Neonatal Medicine [Internet]* (2019.0)
23. Merakou K, Knithaki A, Karageorgos G, Theodoridis D, Barbouni A. **Group patient education: effectiveness of a brief intervention in people with type 2 diabetes mellitus in primary health care in Greece: a clinically controlled trial.**. *Health Educ Res* (2015.0) **30** 223-32. DOI: 10.1093/her/cyv001
24. Rickheim PL, Weaver TW, Flader JL, Kendall DM. **Assessment of Group Versus Individual Diabetes Education: A randomized study**. *Diabetes Care* (2002.0) **25** 269-74. DOI: 10.2337/diacare.25.2.269
25. Torres H de C, Franco LJ, Stradioto MA, Hortale VA, Schall VT. **Evaluation of group and individual strategies in a diabetes education program**. *Rev Saude Publica* (2009.0) **43** 291-8. PMID: 19225700
26. 26Ministry of Health and Family Welfare. National Family Health Survey-4, State Factsheet Karnataka [Internet]. 2017 [cited 2021 Nov 19]. Available from: http://rchiips.org/NFHS/pdf/NFHS4/KA_FactSheet.pdf
27. O’Brien BC, Harris IB, Beckman TJ, Reed DA, Cook DA. **Standards for Reporting Qualitative Research: A Synthesis of Recommendations**. *Academic Medicine* (2014.0) **89** 1245-51. DOI: 10.1097/ACM.0000000000000388
28. Kittle B.. (2013.0)
29. 29Dedoose [Internet]. Los Angeles, California: SocioCultural Research Consultants, LLC; 2020 [cited 2021 Jun 17]. Available from: https://www.dedoose.com/
30. Kashyap S, Spielman AF, Ramnarayan N, SD S, Pant R, Kaur B. **Impact of family-centred postnatal training on maternal and neonatal health and care practices in district hospitals in two states in India: a pre–post study**. *BMJ Open Qual* (2022.0) **11** e001462. DOI: 10.1136/bmjoq-2021-001462
31. Lunkenheimer HG, Burger O, Akhauri S, Chaudhuri I, Dibbell L, Hashmi FA. **Tradition, taste and taboo: the gastroecology of maternal perinatal diet**. *BMJ Nutrition, Prevention & Health* (2021.0) bmjnph. DOI: 10.1136/bmjnph-2021-000252
32. Aubel J.. **Grandmothers—a neglected family resource for saving newborn lives**. *BMJ Global Health* (2021.0) **6** e003808. DOI: 10.1136/bmjgh-2020-003808
33. Aubel J.. **The role and influence of grandmothers on child nutrition: culturally designated advisors and caregivers**. *Matern Child Nutr* (2011.0) **8** 19-35. DOI: 10.1111/j.1740-8709.2011.00333.x
34. Langer A, Meleis A, Knaul FM, Atun R, Aran M, Arreola-Ornelas H. **Women and Health: the key for sustainable development**. *The Lancet* (2015.0) **386** 1165-210. DOI: 10.1016/S0140-6736(15)60497-4
|
---
title: Caesarean section or vaginal delivery for low-risk pregnancy? Helping women
make an informed choice in low- and middle-income countries
authors:
- Alexandre Dumont
- Myriam de Loenzien
- Hung Mac Quo Nhu
- Marylène Dugas
- Charles Kabore
- Pisake Lumbiganon
- Maria Regina Torloni
- Celina Gialdini
- Guillermo Carroli
- Claudia Hanson
- Ana Pilar Betrán
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10022020
doi: 10.1371/journal.pgph.0001264
license: CC BY 4.0
---
# Caesarean section or vaginal delivery for low-risk pregnancy? Helping women make an informed choice in low- and middle-income countries
## Abstract
Women’s fear and uncertainty about vaginal delivery and lack of empowerment in decision-making generate decision conflict and is one of the main determinants of high caesarean section rates in low- and middle-income countries (LMICs). This study aims to develop a decision analysis tool (DAT) to help pregnant women make an informed choice about the planned mode of delivery and to evaluate its acceptability in Vietnam, Thailand, Argentina, and Burkina Faso. The DAT targets low-risk pregnant women with a healthy, singleton foetus, without any medical or obstetric disorder, no previous caesarean scarring, and eligibility for labour trials. We conducted a systematic review to determine the short- and long-term maternal and offspring risks and benefits of planned caesarean section compared to planned vaginal delivery. We carried out individual interviews and focus group discussions with key informants to capture informational needs for decision-making, and to assess the acceptability of the DAT in participating hospitals. The DAT meets 20 of the 22 Patient Decision Aid Standards for decision support. It includes low- to moderate-certainty evidence-based information on the risks and benefits of both modes of birth, and helps pregnant women clarify their personal values. It has been well accepted by women and health care providers. Adaptations have been made in each country to fit the context and to facilitate its implementation in current practice, including the development of an App. DAT is a simple method to improve communication and facilitate shared decision-making for planned modes of birth. It is expected to build trust and foster more effective, satisfactory dialogue between pregnant women and providers. It can be easily adapted and updated as new evidence emerges. We encourage further studies in LMICs to assess the impact of DAT on quality decision-making for the appropriate use of caesarean section in these settings.
## Background
Decision-making in relation to the mode of delivery is increasingly done together with women as opposed to medical personnel deciding alone. Participatory and joint decision-making demands are sufficient and help to understand information and decision aids [1]. While previous childbirth experience can strongly influence this decision, women who have never given birth before are often uncertain about vaginal delivery [2]. Women’s belief and cultural factors mainly determine women’s preference for a planned mode of delivery [3]. Moreover, a lack of complete and reliable information about the potential risks and benefits of a planned caesarean section compared to a planned vaginal delivery contributes to many misconceptions regarding the pros and cons of both options, and reduces women’s ability to make an informed choice [4, 5]. For some women, fear of pain is becoming so widespread globally that preference for caesarean delivery is growing even in the absence of medical indications [3]. However, maternal request for caesarean section may place health professionals in a situation of ethical tension between the duty to promote the safest way to give birth (thus physiological birth in the absence of medical indications), and the need to respect the patient’s choice [6]. In any case, anxiety and decisional conflict can emerge from a non-shared medical decision when women’s values and expectations are not met [7].
Women’s uncertainty about the planned mode of delivery affects not only high-income countries, but also low- and middle-income countries (LMICs) [8, 9]. In some Asian countries, women think that a caesarean section is safer than vaginal delivery for the baby, and that a planned caesarean offers the possibility to schedule the delivery on auspicious birth dates [10, 11]. For women in Latin America and sub-Saharan Africa who prefer vaginal birth, some express that a planned caesarean is a medical decision because they lack reliable information and empowerment in the decision-making process [10–13]. As a consequence, caesarean rates continue to rise in LMICs and, in many cases, to levels well above possible medical needs, resulting in overuse of the procedure and an increase in the risks without clear benefits [14].
According to the Ottawa Decision Support framework, decision aids can improve decision-making by informing patients about the risks and benefits of different options for their health [15]. We developed implementation research to design and evaluate a strategy, called Quality Decision-Making by Women and Providers (QUALI-DEC), to implement interventions targeted simultaneously at women, health care providers, and health systems in order to improve decision-making for planned modes of birth in Argentina, Burkina Faso, Thailand, and Vietnam [16]. The QUALI-DEC strategy combines four active components: [1] opinion leaders to carry out evidence-based clinical guidelines; [2] caesarean audits and feedback to help providers identify potentially avoidable caesarean sections; [3] a decision-analysis tool (DAT) to help women make an informed decision on mode of birth; and [4] companionship during labour to support women during this time, as well as vaginal delivery. We assume that the DAT could enhance women’s knowledge on the risks and benefits of both modes of birth, raise providers’ awareness about women’s attitudes and preferences regarding delivery, facilitate shared decision-making about mode of birth, and reduce maternal requests for a planned caesarean section in participating hospitals.
The aim of this report is to describe the development of a DAT tailored to the context of LMICs, and to evaluate its acceptability from the perspectives of women and health care providers in QUALI-DEC participating countries.
## Ethics statement
We received authorisation from the Department of Reproductive Health of the Ministry of Health in Vietnam, and the four participating hospitals ethically approved the research. We obtained ethical clearance for the study from the local and institutional review boards from the Centro Rosarino de Estudios Perinatales of Rosario, Argentina (Record Notice No. $\frac{1}{20}$), Khon Kaen University in Thailand, the Ethics Committee for Health Research of Burkina Faso (Decision No. 2020-3-038), the Research Project Review Panel (RP2) in the UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction (WHO study No. A66006), and the French Research Institute for Sustainable Development (coordinator). For all individual interviews or focus group discussions, formal written consent was obtained from participants.
We designed the DAT using the Ottawa Decision Support framework [7]. We devised a 3-step approach to generate and evaluate the tool. We used data collected in hospitals in Vietnam to identify needs for decision support. Then, we conducted an overview of the literature to provide evidence for decision support, while we assessed acceptability in hospitals in Argentina, Burkina Faso, and Thailand. Table 1 shows the characteristics of the participating hospitals in the four countries.
**Table 1**
| Characteristic | Vietnam | Burkina Faso | Thailand | Argentina |
| --- | --- | --- | --- | --- |
| Characteristic | N = 4 | N = 8 | N = 8 | N = 8 |
| Type of hospital | | | | |
| Public without a private ward | 3 | 8 | 0 | 8 |
| Public with a private ward | 1 | 0 | 8 | 0 |
| Private | 0 | 0 | 0 | 0 |
| Level of care | | | | |
| Tertiary | 1 | 2 | 7 | 8 |
| Secondary | 2 | 4 | 1 | 0 |
| Primary | 1 | 2 | 0 | 0 |
| Teaching hospital | | | | |
| Yes | 2 | 3 | 8 | 8 |
| No | 2 | 5 | 0 | 0 |
| Range of annual births | 2800–42000 | 2500–6000 | 2500–7500 | 614–4945 |
| Range of caesarean rates | 23%–54% | 21%–48% | 36%–56% | 30%–45% |
The target audience of the DAT is low-risk women with a healthy, singleton foetus, without any known medical or obstetric disorders, no previous caesarean section, and eligible for trial of labour at the time of the anatenatal care visits. Women with previous caesarean section, breech or abnormal presentation, twin pregnancy, or any indication for elective caesarean section (pre-labour) are not the target of the DAT because they are at high-risk for caesarean delivery.
## Step 1: Identifying needs for improved decision-making on mode of birth
We carried out qualitative research in August 2018 and March 2019 in four hospitals in Vietnam, purposively selected by the Ministry of Health to reflect a range of contexts (Table 1).
We held individual interviews and focus group discussions among different key informants, including postpartum women, their relative or companion, and health care providers. Maximum variation sampling was used to achieve a diverse sample of providers, including hospital or service managers, clinicians of different qualification (obstetricians, midwives, nurses), sex and seniority. The same method was used to achieve a diverse sample of postpartum women in terms of age, religion, ethnicity and mode of birth (caesarean or vaginal delivery). We recruited and interviewed women and their companion separately, immediately after delivery and before discharge from the study hospitals. Each interview and focus group was facilitated in the participants’ respective languages by a female data collector with experience in conducting in-depth interviews or focus group discussions, and audio-recorded if consent was obtained. We conducted focus group discussions separately with obstetricians and midwives to encourage the expression of opinions outside of the clinical hierarchy. We asked women and providers about the possible reasons for the high rates of caesarean section in Vietnam and what they needed to prepare them to discuss the most appropriate planned mode of delivery. Based on the ecological model to understand factors influencing caesarean rates [17], we analysed the recordings and interpreted the data using a thematic analysis approach.
## Step 2: Providing evidence for a holistic decision support tool
We designed the DAT in two sections (S1 Text). The first section aims to inform women during antenatal care (ANC) visits about the risks and benefits of caesarean section and vaginal delivery. The second section aims to help women clarify their values and thus prepare them to discuss their preferences with a health care professional during following visits.
We performed an overview of the literature to provide evidence-based information on the risks and benefits of both modes of delivery. We included systematic reviews (SRs), overviews, or agency statements/reports that provide risk estimates for short- and long-term maternal and child outcomes of women who planned for a caesarean section (but in few cases may have had vaginal delivery instead) compared to women who planned for a vaginal delivery (but ended up with either a vaginal delivery or an emergency caesarean section). We took this approach to ensure the inclusion of studies that reflect the relevant risks for pregnant women who were planning modes of delivery during the antenatal period. In cases where there was no information about planned modes of delivery, we present the evidence about the comparison between all types of caesarean (elective or intrapartum) and all types of vaginal delivery (planned or not, spontaneous or assisted). We ran the search in MEDLINE on 18 April 2018 and updated it on 30 April 2020 using the terms ‘delivery, obstetric/adverse effects’[Mesh]) OR (‘Caesarean Section/adverse effects’[Mesh]) and filters (Books and Documents, Meta-Analysis, Review, Systematic Review from 2000–2020). If we identified more than one source of evidence for the same outcome, we used the most recent source document (systematic review or overview or agency report) or the source document with the most recent date of search for its evidence base. We included articles in all languages except in Chinese. The search was complemented by the snowball technique; that is, looking for potentially relevant studies on the same subject going backwards (reviewing citations of the key study) and forwards (identifying articles citing the key study). We assessed the quality of individual observational studies included in the SRs using the Newcastle–Ottawa Scale (NOS) or the Scottish Intercollegiate Guideline Network (SIGN) tools. To analyse the certainty of evidence for each outcome, we used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) scale.
## Step 3: Evaluating the decision support provided by the DAT
We prepared a first draft of the DAT (a paper booklet) based on the first overview of the literature in 2018 to show it as an example to the women and providers so they would better understand what we were proposing. We used the theoretical framework of acceptability (TFA) to assess the prospective acceptability of the DAT booklet among women and providers [18]. We defined acceptability as the extent to which women and providers considered the DAT to be appropriate based on anticipated cognitive and emotional responses to the booklet.
As part of the formative research of the QUALI-DEC project, we held individual interviews in February 2020 in Burkina Faso and in May 2020 in Thailand. In Argentina, due to the COVID-19 situation, individual interviews were not possible. We held two virtual discussion groups in July 2020, one with obstetricians (Heads of the Obstetric Services of the eight participating hospitals) and another with midwives from those same hospitals. One of the objectives of the formative research was to understand how and why a decision aid might help inform preferences and improve decision-making processes for women and providers. We selected a total of 24 hospitals with high caesarean section rates in the three participating countries to reflect a range of contexts, such as district hospitals, regional or provincial hospitals, private clinics, and tertiary/academic hospitals (Table 1).
Recruitment and sampling of women and providers were conducted as specified under step 1 described before. Interviews were also conducted with pregnant women. In each selected facility, researchers facilitated contact with women during their antenatal care visit. Pregnant women who participated in the interviews identified either their partner or the person who they would prefer as a labour companion to participate in the study. Women facilitated contact with potential participants, and researchers followed up to schedule an interview. Female researchers was on site to facilitate recruitment, and was not be involved in clinical care of the patient.
We first transcribed all digitally recorded, qualitative data verbatim in the original language used for collection. We analysed and interpreted the qualitative data using a thematic analysis approach. We performed the analyses in the local or contextually relevant language under the supervision of a social scientist of the QUALI-DEC research team in each country. The country-level analysis involved a combined inductive and deductive approach to allow themes to emerge naturally from the data while also synthesising themes based on questions in a semistructured interview guide. We organised the analyses as a stepwise process: Each country prepared a report in English language interpreting country-level analyses, which we then compared between countries. The main findings of this higher-level analysis were shared and triangulated with researchers during several online workshops.
According to our data management plan, qualitative interview transcripts in original language have been made available for QUALI-DEC researchers only. The translation into English of the thematic analysis (country-level report) will be made available to the public at the end of the project.
## Perceived needs and gaps for quality decision-making
In Vietnam, we interviewed 28 women and 16 health care providers. We also held four focus group discussions with obstetricians and midwives (Table 2).
**Table 2**
| Unnamed: 0 | Development (step 1) | Assessment (step 3) | Assessment (step 3).1 | Assessment (step 3).2 | Unnamed: 5 |
| --- | --- | --- | --- | --- | --- |
| | Vietnam (Aug. 2018 & March 2019) | Burkina Faso (Feb. 2020) | Thaïland (May 2020) | Argentina (July 2020) | Total |
| Pregnant women | - | 22 | 27 | - | 49 |
| Post-partum women | 28 | 16 | 25 | - | 69 |
| Relatives/companions | - | 14 | 16 | - | 30 |
| Hospital director and department heads | 9 | 8 | 8 | - | 25 |
| Gynaecologists-obstetricians | 5 + 20* | 10 | 18 | 18* | 71 |
| Midwifes | 2 + 30* | 9 | - | 15* | 56 |
| Nurses | - | 6 | 33 | - | 39 |
| Total | 94 | 85 | 127 | 33 | 339 |
The main findings highlighted agreement on the overuse of caesarean section and the multifactorial nature of its overuse in Vietnam. Obstetricians claimed organisational gains and explained that defensive medicine promoted caesarean delivery in their context, while the women expressed a strong preference for vaginal delivery during their interviews. According to the women, the lack of dialogue between them and health care providers, the lack of preparation for childbirth and pharmacological methods to control pain during labour were identified as strong obstacles to planned vaginal delivery. Women and their caregivers expressed the need to be better informed about the risks and benefits of planned caesarean section and planned vaginal delivery.
## Risks and benefits of both modes of delivery
The literature review identified 984 unique references. After screening the titles and abstracts, we selected 30 records for full text reading and identified an additional 7 references through other sources. At the end of the process, we included 15 documents (mostly SRs) that compare the risks and benefits of both modes of delivery (S1 Fig). The characteristics of the reviews are presented in S1 Table. The majority of the reviews include observational studies (cohort, case–control, or cross-sectional surveys), which were mainly conducted in high-income countries. Only two reviews [19, 20] exclusively include women with low obstetric risk, and two reviews compare planned caesarean section to planned vaginal delivery for short-term outcomes using an ‘intention-to-treat’ analysis [20, 21]. All other reviews include women who had any type of caesarean section (emergency/elective) or any type of vaginal delivery (planned/actual). Certainty of evidence was moderate for four maternal outcomes (hospital stay, hysterectomy, breastfeeding initiation, complications during future pregnancy) and two infant outcomes (obesity and allergies in adulthood). For the other outcomes, the certainty of evidence was low (Tables 3 and 4).
The advantages of a planned vaginal delivery over a planned caesarean section include a shorter hospital stay, faster recovery, increased chances of starting breastfeeding inmediately after delivery, reduced risks associated with surgery (cardiac arrest), and reduced risk of complications in future pregnancies (uterine rupture, placental abruption, placenta previa or accreta) [19–22, 24]. The disadvantages include possible risk of brachial plexus injury for the baby, increased risk of pain in the perineum and abdomen in the immediate postpartum period, and increased risk of temporary urinary incontinence during the first 2 years after delivery [20–22, 24, 26, 27].
The advantages of a planned caesarean section include less pain in the perineum after delivery and in the first 3 months after delivery, and a reduced risk of urinary incontinence during the first 2 years after delivery [20–22, 26]. The disadvantages include a longer hospital stay, more difficulty in resuming regular life after surgery, more abdominal pain in the first 3 months after birth (including persistent wound pain for 12 or more months), reduced chances of starting to breastfeed after delivery, increased risk of hysterectomy due to postpartum haemorrhage, complications during a future pregnancy, and cardiorespiratory complications for the baby and respiratory disorders after birth when delivery is earlier than 39–40 weeks of pregnancy [19–25, 27, 28, 33]. Caesareans are also associated with possible risks of obesity in childhood or adolescence and allergies/asthma later in life [20, 24, 29–32].
There is no or insufficient/conflicting evidence about the risk with caesarean section or vaginal delivery for the following outcomes: (i) for women: thromboembolic disease, major obstetric haemorrhage, postnatal depression, sexuality, faecal incontinence, and infertility; (ii) for the children: admission to the neonatal unit, infection, persistent verbal delay, and infant mortality (up to 1 year).
## Assessing acceptability of decision-analysis tools by women and providers
Table 2 presents the number of participants to the individual interviews and focus group discussions in participating hospitals at step 3. The findings of the qualitative analysis show that most women prefer a vaginal delivery over a caesarean section. However, safety is a concern, and women preferring a caesarean section gave, as a reason for this preference, the feeling of being safer than if having a vaginal delivery. The women reported that a DAT booklet would be very useful to avoid misinformation and misunderstandings, and confirmed the need for comprehensive information. Furthermore, they reported that it would help them to acquire information that they would otherwise have difficulty obtaining from health care professionals, either because they do not dare to ask for it, or because doctors or midwives do not necessarily take the time to inform them during ANC visits. Most women indicated that the last trimester is the appropriate time to provide information on childbirth preparation and delivery methods. The women identified several other sources of information such as family members, social networks or the internet, but they are often pro-caesarean and women consider these sources to be unreliable. Our qualitative analysis also indicates that women would like to be asked about their preference for mode of delivery since they believe they have the right to choose. Women in Thailand and Argentina (according to the midwives) suggested using the DAT in a digital format, especially on a smartphone, whereas in Burkina Faso, the paper format would be more suitable. Women in Burkina Faso who could not read suggested videos, podcasts, or audio messages broadcast on the radio or in the antenatal waiting room.
Among health care professionals, there was consensus in the three countries that strengthening informational spaces for women and actively involving them in decisions during ANC visits and other non-clinical spaces was relevant. They saw the DAT as complementary to other ongoing actions to strengthen communication between health care providers and pregnant women. In Thailand, some doctors did not see the value of using the DAT for women who have already planned to attempt a vaginal delivery, while they acknowledged that the DAT could benefit women who request a planned caesarean section, particularly in the private sector. The Thai health care providers proposed distributing the DAT to pregnant women during childbirth preparation classes. This would allow them to discuss it with their doctor during subsequent visits. Providers in Thailand also recommended that the DAT include information on pain control methods, as very few women have access to epidural anaesthesia in this country. In Argentina, providers suggested the DAT include information on the expected timing and course of labour and birth, and the roles of the health care team and the companion, especially during labour. Based on the interviews with key informants, each country’s team proposed adaptations of the DAT to fit with their own context and issued recommendations to facilitate its implementation (Table 5). Communication via social networks was highly recommended in Argentina and Thailand.
**Table 5**
| Thailand | Argentina | Burkina Faso |
| --- | --- | --- |
| • Mobile app in addition to the booklet• Flip chart with QR code to access the booklet and the app• Distribute the DAT in ANC parenting school• Include information on the companion’s role and labour pain management.• Prmoting the DAT using various social media. | • Paper-based and mobile app should be available (consider social networks)• Include information on women’s rights and needs for a positive birth experience• Include as many images (or other visual resources) as possible | • Use the ANC booklet for women who can read• Use videos, podcasts, or audio messages broadcast on the radio or in the prenatal waiting room for women who cannot read |
## Development process of the decision analysis tool
The first draft of the DAT booklet was written in January 2020 and reviewed by a committee of eight experts of the QUALI-DEC team who were not involved in its development. The second draft of the booklet was available in December 2020 and discussed via video conferences with each country’s principal investigators of the QUALI-DEC project in Argentina, Vietnam, Thailand, and Burkina Faso. The adaptations that emerged from the individual interviews (Table 5) were discussed and agreed to fit the DAT to local contexts. The third version of the DAT was registered in the Decision Aid Library Inventory (DALI) of the Ottawa Hospital Research Institute: Decision Aid #1959 ‘Caesarean section or vaginal birth: Making an informed choice’ in September 2021. It meets 20 of the 22 Patient Decision Aid Standards for decision support (S2 Text). The two missed standards include: (i) the chances for maternal and infant outcomes; and (ii) readability levels of the target population. The decision aid is available online on the QUALI-DEC website: www.qualidec.com. As requested by women in some countries, we created an application called QUALI-DEC (Fig 1) available in both the Apple and Google app stores in participating countries: (i) Apple (iOS): https://apps.apple.com/bg/app/quali-dec/id1590535948; (ii) Google (Android): https://play.google.com/store/apps/details?id=com.out2bound.whodat.
**Fig 1:** *Decision-analysis tool application for smartphones.*
## Discussion
Based on the Ottawa Decision Support framework, we developed a decision aid to help low-risk pregnant women in LMICs make an informed choice on their planned mode of birth. The DAT was identified as an unmet need in Vietnam and welcomed by women and health care providers in Argentina, Burkina Faso, and Thailand. Providers recognised the need for tools to better equip pregnant women to participate in discussions and decisions during ANC visits. The DAT can be useful in improving communication between providers and women, and addresses the needs of pregnant women for reliable information about childbirth. While most interviewed women would prefer a vaginal delivery, women appreciate that providers ask about their reasons for choosing a vaginal delivery and discuss the pros and cons of both options with them.
Importantly, the DAT is not intended to replace discussion with health care providers, but to better equip women and provide the basis for more informed dialogue with them [34]. This dialogue is important to build a trusting relationship, which can prevent a non-clinical decision for a caesarean section [34, 35]. In addition, in settings where health care providers have limited time to dedicate to each woman during ANC visits, the DAT can promote a more efficient use of this time.
Our process to develop a decision aid for pregnant women has several strengths. First, our DAT is the first decision aid targeting low-risk women without a previous caesarean section in LMICs. Two studies have previously assessed the impact of a decision aid among women with a prior caesarean section [36, 37], showing improved knowledge among women on the risks and benefits of a trial of labour versus a repeat caesarean section and decreased decisional conflict. Second, the DAT meets 20 of the 22 International Patient Decision Aid Standards for decision support. It includes clinical evidence on the outcomes of both modes of delivery, mainly based on large cohort studies, and helps pregnant women to clarify their personal values. Relevant positive and negative features and outcomes of both options are presented. Finally, health care providers, women, and their companions in various settings have positive perceptions of the tool.
However, we faced some limitations in the development process of the DAT. First, the DAT does not include numerical risk estimates for each of the outcomes according to route of delivery. Because the actual probability of each complication will vary between and within countries depending on multiple factors, we chose not to describe the detailed statistics of each outcome. Instead, we presented, in a single table, the advantages and disadvantages of each mode of birth so that the women would have an overview of the pros and cons of each option in an intelligible, appropriate manner (S1 Text). We felt that this approach is more effective for quality decision-making in LMICs. Indeed, the information given in the DAT was well understood and widely appreciated by the women. Second, the DAT cannot be used by illiterate women. For these populations, we chose to develop other communication media, such as video or audio messages. Third, the clinical evidence on the risks and benefits of both modes of delivery is mainly based on indirect comparisons (e.g., pre-labour or emergency caesarean section for various maternal and foetal indications versus planned or actual vaginal delivery) and rarely relies on the ‘intention-to-treat’ approach. Moreover, most of the studies did not adjust for confounding factors that could affect maternal and foetal outcomes such as maternal age, parity, smoking, and body mass index, as well as clinical or obstetric disorders that may have been the primary reason for caesarean section. The inability to disentangle these factors makes it almost impossible to accurately assess risk. Due to the low certainty of the evidence for 8 out of 14 outcomes, the results of our overview of the literature must be interpreted with caution. Moreover, interviews could not be done individually in Argentina due to the COVID-19 pandemic. This undermines the qualitative analysis to assess acceptability of the DAT since the population of that country may be culturally significantly different from the others evaluated. The lack of private hospitals participating in the study is another limitation for findings generalization. Finally, our study provides information about the acceptability of the DAT, but we did not assess the effectiveness or outcomes based on attributes related to the choice made and the decision-making process [15].
Even if causality cannot be confirmed, low-risk pregnant women should be informed that delivery via a planned caesarean section is associated with short- and long-term risks for the mother and the baby, as well as for subsequent pregnancies. The risk of complications with caesarean section, including maternal and perinatal mortality, are probably higher in LMICs than in high-income countries where most of the SRs were conducted, especially regarding risks for future pregnancies [38–41]. For example, the surgical expertise and logistical support (including blood supply) required for a safer caesarean section in cases of abnormal placentation are less likely to be available in low-resource settings [42]. Therefore, women in LMICs could be informed that a planned caesarean section without medical indications can have favourable short-term outcomes, but expose them to potentially severe (and even life-threatening) complications in subsequent pregnancies.
The DAT resonated with the women and health care providers in the four countries of the QUALI-DEC study and is easily adaptable to the specificities of each setting. This tool is designed as a dynamic resource. The booklet and the DAT app can easily be updated as more scientific evidence emerges. If the format is useful and effective, we envisage that important lessons could be drawn on how to implement such a tool in any country or setting.
## Conclusion
Our decision aid for low-risk pregnant women is grounded in the most recent scientific evidence, which does not allow for any doubt about the safety of a planned vaginal delivery compared to a planned caesarean section in low-resource settings. It is expected to build trust and foster more effective, satisfactory dialogue between pregnant women and providers. It can be easily adapted and updated as new evidence emerges. Further studies are needed to improve the quality of the evidence regarding maternal and child outcomes by planned modes of delivery, and to assess the DAT’s impact on quality decision-making for appropriate use of caesarean section in LMICs.
## Participating institutions and staff (QUALI-DEC consortium)
Karolinska Institutet (Sweden): Claudia Hanson, Helle Molsted-Alvesson, Kristi Sidney Annerstedt; University College Dublin, National University of Ireland (Ireland): Michael Robson; World Health Organization (Switzerland): Ana Pilar Betrán, Newton Opiyo, Meghan Bohren; Centro Rosario de Estudios Perinatales Asociacion (Argentina): Guillermo Carroli; Liana Campodonico; Celina Gialdini; Berenise Carroli; Gabriela Garcia Camacho; Daniel Giordano; Hugo Gamerro; CEDES (Argentina): Mariana Romero; Khon Kaen University (Thailand): Pisake Lumbiganon, Dittakarn Boriboonhirunsarn, Nampet Jampathong, Kiattisak Kongwattanakul, Ameporn Ratinthorn, Olarik Musigavong; Fundacio Blanquerna (Spain): Ramon Escuriet, Olga Canet; Centre national de recherche scientifique et technologique—Institut de Recherche en sciences de la sante (Burkina Faso): Charles Kabore, Yaya Bocoum Fadima, Simon Tiendrébéogo, Zerbo Roger; Pham Ngoc Thach University of Medicine (Vietnam): Mac Quoc Nhu Hung, Thao Truong, Tran Minh Thien Ngo, Bui Duc Toan, Huynh Nguyen Khanh Trang, Hoang Thi Diem Tuyet; Research Institute for Sustainable Development (France): Alexandre Dumont, Laurence Lombard, Myriam de Loenzien, Marion Ravit, Delia Visan, Karen Zamboni.
## Transfer Alert
This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.
## References
1. Megregian M, Emeis C, Nieuwenhuijze M. **The Impact of Shared Decision‐Making in Perinatal Care: A Scoping Review**. *J Midwifery Womens Health* (2020.0) **65** 777-788. DOI: 10.1111/jmwh.13128
2. Wagner M.. **Choosing caesarean section**. *The Lancet* (2000.0) **356** 1677-1680. DOI: 10.1016/S0140-6736(00)03169-X
3. Colomar M, Opiyo N, Kingdon C, Long Q, Nion S, Bohren MA, Ortiz-Panozo E. **Do women prefer caesarean sections? A qualitative evidence synthesis of their views and experiences**. *PLOS ONE* (2021.0) **16** e0251072. DOI: 10.1371/journal.pone.0251072
4. Fioretti B, Reiter M, Betrán A, Torloni M. **Googling caesarean section: a survey on the quality of the information available on the Internet**. *BJOG Int J Obstet Gynaecol* (2015.0) **122** 731-739. DOI: 10.1111/1471-0528.13081
5. Torloni MR, Daher S, Betran AP, Widmer M, Montilla P, Souza JP. **Portrayal of caesarean section in Brazilian women’s magazines: 20 year review**. *BMJ* (2011.0) **342** d276-d276. DOI: 10.1136/bmj.d276
6. Schantz C, Lhotte M, Pantelias A-C. *Dépasser les tensions éthiques devant les demandes maternelles de césarienne: Santé Publique* (2021.0) **32** 497-505. DOI: 10.3917/spub.205.0497
7. Stacey D, Légaré F, Boland L, Lewis KB, Loiselle M-C, Hoefel L. **20th Anniversary Ottawa Decision Support Framework: Part 3 Overview of Systematic Reviews and Updated Framework**. *Med Decis Making* (2020.0) **40** 379-398. DOI: 10.1177/0272989X20911870
8. Mazzoni A, Althabe F, Liu N, Bonotti A, Gibbons L, Sánchez A. **Women’s preference for caesarean section: a systematic review and meta-analysis of observational studies: Women’s preference for caesarean section: systematic review**. *BJOG Int J Obstet Gynaecol* (2011.0) **118** 391-399. DOI: 10.1111/j.1471-0528.2010.02793.x
9. Reiter M, Betrán AP, Marques FK, Torloni MR. **Systematic review and meta-analysis of studies on delivery preferences in Brazil**. *Int J Gynecol Obstet* (2018.0) **143** 24-31. DOI: 10.1002/ijgo.12570
10. Takegata M, Smith C, Nguyen HAT, Thi HH, Thi Minh TN, Day LT. **Reasons for Increased Caesarean Section Rate in Vietnam: A Qualitative Study among Vietnamese Mothers and Health Care Professionals**. *Healthcare* (2020.0) **8** 41. DOI: 10.3390/healthcare8010041
11. Suwanrath C, Chunuan S, Matemanosak P, Pinjaroen S. **Why do pregnant women prefer cesarean birth? A qualitative study in a tertiary care center in Southern Thailand**. *BMC Pregnancy Childbirth* (2021.0) **21** 23. DOI: 10.1186/s12884-020-03525-3
12. Liu NH, Mazzoni A, Zamberlin N, Colomar M, Chang OH, Arnaud L. **Preferences for mode of delivery in nulliparous Argentinean women: a qualitative study**. *Reprod Health* (2013.0) **10** 2. DOI: 10.1186/1742-4755-10-2
13. Richard F, Ouattara F, Zongo S. **Fear, guilt, and debt: an exploration of women’s experience and perception of cesarean birth in Burkina Faso, West Africa**. *Int J Womens Health* (2014.0) 469. DOI: 10.2147/IJWH.S54742
14. Betran AP, Ye J, Moller A-B, Souza JP, Zhang J. **Trends and projections of caesarean section rates: global and regional estimates**. *BMJ Glob Health* (2021.0) **6** e005671. DOI: 10.1136/bmjgh-2021-005671
15. Stacey D, Légaré F, Lewis K, Barry MJ, Bennett CL, Eden KB. **Decision aids for people facing health treatment or screening decisions**. *Cochrane Database Syst Rev* (2017.0) 2017. DOI: 10.1002/14651858.CD001431.pub5
16. Dumont A, Betrán AP, Kaboré C, de Loenzien M, Lumbiganon P, Bohren MA. **Implementation and evaluation of nonclinical interventions for appropriate use of cesarean section in low- and middle-income countries: protocol for a multisite hybrid effectiveness-implementation type III trial**. *Implement Sci* (2020.0) **15** 72. DOI: 10.1186/s13012-020-01029-4
17. Betrán AP, Temmerman M, Kingdon C, Mohiddin A, Opiyo N, Torloni MR. **Interventions to reduce unnecessary caesarean sections in healthy women and babies**. *The Lancet* (2018.0) **392** 1358-1368. DOI: 10.1016/S0140-6736(18)31927-5
18. Sekhon M, Cartwright M, Francis JJ. **Acceptability of healthcare interventions: an overview of reviews and development of a theoretical framework**. *BMC Health Serv Res* (2017.0) **17** 88. DOI: 10.1186/s12913-017-2031-8
19. Mascarello KC, Horta BL, Silveira MF. **Maternal complications and cesarean section without indication: systematic review and meta-analysis**. *Rev Saúde Pública* (2017.0) **51** 105. DOI: 10.11606/S1518-8787.2017051000389
20. 20National Institute for Health and Care Excellence. Caesarean section: clinical guideline CG132. Available at: https://www.nice.org.uk/guidance/cg132 (Acessed March 20, 2020). London, UK.: Royal College of Obstetricians and Gynaecologists.; 2019.
21. Hofmeyr GJ, Hannah M, Lawrie TA. **Planned caesarean section for term breech delivery**. *Cochrane Database Syst Rev* (2015.0). DOI: 10.1002/14651858.CD000166.pub2
22. Visco A. G., Viswanathan M., Lohr K. N., Wechter M. E., Gartlehner G., Wu J. M.. **Cesarean delivery on maternal request: maternal and neonatal outcomes**. *Obstetrics and gynecology* (2006.0) **108** 1517-1529. DOI: 10.1097/01.AOG.0000241092.79282.87
23. Weibel S, Neubert K, Jelting Y, Meissner W, Wöckel A, Roewer N. **Incidence and severity of chronic pain after caesarean section: A systematic review with meta-analysis**. *Eur J Anaesthesiol* (2016.0) **33** 853-865. DOI: 10.1097/EJA.0000000000000535
24. Keag OE, Norman JE, Stock SJ, Myers JE. **Long-term risks and benefits associated with cesarean delivery for mother, baby, and subsequent pregnancies: Systematic review and meta-analysis**. *PLOS Med* (2018.0) **15** e1002494. DOI: 10.1371/journal.pmed.1002494
25. Prior E, Santhakumaran S, Gale C, Philipps LH, Modi N, Hyde MJ. **Breastfeeding after cesarean delivery: a systematic review and meta-analysis of world literature**. *Am J Clin Nutr* (2012.0) **95** 1113-1135. DOI: 10.3945/ajcn.111.030254
26. Tähtinen RM, Cartwright R, Tsui JF, Aaltonen RL, Aoki Y, Cárdenas JL. **Long-term Impact of Mode of Delivery on Stress Urinary Incontinence and Urgency Urinary Incontinence: A Systematic Review and Meta-analysis**. *Eur Urol* (2016.0) **70** 148-158. DOI: 10.1016/j.eururo.2016.01.037
27. Hankins GDV, Clark SM, Munn MB. **Cesarean Section on Request at 39 Weeks: Impact on Shoulder Dystocia, Fetal Trauma, Neonatal Encephalopathy, and Intrauterine Fetal Demise**. *Semin Perinatol* (2006.0) **30** 276-287. DOI: 10.1053/j.semperi.2006.07.009
28. Hansen AK, Wisborg K, Uldbjerg N, Henriksen TB. **Elective caesarean section and respiratory morbidity in the term and near-term neonate**. *Acta Obstet Gynecol Scand* (2007.0) **86** 389-394. DOI: 10.1080/00016340601159256
29. Darmasseelane K, Hyde MJ, Santhakumaran S, Gale C, Modi N, Dewan A. **Mode of Delivery and Offspring Body Mass Index, Overweight and Obesity in Adult Life: A Systematic Review and Meta-Analysis**. *PLoS ONE* (2014.0) **9** e87896. DOI: 10.1371/journal.pone.0087896
30. Bager P, Wohlfahrt J, Westergaard T. **Caesarean delivery and risk of atopy and allergic disesase: meta-analyses**. *Clin Exp Allergy* (2008.0) **38** 634-642. DOI: 10.1111/j.1365-2222.2008.02939.x
31. Darabi B, Rahmati S, HafeziAhmadi MR, Badfar G, Azami M. **The association between caesarean section and childhood asthma: an updated systematic review and meta-analysis**. *Allergy Asthma Clin Immunol* (2019.0) **15** 62. DOI: 10.1186/s13223-019-0367-9
32. Huang L, Chen Q, Zhao Y, Wang W, Fang F, Bao Y. **Is elective cesarean section associated with a higher risk of asthma? A meta-analysis**. *J Asthma* (2015.0) **52** 16-25. DOI: 10.3109/02770903.2014.952435
33. Gurol-Urganci I, Bou-Antoun S, Lim CP, Cromwell DA, Mahmood TA, Templeton A. **Impact of Caesarean section on subsequent fertility: a systematic review and meta-analysis**. *Hum Reprod* (2013.0) **28** 1943-1952. DOI: 10.1093/humrep/det130
34. Kingdon C, Downe S, Betran AP. **Women’s and communities’ views of targeted educational interventions to reduce unnecessary caesarean section: a qualitative evidence synthesis**. *Reprod Health* (2018.0) **15** 130. DOI: 10.1186/s12978-018-0570-z
35. Kingdon C, Downe S, Betran AP. **Interventions targeted at health professionals to reduce unnecessary caesarean sections: a qualitative evidence synthesis**. *BMJ Open* (2018.0) **8** e025073. DOI: 10.1136/bmjopen-2018-025073
36. Shorten A, Shorten B, Keogh J, West S, Morris J. **Making Choices for Childbirth: A Randomized Controlled Trial of a Decision-aid for Informed Birth after Cesareana**. *Birth* (2005.0) **32** 252-261. DOI: 10.1111/j.0730-7659.2005.00383.x
37. Montgomery AA, Emmett CL, Fahey T, Jones C, Ricketts I, Patel RR. **Two decision aids for mode of delivery among women with previous caesarean section: randomised controlled trial**. *BMJ* (2007.0) **334** 1305. DOI: 10.1136/bmj.39217.671019.55
38. Sobhy S, Arroyo-Manzano D, Murugesu N, Karthikeyan G, Kumar V, Kaur I. **Maternal and perinatal mortality and complications associated with caesarean section in low-income and middle-income countries: a systematic review and meta-analysis**. *The Lancet* (2019.0) **393** 1973-1982. DOI: 10.1016/S0140-6736(18)32386-9
39. Souza J, Gülmezoglu A, Lumbiganon P, Laopaiboon M, Carroli G. **Caesarean section without medical indications is associated with an increased risk of adverse short-term maternal outcomes: the 2004–2008 WHO Global Survey on Maternal and Perinatal Health**. *BMC Med* (2010.0) **8** 71. DOI: 10.1186/1741-7015-8-71
40. Heitkamp A, Meulenbroek A, van Roosmalen J, Gebhardt S, Vollmer L, I de Vries J. **Maternal mortality: near-miss events in middle-income countries, a systematic review**. *Bull World Health Organ* (2021.0) **99** 693-707F. DOI: 10.2471/BLT.21.285945
41. Bishop D, Dyer RA, Maswime S, Rodseth RN, van Dyk D, Kluyts H-L. **Maternal and neonatal outcomes after caesarean delivery in the African Surgical Outcomes Study: a 7-day prospective observational cohort study**. *Lancet Glob Health* (2019.0) **7** e513-e522. DOI: 10.1016/S2214-109X(19)30036-1
42. Escobar MF, Gallego JC, Nasner D, Gunawardana K. **Management of abnormal invasive placenta in a low- and medium-resource setting**. *Best Pract Res Clin Obstet Gynaecol* (2021.0) **72** 117-128. DOI: 10.1016/j.bpobgyn.2020.08.004
|
---
title: 'Know-do gaps for cardiovascular disease care in Cambodia: Evidence on clinician
knowledge and delivery of evidence-based prevention actions'
authors:
- Nikkil Sudharsanan
- Sarah Wetzel
- Matthias Nachtnebel
- Chhun Loun
- Maly Phy
- Hero Kol
- Till Bärnighausen
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10022025
doi: 10.1371/journal.pgph.0000862
license: CC BY 4.0
---
# Know-do gaps for cardiovascular disease care in Cambodia: Evidence on clinician knowledge and delivery of evidence-based prevention actions
## Abstract
Cardiovascular diseases (CVD) are the leading cause of death in Cambodia. However, it is unknown whether clinicians in Cambodia provide evidence-based CVD preventive care actions. We address this important gap and provide one of the first assessments of clinical care for CVD prevention in an LMIC context. We determined the proportion of primary care visits by adult patients that resulted in evidence-based CVD preventive care actions, identified which care actions were most frequently missed, and estimated the know-do gap for each clinical action. We used data on 190 direct clinician-patient observations and 337 clinician responses to patient vignettes from 114 public primary care health facilities. Our main outcomes were the proportion of patient consultations and responses to care vignettes where clinicians measured blood pressure, blood glucose, body mass index, and asked questions regarding alcohol, tobacco, physical activity, and diet. There were very large clinical care shortfalls for all CVD care actions. Just $6.4\%$ ($95\%$ CI: $3.0\%$, $13.0\%$) of patients had their BMI measured, $8.0\%$ ($4.6\%$, $13.6\%$) their blood pressure measured at least twice, only $4.7\%$ ($1.9\%$, $11.2\%$) their blood glucose measured. Less than $21\%$ of patients were asked about their physical activity ($11.7\%$ [$7.0\%$, $18.9\%$]), smoking ($18.0\%$ [$11.8\%$, $26.5\%$]), and alcohol-related behaviors ($20.2\%$ [$13.7\%$, $28.9\%$]). We observed the largest know-do gaps for blood glucose and BMI measurements with smaller but important know-do gaps for the other clinical actions. CVD care did not vary across clinician cadre or by years of experience. We find large CVD care delivery gaps in primary-care facilities across Cambodia. Our results suggest that diabetes is being substantially underdiagnosed and that clinicians are losing CVD prevention potential by not identifying individuals who would benefit from behavioral changes. The large overall and know-do gaps suggest that interventions for improving preventive care need to target both clinical knowledge and the bottlenecks between knowledge and care behavior.
## Introduction
Reducing preventable mortality from cardiovascular diseases (CVD) is a major global health priority [1]. In Cambodia, the site of our study, CVDs are estimated to be the leading cause of death [2]. Major risk factors for CVD are also poorly controlled: $12\%$ of all adults ages 40+ have uncontrolled hypertension [3], $32\%$ of men smoke cigarettes [4], and $40\%$ of all individuals with diabetes are untreated [5]. Improving preventive care will thus be essential for meeting CVD mortality reduction targets. This need is especially pressing in middle-income countries like Cambodia, where rapid population aging is expected to dramatically increase the number of individuals in need of preventive care [6].
Conducting a CVD risk assessment where clinicians screen for hypertension, diabetes, and other key risk factors such as physical inactivity and tobacco use is the first essential clinical action for preventing CVD [7]. To address low health-seeking behavior for preventive care [8], major CVD prevention guidelines—including those in Cambodia—recommend that clinicians opportunistically screen and conduct CVD risk assessments for all adult patients that come to their clinics [7, 9]. Whether such efforts translate into improved patient outcomes, however, depends on whether clinicians actually provide guideline-recommended actions. The low levels of diagnosed hypertension and diabetes in Cambodia and other LMICs suggest that clinicians may not be consistently delivering the care actions specified in evidence-based prevention guidelines. To our knowledge, however, there are no studies that have measured clinician behavior regarding CVD prevention care in an LMIC context like Cambodia.
In addition to measuring shortfalls in clinical care for CVD, it is important to identify whether gaps are due to a lack of clinical knowledge or a gap between knowledge and care behavior. Clinicians may simply not be aware of screening guidelines or what actions they should take. This hypothesis is consistent with emerging evidence across Asian countries that finds low levels of clinician care knowledge for common conditions such as diabetes, tuberculosis, and child diarrhea [10–12]. Alternatively, physicians may know that they should conduct opportunistic screenings, but not do so in actual clinical care (this discrepancy between clinicians’ knowledge and behavior is referred to as the "know-do" gap [11, 13, 14]). Distinguishing between these two sources is important for crafting effective policy, as interventions to address poor clinical knowledge likely differ substantially from those designed to encourage clinicians with accurate knowledge of evidence-based care to change their care behaviors.
Here we provide one of the first assessments of clinical care for CVD prevention in an LMIC context. Our analysis benefits from data on observations of real patient consultations and clinicians’ responses to hypothetical patient vignettes. This allows us to investigate whether care gaps are driven by low levels of clinical knowledge (based on responses to the vignettes) or by a barrier between knowledge and behavior (the difference between responses to the vignettes and care for real patients). We also investigate if clinician knowledge and behavior are related to clinician cadre and overall clinician experience. Overall, our study results are important for assessing the quality of CVD care in Cambodia and informing interventions to improve clinician delivery of evidence-based CVD prevention care.
## Setting
Our study takes place in public Health Centres across Cambodia. Health Centres are responsible for delivering primary care throughout the country. The Health Centres provide a range of basic preventive and curative services including care for infectious and acute conditions (e.g. malaria, respiratory disease, and tuberculosis), family planning and ante and postnatal care, and basic care and screening for chronic non-communicable diseases including screening for hypertension and diabetes. The full range of services that the Health Centres provide is specified in the Cambodian Minimum Package of Activities [15]. Care in the Health *Centres is* free for poorer families through Cambodia’s Health Equity Fund [16]; families that are not eligible for the Health Equity Fund have to pay for their care based on a pre-specified fee schedule [16].
Clinical care in the Health *Centres is* provided by nurses, midwives, and medical doctors (physicians), with nurses and midwives providing the majority of care. While some of the care responsibilities are differentiated by clinician cadre (e.g. midwives focus more heavily on ante and postnatal care while physicians focus more heavily on the outpatient department), there is substantial overlap in the care responsibilities by cadre, and importantly, all three clinician types are expected to provide CVD preventive care that includes screening for hypertension and diabetes and assessing a range of behavioral risk factors.
Patients that arrive at the Health Centres are first directed to a reception area, after which they are directed to either an outpatient department for general medicine (primary care) or to a maternal child health/delivery department. After their consultation, patients are then directed to the station for wound dressing (if needed following minor surgeries) and pharmacy. If no additional care is required at that time the patient is then sent home; otherwise, the patient is admitted and referred to a public Referral Hospital [15].
The Government of Cambodia established evidence-based care guidelines for cardiovascular disease prevention (in the form of guidelines for nutrition, physical activity, hypertension, and diabetes care) as part of the first National Plan for Prevention and Control of Non-Communicable Diseases 2007–2010. These guidelines were updated for the second, 2013–2020, and third, 2018–2027 plans [17, 18]. Unfortunately, there is no definitive evidence on how thoroughly clinicians in Health Centres were trained on these guidelines; this is an important point we return to when contextualizing our findings in the Discussion section.
## Data
We use data collected from 114 public primary Health Centres across Cambodia between July and September 2020. These data were collected as part of an ongoing evaluation of a country-wide non-communicable disease quality improvement effort. All the study data were collected before the program implementation began.
We used a non-random procedure to select health facilities that was designed to enable an evaluation of the larger quality-improvement intervention. The Government of Cambodia selected 20 out of Cambodia’s 94 operational districts to implement the intervention. Each operational district contains on average 13 primary Health Centres which cater to a population between 100,000–200,000 individuals. Within each operational district, the government selected between 2–3 Health Centres to receive the non-communicable disease intervention. We collected data from the 2–3 Health Centres selected by the Government and a control set of 2–3 facilities, chosen to be similar to the intervention facilities in terms of geography and patient caseloads. The median number of patients seen in the facilities per month was 773, ranging from a minimum of 114 to a maximum of 2820 patients.
## Measurements
We collected two sources of data within each selected Health Centre: [1] direct observations of patient-clinician consultations and [2] clinical vignettes. We use the standard definition of “clinician” to mean all individuals who provide clinical patient care. In the Health Centres in Cambodia, this includes midwives, nurses, and physicians/medical doctors. We collected data from approximately 2 clinical consultations of patients ages 40+ per facility for a total sample of 190 patient observations. To do so, we stationed an enumerator in the clinician’s consultation room (after obtaining consent from both the patient and clinician). The observer had a checklist of clinician action items and marked items corresponding to what they observed during the consultation. Importantly, the observer did not directly interact with the patient or the clinician in any way. We observed a unique set of patients in each facility (e.g. it was not the case that the same patient met a nurse, midwife, and physician and was counted as three observations).
We complemented the direct observations with data on clinician responses to a hypothetical patient vignette. Administering clinical vignettes involves providing clinicians descriptions of hypothetical patients and asking them how they would proceed with care. Vignettes are often used as a measure of clinician knowledge and are included as part of many large health facility surveys, including in the Indonesian Family Life Survey [19]. For our study, enumerators presented clinicians with a short description of a hypothetical patient and asked how they would proceed with care. Enumerators described the patient as a 40-year-old who came to the clinic complaining of lower back pain. Clinicians were then asked what questions they would ask the patient, what tests they would run, what medicines or treatments they would prescribe, and what advice or counseling they would provide. Clinicians were not prompted beyond these questions and were only provided responses to the questions or tests they stated they would ask or conduct. We structured the vignette such that the hypothetical patient had uncontrolled blood pressure (values of $\frac{160}{90}$ mmHg, $\frac{159}{92}$ mmHg, and $\frac{159}{93}$ mmHg per measurement respectively), uncontrolled blood glucose (fasting plasma glucose > 126 mg/dl), and a body mass index of 29.4 kg/m^2. The hypothetical patient was also a non-smoker, with no reported alcohol habit, and no reported regular exercise. If the clinician asked a question or stated that they would conduct a test that we did not have pre-determined answers for, enumerators told the clinician that "the patient is unsure or does not know" or "the test came back normal." The vignette was administered to approximately 3 clinicians per facility, for a total sample of 337 vignette responses.
We collected data on the following CVD care items. Measurement items included whether the clinician measured BP at all, whether they measured BP at least twice, whether they measured blood glucose, and whether they measured body mass index. Risk assessment items included whether they asked about the patients’ smoking, alcohol, diet, and physical activity behaviors. We measured both whether clinicians conducted these actions with real patients and whether they stated that they would ask/measure these items in their responses to the hypothetical patient vignettes.
We also used data on clinician cadre (whether the clinician was a nurse, midwife, or medical doctor) and the number of years that the clinician reported practicing medicine.
## Statistical analyses
We first estimated the proportion of direct observations where clinicians measured or asked about each CVD care item. We then estimated these same proportions among clinical vignettes and constructed the "know-do" gap as the difference in each of the items between the patient observations and clinical vignettes.
Next, we examined whether there were differences in CVD care across clinician cadre by estimating Poisson regression models for each item with indicator variables for midwives and nurses (with physicians as the reference category). We examined the association with years of experience using similar regression models, this time with a continuous measure of years practicing medicine as the independent variable. We chose to estimate Poisson, rather than logistic, regression models so that our coefficients can be interpreted as prevalence, rather than odds, ratios. All our standard errors were clustered at the health facility level.
This study received ethics approval from the Ethics Commission of the Medical Faculty of Heidelberg University and the Cambodian National Ethics Committee for Health Research. We obtained written informed consent from health administrators to collect data within health centres and verbal informed consent from clinicians and patients before collecting any data. We used verbal, rather than written consent, to avoid disrupting regular clinical services and to avoid literacy and comprehension challenges with patients. Participant consent was noted by the observer in the data collection form. This procedure was approved by the ethics committees. We conducted all analyses using R Version 4.1.0.
## Inclusivity in global research
Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the S1 Checklist.
## Results
Nurses and midwives provided the most medical care (Table 1): 152 ($45\%$) of the vignette sample and 33 ($17\%$) of patient observations were midwives, 151 ($45\%$) of the vignette sample and 129 ($68\%$) of the patient observations were nurses, and less than $15\%$ of both samples were medical doctors (vignette sample: $10\%$, patient observations: $15\%$). On average, providers had around 11–12 years of experience practicing medicine Vignette sample: 10.8 (9.2), patient observation sample: 12.1 (9.9).
**Table 1**
| Unnamed: 0 | Clinicians who responded to the clinical vignettes (N = 337) | Clinicians providing care to the patients whose consultations were observed (N = 190) |
| --- | --- | --- |
| Qualification of clinicians in each sample | | |
| medical doctor | 34 (10.1%) | 28 (14.7%) |
| midwife | 152 (45.1%) | 33 (17.4%) |
| nurse | 151 (44.8%) | 129 (67.9%) |
| Years practicing among clinicians in each sample | | |
| Mean (SD) | 10.8 (9.22) | 12.1 (9.89) |
| Median [Min, Max] | 8.00 [0, 39.0] | 8.00 [0, 34.0] |
There were very large clinical care gaps for nearly all CVD care actions (Fig 1). Just $6.4\%$ ($95\%$ CI: $3.0\%$, $13.0\%$) of patients had their BMI measured, $8.0\%$ ($95\%$ CI: $4.6\%$, $13.6\%$) their blood pressure measured at least twice, only $4.7\%$ ($95\%$ CI: $1.9\%$, $11.2\%$) their blood glucose measured. Similarly, less than $21\%$ of patients were asked about their physical activity ($11.7\%$, $95\%$ CI: $7.0\%$, $18.9\%$), smoking ($18.0\%$, $95\%$ CI: $11.8\%$, $26.5\%$), and alcohol-related behaviors ($20.2\%$, $95\%$ CI: $13.7\%$, $28.9\%$).
**Fig 1:** *Know-do gaps for non-communicable disease screening measurement actions and behavioral risk assessments.Error bars represent 95% confidence intervals.*
Contrasting these results with responses to the clinical vignettes, we observed very large know-do gaps for blood glucose and BMI measurement. Clinicians stated that they would measure blood glucose among $86.6\%$ ($95\%$ CI: $82.1\%$, $90.2\%$) of vignettes and BMI among $73.3\%$ ($95\%$ CI: $68.0\%$, $78.0\%$) of vignettes. There were smaller but still important know-do gaps for the remaining care items. Based on vignette responses, $22.6\%$ ($95\%$ CI: $17.8\%$, $28.1\%$) of clinicians stated they would collect at least two blood pressure measurements, $20.8\%$ ($95\%$ CI: $16.7\%$, $25.5\%$) stated they would ask about physical activity, $33.8\%$ ($95\%$ CI: 28.6, $39.4\%$) stated they would ask about smoking, and $34.4\%$ ($95\%$ CI: $29.5\%$, $39.7\%$) stated they would ask about alcohol.
We did not find strong evidence of differences in the size of CVD care gaps by clinician cadre (Fig 2). The only exception was that midwives (PR: 0.19 [0.056, 0.760]) and nurses (PR: 0.33 [0.122, 0.906]) were less likely to ask patients about their physical activity compared to physicians. We also did not find strong evidence of an association between the number of years a clinician reported practicing medicine and any of the CVD care items (Fig 3). The detailed regression results for Figs 2 and 3 are available in Tables B-Q in S2 Text.
**Fig 2:** *Association between clinician cadre and CVD care actions based on univariate Poisson regression.Coefficients are presented as prevalence ratios relative to physicians. Error bars represent 95% confidence intervals. There is no 95% confidence interval for the coefficient on midwives measuring blood glucose since no midwives in the sample measured blood glucose (resulting in perfect prediction and a confidence interval of 0 width).* **Fig 3:** *Association between years practicing and CVD care actions based on univariate Poisson regression.Coefficients are presented as prevalence ratios associated with a 1-year increase in experience. Error bars represent 95% confidence intervals.*
## Discussion
We found evidence of large CVD care delivery gaps by clinicians in public primary-care facilities in Cambodia. Based on direct observations of clinical consultations, virtually no patients had their blood glucose measured and only small shares had their blood pressure measured correctly and were asked about important risk factors. These alarming findings have three important implications. First, due to the low rates of blood sugar measurement, it is likely that many individuals with diabetes are not being detected. Second, clinicians are losing a large and important source of CVD prevention potential by not identifying individuals who would benefit from behavioral changes. Lastly, while many clinicians measured blood pressure once, few measured it at least twice as recommended by most care guidelines [7, 20]. This suggests that clinical decisions around hypertension in Cambodia may be made with substantial error due to potential "white coat effects" [21].
Whether these gaps are driven by a lack of clinical knowledge or a discrepancy between knowledge and action depended on the specific care actions. There were large know-do gaps for blood glucose and BMI measurements, suggesting that knowledge is unlikely to be the main driver of care gaps. We observed smaller know-do gaps for the other CVD care items; for these care behaviors, both improving knowledge and addressing the gap between knowledge and action are likely to be needed to close overall care gaps. To our knowledge, this is the first study to measure know-do gaps for CVD care in an LMIC context; however, our results are consistent with several studies from LMICs that find large overall and know-do gaps for a range of other clinical areas, including tuberculosis, child diarrhea, and sick child care [10–12, 14].
An important question is whether the low shares of BMI, blood glucose, and blood pressure measurement is due to clinicians not having access to measurement devices. However, every facility in our study reported having a scale and height board and a sphygmomanometer, making this explanation unlikely for BMI and blood pressure. We did find that only $21\%$ of facilities had access to glucometers and test strips and that clinicians in facilities with these materials were 11 percentage points more likely to measure blood glucose although the $95\%$ confidence interval for this estimate was wide and overlapped the null. Importantly, even among facilities with access to blood glucose measurement devices, just $14\%$ of adult patients had their blood glucose screened, indicating a large remaining gap that is not attributable to equipment availability (S1 Text and Table A in S1 Text).
Our results implicate several areas where clinical knowledge improvements are needed. While the care actions we study are all outlined in the official Cambodian care guidelines, there is limited evidence on whether clinicians practicing in health facilities are aware of and have been trained on these guideline actions. Therefore, measuring what share of clinicians have been trained on guideline actions and subsequently, improving clinician training are likely an important first step for improving clinical care in the country. However, there is mixed evidence on whether education interventions alone are sufficient for meaningfully changing clinician behavior [22–25]. Reconciling these differences and understanding under which circumstances educational interventions are effective will be essential for improving outcomes in Cambodia and similar contexts. Further, our results suggest that even with effective education interventions, there is likely to be a gap between clinician knowledge and behavior. This discrepancy may be related to several factors. For example, clinicians working in public facilities in Cambodia do not have a direct financial—or extrinsic—incentive for providing proper CVD care. Intrinsically, clinicians may not be motivated or find personal satisfaction in providing CVD care if patients do not expect CVD care and thus do not value it in the same way that they value care for acute conditions. Developing and testing strategies that leverage both extrinsic and intrinsic motivators will be crucial for improving clinician behavior [26]. Clinicians may also be habituated to providing acute care and may require significant attention and focus to change their usual behavior towards CVD care. Therefore, clinicians who know and even intend on providing CVD care may face effort barriers and thus opt for the forms of care that they are most accustomed to providing. In such circumstances, providing clinical support tools and checklists that ease the attention and effort needed for clinicians to provide new forms of care may help to improve clinician CVD care behavior [27]. Clinicians may also simply lack the equipment (glucometers) needed to correctly conduct CVD risk assessments.
One of the primary limitations of our paper is that it is based on a non-random selection of public health facilities from 20 out of Cambodia’s 94 operational districts. Based on discussions with Government officials and implementers, the 20 operational districts that were selected to receive the intervention are more likely to have the capacity to provide diabetes and hypertension care compared to the remaining 74 districts. Within districts, the Government selected health facilities with higher patient caseloads and better baseline levels of care quality compared to non-selected facilities. This non-random selection, however, would imply that clinical care behavior among a representative set of facilities would be potentially worse than the levels we document here. Our measures of clinician behavior are based on direct observations and may be biased if clinicians change their behavior when they know they are being watched. This bias, however, would also imply that the already low levels of CVD care behavior observed here may be an overestimate. Our study design only allowed us to measure CVD care at the screening stage and did not allow us to investigate CVD management behaviors by clinicians. Future studies that leverage longitudinal tracking of patients will be essential for measuring care gaps among individuals who have already screened positive for CVD care needs. While we measure many key aspects of preventive care, our analysis does not examine all actions, including for example an assessment of family history. However, as family history is not a substitute for the other actions, this omission would not bias our assessment of the share of clinicians that are not meeting the other clinical actions. Lastly, our study did not investigate whether some types of patients were more likely to receive screening actions compared to others. This is an important question and future research should collect more detailed information on patient characteristics to determine whether certain population groups are being disproportionately overlooked by clinician screening.
Clinician behavior forms a key component of effective CVD prevention care. Yet, there have been few studies that have examined to what extent clinicians in an LMIC context are providing high-quality CVD care and whether care gaps reflect knowledge or other behavioral barriers. Our results reveal large care gaps for key CVD prevention actions, such as measuring blood sugar and blood pressure. Developing interventions that both improve clinician knowledge and address the barriers between knowledge and action will be crucial for closing care gaps and improving CVD prevention in Cambodia and similar contexts.
## References
1. Bennett JE, Stevens GA, Mathers CD, Bonita R, Rehm J, Kruk ME. **NCD Countdown 2030: worldwide trends in non-communicable disease mortality and progress towards Sustainable Development Goal target 3.4**. *The Lancet* (2018.0) **392** 1072-88. DOI: 10.1016/S0140-6736(18)31992-5
2. 2GBD Compare. cited 2021 Nov 30. Institue for Health Metrics and Evaluation. Available from: https://vizhub.healthdata.org/gbd-compare/.
3. Sudharsanan N, Theilmann M, Kirschbaum TK, Manne-Goehler J, Azadnajafabad S, Bovet P. **Variation in the Proportion of Adults in Need of Blood Pressure-Lowering Medications by Hypertension Care Guideline in Low- and Middle-Income Countries: A Cross-Sectional Study of 1 037 215 Individuals From 50 Nationally Representative Surveys**. *Circulation* (2021.0) **143** 991-1001. DOI: 10.1161/CIRCULATIONAHA.120.051620
4. 4National Institute of Statistics/Cambodia, Directorate General for Health/Cambodia, ICF International. Cambodia Demographic and Health Survey 2014. Phnom Penh, Cambodia; 2015.
5. Flood D, Seiglie JA, Dunn M, Tschida S, Theilmann M, Marcus ME. **The state of diabetes treatment coverage in 55 low-income and middle-income countries: a cross-sectional study of nationally representative, individual-level data in 680 102 adults**. *The Lancet Healthy Longevity* (2021.0) **2** e340-e351. DOI: 10.1016/s2666-7568(21)00089-1
6. Sudharsanan N, Geldsetzer P. **Impact of Coming Demographic Changes on the Number of Adults in Need of Care for Hypertension in Brazil, China, India, Indonesia, Mexico, and South Africa**. *Hypertension* (2019.0) **73** 770-6. DOI: 10.1161/HYPERTENSIONAHA.118.12337
7. 7World Health Organization. HEARTS technical package for cardiovascular disease management in primary health care: risk based CVD management. cited 2021 Dec 05. Geneva: 2020. Available from: https://apps.who.int/iris/bitstream/handle/10665/333221/9789240001367-eng.pdf.
8. Dupas P.. **Health Behavior in Developing Countries**. *Annu Rev Econ* (2011.0) **3** 425-49. DOI: 10.1146/annurev-economics-111809-125029
9. Arnett Donna K., Blumenthal Roger S., Albert Michelle A., Buroker Andrew B., Goldberger Zachary D., Hahn Ellen J.. **2019 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease**. *Journal of the American College of Cardiology* (2019.0) **74** e177-e232. DOI: 10.1016/j.jacc.2019.03.010
10. Kwan A, Daniels B, Saria V, Satyanarayana S, Subbaraman R, McDowell A. **Variations in the quality of tuberculosis care in urban India: A cross-sectional, standardized patient study in two cities**. *PLoS Med* (2018.0) **15** e1002653. DOI: 10.1371/journal.pmed.1002653
11. Mohanan M, Vera-Hernández M, Das V, Giardili S, Goldhaber-Fiebert JD, Rabin TL. **The know-do gap in quality of health care for childhood diarrhea and pneumonia in rural India**. *JAMA Pediatr* (2015.0) **169** 349-57. DOI: 10.1001/jamapediatrics.2014.3445
12. Stein DT, Sudharsanan N, Dewi S, Manne-Goehler J, Witoelar F, Geldsetzer P. **Change in clinical knowledge of diabetes among primary healthcare providers in Indonesia: repeated cross-sectional survey of 5105 primary healthcare facilities**. *BMJ open diabetes research & care* (2020.0). DOI: 10.1136/bmjdrc-2020-001415
13. Das J, Hammer J. **Money for nothing: The dire straits of medical practice in Delhi, India**. *Journal of Development Economics* (2007.0) **83** 1-36. DOI: 10.1016/j.jdeveco.2006.05.004
14. Gage AD, Kruk ME, Girma T, Lemango ET. **The know-do gap in sick child care in Ethiopia**. *PLoS One* (2018.0) **13** e0208898. DOI: 10.1371/journal.pone.0208898
15. 15Operational Guidelines On Minimum Package of Activities. Ministry of Health Cambodia 2017 [cited 29 Mar 2022].
16. Annear P, Eang R, Jacobs B. **Providing access to health services for the poor: Health equity in Cambodia**. *Studies in Health Services Organisation and Policy* (2008.0) **23**
17. 17National Strategic Plan for the Prevention and Control of Noncommunicable Diseases (2013–2020). Cardiovascular Disease, Cancer, Chronic Respiratory Disease and Diabetes. Ministry of Health Cambodia 2013.
18. 18National Multisectoral Action Plan for the Prevention and Control of Noncommunicable Diseases 2018–2027. Royal Government of Cambodia 2018 [cited 29 Mar 2022]. Available from: http://moh.gov.kh/content/uploads/2017/05/NMAP-NCD_-13-06-2018-Signed_En.pdf.
19. Strauss J, Witoelar F, Sikoki B. *Community-Facility Survey Questionnaire for the Indonesia Family Life Survey* (2016.0)
20. Franklin SS, Thijs L, Hansen TW, O’Brien E, Staessen JA. **White-coat hypertension: new insights from recent studies**. *Hypertension* (2013.0) **62** 982-7. DOI: 10.1161/HYPERTENSIONAHA.113.01275
21. Tanner RM, Shimbo D, Seals SR, Reynolds K, Bowling CB, Ogedegbe G. **White-Coat Effect Among Older Adults: Data From the Jackson Heart Study**. *J Clin Hypertens (Greenwich)* (2016.0) **18** 139-45. DOI: 10.1111/jch.12644
22. Chauhan BF, Jeyaraman MM, Mann AS, Lys J, Skidmore B, Sibley KM. **Behavior change interventions and policies influencing primary healthcare professionals’ practice-an overview of reviews**. *Implement Sci* (2017.0) **12** 3. DOI: 10.1186/s13012-016-0538-8
23. Seidu S, Walker NS, Bodicoat DH, Davies MJ, Khunti K. **A systematic review of interventions targeting primary care or community based professionals on cardio-metabolic risk factor control in people with diabetes**. *Diabetes Res Clin Pract* (2016.0) **113** 1-13. DOI: 10.1016/j.diabres.2016.01.022
24. Mostofian F, Ruban C, Simunovic N, Bhandari M. **Changing physician behavior: what works**. *Am J Manag Care* (2015.0) **21** 75-84. PMID: 25880152
25. Thepwongsa I, Kirby C, Schattner P, Shaw J, Piterman L. **Type 2 diabetes continuing medical education for general practitioners: what works? A systematic review**. *Diabet Med* (2014.0) **31** 1488-97. DOI: 10.1111/dme.12552
26. Judson TJ, Volpp KG, Detsky AS. **Harnessing the Right Combination of Extrinsic and Intrinsic Motivation to Change Physician Behavior**. *JAMA* (2015.0) **314** 2233-4. DOI: 10.1001/jama.2015.15015
27. Shah MK, Kondal D, Patel SA, Singh K, Devarajan R, Shivashankar R. **Effect of a multicomponent intervention on achievement and improvements in quality-of-care indices among people with Type 2 diabetes in South Asia: the CARRS trial**. *Diabet Med* (2020.0) **37** 1825-31. DOI: 10.1111/dme.14124
|
---
title: 'Analysis of intermunicipal journeys for cardiac surgery in Brazilian Unified
Health System (SUS): an approach based on network theory'
authors:
- Ludmilla Monfort Oliveira Sousa
- Hernane Borges de Barros Pereira
- Edna Maria de Araújo
- José Garcia Vivas Miranda
journal: International Journal for Equity in Health
year: 2023
pmcid: PMC10022046
doi: 10.1186/s12939-023-01857-y
license: CC BY 4.0
---
# Analysis of intermunicipal journeys for cardiac surgery in Brazilian Unified Health System (SUS): an approach based on network theory
## Abstract
### Introduction
The transformation of data into information is important to support decision making and, thus, to induce improvements in healthcare services. The regionalized organization of healthcare systems is necessary to ensure the integrity of citizen care. From this perspective, the creation of mechanisms to guide and assess the behavior of a healthcare services network becomes necessary. However, these mechanisms must consider the interaction between different municipalities. The objective of this study is to apply network analysis as a supporting tool in the Brazilian Unified Health System (Sistema Único de Saúde—SUS) management.
### Methods
The stages of the proposed method are described and applied in a real situation, analyzing the intermunicipal interaction network for cardiovascular surgery in the municipality of Vitória da Conquista, Bahia, Brazil, from 2008 to 2020. The metrics analyzed were journeys indices, flow of patients and distance of the journeys, considering the journeys from and to the municipality in focus.
### Result
There was an increase of the incoming flow and in-degree indices combined with a decrease in outgoing flow, showing the growing importance of this municipality as a provider of these services.
### Conclusion
The method used in the study has potential to be adopted as a management tool to assess the behavior of the interactions network of the selected service, aiding the regionalized organization of the healthcare system.
## Introduction
In Brazil, various databases that can be used by healthcare managers to support decision making are freely and openly available [1]. However, these secondary data sources are underused, and some of the reasons are absence of informational culture and of trained personnel to work with the data [2]. Professionals that work in healthcare management can benefit from the incorporation of information from such databases in managing [2, 3]. On this account, developing mechanisms to facilitate the transformation of data into information and then using them in healthcare monitoring and assessment, for decision making, can potentialize healthcare managers’ actions.
The Brazilian public healthcare system, known as SUS (Sistema Único de Saúde—Unified Health System), needs to ensure integral assistance to citizens, thus, when necessary citizens must be referred to other healthcare units, including to other municipalities. That way, healthcare institutions that comprise SUS’s services network should be decentralized and regionally organized [4, 5] composing a hierarchical network that can articulate from the simplest units to the most complex ones, through a system of reference and contra reference to different degrees of complexity. In Brazil, this healthcare regionalization is done with a division of the states’ territories into Healthcare Regions [6, 7].
For building a regionalized and hierarchical network, the responsibility for managing and SUS financing are shared between the three levels of government, namely: the Union, the states, and the municipalities. However, the SUS has as one of its organizational principles the decentralization of management and public health policies, which, in turn, have the purpose of transferring responsibility and resources to municipalities. In this way, municipal managers are encouraged to be attentive to the needs of the population residing in the municipality in which they operate and to offer the appropriate health services. However, most municipalities are not able to provide all health services, and that is why the regionalization of these services is so important to guarantee the integrality of care for citizens. And to build this regionalized network of health services, there are the commissions of managers between levels of government, which are intergovernmental spaces for planning, negotiating, and implementing public health policies [6, 7].
Network approaches are used in decision-making processes to improve healthcare efficiency. The network analysis was also employed as one of the processes to constructing models to be used in healthcare [8, 9]. Some studies were applied for prediction and diagnosis objectives [10], medical effect prediction [11], improvement of medicine distribution logistics [12], and other reasons.
For this reason, creating mechanisms to assess and guide the intermunicipal networks that provide citizens with access to healthcare services is necessary, considering the interrelation between municipalities, and verifying how a location influences and is influenced by those around it. The Hospital Information System (Sistema de Informação Hospitalar—SIH-SUS) enables the verification of the municipality in which SUS patients live and the municipality in which they are hospitalized [13], thus enabling the analysis of the incoming and outgoing flows of patients for hospitalizations. In this way, the municipal manager would be able to carry out the following analyses: where do more patients come from for the municipality under my management? Where do patients residing in the municipality under my management go? Do they make big displacements? Could it be that with this volume and the distance covered, would it not be better to offer certain services in strategically located municipalities? How to regionalize health services to offer them more efficiently and safely to patients? Patients' incoming and outgoing flow is a phenomenon that can be modeled with a network, enabling the analysis of how this network is organized.
*In* general terms, relationships between entity pairs can be represented through networks. Mathematically, a network is represented by a graph. A graph is composed of a set of vertices (or nodes, or points), and edges or arcs (or links, or lines). If two vertices are linked by a line it means that they are connected because they keep some relation of interest. In building a network it is possible to study its entities from a local, global, topological, temporal and spatial perspective and it is possible to investigate the mechanisms of interaction between the components that form the system [14–19].
Aiming at supporting decision making in healthcare systems, specifically in the assessment of SUS patients intermunicipal flow, the objective of the study reported in this article is to propose a method for the analysis of intermunicipal journeys for hospitalizations in SUS, using network analysis as a management tool for SUS. To describe the application of the method, the networks of cardiovascular surgeries hospitalization, which represent a high complexity healthcare service, were investigated.
## The proposed method
In this article we describe the network analysis method to analyze SUS patients’ intermunicipal journeys for healthcare services. Similar analyses had been conducted in other studies [20, 21], however, these studies focus on analysis at the State level. Moreover, the incoming and outgoing flows of patients have been used to analyze citizens' access to health services and strengthening of regionalization. Initial studies have already demonstrated this type of displacement in general in Brazil [22, 23] and focused on these types of patient flows. More recent research demonstrated the incoming and outgoing flows and added some other network indices to perform this analysis [20, 21, 24, 25].
The difference of this article is the application of the network analysis method at the municipality level, with a detailed description of the method, so that it can be replicated by healthcare managers and other professionals working with healthcare data. After all, the positive impacts observed in State and *National data* is a result of the organization of the healthcare system in the municipalities. This method can be used for different types of healthcare systems.
To detail the proposed method, the flowchart (Fig. 1) describes all the steps followed in this study. Fig. 1Flowchart describing the proposed method for network analysis
## Definition of the network of interest
There are various Healthcare Information System (Sistema de Informação em Saúde—SIS) with many different databases that provide us an infinitive of health data with large potential. This variety of databases, such as the Hospital Information System (Sistema de Informação Hospitalar—SIH), Ambulatorial Information System (Sistema de Informação Ambulatorial—SIA), Minimum Data Set (Conjunto Mínimo de Dados—CMD), among others, can be used in network analyses.
First, it is necessary to define which healthcare service to analyze, because the ideal spatial distribution of the healthcare units in the territory may be different according to the type of service under study. It is also important to know if the information system used has the necessary data to generate a network. The system must have at least the municipality of origin and the destination of SUS patients.
To demonstrate the application of the network theory in the present study, the journeys for intermunicipal cardiac hospitalizations were analyzed, thus representing a high complexity modality of hospital care. Cardiac problems deserve emphasis because of the relevance of cardiovascular disease in the national scenario, considering that it is the main cause of death in Brazil [22–29] and the cause of hospitalizations with the highest expenses.
Only one municipality was selected to facilitate the demonstration of the proposed method. Therefore, the results found in one municipality may be different in other municipalities. Within this context, it is important to analyze the results and contextualize them in the territory where the information was recorded, because what happens in one territory can be different in another territory. To demonstrate the cardiovascular surgery network, the municipality of Vitória da Conquista [26], Bahia, Brazil, was selected. This municipality is located in the Southwest macroregion of the state of Bahia (Fig. 2), being the municipality of reference of the macroregion that showed the highest decrease in the number of cardiovascular surgeries as well as the lowest proportion of municipalities without records about cardiovascular surgery. Fig. 2a Spatial location of Bahia in Brazil. b Spatial location of Vitória da Conquista in Bahia
## Data collection
To construct the networks, it is necessary to prepare two types of files, one for the arcs, and other for the vertices. The file with the arcs contains the patients' journeys, from the municipality of origin to the destination in which the procedure would be performed. The file with information about the vertices contains the spatial locations of the municipalities in the territory that will be analyzed. The documents and databases are on a server which can be accessed by the URL https://doi.org/10.5281/zenodo.7470351.
## SIH / List of hospitalizations
To create the file containing the arcs, data from the Hospital Information System (SIH/SUS) were used. The SIH/SUS is the system in which the records of the patients hospitalized in the units that are part of SUS (public or associated units) are processed.1 These units send hospitalization information collected through the Hospitalization Authorization Form (Autorização de Internação Hospitalar—AIH). After data is sent to Datasus, SUS Department of Informatics, this information becomes part of the national database, for further dissemination [13]. These files about the AIH are generated monthly by Brazilian municipalities. In the present study the files from Bahia (RDBA) from 2008 to 2020 were downloaded. The downloaded files are organized in the format RDBAaamm.dbc, in which “aa” is the year of the file and “mm” is the month. The variables “municipality of origin” and “destination” are some of the information available in the hospitalization list generated by the SIH/SUS, which are necessary to analyze the flow of patients.
## INDE / Coordinates
Data of the National Infrastructure of Spatial Data (Infraestrutura Nacional de Dados Espaciais2 - INDE) were used for the creation of the file containing the information about the network vertices. This web page makes geospatial data available through a network of servers connected to the internet. This web page provides files of type “shape” for download of the cities in the 417 municipalities in Bahia. To do so, the option “Geo serviços” (Geo services) was selected, then the INDE visualizer (VINDE), then “Adicionar camadas” (Add layers) in the option “Temas” (Themes). In this stage, two maps were selected: one to get the location of the city of Salvador and the other to get the locations of the cities in Bahia municipalities. The map containing Brazilian capitals (BCIM Capital—Ponto) was used to get Salvador’s location and the map of cities (BCIM Cidade—Ponto) was used to get the locations of the other cities in Bahia municipalities.
## List of municipalities of origin and destinations
The intermunicipal journeys by SUS patients were organized as matrices of municipality of origin and destination. This matrix is analogous to the matrix of migration, or cost matrix, used to represent a graph/network. In this matrix, the municipalities in Bahia were organized in a way so that the lines contained the municipalities of residence (municipalities of origin) and the columns contained the municipalities of occurrence of the group of procedures under study. The matrices of origin and destination were constructed with the software Tabwin [30].
*To* generate the matrix of cardiovascular surgery, SUS codes of procedures listed in the Covenanted and Integrated Plan (Programação Pactuada e Integrada—PPI) about cardiovascular surgery were filtered, in the intermunicipal hospital plan of high complexity [31], along with the municipalities of residence and occurrence in Bahia. To visualize the temporal evolution of the municipalities’ intermunicipal networks, a matrix of origin and destination was generated for each year. Subsequently, these data were reorganized so that they could be imported in a network analysis tool. In this study, the tool used was Gephi. Each matrix of origin and destination generated another file, here called list of origin and destination, containing the columns origin, destination and weight. In this newly generated file, the origin column is the municipality of residence, the destiny column is the municipality in which the procedure was conducted, and the weight column represents the number of people that traveled from the municipality of residence to the municipality of the procedure.
## Coordinates in Lat/Long and UTM
The software ArcGis was used to reorganize the data for the construction of the file with the vertices of the network. The file with the vertices was generated to contain two different formats of coordinates: [1] latitude and longitude that must be in decimal degrees (lat/long), and [2] coordinates in UTM (Universal Transverse Mercator). To do the spatialization of the municipalities’ cities in Gephi, the coordinates used were in lat/long, and to calculate the mean length of the arcs it is important that the coordinates used are in UTM.
## Construction of the networks
With the data organized, two types of report were generated with the software Gephi and R.
With Gephi, the spatial distributions of the networks were generated and the network indices were calculated, which in this study were called migration indices, as they are linked to people's journeys. The two migration indices generated were in-degree and out-degree. With R, four statistical indices were generated, namely: incoming flow, outgoing flow, mean length of the incoming arcs and mean length of the outgoing arcs. Based on the indices generated from the networks’ spatial distributions, it is possible to characterize the hospitalization network of a municipality and to know how many people and from how many different municipalities people are looking for healthcare services in the municipality under analysis. On the other hand, it is also possible to verify how many people are traveling out of the municipality, the municipality from which they are traveling and the number of different municipalities to which they are traveling for procedures.
## Analyses of the networks
The intermunicipal journeys organized as networks can be considered as a directed graph. A graph is a structure \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G=(V,A)$$\end{document}G=(V,A) in which \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V$$\end{document}V is the set of vertices or nodes and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$A$$\end{document}A is the set of arcs of the network, considering that the networks in this study are directed [14].
The indices used in the analyses of the networks in this study were: in-degree, out-degree, incoming flow, outgoing flow, mean length of incoming arcs and mean length of outgoing arcs [20, 21].
In the present study, the in-degree index is the number of arcs coming into a vertex (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{I}$$\end{document}kI). In this study, the interpretation of this index is similar to the immigration of patients to a given municipality. The municipality in which the procedure is conducted is the reference, thus this index quantifies the number of different municipalities from which people traveled in search for a procedure in the municipality under analysis. The out-degree is the number or arcs going outwards of a vertex (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{O}$$\end{document}kO), in this study its interpretation is similar to the emigration of patients, representing the municipality of residence and thus quantifies the number of different municipalities to which people travel in search of the procedure. The incoming flow is the weight of the incoming arcs of a vertex (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{I}$$\end{document}FI). The interpretation of this index in this study is the number of people that arrive at a municipality for the procedure in question. The outgoing flow is the weight of the outgoing arcs (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{O}$$\end{document}FO), and in this study it measures the number of people that traveled out of their municipality of residence for a procedure in another municipality.
The mean length of the outgoing arcs, initially presented by Sousa et al. [ 2017] [20] and used in Sousa et al. [ 2020] [21] is represented by Eq. 1:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\overline D=\frac1{\sum_{$a = 1$}^{k_0}F_a}\sum_{$a = 1$}^{k_0}D_aF_a$$\end{document}D¯=1∑$a = 1$k0Fa∑$a = 1$k0DaFain which \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{O}$$\end{document}kO is the out-degree, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${D}_{a}$$\end{document}*Da is* the distance of the journey, in kilometers, from the municipality of residence to the destination municipality, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{a}$$\end{document}*Fa is* the number of people that traveled.
The equation above represents the weighted average of the distances between the municipalities in which outgoing flow of patients were observed [20]. It is important to emphasize that this index considers the straight-line distance, in other words, the shortest distance between two municipalities, with the possibility of longer real distances depending on the road layout and the means of transportation used. Notwithstanding, considering the processes of analysis and decision making, this index turned out to be appropriate to measure the distances that citizens are traveling to have access to the healthcare service in question.
Finally, the mean length of the incoming arcs, whose value represents the weighted average of the distances between the municipalities in which incoming flow of people for the procedure under analysis was observed, is similar to Eq. 1, replacing the out-degree (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{O}$$\end{document}kO) with the in-degree (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{I}$$\end{document}kI).
## Results
Applying the proposed method of network analysis to Vitória da Conquista showed that this municipality started conducting cardiac surgery procedures in 2010 with a constant increase in the number of such procedures each year (Fig. 3). From 2015, it became the municipality with the second highest number of this type of surgery in the state of Bahia, only behind Salvador. Fig. 3Number of hospitalizations for cardiovascular surgery in Bahia, by service provider municipality, 2008 to 2020. Data Source: SUS Hospital Information System (Sistema de Informações Hospitalares do SUS—SIH/SUS) Figure 4 shows the evolution of the network of intermunicipal hospitalization for cardiovascular surgery in the municipality of Vitória da Conquista. In 2010, there were records of hospitalizations only in municipalities of the Southwest macroregion, and from 2014 hospitalizations of patients from 5 different macroregions were registered. Fig. 4Evolution of Vitória da Conquista intermunicipal network of cardiovascular surgery, vertices geographically localized. Bahia, 2010 to 2020 The indices shown in Table 1 enables the analysis of the evolution of the intermunicipal network of cardiovascular surgery. From the in-degree, it is possible to observe that in 2010 Vitória da Conquista had records of patients coming from 8 different municipalities, and in 2020 this number increased to 115. From these 115 municipalities, 71 (61,$7\%$) were from the same Southwest macroregion, 23 ($20\%$) from the West macroregion and 18 (15,$7\%$) from the South macroregion. In relation to the out-degree, however, in 2012 and 2014 there were records of cardiovascular surgery in Vitória da Conquista citizens in two different municipalities, namely Salvador and Itabuna, and in the other years only in Salvador. Table 1Indices of the Vitória da Conquista intermunicipal network of cardiovascular surgery. Bahia, 2008 to 2020IndicesNetwork indicesStatistical indicesIn-degreeaOut-degreebIncoming FlowcOutgoing FlowdMean length of incoming arcs (km)eMean length of outgoing arcs (km)f200801035NA326.52200901043NA326.522010811325131.57326.522011351665103.33326.5220125921626127.25274.3620137111831167.57326.5220147522573146.33274.3620159013074156.54326.5220168913625159.52326.5220179313986147.80326.52201810314672169.32326.522019104144510200.45326.52202011514564194.98326.52NA [not applicable]aIn-degree: number of different municipalities from which patients traveled for the procedurebOut-degree: number of different municipalities to which patients traveled for the procedurecIncoming Flow: number of patients that arrived at the municipality for the proceduredOutgoing Flow: number of residents that traveled to other municipalities for the procedureeMean length of incoming arcs (km): mean distance traveled by the patient to arrive at the municipality in which the procedure is conductedfMean length of outgoing arcs (km): mean distance traveled by the patient to go to the municipality in which the procedure is conducted The incoming flow measures the number of patients from other municipalities that went to Vitória da Conquista for cardiovascular surgery. Table 1 shows evidence that this number continually increases, since it began in 2010, with 13 hospitalization records, and in 2020 with 456 hospitalization records of patients from other municipalities. The analysis of the outgoing flow enables the observation of a decrease in the number of patients that traveled out of Vitória da Conquista for such procedures in other municipalities from 2010. In 2008 there were 35 hospitalization records of Vitória da Conquista residents in other municipalities, but in 2020 this number decreased to 4 patients.
The mean length of the incoming arcs enables the estimate of the straight-line distance of the journey by SUS patients to the municipality in which the procedure was conducted. The analysis of this index in Table 1 shows that it has been continually increasing. In 2010, this mean distance was about 131 km, but in 2019 the longest distance in the historic series was about 200 km. Regarding the mean length of the outgoing arcs, the distance was about 327 km, but in the years 2012 and 2014 there was a reduction to approximately 275 km.
## Discussion
In this study, Vitória da Conquista was considered increasingly important for the access to cardiovascular surgery in the state of Bahia. The crescent number of hospitalizations for cardiac surgery along the historic series combined with the increase in this type of hospitalization of Vitória da Conquista residents and with the decrease in the outgoing flow suggests that Vitória da Conquista residents do not need to travel to other municipalities to access this healthcare service. Vitória da *Conquista is* the third largest municipality in Bahia [32] and has the highest gross domestic product (GDP) and the highest increase in this index in the Southwest macroregion [33]. This result is in line with the study about geographical and social inequalities in the access of healthcare services in Brazil [34], which shows evidence that the citizen’s location of residence influences the access to healthcare services, and that this access improves according to the socioeconomic development of the region.
The incoming arcs, or mean distance of the journey to arrive at the municipality in which the procedure is conducted, increased continuously during the historic series. Since there was an increase in the in-degree, the increase in the mean length of the arcs can be a consequence of the increase in hospitalizations of people from other municipalities, especially municipalities from distant macroregions. In 2020, the mean length of the incoming arcs was approximately 195 km. A study that mapped the hospital care network in Brazil [35] observed that only $3\%$ of the patients that were hospitalized for cardiac surgery were residents in municipalities more than 60 km apart from the municipality in which the procedure was conducted, and about $40\%$ of Brazilian population live in those municipalities that are more than 60 km apart from the healthcare service. The regionalization of healthcare services is important to ensure the equity in the access and use of healthcare services, however, the studies suggest that regionalization is still a challenge to overcome. Since distance can be a factor that interferes in the access and use of healthcare service, the definition of the arc length can be an important index in planning the municipalities to which patients traveled in search of the procedure, which serve as municipalities of reference.
In two years, 2012 and 2014, there was a decrease in the mean length of the outgoing arcs, or mean distance of the journey to arrive at the municipality in which the procedure is conducted. This variation occurred according to the increase in the out-degree. In these two years there were records of cardiovascular surgery in Salvador and Itabuna. Since *Itabuna is* closer from Vitória da Conquista in comparison with Salvador, the mean length of the outgoing arcs decreased reflecting this smaller distance.
Looking at the sharp increase in cardiovascular surgeries in Vitória da Conquista in the period under analysis is important, especially involving patients from municipalities in different macroregions. Because of the Covenanted and Integrated Plan (Programação Pactuada e Integrada—PPI) [31], every municipality has defined which other municipality is a reference in conducting a given healthcare service. The data from this method of analysis can support the development of PPI, supporting in the definition of the municipalities of reference in conducting cardiac surgery services in the state of Bahia and, particularly, in Vitória da Conquista.
Despite hospitalizations for cardiovascular surgery being selected in this study, the proposed method can be used to analyze any other type of hospitalization. However, it is relevant not to combine different types of hospitalizations in a single network, since each type of hospitalization can inform different layouts of regionalization.
To facilitate the application of the proposed method, only the hospitalizations in Bahia were considered. However, this method can be used to verify hospitalizations between municipalities in different states.
Secondary data from healthcare information databases available online were used to demonstrate the application of the method. Because of the great volume of data, the availability of information and the easy access to databases about hospitalizations, the use of such information increased in Collective Health [36]. Thus, the use of secondary databases is important in assessment studies, because of their low cost, the wide availability of data, the speed of data collection and results, and for their possible impact in healthcare services [37].
Considering that secondary databases from the Healthcare Information System were used, it was and it is important to verify the reliability of the data used in the system, to make a robust and adequate healthcare planning. From this perspective, the field that records the data “conducted procedure” (“procedimento realizado”) has been considered satisfactorily reliable by some authors [38–40]. On the other hand, regarding the field “municipality of residence” (“município de residência”), a study that analyzed its quality for woman hospitalized because of breast or cervical cancer in Rio de Janeiro [41] found an accuracy of $82.9\%$ when compared to records from the Information System about Mortality (Sistema de Informação sobre Mortalidade—SIM) and SIH/SUS, suggesting high reliability; however, this reliability decreased when a woman lived outside the municipality in which hospitalization was required. Since the proposed method considered only the patients whose municipality of residence was different from the municipality in which they were hospitalized, it is important to consider some limitations, such as the omission of the true municipality of residence. Even with this limitation, the studies conducted in Rio de Janeiro showed enough quality in the field “municipality of residence” (“município de residência”) to support the decision-making process in healthcare management [41].
There are various challenges in the construction of a system that must be unified and in compliance with a regionalized network of services [4, 5, 42]. Therefore, the creation of permanent mechanisms of network assessment that can support the construction of regional planning becomes indispensable. Network analysis, as presented in this article, can reveal general aspects of the phenomenon under assessment, and the combination with a qualitative approach is advised, because they are complimentary, enabling the assessment of different peculiarities of the same phenomenon [43].
Since the cutoff of this paper was for only one municipality, the network approach loses its strength. In this case, the same conclusions could have been obtained using a statistical based on the flow databases analysis of the municipalities together with their geolocation. However, the network approach, in addition to simplifying the algorithms for calculating the indices, allowed us to use a schematic view of the temporal evolution of the demand coming from other regions of the state. Future work could extend the proposed approach here to evaluate the topological properties of the network formed by sets of municipalities, characterizing the relationship between the demands of the regions and their temporal evolution.
## Conclusions
The proposed method for monitoring and assessment of the intermunicipal hospitalization network can be used as a management tool to analyze the behavior of this and other networks according to the selected services of interest. Assessment methods are important in the management process, because they provide elements that assist decision making.
The practical application of the proposed method enabled the verification of the number of patients that traveled from their municipality of residence to the municipality in which the service was conducted, as well as the number of different municipalities for which those citizens traveled, and the traveled distance to arrive at the municipality with the objective of having access to healthcare assistance. In this way, from another perspective, it is possible to verify the number of people arriving at a municipality that provides the healthcare service, their municipality of residence, and the distance they travel to arrive at the destination.
The results show evidence of the possible wide use of this methodology to the assessment of the hospitalization network systems. Therefore, the proposed method can assist the planning of an intermunicipal hospitalization network, as well as support the decisions about the best way of regionalizing healthcare services.
The results also suggest that to improve access to this essential medical service in a particular location, regionalization techniques need to be reviewed. To lower the death rates from cardiac surgery, it is essential to guarantee access to high-quality care. Regionalization of care can aid in the concentration of resources and knowledge in specialized health centers, but it must be properly planned to guarantee that all cardiovascular patients, regardless of their location, have equitable access to care. Cardiovascular patients who lack proper access to care may experience unfavorable results, hence it is crucial to address this problem through efficient regionalization policies and initiatives.
## References
1. 1.Brasil. Datasus. Ministério da Saúde. 2020. http://www2.datasus.gov.br/DATASUS/index.php?area=0901. Accessed 21 Nov 2020.
2. Quites HF de O. **Barreiras do uso da Informação em Saúde na tomada de decisão municipal: uma Revisão de Literatura**. *Rev Eletrônica Gestão Saúde.* (2016.0) **07 Supl. 7** 1011-22. DOI: 10.18673/gs.v0isupl..22004
3. Cohn A, Westphal MF, Elias PE. **Informação e decisão política em saúde**. *Rev Saude Publica* (2005.0) **39** 114-121. DOI: 10.1590/S0034-89102005000100015
4. Santos L, Campos GW de S. **SUS Brasil: a região de saúde como caminho**. *Saúde e Soc* (2015.0) **24** 438-46
5. Santos L, De ALOM. **Redes interfederativas de saúde: um desafio para o SUS nos seus vinte anos**. *Cien Saude Colet* (2011.0) **16** 1671-1680. DOI: 10.1590/S1413-81232011000300002
6. 6.BrasilDecreto no 7.508, de 28 de junho de 2011. Regulamenta a Lei no 8.080. Brasília: Casa Civil2011. *Decreto n* (2011.0)
7. 7.BrasilPortaria no 1.559, de 1 de agosto de 2008. Institui a Política Nacional de Regulação do Sistema Único de Saúde - SUS2008. *Portaria n* (2008.0)
8. Ramzan F, Khan MUG, Rehmat A, Iqbal S, Saba T, Rehman A. **A Deep learning approach for automated diagnosis and multi-class classification of Alzheimer’s disease stages using resting-State fMRI and Residual Neural Networks**. *J Med Syst* (2020.0) **44** 1-6. DOI: 10.1007/s10916-019-1475-2
9. Li J, Tian Y, Zhu Y, Zhou T, Li J, Ding K. **A multicenter random forest model for effective prognosis prediction in collaborative clinical research network**. *Artif Intell Med* (2019.0) **2020** 101814
10. Beghriche T, Djerioui M, Brik Y, Attallah B, Belhaouari SB. **An efficient prediction system for diabetes disease based on deep neural network**. *Complexity* (2021.0) **2021** 1-4. DOI: 10.1155/2021/6053824
11. Polak S, Skowron A, Brandys J, Mendyk A. **Artificial neural networks based modeling for pharmacoeconomics application**. *Appl Math Comput* (2008.0) **203** 482-492
12. Zarei L, Moradi N, Peiravian F, Mehralian G. **An application of analytic network process model in supporting decision making to address pharmaceutical shortage**. *BMC Health Serv Res* (2020.0) **20** 1-11. DOI: 10.1186/s12913-020-05477-y
13. 13.BrasilMinistério da Saúde. Secretaria de Atenção à Saúde. Departamento de Regulação, Avaliação e Controle. Coordenação Geral de Sistema de Informação. SIH – Sistema de Informação Hospitalar do SUS: Manual Técnico Operacional do Sistema2015BrasiliaMinistério da Saúde. *Ministério da Saúde. Secretaria de Atenção à Saúde. Departamento de Regulação, Avaliação e Controle. Coordenação Geral de Sistema de Informação. SIH – Sistema de Informação Hospitalar do SUS: Manual Técnico Operacional do Sistema* (2015.0)
14. 14.Boaventura Netto POGrafos: Teoria, Modelos, Algoritmos. 5a. São Paulo: Blucher2012. *Grafos: Teoria, Modelos, Algoritmos. 5* (2012.0)
15. Albert R, Barabási A-L. **Statistical mechanics of complex networks**. *Rev Mod Phys* (2002.0) **74** 47. DOI: 10.1103/RevModPhys.74.47
16. Fadigas IS, Pereira HBB. **A network approach based on cliques**. *Phys A Stat Mech its Appl* (2013.0) **392** 2576-2587. DOI: 10.1016/j.physa.2013.01.055
17. Monteiro RLS, Fontoura JRA, Carneiro TKG, Moret MA, Pereira HBB. **Evolution based on chromosome affinity from a network perspective**. *Phys A Stat Mech its Appl* (2014.0) **403** 276-283. DOI: 10.1016/j.physa.2014.02.019
18. Monteiro RLS, Carneiro TKG, Fontoura JRA, Da Silva VL, Moret MA, Pereira HBB. **A model for improving the learning curves of artificial neural networks**. *PLoS One* (2016.0) **11** e0149874. DOI: 10.1371/journal.pone.0149874
19. Pereira EJ de AL, Ferreira PJS, da Silva MF, Miranda JGV, Pereira HBB. **Multiscale network for 20 stock markets using DCCA**. *Phys A Stat Mech its Appl.* (2019.0) **529** 121542. DOI: 10.1016/j.physa.2019.121542
20. 20.Sousa LMO, Araújo EM de, Miranda JGV. Caracterização do acesso à assistência ao parto normal na Bahia, Brasil, a partir da teoria dos grafos. Cad Saude Publica. 2017;33(12):e00101616.
21. 21.Sousa LMO, Araújo EM de, Vivas JGM, Pereira HBB. Cirurgia cardiovascular no estado da Bahia: Informação em Pauta. 2020;5 Especial 1:84–103.
22. Oliveira EXG, Melo ECP, Pinheiro RS, Noronha CP, Carvalho MS. **Acesso à assistência oncológica: mapeamento dos fluxos origem-destino das internações e dos atendimentos ambulatoriais. O caso do câncer de mama**. *Cad Saude Publica* (2011.0) **27** 317-26. DOI: 10.1590/S0102-311X2011000200013
23. Grabois MF, Oliveira EXG, Carvalho MS. **Assistência ao câncer entre criancas e adolescentes: mapeamento dos fluxos origem-destino no Brasil**. *Rev Saude Publica* (2013.0) **47** 368-378. DOI: 10.1590/S0034-8910.2013047004305
24. 24.Xavier DR, Dantas De Oliveira RA, Barcellos C, De Freitas Saldanha R, Ramalho WM, Laguardia J, et al. Health Regions in Brazil based on hospital admissions: a method to support health regionalization, Cad. Saúde Pública 35 (Suppl 2), 2019. 10.1590/0102-311X00076118.
25. Saldanha R de F, Xavier DR, Carnavalli K de M, Lerner K, Barcellos C. **Estudo de análise de rede do fluxo de pacientes de câncer de mama no Brasil entre 2014 e 2016**. *Cad Saude Publica* (2019.0) **35** e00090918. DOI: 10.1590/0102-311x00090918
26. 26.IBGE. Pesquisa Nacional de Saúde 2013: Percepção do estado de saúde, estilos de vida e doenças crônicas. Rio de Janeiro: Instituto Brasileiro de Geografia e Estatísticas; 2014. p. 1–181.
27. 27.Barata RB. Condições de Saúde da População Brasileira. In: Giovanella L, Escorel S, Lobato L de VC, Noronha JC de, Carvalho AI de (editors). Políticas e Sistema de Saúde no Brasil. 2nd edition. Rio de Janeiro: FIOCRUZ; 2012. p. 143–81.
28. Guimarães RM, Andrade SSC de A, Machado EL, Bahia CA, Oliveira MM de, Jacques FVL. **Diferenças regionais na transição da mortalidade por doenças cardiovasculares no Brasil, 1980 a 2012**. *Rev Panam Salud Pública* (2015.0) **37** 83-9. PMID: 25915012
29. Schmidt MI, Duncan BB, e Silva GA, Menezes AM, Monteiro CA, Barreto SM. **Chronic non-communicable diseases in Brazil: burden and current challenges**. *Lancet* (2011.0) **377** 1949-61. DOI: 10.1016/S0140-6736(11)60135-9
30. 30.Datasus. TABWIN. Brasília: Ministério da Saúde: Departamento de Informática do SUS; 2014.
31. 31.Secretaria da Saúde do Estado da Bahia (SESAB). Planejamento Regional Integrado - Observatório Baiano de Regionalização. 2014. http://www1.saude.ba.gov.br/obr/?id=3. Accessed 15 Dec 2014.
32. 32.Datasus. População Residente - Estimativas para o TCU - Bahia. Departamento de Informática do SUS. 2017. http://tabnet.datasus.gov.br/cgi/tabcgi.exe?ibge/cnv/poptba.def. Accessed 29 Jan 2017.
33. 33.Superintedência de Estudos Econômicos e Sociais da Bahia. PIB Municipal. http://www.sei.ba.gov.br/index.php?option=com_content&view=article&id=561&Itemid=335. Accessed 17 Feb 2018.
34. Travassos C, Oliveira EXG, Viacava F. **Desigualdades geográficas e sociais no acesso aos serviços de saúde no Brasil: 1998 e 2003**. *Cien Saude Colet* (2006.0) **11** 975-986. DOI: 10.1590/S1413-81232006000400019
35. Oliveira EXG, Carvalho MS, Travassos C. **Territórios do Sistema Único de Saúde: mapeamento das redes de atenção hospitalar**. *Cad Saúde Pública, Rio Janeiro* (2004.0) **20** 386-402. DOI: 10.1590/S0102-311X2004000200006
36. Bittencourt SA, Camacho LAB, Leal MDC. **O Sistema de Informação Hospitalar e sua aplicação na saúde coletiva**. *Cad Saude Publica* (2006.0) **22** 19-30. DOI: 10.1590/S0102-311X2006000100003
37. Tanaka OY, Tamaki EM. **O papel da avaliação para a tomada de decisão na gestão de serviços de saúde**. *Cienc e Saude Coletiva* (2012.0) **17** 821-828. DOI: 10.1590/S1413-81232012000400002
38. Veras CMT, Martins MS. **A confiabilidade dos dados nos formulários de Autorização de Internação Hospitalar (AIH), Rio de Janeiro**. *Brasil Cad Saude Publica* (1994.0) **10** 339-355. DOI: 10.1590/S0102-311X1994000300014
39. Escosteguy CC, Portela MC, Medronho RA, Vasconcellos MTL. **O Sistema de Informações Hospitalares e a assistência ao infarto agudo do miocárdio**. *Rev Saúde Pública* (2002.0) **36** 491-499. DOI: 10.1590/S0034-89102002000400016
40. Mathias TA de F, Soboll ML de MS. **Confiabilidade de diagnósticos nos formulários de autorização de internação hospitalar**. *Rev Saude Publica* (1998.0) **32** 526-32. DOI: 10.1590/S0034-89101998000600005
41. Aguiar FP, Melo ECP, Oliveira EXG, Carvalho MS, Pinheiro RS. **Confiabilidade da informação sobre município de residência no Sistema de Informações Hospitalares - Sistema Único de Saúde para análise do fluxo de pacientes no atendimento do câncer de mama e do colo do útero**. *Cad Saúde Coletiva* (2013.0) **21** 197-200. DOI: 10.1590/S1414-462X2013000200015
42. da Silva SF. **Organização de redes regionalizadas e integradas de atenção à saúde: Desafios do Sistema ÚNico de Saúde (Brasil)**. *Cienc e Saude Coletiva* (2011.0) **16** 2753-2762. DOI: 10.1590/S1413-81232011000600014
43. Tanaka OY, Melo C, Bosi MLM, Mercado FJ. **Reflexões sobre a avaliação em serviços de saúde e a adoção das abordagens qualitativa e quantitativa**. *Pesquisa qualitativa de serviços de saúde* (2004.0) 121-136
|
---
title: Prevalence and factors associated with hyperglycemia in a rural population
of Tanvè and Dékanmey in Benin in 2019
authors:
- Nicolas Hamondji Amegan
- Ariyoh Salmane Amidou
- Corine Yessito Houehanou
- Helene Robin
- Gwladys Nadia Gbaguidi
- Corine Agathe Lucresse Fassinou
- Kuassi Daniel Amoussou-Guenou
- Pierre-Marie Preux
- Philippe Lacroix
- Stephan Dismand Houinato
journal: PLOS Global Public Health
year: 2022
pmcid: PMC10022052
doi: 10.1371/journal.pgph.0000471
license: CC BY 4.0
---
# Prevalence and factors associated with hyperglycemia in a rural population of Tanvè and Dékanmey in Benin in 2019
## Abstract
### Background
Hyperglycemia leads to serious damage to the body, especially the blood vessels and nerves. This study aimed to determine the prevalence and factors associated with hyperglycemia in a rural population of Tanvè and Dékanmey in Benin in 2019.
### Materials and methods
This was a cross-sectional, descriptive and analytical study, nested in the Tanvè Health Study (TAHES) cohort. It covered all residents of the villages of Tanvè and Dékanmey, aged 25 years and above, and having given their written consent. Data were collected in the households during the fourth annual monitoring visit in 2019 using the WHO STEPS Wise approach. Hyperglycemia was defined as a fasting capillary blood glucose value ≥ 110 mg/dL. Data were analyzed with R Studio software version 3.5.1.
### Results
A total of 1331 subjects were included in the study with a $60\%$ female predominance and a sex ratio (male/female) of 0.7. The median age was 40 years (Q1 = 32 years; Q3 = 53 years) with a range of 25 and 98 years. The prevalence of hyperglycemia was $4.6\%$. In multivariate analysis, advanced age (AOR = 1.03; $95\%$CI = 1.02–1.73; $$p \leq 0.004$$), male sex (AOR = 2.93; $95\%$CI = 1.49–5.84; $$p \leq 0.023$$), monthly income> 105,000 FCFA (AOR = 2.63; $95\%$CI = 1.24–5.63; $$p \leq 0.030$$), abdominal obesity (AOR = 2.80; $95\%$CI = 1.29–6.16; $$p \leq 0.007$$, and obesity (AOR = 1.68; $95\%$CI = 0.75–3.59; $$p \leq 0.004$$) were statistically associated with hyperglycemia.
### Conclusion
The prevalence of hyperglycemia is not negligible in rural areas in Benin. Our study found that older age, male gender, high income, abdominal obesity, and obesity are determining factors in its occurrence.
## Background
Hyperglycemia is a precursor stage of diabetes, which corresponds to an increase in fasting glucose from 110 mg/dL [1].
The International Diabetes Federation (IDF) estimated in 2013 the number of adults suffering from diabetes to around 382 million in the world, meaning a global proportion of $8.3\%$, with around $80\%$ of these people living in Low and Middle Income Countries (LMICs). If nothing is done, by 2035, the number of persons living with diabetes will rise to around 592 million; meaning one in ten adults. Estimates based on these same figures, predict about three new cases every ten seconds, which corresponds to nearly 10 million per year [2].
The World Health Organization (WHO) warned in 2012 that hyperglycemia caused around 2.2 million deaths in the world while diabetes was the direct cause of 1.6 million deaths a few years later in 2015 [1]. In addition, almost half of all deaths due to diabetes have been recorded in population under 60 years of age, and mostly in less developed regions such as Sub-Saharan Africa (SSA), where $75\%$ of the global deaths occurred [2].
*In* general, hyperglycemia and diabetes result in disabilities due to their manifestations and the expensive cost of treatment. The IDF estimated in 2017 that two-thirds of people ($327 million) living with diabetes are of working age; adding that in the same year, diabetes was the cause of $727 billion in health spending [3]. Hyperglycemia has a serious impact on human health and constitutes an independent cardiovascular risk factor [4]. It causes as many deaths, all causes combined, as smoking or increased cholesterol and blood pressure [5]. Moreover, considered as a potentially modifiable factor, it is associated with an unfavorable progression of stroke in acute phase, even in the absence of known diabetes (OR = 2.99; $95\%$CI = 1.37–6.50), also contributing to vascular lesions and complications such as coma, blindness, kidney failure, coronary artery disease and stroke [6].
In the past, high prevalence of hyperglycemia was considered to be the prerogative of high-income urban populations; but today, studies show significant prevalence of hyperglycemia in rural areas and according to IFD in 2013, more people with diabetes live in rural areas, meaning 136 million compared with 246 million in urban areas [2]. LMICs are not on the sidelines of the question. In fact, in 2013, the number of people living with diabetes in rural areas was 122 million compared to 181 million in urban areas and by 2035, estimates predict 145 million diabetics in rural areas versus 347 million people in urban areas [2].
The situation is no less worrying in rural areas in Benin as evidenced by the Tanvè Health Study (TAHES) cohort erected as a rural population in South Benin since 2015 where prevalence of hyperglycemia has been estimated to respectively $3.5\%$ and $4.5\%$ in 2015 and 2017 [7, 8]. In light of this statistics, it is quite imperative to act quickly to curb the increase in the burden of hyperglycemia in order to prevent its harmful consequences on the health of rural populations. Our research aimed then, to study the factors associated with hyperglycemia in rural areas in Benin in order to guide the implementation of possible consequent control actions.
## Materials and methods
Our study was a cross-sectional, descriptive and analytical, conducted during the fourth annual follow-up visit of the TAHES cohort (Tanvè Health Study). The study population included persons aged 25 years or older, residing in the villages of Tanvè and Dékanmey, and having given their free written consent after clarification to participate in the TAHES cohort.
Fasting hyperglycemia was considered as the dependent variable and was defined by fasting capillary glycemia ≥110 mg/dL. The glycemia was defined as a diabetic-type in case of fasting capillary glycemia ≥126 mg/dL The independent variables were socio-demographic, economic, cultural, behavioural and biological variables. The definition of risk factors used in our study was those described in the WHO STEPS Survey Manual. Only current and former smokers were considered to be smokers. Consumption of less than five servings (400 grams) of fruits and vegetables (FV) per day was defined as insufficient consumption of fruits and vegetables. The practice of physical activity was categorized in 2 ways: sufficient and insufficient. Subjects were reported to have insufficient physical activity when they had moderate physical activity of at least 30 minutes or intense activity of at least 15 minutes during their activities or recreation for less than 5 days per week. This practice was considered as sufficient when it was done for at least 5 days a week. Alcohol consumption has been defined as a consumption on at least one occasion in the past 12 months. Without any notion of follow-up, systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥90 mm Hg was defined as High Blood Pressure (HBP). Hypertension has been defined as either having HBP or being on antihypertensive treatment. The body mass index (BMI) was calculated by dividing the weight in kilograms (Kg) by the square of the height in meters (m). Under-weight was defined as a BMI <18.5kg/m2, overweight by a BMI comprised between 25kg/m2 and <30kg/m2 and obesity as BMI ≥ 30kg/m2. Abdominal obesity was defined by a waist circumference of at least 80 cm in women and 94 cm in men according to IFD criteria.
The data were collected from 26th January to 17th February 2019 in the habitations using the WHO STEPS Wise approach. An individual door-to-door interview was done with each respondent (STEP1), with anthropometric measurements (weight, height, waist circumference) and blood pressure that were after the interview, taken according to WHO standards with validated materials (STEP2). Capillary glucose was measured by finger on appointment the next morning after at least 8 hours of fasting using a glucose meter. The data were recorded using a standardized sheet adapted from the WHO STEPS instrument and integrated into the smartphone via the Kobo-collect application. The study was approved by the National Ethics Committee for Health Research and pregnant women were excluded from the study.
The collected data were cleared and analyzed with R Studio 3.5.1 software. The proportions were compared with the Chi-two test; the means with the Student test and the medians with the non-parametric Kruskal-Wallis test. A step-by-step descending binary logistic regression of Wald was performed using a significance threshold $p \leq 20$% for initial model and $p \leq 5$% for final model. Prior to model execution, the interactions were verified and the suitability of the model was made by the Akaike Information Criterion (AIC).
## Results
A total of 1331 subjects were included in the study. The median age was 40 years (Q1 = 32 years; Q3 = 53 years) with extreme values of 25 and 98 years. There was a female predominance ($60\%$) with a sex ratio of 0.7. In the sample, $63.4\%$ had no educational level, $40.1\%$ had a monthly income of 40,000 FCFA. Poor fruit and vegetable consumption was found in $96.4\%$, alcohol consumption in $63.9\%$, tobacco consumption in $5.0\%$ and insufficient physical activity in $62.5\%$. Concerning the biological risk factors, the average waist circumference was specifically 83.4 ±10.8 cm (extremes: 53 cm and 136 cm) in male subjects and 86.8 ±12.6 cm (extremes: 52 cm and 162 cm) in female subjects. The average BMI was 23.3 ±5.2 Kg/m2.
Abdominal obesity was found in $49.6\%$ of the population. Almost one in ten persons ($9.8\%$) was found to be obese and $19.5\%$ of them were overweight. High blood pressure was found in $29.5\%$ and hypertension was found in $32.1\%$. In addition, around 2 in 10 persons ($21.1\%$) reported taking a glycemic control less than one year ago while $78.9\%$ of them reported doing so within a year or more. A total of 61 subjects had fasting hyperglycemia, meaning a prevalence of $4.6\%$ (IC$95\%$: 3.6–5.8). The average fasting blood glucose value was 93.5 ±33.8 mg/dL with extreme values of 51.0 mg/dL and 333.0 mg/dL. This value was 85.6 ±36.3 mg/dL in male subjects and 82.2 ±31.9 mg/dL in female subjects ($$p \leq 0.073$$). On the other hand, $2.6\%$ had diabetic-type hyperglycemia and $3.8\%$ had a known history of diabetes. The Behavioral and Biological Risk Factors are described in Tables 1 and 2 below. After binary logistic regression, factors independently associated with hyperglycemia were: advanced age, male sex, abdominal obesity, weight status and monthly income above 105,000 FCFA (Table 3).
## Prevalence of hyperglycemia and diabetes
We found in our study a prevalence of fasting hyperglycemia of $4.6\%$ ($95\%$CI: 3.6–5.8). This prevalence corroborates those of Amidou et al and Houehanou et al who respectively found within the same cohort in 2017 and 2015, a prevalence of $4.5\%$ and $3.5\%$ [7, 8]. Also, the STEPS survey realized in Burkina Faso in 2013 showed a similar result with a prevalence of hyperglycemia of $4.9\%$ [9]. In contrast to the results found in our study, studies realized through the STEPS surveys in Senegal in 2015 and Congo Brazzaville in 2014 showed respectively a prevalence of $2.1\%$ and $7.1\%$ [10, 11]. Moreover, in Uganda in 2014, Bahendeka et al, according to a national survey, reported a prevalence of fasting hyperglycemia of $1.9\%$ (IC$95\%$: 1.3–2.4) in rural areas, a result that is lower than ours [12].These differences could be first explained by a disparity in dietary and life behaviors of rural populations in these countries. In addition, the differences observed in the inclusion criteria such as age could be part of these disparities as it has been established that the prevalence of hyperglycemia increases significantly with age. In fact, in some studies, the minimum age required for inclusion was 18 while it was 25 in others.
The prevalence of diabetic-type hyperglycemia in our study was $2.6\%$ and that of diabetes was $3.8\%$. This result is similar to that obtained by Fen L et al in China in a study of 2,967 residents of the Han ethnic group aged 20 to 80 years, selected by multi-stage cluster sampling in Guizhou and which reported a prevalence of $3.77\%$ among rural residents [13]. Bahendeka et al also reported in their study a prevalence of $1\%$ ($95\%$CI: 0.5–1.6) for diabetic hyperglycemia in rural settings [12]. The prevalence of diabetes found in our study was $3.8\%$; which was lower than that of Mohamed et al in Kenya who reported a prevalence of $1.9\%$ in rural areas in 2015. This slight difference could be explained by the socio-cultural disparities and lifestyle patterns of the two study environments [14].
Our study also revealed that 281 ($21.1\%$) persons reported having taken a glycemic control less than a year ago whereas 1050 ($78.9\%$) of them reported having taken it more than a year ago. These results showed that this rural population is becoming more aware of its health. In fact, the $21.1\%$ who did this check-up, were supposed to have done it in the absence of an annual check-up visit as part of the TAHES Cohort since the last visit before that of our study had been done in 2017. Similar to our findings, BeLue et al in Senegal reported in their study a proportion of $24.8\%$ in type 2 diabetic patients that were regular in taking their glycemic control at the M’Bour Hospital Complex [15]. On the contrary, Mohamed et al in Kenya in 2015, based on STEPS data, reported that out of a total of 4,069 respondents aged 18 to 69, only $43.7\%$ knew their glycemic status [14]. The difference with our results could be explained by the fact that their study was not specific to the rural environment. Moreover, the results of the STEPS survey of Burkina Faso in 2013, showed a prevalence of $5.8\%$ for glycemic control [9]. Another factor that could influence the glycemic control is pregnancy. In fact, pregnant women often benefit in the context of their prenatal consultations, from a series of routine tests that include glycemic control. As a result, when these women are part of the next survey concerning the cohort TAHES as they are no longer pregnant, they will report having carried out a glycemic control, although this control was not a voluntary control, what could over-estimate the prevalence of voluntary glycemic control in the population. Also, the exclusion of pregnant women from the study could underestimate the prevalence of glycemic control as these women that were not pregnant the previous year, were subject to a glycemic control during the previous survey and could have reported a glycemic control if considered in the current study.
## Factors associated with fasting hyperglycemia
In our study, factors associated with fasting hyperglycemia in multivariate analysis were: advanced age, male sex, abdominal obesity, weight status and high monthly income.
The prevalence of hyperglycemia increased significantly with age ($$p \leq 0.004$$). In fact, the risk of being hyperglycemic increased by 1.03 with each increase of one unit of age ($95\%$CI: 1.02–1.73). Bahendeka et al had reached out to the same conclusion in their study where they reported that the prevalence of hyperglycemia increased significantly with age ($$p \leq 0.003$$) [12]. Furthermore, our study revealed that men were at a higher risk of developing hyperglycemia compared to women ($$p \leq 0.023$$), AOR = 2.93 (1.49–5.84). Okumiya et al also found a significant association between hyperglycemia and sex in Tibetans in Asia. In their study, hyperglycemia was 2.17 times more frequent in men than in women ($95\%$CI AOR: 1.57–2.99; $p \leq 0.0001$) [16]. The same observation was made by Saleem et al in the United States [17]. Feng et al had come to the same conclusion in their study by comparing the mean fasting blood sugar level where they reported that the mean fasting blood sugar level was higher in men than in women (5.23 mmol/L versus 5.09 mmol/L, $$p \leq 0.003$$) [13]. In contrary, Millogo et al in Burkina Faso found that the proportion of hyperglycemia increased significantly in women compared to men ($10.4\%$ vs. $6.9\%$, $p \leq 0.05$) [18]. This difference could be explained by the fact that their study took place in urban area where women are less engaged in activities requiring moderate or intense physical activity, contrary to our study environment, which is an almost agricultural environment where the majority of populations, carry out daily rural work and small trade on foot while walking in the surrounding villages, thus deploying a lot of energy. In contrast, Bahendeka et al in Uganda did not find a significant association between hyperglycemia and sex ($$p \leq 0.421$$) [12].
Our study showed that abdominal obesity was associated with hyperglycemia ($$p \leq 0.007$$; AOR = 2.80 (1.29–6.16)); that is consistent with the finding of Aynalem et al in Ethiopia in 2016 who also found an association between hyperglycemia and abdominal obesity. In their study, people with abdominal obesity were more exposed to hyperglycemia compared to people with a normal waist circumference (AOR = 4.107 (1.108–15.231)) [19]. Similarly, Islam et al had the same association in Bangladesh in a study with a representative sample of 1,843 adults aged at least 18 [20].
Obesity is established as an associated factor of hyperglycemia. In our study we found a significant association between hyperglycemia and obesity in univariate analysis (COR = 1.97(1.05–3.69); $$p \leq 0.030$$)). After adjustment, this significant association remained, though unstable (AOR = 1.68 (0.75–3.59); $$p \leq 0.004$$). Okumiya et al reported in their study that obese people were 1.77 times more exposed to hyperglycemia compared to people with normal BMI ($95\%$CI: 1.33–2.36; $p \leq 0.0001$) [16]. Millogo et al also reported in their study that hyperglycemia was two times more frequent in obese or overweight subjects compared to normal or under-weight subjects ($13.3\%$ vs $5.9\%$, $p \leq 0.00001$) [18]. Similarly, Binh et al in rural population in Vietnam; out of a sample of 2710 people, had come to the conclusion that obesity is a factor associated with hyperglycemia (OR = 2.41, $95\%$CI:1.41 4.11) [21].
People with relatively high incomes tend in a given society to adopt eating habits or lifestyle habits in general, similar to a life of ease. The latter can sometimes lead to the occurrence of hyperglycemia which can progress to diabetes. Our study found that the prevalence of hyperglycemia was significantly higher among participants with a monthly income above 105,000 FCFA, compared to those with a monthly income below the Beninese’s minimum wage (< 40,000 FCFA). Furthermore, hyperglycemia was 2.63 times more frequent in people monthly earning at least the minimum wage, compared to those earning below this amount ($$p \leq 0.030$$; $95\%$CI AOR: 1.24–5.63). Millogo et al had made the same observation by pointing out in their study that subjects of high socio-economic level were more affected by hyperglycemia ($13.6\%$) than others ($7.6\%$; $$p \leq 0.0098$$) [18]. Harshfield et al’s findings in Bangladesh reinforce this evidence by establishing that the risk of hyperglycemia was four times higher among the richest individuals than among the poorest (OR = 6.48, $95\%$ CI: 5.11–8.22 for men versus 4.77, $95\%$ CI: 3.72–6.12 for women) [22]. Fu et al with 4,506 subjects aged between 18 and 64 concluded that high household income was significantly associated with an increased risk of fasting hyperglycemia in rapidly changing rural Chinese communities (AOR = 1.93 (1.26–2.96); $$p \leq 0.002$$) [23]. On the contrary, Drame et al in Benin in 2015 did not find an association between hyperglycemia and socio-economic status ($$p \leq 0.62$$); which could be explained by the fact that their study area was extended to two administrative departments in Benin [24].
The results of this research must challenge us and draw our attention to a phenomenon that begins to emerge gradually even in the rural regions.
## Strengths and weaknesses of the study
We conducted a comprehensive recruitment of subjects who met the inclusion criteria with the benefit of a large enough sample size. The data were collected by investigators accustomed to the STEPS surveys with the STEPS wise tool of the WHO transcribed for needs in the main language of the medium "fon". Measurements of blood glucose, blood pressure and anthropometric parameters were made by trained investigators using WHO recommended methods and with valid and standardized equipments. All this has contributed to the minimisation of measurement errors and potential information biases.
However, blood sugar and blood pressure measurements being taken during a single visit, are likely to be influenced by external factors such as infection, ongoing treatment, stress or lack of rest; which may overestimate our results.
## Conclusion
The prevalence of hyperglycemia was not negligible in the rural populations of Tanvè and Dékanmey in Benin. Advanced age, males, abdominal obesity, obesity and high income were significantly associated with its occurrence. Effective action against this health problem requires concerted political and health action at a much more multi-sectoral level considering the diverse origins of its risk factors.
## References
1. 1Anaes. Service d’évaluation des technologies, Service évaluation économique, février 2003. Disponible sur: http://www.has-sante.fr/portail/jcms/c_464100/principes-de-depistage-du-diabete-de-type-2.
2. **ATLAS du DIABÈTE de la FID 6e édition.**. *En ligne*
3. **Le diabète concerne chaque famille**. *Guide de la campagne* (2018.0)
4. Levitan EB, Song Y, Ford ES, Liu S. **Is non diabetic hyperglycemia a risk factor for cardiovascular disease? A meta-analysis of prospective studies**. *Arch Intern Med* (2004.0) **164** 2147-55. PMID: 15505129
5. Danaei G, Lawes CMM, Vander Hoorn S, Murray CL, Ezzati M. **Global and regional mortality from ischaemic heart disease ans stroke attribuable to higher-thanoptimum blood glucose concentration: comparative risk assessment**. *Lancet* (2006.0) **368** 1651-9. DOI: 10.1016/S0140-6736(06)69700-6
6. Gnonlonfoun D, Adjien C, Kerekou A, Ossou-Nguiet PM, Agbetou M, Adoukonou T. **Hyperglycémie à la phase aiguë des accidents vascuclaires cérébraux au CNHU-HKM de Cotonou (Bénin)**. *Rev.CAMES-Série A* (2012.0) **13** 59-62
7. 7Houehanou Sonou YCN. Épidémiologie des facteurs de risque cardiovasculaire en population tropicale—cas du Bénin. [Thèse Epidémiologie]. Médecine humaine et pathologie. Université de Limoges; Université d’Abomey-Calavi (Bénin), 2015. Français. ⟨NNT: 2015LIMO0135⟩. ⟨tel-01332830⟩.
8. 8Amidou SA. Epidémiologie des maladies cardiovasculaires en population générale rurale: Cohorte de Tanvè Health Study (TAHES). Thèse de Santé publique (Epidémiologie). Université de Limoge cotutelle université d’Abomey-Calavi;2018,253p. [En ligne]. http://creativecommons.org/licens/by-nc-nd/3.0/fr/. Consulté le 03 mars 2019.
9. **Rapport Enquête STEPS**. (2013.0)
10. 10Ministère de la santé et de l’action sociale du Sénégal. Enquête nationale sur les facteurs de risque des maladies non transmissibles STEPS 2015. «Rapport préliminaire: les indicateurs-clés». 2016.
11. **Hypertension artérielle et les autres facteurs de risque cardio-vasculaire à Brazaville**. *Rapport d’enquête* (2014.0)
12. Bahendeka S, Wesonga R, G Mutungi G, Muwonge J, Neema S, Guwatudde D. **Prevalence and correlates of diabetes mellitus in Uganda: apopulation-based national surveySilver**. *Tropical Medicine and International Health* (2016.0) **21** 405-16. DOI: 10.1111/tmi.12663
13. Feng Y, Wang K, Wang D, Dong F, Yu Y, Pan L. **Prevalence and associated risk factors of diabetes among ethnic Han residents in Guizhou.**. *Zhonghua Liu Xing Bing Xue Za Zhi* (2015.0) **36** 1220-5. PMID: 26850240
14. Mohamed SF, Mwangi M, Mutua MK, Kibachio J, Hussein A, Ndegwa Z. **Prevalence and factors associated with pre-diabetes and diabetes mellitus in Kenya: results from a national survey**. *BMC Public Health* (2018.0) **18** 1215. DOI: 10.1186/s12889-018-6053-x
15. BeLue R, Ndiaye K, NDao F, Ba FN, Diaw M. **Glycemic Control in a Clinic-Based Sample of Diabetics in M’Bour Senegal.**. *Health Educ Behav* (2016.0) **43** 112S-6S. DOI: 10.1177/1090198115606919
16. Okumiya K, Sakamoto R, Ishimoto Y, Kimura Y, Fukutomi E, Ishikawa M. **Glucose intolerance associated with hypoxia in people living at high altitudes in the Tibetan highland**. *BMJ Open* (2016.0) **6** e009728. DOI: 10.1136/bmjopen-2015-009728
17. Saleem SM, Jan SS, Haq I, Khan SMS. **Identification des facteurs de risque affectant le métabolisme du glucose avec facultés affaiblies dans la population adulte du district de Srinagar**. *J Elsevier* (2019.0) **13** 1047-1051. DOI: 10.1016/j.dsx.2019.01.023
18. Millogo GRC, Yaméogo C, Samandoulougou A, Yaméogo NV, Kologo KJ, Toguyeni JY. **Diabète en milieu urbain de Ouagadougou au Burkina Faso: profil épidémiologique et niveau de perception de la population adulte**. *Pan African Medical J* (2015.0) **20** 146. DOI: 10.11604/pamj.2015.20.146.3249
19. Aynalem SB, Zeleke AJ. **Prevalence of Diabetes Mellitus and Its Risk Factors among Individuals Aged 15 Years and Above in Mizan-Aman Town, Southwest Ethiopia, 2016: A Cross Sectional Study**. *International Jof Endocrinology 2018* (2018.0) **7** 9317987. DOI: 10.1155/2018/9317987
20. Islam JY, Zaman MM, Bhuiyan MR, Haq SA, Ahmed S, Al-Qadir AZ. **Prevalence and determinants of hyperglycaemia among adults in Bangladesh: results from a population-based national survey**. *BMJ Open* (2019.0) **9** e029674. DOI: 10.1136/bmjopen-2019-029674
21. Binh TQ, Phuong PT, Nhung BT, Thoang DD, Thang PV, Long TK. **Prevalence and correlates of hyperglycemia in a rural population, Vietnam: implications from across–sectional study**. *BMC Public Health* (2012.0) **12** 939. PMID: 23114020
22. Harshfield E, Chowdhury R, Harhay NM, Bergquist H, Harhay OM. **Association of hypertension and hyperglycaemia with socioeconomic contexts in resource-poor settings: the Bangladesh Demographic and Health Survey**. *Int J Epidemiol* (2015.0) **44** 1625-36. DOI: 10.1093/ije/dyv087
23. Fu C, Chen Y, Wang F, Wang X, Song J, Jiang Q. **High prevalence of hyperglycaemia and the impact of high household income in transforming Rural China.**. *BMC Public Health* (2011.0) **11** 862. DOI: 10.1186/1471-2458-11-862
24. Drame ML, Mizéhoun-Adissoda C, Amidou S, Sogbohossou P, Paré R, Ekambi A. **Diabetes and its traitement quality in Benin (west Africa): Analysis of data from the STEPS survey 2015**. *Open J of epidemiol* (2018.0) **8** 242-58. DOI: 10.4236/ojepi.2018.8419
|
---
title: The effect of high polycyclic aromatic hydrocarbon exposure on biological aging
indicators
authors:
- Manuela Campisi
- Giuseppe Mastrangelo
- Danuta Mielżyńska-Švach
- Mirjam Hoxha
- Valentina Bollati
- Andrea A. Baccarelli
- Angela Carta
- Stefano Porru
- Sofia Pavanello
journal: Environmental Health
year: 2023
pmcid: PMC10022060
doi: 10.1186/s12940-023-00975-y
license: CC BY 4.0
---
# The effect of high polycyclic aromatic hydrocarbon exposure on biological aging indicators
## Abstract
### Background
Aging represents a serious health and socioeconomic concern for our society. However, not all people age in the same way and air pollution has been shown to largely impact this process. We explored whether polycyclic aromatic hydrocarbons (PAHs), excellent fossil and wood burning tracers, accelerate biological aging detected by lymphocytes DNA methylation age (DNAmAge) and telomere length (TL), early nuclear DNA (nDNA) hallmarks of non-mitotic and mitotic cellular aging, and mitochondrial DNA copy number (mtDNAcn).
### Methods
The study population consisted of 49 male noncurrent-smoking coke-oven workers and 44 matched controls. Occupational and environmental sources of PAH exposures were evaluated by structured questionnaire and internal dose (urinary 1-pyrenol). We estimated Occup_PAHs, the product of 1-pyrenol and years of employment as coke-oven workers, and Environ_PAHs, from multiple items (diet, indoor and outdoor). Biological aging was determined by DNAmAge, via pyrosequencing, and by TL and mtDNAcn, via quantitative polymerase chain reaction. Genomic instability markers in lymphocytes as target dose [anti-benzo[a]pyrene diolepoxide (anti-BPDE)–DNA adduct], genetic instability (micronuclei), gene-specific (p53, IL6 and HIC1) and global (Alu and LINE-1 repeats) DNA methylation, and genetic polymorphisms (GSTM1) were also evaluated in the latent variable nDNA_changes. Structural equation modelling (SEM) analysis evaluated these multifaceted relationships.
### Results
In univariate analysis, biological aging was higher in coke-oven workers than controls as detected by higher percentage of subjects with biological age older than chronological age (AgeAcc ≥ 0, $$p \leq 0.007$$) and TL ($$p \leq 0.038$$), mtDNAcn was instead similar. Genomic instability, i.e., genotoxic and epigenetic alterations (LINE-1, p53 and Alu) and latent variable nDNA_changes were higher in workers ($p \leq 0.001$). In SEM analysis, DNAmAge and TL were positively correlated with Occup_PAHs ($p \leq 0.0001$). Instead, mtDNAcn is positively correlated with the latent variable nDNA_changes ($p \leq 0.0001$) which is in turn triggered by Occup_PAHs and Environ_PAHs.
### Conclusions
Occupational PAHs exposure influences DNAmAge and TL, suggesting that PAHs target both non-mitotic and mitotic mechanisms and made coke-oven workers biologically older. Also, differences in mtDNAcn, which is modified through nDNA alterations, triggered by environmental and occupational PAH exposure, suggested a nuclear-mitochondrial core-axis of aging. By decreasing this risky gerontogenic exposure, biological aging and the consequent age-related diseases could be prevented.
## Introduction
Aging is associated with higher risk of chronic conditions, such as cardiovascular, cancer and neurodegenerative diseases, and represents a serious health and socioeconomic societal concern [1]. However, not all people become elderly in the same manner [2]. One of the major issues arisen in recent years is that environmental pollution can accelerate aging [3]. Air pollution, jointly with genetic and life-style risk factors, including tobacco smoking, alcohol drinking and unsafe diet, contributes significantly to aging [4, 5]. Air pollution exposure can shorten life expectancy even among people with the best genetic makeup.
Oxidative stress and the consequent chronic inflammation have been identified as pathogenic factors in aging and age-related diseases, partly caused by chemical metabolism, which results in excessive production of oxygen-derived free radicals and low-grade inflammation [6]. The toxicology screening program Tox21, testing biological responses to more than 9,000 chemicals, showed that numerous organic chemicals, including polycyclic aromatic hydrocarbons (PAHs), alter molecular signaling pathways related to inflammation which contribute to the pathogenesis of many age-related diseases [7]. Benzo(a)pyrene (B[a]P) metabolic activation, via aldo–keto reductase and/or manganese superoxide dismutase [8], produces reactive oxygen species (ROS) that can generate high levels of oxidized guanine in nDNA [9]. In our previous work, the exposure to PAHs of coke-oven workers, amongst the highest exposure to these carcinogenic compounds, monitored by internal dose (urinary 1-pyrenol) and by genotoxic measures at target dose [anti-benzo[a]pyrene diolepoxide (anti-BPDE)–DNA adduct], was found to increase a systemic inflammatory response [10] and induced gene-specific (p53, IL6 and HIC1) and global (extrapolated from Alu and LINE-1 repeats) DNA methylation changes [11]. Furthermore, those coke-oven workers with shorter telomere length (TL) [12] showed lower mitochondrial DNA copy number (mtDNAcn), thus linking PAHs exposure and mitochondrial dysfunction with cellular aging [13]. Moreover, an important link between telomeres and mitochondria was made by Sahin et al. [ 14] with the suggestion of a telomere-mitochondrial core-axis of aging, including tumor suppressor p53 as key regulator.
In the present study, we hypothesized that exposure to higher levels of PAHs in coke-oven workers may be also associated with increased biological aging detected by lymphocyte DNA methylation age (DNAmAge). DNAmAge is an emerging and most robust epigenetic marker of non-mitotic cellular aging [15, 16], assessed through the analysis of methylation at a specific subset of cytosine-guanine dyads (CpG), which showed a strong correlation with the chronological age [17–20]. This biomarker was linked to the “epigenetic clock” theory of aging [21] according to which an increase in DNAmAge is indicative of altered biological functions [16] and an elevated risk for morbidity and mortality [22]. Furthermore, the close nexus we found between DNAmAge of the pulmonary cells and blood lymphocytes, advise that blood lymphocytes could be a surrogate tissue for lung age status [23].
To this purpose, we investigated the effects of high chronic exposure to PAHs on DNAmAge of Polish male, non-current smoking, coke-oven workers and matched controls. We also investigated the relative magnitude of other measures of aging, including TL, which measures mitotic or replicative cellular aging, and mtDNAcn. These multifaceted relationships are evaluated using the structural equation models (SEM) analysis. SEM is a statistical technique linking observed data with qualitative causative assumptions and testing whether variables are interdependent, and if so, the details of their relationships. This methodology is appropriate for the investigation of complex interrelationships, as it tests causative relationships instead of mere correlations [24].
## Study design
This is a cross-sectional study comparing a group with high occupational exposure to PAHs and a reference group without such exposure. The study population consisted of $$n = 49$$ coke-oven workers in 3 Polish cokeries and $$n = 45$$ controls matched by gender and ethnicity, who were part of a group of 94 study individuals examined in our previous work [12] for whom DNA was still available. All participants were males and noncurrent smokers defined as either never-smokers or former smokers who had quit smoking at least 1 year before sample collection, as confirmed by analysis of nicotine and its metabolites [25]. All coke-oven workers performed tasks (i.e., charging, coking, and pushing operations at the coke-oven battery section) involving exposure to high levels of PAHs. Individuals whose work involved exposure to benzene (i.e., workers in byproduct operations) were excluded. Controls were clerks of the Institute of Occupational Medicine and Environmental Health in Sosnowiec, recruited during their periodic check-ups at the Preventive Health Services of the Institute.
PAH exposure was assessed by measuring 1-pyrenol in a urine sample (50 mL) collected from each of the workers at the end of their work shift (after at least 3 consecutive working days) and in the late afternoon from controls. Simultaneously, blood samples were collected in EDTA (20 mL) and heparin tubes (10 mL) for adduct and micronuclei analyses, as described previously [25], and further genetic [12, 26] and epigenetic analyses. All samples were transported at the Institute of Occupational Medicine and Environmental Health in Sosnowiec where: (i) lymphocytes cultures for micronuclei detection were prepared and micronuclei analyses were conducted; (ii) lymphocytes for adduct analyses were isolated in Ficoll solution (Seromed) within 4 h after blood collection and kept frozen at − 80 °C until shipment to the Occupational Medicine section of the University of Padova, Italy, where DNA was extracted. Structured questionnaires were administered to collect data on other non-occupational PAHs exposure (indoor and outdoor exposure and diet) and daily consumption of fruit or vegetables, as well as information on chronic diseases, as described previously [12, 25].
The Ethics Committee of the Institute of Occupational Medicine and Environmental Health in Sosnowiec reviewed the study. All study participants gave their written informed consent before recruiting.
## Estimation of PAHs exposure from the questionnaire
Structured questionnaires were administered to collect data on non-occupational PAHs exposure, focusing on the following categories. Diet. Individuals who declared to consummate PAHs-rich meals (charcoaled meat and pizza) more than once a week were recorded as individuals with high dietary intake of PAHs;Indoor. Individuals who declared to use wood or coal heating at home were considered as individuals with indoor exposure to PAHs. Outdoor. Individuals who declared to live (home residence) in areas with intense local traffic and/or the presence of industries were considered as individuals with environmental exposure to PAHs. Home. Individuals who declared to have the residence in country or town.
## Internal exposure: 1-pyrenol analysis
Exposure to PAHs was determined as previously described [25] by measuring 1-pyrenol in urine samples by high-performance liquid chromatography/fluorescence. The urine sample is pretreated by enzymatic hydrolysis overnight in the dark at 37 °C with 1.25 µl/ml urine of the enzymatic mixture β-glucuronidase (134.8U/ml) and aryl sulfatase (4.0U/ml) (Sigma-Aldrich). 1-pyrenol level in each urine sample was expressed as micromoles per mole of creatinine, determined using a colometric test, based on the Jaffé reaction between creatinine and sodium picrate. 1-pyrenol was multiplied per years of work in the cokery.
## Target dose: anti-BPDE–DNA adduct
Anti-BPDE–DNA adduct formation was detected after DNA isolation with a Promega Wizard genomic DNA purification kit (Promega) by high-performance liquid chromatography/fluorescence analysis of BP-tetrol-I-1 (r-7,c-10,t-8,t-9-tetrahydroxy-7,8,9,10-tetrahydro-benzo[a]pyrene) released after acid hydrolysis of DNA samples, as described previously [25]. The mean coefficient of variation (CV) for analyses of a standard curve repeated five times on five different days was $10\%$. The highest CV value was $5.70\%$ for those samples ($$n = 8$$) with more than 200 μg DNA, repeated twice.
## Genetic instability: micronuclei
Micronuclei analysis was conducted on coded slides scored by light microscopy at × 400 magnification, as described previously [25]. To exclude artifacts, the identification of micronuclei was confirmed at × 1,000 magnification in $10\%$ of samples. The scoring of bi-, tri- and tetra-nucleate cells and micronuclei analysis was done and the cytokinesis block proliferation index was calculated as being equal in both groups ($$p \leq 0.60$$).
## Analyses of DNA methylation states of p53, p16, HIC1 and IL-6 gene-specific promoters and of Alu and LINE-1 repetitive element
DNA methylation status was quantified as previously described [11], using bisulfite-PCR and pyrosequencing. The degree of methylation was expressed as the percentage of methylated cytosines divided by the amount of both methylated and unmethylated cytosines. All samples were analyzed 3 times for each marker to verify the reproducibility of our measurements and their average was used in the statistical analysis.
## mtDNAcn analysis
MtDNAcn was measured in DNA using real-time quantitative PCR (qPCR) as previously described [13, 27]. This assay measures relative mtDNAcn by determining the ratio of mitochondrial (MT) copy number to single copy gene (S) copy number (human β-globin, hbg) in experimental samples relative to the MT/S ratio of a reference pooled sample. The primers for qPCR analysis of mtDNA and hbg were previously described [28]. All qPCRs were performed on 7900HT Fast Real-Time PCR System (Applied Biosystems, Monza, Italy). The average of the three MT measurements was divided by the average of the three S measurements to calculate the MT/S ratio for each sample. The CV for the MT/S ratio in duplicate samples analyzed on two different days was $7.8\%$.
## TL analysis
TL was measured in DNA using the qPCR method as described previously [29, 30]. This method appraises the relative telomere length in genomic DNA by establishing the ratio of telomere repeat copy number (T) to single copy gene (S) copy number (T/S ratio) in experimental samples relative to the T/S ratio of a reference pooled sample [29, 30]. The single-copy gene used in this study was human β-globin (hbg). All samples and standards were run in triplicate. The qPCR runs were conducted in triplicate on a 7900HT Fast Real Time PCR System (Applied Biosystems, Monza, Italy) and the average of the 3 T/S ratio measurements was used in the statistical analyses. To test the reproducibility of telomere length measurements, we replicated TL analysis 3 times in 3 different days for 15 samples. The within-sample CV for the average T/S ratio over the 3 consecutive days was $8.7\%$, which was similar to the CV reported for the original protocol [31].
## DNAmAge analysis
DNAmAge was assessed by analyzing the methylation levels from specific CpG sites using bisulfite-PCR and Pyrosequencing® methodology as previously reported [32, 33]. This method is based on the determination of the percentage of methylation level measured in five selected markers (ELOVL2, C1orf132, KLF14, TRIM59 and FHL2) in genomic DNA, as described by Zbieć-Piekarska et al. [ 19] with some modifications based on the fact that the method was almost completely automated using the PyroMark Q48 Autoprep (QIAGEN, Milano, Italy). Briefly, 1 μg DNA was bisulfite treated using Epitect Fast® DNA Bisulfite Kit (QIAGEN, Milano, Italy) following the manufacturer’s instructions. An aliquot of template DNA was used for PCR amplification of five specific markers using PCR primers of the AgePlex Mono kit (Biovectis, Warszawa, Poland). PCR reactions were performed in 25 μL, comprising 0.2 μM of each primer, 20 ng of template DNA, and PyroMark PCR Master Mix holding HotStarTaq DNA Polymerase, 1X PyroMark PCR Buffer and dNTPs. The amplification plan involved a preliminary denaturation step at 95 °C for 10 min, followed by 45 cycles of denaturation (94 °C for 30 s), annealing (56 °C for 60 s) and extension (72 °C for 90 s), with a final extension of 72 °C for 10 min. Every PCR amplification contained negative PCR controls. In total, 10 µL of PCR product was used for each pyrosequencing primer (2 µL) included in the AgePlex Mono kit (Biovectis, Warszawa, Poland) and loaded into a 48 well-plate (Pyromark Q48 Discs, QIAGEN, Milano, Italy). Pyrosequencing was performed on a Pyromark Q48 Autoprep instrument (QIAGEN, Milano, Italy) using Pyromark Q48 Advanced Reagents (QIAGEN, Milano, Italy) according to the manufacturer’s instructions. The resulting Pyrograms® generated by the instrument were automatically analyzed using Pyromark Q48 Autoprep Software (QIAGEN, Milano, Italy). The percentages of marker methylation levels were put in an online calculator system accessible at www.agecalculator.ies.krakow.pl, for estimation of DNAmAge. The equation corresponds to a previously developed age prediction model [19]. All samples were tested 3 times for each marker to confirm the reproducibility of our results, and their averages were used in the statistical testing. All samples were analyzed on two different days, and the CV for replicate pyrosequencing runs was $0.5\%$.
## Age acceleration evaluation
Age Acceleration (AgeAcc) was computed for each participant as the difference between the DNAmAge of blood lymphocytes and the chronological age (equation AgeAcc = DNAmAge—Chronological age).
## Analytic strategy
Variables As stated above, chronic exposure to PAHs was twofold, originating from occupation or environmental sources. Occupational exposure to PAHs (Occup_PAHs) was assessed as the product of internal dose (urinary 1-pyrenol) and years of employment as coke workers. This value represents a cumulative occupational exposure that was always zero in control subjects. While occupational PAHs exposure was measured, environmental PAHs exposure was estimated from multiple items (diet and indoor and outdoor living characteristics and activities). The large number of measured covariates called for dimension reduction to avoid over-fitting and collinearity issues in estimating exposure effects [34]. This was obtained by summarizing high-dimensional data in a single summary score, a latent variable named with the descriptive term Environ_PAHs.
There was interest to contrast the effects of environmental and occupational PAH exposure. Therefore, the relative importance of each source was evaluated by forcing both PAH variables into the same model of mediation analysis (see below), where they acted as independent variables.
There were three dependent variables – DNAmAge, TL and mtDNAcn – each being evaluated separately using three models of mediation analysis.
Upon metabolic activation, catalyzed mainly by the cytochrome P450 enzymes, some PAHs become metabolites that react with DNA leading to genotoxic and/or epigenetic alterations. Many biomarkers reflecting the interaction between the external agent and the exposed body are usually included in a broad and heterogeneous category [35]. Therefore, we pooled the available covariates in a second latent variable called "nDNA_changes" (see below). The nature of the latent variable is intrinsically related to the nature of the indicator variables used to define it. The indicator variables were: peripheral blood lymphocytes measures of target dose (anti-BPDE–DNA adducts), genetic instability (micronuclei—MN), DNA methylation (methylation states of p53, p16, HIC1 and IL-6 gene-specific promoters and global methylation estimated in Alu and LINE-1 repeats) and presence of the detoxifying GSTM1. Again, summarizing high-dimensional data in a single summary score allowed for avoiding over-fitting and collinearity issues in regression analyses [34]. That score was particularly useful to act as a mediator variable through which some of the effects of the independent variable pass on to the dependent variable (this is known as the indirect effect).
b)Directed acyclic graph (DAG) Independent, dependent and mediator variables and their relationships can be represented as a graph formed by vertices and edges, where vertices display variables and the relationships between them take the form of lines (or edges) going from one vertex to another. These edges are directed, which means that they have a single arrowhead indicating their effect. A DAG is also acyclic, which means that there are no feedback loops such that following those directions will never form a closed loop. DAG is more specifically concerned with structural causal models displaying causal assumptions about a set of variables. Figures 1, 2 and 3 display three DAGs with a similar polygon: two triangles sharing one edge. The vertices are Occup_PAHs, DNAmAge (or TL or mtDNAcn) and nDNA_changes on the left triangle, Environ_PAHs, DNAmAge (or TL or mtDNAcn) and nDNA_changes on the right triangle. The arrows specify direct effects and indirect path. The latter is an effect of PAHs mediated through nDNA_changes. Fig. 1Path diagram of mediation model for DNAmAge. Path diagram of a mediation model showing the dependent variable «DNAmAge», the mediator variable «nDNA_changes» and two independent variables «Occup_PAHs» and «Environ_PAHs». An oval indicates the latent variable, square boxes indicate the observed variables, arrows specify the direction of causal flow, an arrowed route is a path, and the estimated beta coefficients with p-values appeared along the paths. The effect of one variable on another is called direct. Paths from each latent variable (nDNA_changes, Environ_PAHs) to each endogenous variable are also drawn. Abbreviations: DNAmAge = DNA methylation age; nDNA_changes = Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, p53, p16, IL-6, HICI, LINE-1, Alu, GSTM1; Environ_PAHs = Latent variable estimated by SEM (environmental exposure to PAHs) based on diet, indoor and outdoor activities; MN = micronuclei; PAHs = Polycyclic aromatic hydrocarbons; Occup_PAHs = (years of work in the cokery) × (1-pyrenol in µmoles/mol creatinine)Fig. 2Path diagram of mediation model for TL. Path diagram of a mediation model showing the dependent variable «TL», the mediator variable «nDNA_changes» and two independent variables «Occup_PAHs» and «Environ_PAHs». An oval indicates the latent variable, square boxes indicate the observed variables, arrows specify the direction of causal flow, an arrowed route is a path, and the estimated beta coefficients with p-values appeared along the paths. The effect of one variable on another is called direct. Paths from each latent variable (nDNA_changes, Environ_PAHs) to each endogenous variable are also drawn. Abbreviations: TL = telomere length; nDNA_changes = Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, p53, p16, IL-6, HICI, LINE-1, Alu, GSTM1; Environ_PAHs = Latent variable estimated by SEM (environmental exposure to PAHs) based on diet, home, indoor and outdoor activities; MN = micronuclei; PAHs = Polycyclic aromatic hydrocarbons; Occup_PAHs = (years of work in the cokery) × (1-pyrenol in µmoles/mol creatinine)Fig. 3Path diagram of mediation model for mtDNAcn. Path diagram of a mediation model showing the dependent variable «mtDNAcn», the mediator variable «nDNA_changes» and two independent variables «Occup_PAHs» and «Environ_PAHs». An oval indicates the latent variable, square boxes indicate the observed variables, arrows specify the direction of causal flow, an arrowed route is a path, and the estimated beta coefficients with p-values appeared along the paths. The effect of one variable on another is called direct. Paths from each latent variable (nDNA_changes, Environ_PAHs) to each endogenous variable are also drawn. Abbreviations: mtDNAcn = mitochondrial DNA copy number; nDNA_changes = Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, p53, p16, IL-6, HICI, LINE-1, Alu, GSTM1; Environ_PAHs = Latent variable estimated by SEM (environmental exposure to PAHs) based on diet, indoor and outdoor activities; MN = micronuclei; PAHs = Polycyclic aromatic hydrocarbons; Occup_PAHs = (years of work in the cokery) × (1-pyrenol in µmoles/mol creatinine) iii)Structural equation models (SEM) Another way to think about DAGs is as non-parametric structural equation models. Therefore, all the above assumptions were converted into three models of structural equation modeling (SEM), one for each final outcome (DNAmAge, TL or mtDNAcn). In the first and second set of parentheses of STATA commands we specified the estimations of the two latent variables: Environ_PAHs (an independent variable); and nDNA_changes (the mediator variable). In the third and fourth sets of parentheses consist of fitting two regression models: the first regression is of nDNA_changes on the two PAHs exposures generated by occupational or environmental sources; the second regression is the model for the final outcome (either DNAmAge, TL, mtDNAcn). Notice that “nDNA_changes” was a dependent variable in the first regression (third set) and an explanatory variable in the second regression (fourth set of parentheses). The exposure coefficient in the second model of regression that includes the mediator is then generally taken as a measure of the direct effect because the effect on the outcome appears to remain even when control has been made for the mediator. The ending command vce(cluster occup) specifies how the VCE (variance–covariance matrix of the estimators) and the standard errors reported by SEM were calculated. As already reported, the study population consisted of 49 blue-collar workers (employed in 3 Polish coking plants) and 45 white-collar workers (office workers of the Institute of Occupational Medicine and Environmental Health in Sosnowiec). Therefore, it would not be unreasonable to assume that the error of one person is correlated with those of others belonging to the same group because each social class tends to be homogeneous. In such cases, where the observations are correlated and the cluster is known, we typed: vce(cluster occup).
According to the above description, the STATA command syntax was the following.
sem (Environ_PAHs -> diet indoor outdoor) (nDNA_changes -> mn adducts p53 p16 il6 hici line alu gstm) (nDna_changes <- Environ_PAHs Occup_PAHs) (DNAmAge <- chron_d nDNA_changes Environ_PAHs Occup_PAHs), vce(cluster occup).sem (Environ_PAHs -> diet indoor outdoor home) (nDNA_changes -> mn adduct p53 p16 il6 hici line alu gstm) (nDNA_changes <- Environ_PAHs Occup_PAHs) (TL <- nDNA_changes Environ_PAHs Occup_PAHs), vce(cluster occup).sem (Environ_PAHs -> diet indoor outdoor) (nDNA_changes -> mn adduct p53 p16 il6 hici line alu gstm) (nDNA_changes <- Environ_PAHs Occup_PAHs) (mtDNAcn <- chron_d nDNA_changes Environ_PAHs Occup_PAHs), vce(cluster occup).
Some variables, particularly those with dichotomous values, were removed because they prevented SEM algorithm to converge.
We used three SEM goodness-of-fit statistics: [1] Standardized root mean squared residual (SRMR); [2] the coefficient of determination (CD); and [3] Stability index. Tables 2, 3 and 4 show the three groups of SEM results (Structural Equations, Measurement, Covariates). Structural equations include the beta coefficients (with a “minus” sign indicating an inverse relationship), $95\%$ confidence intervals ($95\%$CI) and p-values for each of two structural equation models. Measurement including the beta coefficients for this measurement model can be interpreted as correlation coefficients describing the direction (positive or negative) and degree (strength) of the relationship between each indicator and the latent variables Environ_PAHs and nDNA_changes. Covariances representing set of concurrent regression equations to yield coefficient estimators and the covariance is a measure of cross-equation correlation. All coefficients specifying the effects are expressed in the own variable’s scale of units.
SEM results were also presented graphically (Figs. 1, 2 and 3) using the graphical interface of SEM. In Figs. 1, 2 and 3, square boxes stand for variables, arrows specify the direction of causal flow, an arrowed route is a path. The estimated beta coefficients with corresponding p-values appeared along the paths. The figures are a useful synthesis of the findings.
iv)Study size The sample size required for SEM is dependent on model complexity, the estimation method used, and the distributional characteristics of observed variables. The best option is to consider the model complexity (i.e., the number of exogenous variables) and the following rules of thumb: minimum ratio 5:1, with a recommended ratio of 10:1, or a recommended ratio of 15:1 for data with no normal distribution [36, 37]. With three exogenous variables (occupational exposure to PAHs, environmental exposure to PAHs and chronic diseases) used in the SEM model, we should have a minimum of 45 (= 15 × 3) subjects; in total we reached 87 with complete data, thus fulfilling these requirements.
e)Latent variables A numerical value for each latent variable (nDNA_changes and Environ_PAHs) was estimated by SEM program for each subject. Multiple summary statistics were calculated conditioned on a categorical variable that identified the two groups: coke-oven workers and controls. The numerical statistics (median, min, max) were reported in Table 1, together with the p-value of Wilcoxon rank sum test for the equality of the median distribution across the two groups. Table 1Main characteristics of polish male coke-oven workers and non-exposed controlsInterval variables Frequency variablesMedian(min; max) Number(%)Statistical tests and p-values@Coke-oven workers $$n = 49$$Controls $$n = 45$$General characteristics and PAHs§exposure Age (Years)36(20−59)37(21−58)0.911 Dieta8($16\%$)8($18\%$)0.852 Fruit and vegetablesb27($55\%$)22($49\%$)0.549 Indoor PAH exposurec26($53\%$)18($40\%$)0.205 Outdoor PAH exposured17($34\%$)16($36\%$)0.930 Nicotine & metabolites (mg/mmol creatinine)e0($0\%$)0($0\%$) Former smokers20($41\%$)16($36\%$)0.600 Years since cessation6[1-26]7[1-20]0.640 Length of work in the cokery (years)10.5 (1; 40.0)0 (0; 0) 1-pyrenol (μmoles/mol creatinine) f3.1(0.4; 7.5)0.1(0.0; 0.4)0.00001 Occup_PAHsg36.2(3.6; 253.0)0(0; 0)0.00001 Environ_PAHsh0.01313(-0.0471; 0.0607)-0.0131(-0.0970; 0.0378)0.0001 nDNA_changesi1.7377(-0.0006; 4.0372)-0.5360 (-1.0397; 0.9357)0.00001Genotoxic alterations Anti-BPDE-DNA adducts (adducts/108 nucleotides) k5.1 (0.9; 12.2)0.1 (0.1; 5.6)0.00001 MN (MN/1000BN cells)4 (1.0; 11.0)0.01 (0; 4.0)0.00001Epigenetic alterations LINE-1 (%mC)80.3(56.91; 83.7)75.7(55.7; 85.7)0.00001 p53 (%mC)11.8(5.6; 25.1)18.6(6.9; 46.2)0.0003 Alu (%mC)23.6(21.8; 24.5)22.9(20.6; 24.4)0.0007 HICI (%mC)17.2(10.4; 33.0)20.4(7.4; 34.4)0.048 IL-6 (%mC)48.4(34.5; 69.9)44.7(28.7; 61.8)0.095 p16 (%mC)1.9(1.3; 3.6)2.1(0.8; 3.9)0.205GSTM1 genotyping *0/*018($41\%$)25($54\%$)0.204 *1/*1 and *0/*126($59\%$)22($46\%$)Outcomes DNAmAge l37(17.0; 54.0)36(19.0; 52.0)0.567 AgeAcc m-1(-22.0; 15.0)-3(-13.0; 4.0)0.096 Subjects with AgeAcc ≥ 0 N (%)20($40.8\%$)7($15.6\%$)0.007 TL n1.0(0.3; 3.0)1.2(0.4; 2.1)0.038 mtDNAcno1.0(0.4; 2.5)0.9(0.3; 1.7)0.128@ Statistical tests: Wilcoxon rank-sum test for interval variables and Chi-square test for frequency variables§ PAHs = Polycyclic Aromatic Hydrocarbonsa Charcoaled meat consumption ≥ once a weekb Daily consumption of fruit or vegetablesc Wood or coal-based heating at homed High environmental exposure from residence in town, intense traffic and presence of industries near homee Positive subjects with values higher than threshold limit of assay (0.01 mg/mmoles creatinine)f PAHs exposure evaluated by urinary excretion of 1-pyrenolg Occup_PAHs = (years of work in the cokery) × (1-pyrenol in µmoles/mol creatinine)h Environ_PAHs = Latent variable estimated by SEM (environmental exposure to PAHs) based on diet, indoor and outdoor activitiesi nDNA_changes = Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, LINE-1, p53, Alu, HICI, IL-6, p16, GSTM1k A value of 0.125 adducts/108 nucleotides was assigned to subjects with non-detectable adductsl DNAmAge = DNA methylation agem AgeAcc = difference between the DNAmAge and the chronological agen TL = Telomere length in lymphocyteso mtDNAcn = mitochondrial DNA copy number in lymphocytes *The analysis* was conducted with the statistical package STATA 14.
## Descriptive results
Table 1 shows the main characteristics of Polish coke-oven workers and non-exposed controls. We used the Wilcoxon rank-sum test to compare interval variables and Chi-square test for frequency variables; the corresponding p-values are reported in the last column of Table 1. To reduce confounding factors, all participants were males and noncurrent smokers (either never-smokers or former smokers who had quit smoking at least 1 year before sample collection as confirmed by analysis of nicotine and its metabolites) and matched for age (ages of two groups were almost overlapping and the difference was not significant) and ethnicity. Furthermore number of former smokers and years since cessation were equally distributed between workers and controls. The prevalence of indoor PAH exposure was higher, but non-significant, among coke-oven workers while that of diet and outdoor PAH exposure was similar in both groups. The cumulative occupational exposure Occup_PAHs was highly significant ($$p \leq 0.00001$$) in coke-oven workers compared to control subjects, for whom it was always zero, as well as the urinary excretion of 1-pyrenol ($$p \leq 0.00001$$). The latent variable Environ_PAHs, based on arbitrary units, had a negative median indicating lower PAHs exposure among controls and a positive median value suggesting higher exposure to environmental PAHs among coke-oven workers ($$p \leq 0.0001$$). Highly significant differences among groups were observed for anti-BPDE–DNA adducts ($$p \leq 0.00001$$), micronuclei ($$p \leq 0.00001$$), LINE-1 ($$p \leq 0.00001$$), p53 ($$p \leq 0.0003$$) and Alu ($$p \leq 0.0007$$), but not for other variables used as indicators to infer nDNA_changes. The median of this score, based on arbitrary units, was negative among controls and positive among coke-oven workers, suggesting that the nuclear DNA was more “damaged” among the latter than the formers; the difference was highly significant ($$p \leq 0.00001$$). No difference in the distribution of GSTM1 genotyping was found. In univariate analysis biological aging determined by DNAmAge, TL and mtDNAcn showed that the percentage of subjects with AgeAcc ≥ 0 (biological age older than chronological age) was higher in coke-oven workers than controls ($$p \leq 0.007$$), and TL is significantly shorter in coke-oven workers than in controls ($$p \leq 0.038$$), while mtDNAcn was similar in the two groups. In total 6 out of 49 ($10\%$) coke-oven workers and 10 out of 45 ($18\%$) controls declared diabetes or hypertension, no participant had cancer disease. Furthermore, no correlation between years since cessation, in former smokers in both coke-oven workers and controls, and biological aging detected by AgeAcc (r = -0.134, $$p \leq 0.407$$ and r = -0.062, $$p \leq 0.666$$), TL (r = -0.006, $$p \leq 0.654$$ and r = -0.017, $$p \leq 0.111$$) and mtDNAcn ($r = 0.003$, $$p \leq 0.691$$ and r = -0.006, $$p \leq 0.726$$), was found.
## DNAmAge
Table 2 shows three groups of SEM results (Structural Equations, Measurement, Covariates) for the mediation analysis for DNAmAge. In “Structural Equations”, the first model reveals that DNAmAge significantly associated with Chronic diseases ($p \leq 0.0001$) and Occup_PAHs ($p \leq 0.0001$), but not with Environ_PAHs ($$p \leq 0.267$$) or nDNA_changes ($$p \leq 0.079$$). The second model reveals that nDNA_changes significantly increased with Environ_PAHs ($$p \leq 0.008$$), while Occup_PAHs was not significant ($$p \leq 0.094$$). In the section “Measurement” the latent variable nDNA_changes was positively correlated with Anti-BPDE-DNA adducts ($p \leq 0.0001$), LINE-1 ($p \leq 0.0001$), Alu ($p \leq 0.0001$) and IL-6 ($$p \leq 0.005$$); this correlation was negative with p53 ($p \leq 0.0001$), HICI ($p \leq 0.0001$) and p16 ($p \leq 0.0001$). In the section “Covariances” the findings demonstrated that chronic disease was individually correlated with Environ_PAHs ($$p \leq 0.033$$). Using the graphical interface of SEM, the results shown in Table 2 were displayed as a path diagram in Fig. 1. Table 2SEM results of mediation analysis for DNAmAgeEndogenous variablesExogenous variablesBeta Coef$.95\%$CIp-valueLowerUpperStructural EquationsDNAmAgeOccup_PAHs0.070.060.080.000Environ_PAHs38.63.4073.90.267nDNA_changes−0.98−2.080.110.079Chronic diseases4.442.836.060.000nDNA_changesOccup_PAHs0.02−0.0030.050.094Environ_PAHs4.281.147.420.008MeasurementDiet ← Environ_PAHs1(constrained)Indoor ← Environ_PAHs3.43−7.1114.00.523Outdoor ← Environ_PAHs7.86−5.3221.00.242MN ← nDNA_changes1(constrained)Anti-BPDE-DNA adducts ← nDNA_changes1.811.631.990.000LINE-1 (%mC) ← nDNA_changes1.421.241.610.000p53 (%mC) ← nDNA_changes−2.53−2.97−2.100.000Alu (%mC) ← nDNa_changes0.250.220.270.000HICI (%mC) ← nDNA_changes−0.96−1.28−0.630.000IL-6 (%mC) ← nDNA_changes1.660.492.830.005p16 (%mC) ← nDNA_changes−0.05−0.06−0.040.000GSTM1 ← nDNA_changes0.02-0.120.210.578Covariancescov(Occup_PAHs,Environ_PAHs)0.54-1.012.140.483cov(Chron_d,Environ_PAHs)−0.01−0.02−0.0010.033Three groups of SEM results (structural equations; measurement; covariances) for the mediation analysis of biological age. Beta coefficients with “minus” sign indicating inverse relationship. Lower and upper limit of $95\%$ confidence intervals ($95\%$CI) and p-values estimated with standard error adjusted for 2 clusters (coke-oven workers and controls). Number of observations = 87Abbreviations: DNAmAge DNA methylation age, nDNA_changes Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, LINE-1, p53, Alu, HICI, IL-6, p16, GSTM1, Environ_PAHs Latent variable estimated by SEM (environmental exposure to PAHs) based on diet, indoor and outdoor activities, MN Micronuclei, PAHs Exposure to Polycyclic aromatic hydrocarbons, Occup_PAHs (years of work in the cokery) × (1-pyrenol in μmoles/mol creatinine)Goodness of fit statistics: Standardized root mean squared residual (SRMR) = 0.09; Coefficient of determination (CD) = 0.703; Stability index = 0 (SEM satisfies stability condition)
## TL
Table 3 reports three groups of SEM results (Structural Equations, Measurement, Covariates) for the mediation analysis for TL. In “Structural Equations”, the first model shows that TL significantly decreased with Occup_PAHs ($p \leq 0.0001$), but not with Environ_PAHs ($$p \leq 0.074$$) or nDNA_changes ($$p \leq 0.966$$). The second model reveals that nDNA_changes significantly increased with Environ_PAHs ($p \leq 0.0001$) and Occup_PAHs ($$p \leq 0.040$$). In “Measurement”, the first model indicated that the most significant determinants of environmental PAH exposure were “indoor” ($p \leq 0.0001$) and “outdoor” ($p \leq 0.0001$). Their positive coefficients indicate that the latent variable Environ_PAHs tends to increase with increasing values of these variables. In the second measurement model, the results of Table 3 were equal to those shown in Table 2, that the latent variable nDNA_changes tends to increase with rising values of the methylation states (%) of LINE-1, Alu and IL-6 ($p \leq 0.0001$), and of anti-BPDE–DNA adducts ($p \leq 0.0001$) that are the positive coefficients. While, the negative coefficient of the methylation state (%) of p53, HICI, and p16 ($p \leq 0.0001$) means that these factors tend to go in the opposite direction and nDNA_changes (latent variable) increases with their decreasing values. No significant result was obtained for the relationship between Occup_PAHs and Environ_PAHs ($$p \leq 0.683$$). Using the graphical interface of SEM, the results shown in Table 3 were displayed as a path diagram in Fig. 2.Table 3SEM results of mediation analysis for TLEndogenous variablesExogenous variablesBeta Coef$.95\%$CIp-valueLowerUpperStructural EquationsTLOccup_PAHs−0.002−0.003−0.0010.000Environ_PAHs−3.783−7.9280.3630.074nDNA_changes0.008−0.040.040.966nDNA_changesOccup_PAHs0.0210.0020.0400.040Environ_PAHs19.6912.1627.230.000MeasurementDiet ← Environ_PAHs1(constrained)Indoor ← Environ_PAHs7.6085.4669.7500.000Outdoor ← Environ_PAHs11.787.09816.470.000Home ← Environ_PAHs12.09−1.02425.200.071MN← nDNA_changes1(constrained)Anti-BPDE-DNA adducts ← nDNA_changes1.8291.4292.2290.000LINE-1 (%mC) ←nDNA_changes1.4171.4051.4290.000p53 (%mC) ← nDNA_changes−2.519−2.624−2.4140.000Alu (%mC) ← nDNA_changes0.2350.2080.2610.000HICI (%mC) ← nDNA_changes−0.915−1.133−0.6970.000IL-6 (%mC) ← nDNA_changes1.5720.8242.3200.000p16 (%mC) ← nDNA_changes-0.052-0.057-0.0480.000GSTM1 ← nDNA_changes0.016-0.0310.0630.504Covariancescov(Occup_PAHs,Environ_PAHs)0.122-0.4630.7060.683Three groups of (structural equations, measurement and covariances) for the mediation analysis; standardized beta coefficients (with “minus” sign indicating inverse relationship) with lower and upper limit of $95\%$ confidence intervals ($95\%$CI) and p-values. Standard Error adjusted for 2 clusters (coke-oven workers and controls). Number of observations = 87Abbreviations: TL Telomere length, Chronic_d Chronic diseases, nDNA_changes Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, LINE-1, p53, Alu, HICI, IL-6, p16, GSTM1, Environ_PAHs Latent variable estimated by SEM (environmental exposure to PAHs) based on diet, features ofhome, indoor and outdoor activities, MN Micronuclei, PAHs Exposure to Polycyclic aromatic hydrocarbons, Occup_PAHs (years of work in the cokery) × (1-pyrenol in μmoles/mol creatinine)Goodness of fit statistics: Standardized root mean squared residual (SRMR) = 0.092; Coefficient of determination (CD) = 0.914; Stability index = 0 (SEM satisfies stability condition)
## mtDNAcn
Table 4 shows three groups of SEM results (Structural Equations, Measurement, Covariates) for the mediation analysis for mtDNAcn. In “Structural Equations”, the first model shows that mtDNAcn significantly increased with nDNA_changes ($p \leq 0.0001$), but not with Occup_PAHs ($$p \leq 0.598$$) or Environ_PAHs ($$p \leq 0.719$$) or chronic diseases ($$p \leq 0.294$$). The second model reveals that nDNA_changes significantly increased with Environ_PAHs ($p \leq 0.0001$) and Occup_PAHs ($$p \leq 0.009$$). In “Measurement” section, the most significant indicators for the latent variable nDNA_changes are the methylation states (%) of LINE-1 and Alu ($p \leq 0.0001$) and IL-6 ($$p \leq 0.022$$), the presence of anti-BPDE–DNA adducts ($p \leq 0.0001$), that are positively correlated, and the methylation state (%) of p53, HICI and p16 ($p \leq 0.0001$), which are negatively correlated. In “Covariances” section, no significant result was obtained for the relationship between Occup_PAHs and Environ_PAHs ($$p \leq 0.688$$) as well as between Chronic diseases and Occup_PAHs ($$p \leq 0.420$$). Using the graphical interface of SEM, the results shown in Table 4 were displayed as a path diagram in Fig. 3.Table 4SEM results of mediation analysis for mtDNAcnEndogenous variablesExogenous variablesBeta Coef$.95\%$CIp-valueLowerUpperStructural EquationsmtDNAcnOccup_PAHs-0.0005-0.00220.00130.598Environ_PAHs-1.4688-9.47146.53400.719nDNA_changes0.06900.04830.08960.000Chronic diseases0.0338-0.02930.09690.294nDNA_changesOccup_PAHs0.02140.00540.03730.009Environ_ PAHs7.11086.08858.13310.000MeasurementDiet ← Environ_ PAHs1(constrained)Indoor ← Environ_ PAHs4.4960-2.22581.12170.190Outdoor ← Environ_ PAHs5.6125-29.54540.7710.754MN← nDNA_changes1(constrained)Anti-BPDE-DNA adducts ← nDNA_changes1.80011.65201.94830.000LINE-1 (%mC) ← nDNA_changes1.38201.13541.62850.000p53 (%mC) ← nDNA_changes-2.4248-3.0079-1.84170.000Alu (%mC) ← nDNA_changes0.23720.23650.23790.000HICI (%mC) ← nDNA_changes-0.8902-1.1999-0.58050.000IL-6 (%mC) ← nDNA_changes1.78270.25533.31020.022p16 (%mC) ← nDNA_changes-0.0409-0.0557-0.02610.000GSTM1 ← nDNA_changes0.0241-0.01220.06060.193Covariancescov(Occup_PAHs, Environ_PAHs)0.2981-115.551.75170.688cov(Chron_d,Environ_ PAHs)-0.0096-0.03290.01370.420Three groups of SEM results (structural equations, measurement and covariances) for the mediation analysis; standardized beta coefficients (with “minus” sign indicating inverse relationship) with lower and upper limit of $95\%$ confidence intervals ($95\%$CI) and p-values. Standard Error adjusted for 2 clusters (coke-oven workers and controls). SEM’s goodness-of-fit statistics at bottom of table. Number of observations = 87Abbreviations: mtDNAcn Mitochondrial DNA copy number, nDNA_changes Latent variable estimated by SEM (overall change of DNA) based on MN, Anti-BPDE-DNA adducts, LINE-1, p53, Alu, HICI, IL-6, p16, GSTM1, MN Micronuclei, PAHs Exposure to Polycyclic aromatic hydrocarbons, Environ_PAHs Latent variable estimated by SEM (environmental exposure to PAHs), Occup_PAHs (years of work in the cokery) × (1-pyrenol in μmoles/mol creatinine)Goodness of fit statistics: Standardized root mean squared residual (SRMR) = 0.092; Coefficient of determination (CD) = 0.914; Stability index = 0 (SEM satisfies stability condition)
## Discussion
The present study evaluated by SEM analysis the relative magnitude of different pathways by which PAHs exposure may affect the main hallmarks of biological aging – DNAmAge, TL, and mtDNAcn– supposing that some of the effects of environmental and occupational exposure to PAHs also act indirectly by triggering nDNA alterations. Our study reveals that DNAmAge increased with occupational PAH exposure and the presence of chronic diseases; but not with environmental PAH exposure and nDNA alterations. On the same line, TL is confirmed to be directly decreased with occupational PAH exposure; but not with environmental PAH exposure and nDNA alterations. Conversely, mtDNAcn indirectly increased with environmental and occupational PAH exposure acting through nDNA alterations.
The direct positive relationship between DNAmAge and occupational PAH exposure is in line with a previous study by Li et al. [ 38], underscoring the negative impact of high PAH exposure on aging. The work by Li et al. [ 38] was however performed with a different methylation age predictor, specifically built for Chinese populations, and reported that a 1-unit increase in 1-hydroxypyrene, even deriving from smoking behavior, was associated with a 0.53-y increase in AgeAcc. In our work, all study subjects, all non-current smokers and exposed individuals were similar to controls for age, gender and ethnicity minimizing the possibility that the higher biological aging could depend on factors other than PAH exposure. In addition, we evaluated several other potential confounding factors, including dietary PAHs, indoor and outdoor PAH exposures that showed no differences between the two groups. DNA methylation is currently the most promising molecular marker for monitoring biological aging and predicting life expectancy [39]. In humans, DNA methylation changes start early in life, as demonstrated by longitudinal studies of infants' blood [40, 41]. Notably, these early epigenetic profiles continue to accumulate changes with the advancement of age as shown in twins that do not share the same habits and/or environments [42, 43], indicating aging-associated DNA methylation changes depending on environmental factors. In our previous study [23], DNAmAge and AgeAcc blood lymphocytes correlated with those of pulmonary cells, advising that blood lymphocytes could be a validate surrogate tissue for lung aging studies. This suggests that the PAH-related acceleration aging of blood lymphocytes observed in coke-oven worker mirrors what happens in the respiratory tract.
Also we found an increase in DNAmAge related to chronic diseases. This finding is consistent with previous studies that reported a substantial increase in DNAmAge associated with age-related chronic diseases including frailty [44], cancer [45], diabetes [46], cardiovascular diseases (CVD) [47], dementia [48], as well as with chronic obstructive pulmonary disease (COPD) [23]. The latter, from our previous work [23], shows that blood leukocytes DNAmAge and AgeAcc significantly increase (become older) in COPD patients, and with a reduction in lung function (FEV$1\%$). DNAmAge, therefore, seems an accountable signature of the epigenetic aging chronic disease-related.
The direct negative relationship, detected by SEM analysis, between TL and PAH occupational exposure, not influenced by chronic diseases, confirmed the results of our previous study in coke-oven workers showing that telomeres significantly reduced with years of work [12]. The fact that the environmental exposure to PAHs in which the most significant determinants were “indoor” and “outdoor” are not directly correlated depends on the fact that the occupational exposure of coke-oven workers is very high. In our study, the cumulative occupational exposure to PAHs was in fact very much higher in workers compared to controls, given that the majority of the PAH-exposed workers exceeded the Biological Exposure Index proposed by Jongeneelen [49] for urinary 1-pyrenol. The direct negative relationship between TL and PAHs was also observed in everyday-life exposure to PAHs in the general population [27].
Telomeres, repetitive functional complexes of DNA/protein at the ends of chromosomes, preserve DNA integrity that in their absence would be gradually lost with each cell division [50]. Their length measured in blood lymphocytes is considered an indicator of biological aging [50]. Loss of telomere sequence in lymphocytes has been also related to adverse age-related outcomes, in particular CVD [51, 52] and respiratory diseases [53]. Exposure to PAHs may pose a risk not only for lung cancer, but also for CVD, including atherosclerosis, hypertension, thrombosis and myocardial infarction [54]. Since PAH exposure is pervasive and modifiable, it is an appropriate target for age-related disorders, especially CVD, prevention research studies.
The chances that the association with shorter TL could depend on factors other than PAH exposure were minimized because all study subjects were non-current smokers and exposed individuals were similar to controls for age, gender and ethnicity. Furthermore, TL was not affected by chronic diseases.
MtDNAcn indirectly increased with environmental and occupational PAH exposure, acting through nDNA_changes. nDNA_changes is the latent variable estimated by SEM (overall change of DNA) in which adducts, LINE-1, p53, Alu, HICI, IL6, p16, are major determinants. This would suggest a relationship between the number of changes in nDNA and mtDNA. This result confirmed our previous study detecting a significantly higher mtDNAcn in coke-oven workers using another statistical analysis [13]. This observation is in line with previous findings by Sahin and colleagues [55] that showed a potential unifying mechanism connecting the nucleus and mitochondria in cellular aging. In that work, progressive nuclear changing, mediated by the activation of a p53-dependent pathway, was found to determine a reduction of mitochondrial function and mtDNAcn [55].
The present study has several strengths. The enrollment of the study participants was carefully designed to minimize potential confounding and increase the capability to reveal PAH effects by selecting non–current smoking males, all coke-oven workers and controls living in the same residential area, reducing the probability that the observed associations were dependent on factors other than occupational PAH exposure. We also evaluated several other potential sources of PAH exposure, including dietary PAHs, indoor and outdoor PAH exposures, which showed no differences between coke-oven workers and controls. Our study had reliable measurements of PAH(B[a]P) internal and target doses. Also, we measured in the study participants, biomarkers of genetic instability and methylation that allowed for characterizing the inter-correlation between nDNA and mtDNAcn alterations. Furthermore, we applied the method proposed by Zbieć-Piekarska et al. [ 19] to assess DNAmAge, on data from five CpG sites using the locus-specific technology pyrosequencing with some modifications [23, 32]. This model, based on an algorithm developed in a larger sample ($$n = 420$$) and then validated in a smaller one ($$n = 300$$) covering the entire adult life span, shows that DNAmAge highly correlates ($r = 0.94$) to chronological age with a mean deviation (4.5 years) similar to those of Horvath [17] and Hannum et al. [ 18] ($r = 0.96$ and $r = 0.91$) with 3.6 and 4.9 years mean, which are considered the reference methods. We recently automated this method to improve efficiency and speed while maintaining high prediction accuracy [23, 32]. By using this method, we can perform the analyses in a standardized manner reducing errors, and this is another strong point of our study. Moreover, it is interesting to note that pyrosequencing has the potential for multiplexing, which can simplify the protocol and reduce the cost of technical analysis. Finally, the results of this study appear to be biologically plausible and the direction of the effects is consistent with the available literature data on aging mechanisms.
We also recognize limitations to our study. This is a small-sized study and its results need to be confirmed in a larger independent investigation. Its cross-sectional design does not allow for investigating the temporal relationship of PAH exposure with biomarkers of damage, genetic instability, and indicators of biological aging. The absence of air monitoring, as well as repeated biological sampling, are also limitations of the study exposure assessment strategy. However, PAH exposure was assessed using biomarkers of internal dose (urinary 1-pyrenol) and target dose (anti-BPDE–DNA adduct), which may more appropriately represent the effective exposure dose. We also recognize that socio-economic factors, in particular related to education and income, were not part of this study. We cannot therefore exclude that different socio-economic status in particular related to education and income, might have contributed, along with PAH exposure in the increased of biological aging observed in lymphocytes of coke-oven workers. Socio-economic factors are however non-specific factors indirectly linked to lifestyle habits such as cigarette smoking or diet, environmental exposures, housing, which we have considered in our study. The enrollment of the study participants was in fact carefully designed to minimize potential confounding factors. We matched coke-oven workers and controls for their individual characteristics, including age, gender, and ethnicity. Furthermore, the lack of statistical significant differences in outdoor and indoor exposures as well as the fact that all participants live in the same residential area, lead us to assume that they could be exposed to similar levels of environmental factors different from PAHs. In addition, we adjusted the analysis contrasting high-exposed workers, as well as those based on continuous exposure or biomarker variables, for age.
The attractiveness of SEM analysis stems mainly from the fact that researchers have recognized the necessity of grasping the complex interrelations between multiple variables under study. Traditional statistical approaches apply solely to a limited number of variables and thus fail to deal with emerging sophisticated theories. SEM analysis is a statistical technique that links observed data with qualitative causative assumptions and tests whether variables are interdependent, and if so, the details of their interactions. This is achieved through an estimation procedure [56], which uses a set of concurrent regression equations to yield coefficient estimators more efficiently than single-equation estimators. This methodology is appropriate for the investigation of complex interrelationships, as it tests causative relationships instead of mere correlations [24].
## Conclusion
We showed that DNAmAge was positively and significantly correlated with occupational PAH exposure. Our findings indicate that occupational PAH exposure makes coke-oven workers biologically older. We also confirmed that TL was negatively correlated with occupational PAH exposure, suggesting that both mechanisms of biological aging (DNAmAge and TL) are PAH targets. Also, differences in mtDNAcn, which is typically altered through nDNA alterations triggered by environmental and occupational PAH exposure, suggested a nuclear-mitochondrial core-axis of aging. Lowering PAH exposure may prevent biological aging and age-related diseases.
## References
1. 1.Fulop T, Larbi A, Khalil A, Cohen AA, Witkowski JM. Are We Ill Because We Age? Frontiers in Physiology. 2019;10. Available from: https://www.frontiersin.org/articles/, 10.3389/fphys.2019.01508
2. Ahadi S, Zhou W, Schüssler-Fiorenza Rose SM, Sailani MR, Contrepois K, Avina M. **Personal aging markers and ageotypes revealed by deep longitudinal profiling**. *Nat Med* (2020.0) **26** 83-90. DOI: 10.1038/s41591-019-0719-5
3. Hahad O, Frenis K, Kuntic M, Daiber A, Münzel T. **Accelerated Aging and Age-Related Diseases (CVD and Neurological) Due to Air Pollution and Traffic Noise Exposure**. *Int J Mol Sci* (2021.0) **22** 2419. DOI: 10.3390/ijms22052419
4. Wigmann C, Hüls A, Krutmann J, Schikowski T. **Estimating the Relative Contribution of Environmental and Genetic Risk Factors to Different Aging Traits by Combining Correlated Variables into Weighted Risk Scores**. *Int J Environ Res Public Health* (2022.0) **19** 16746. DOI: 10.3390/ijerph192416746
5. Ward-Caviness CK, Nwanaji-Enwerem JC, Wolf K, Wahl S, Colicino E, Trevisi L. **Long-term exposure to air pollution is associated with biological aging**. *Oncotarget* (2016.0) **7** 74510-74525. DOI: 10.18632/oncotarget.12903
6. 6.Sharifi-Rad M, Anil Kumar NV, Zucca P, Varoni EM, Dini L, Panzarini E, et al. Lifestyle, Oxidative Stress, and Antioxidants: Back and Forth in the Pathophysiology of Chronic Diseases. Frontiers in Physiology. 2020;11. Available from: https://www.frontiersin.org/articles/, 10.3389/fphys.2020.00694
7. Zuo L, Prather ER, Stetskiv M, Garrison DE, Meade JR, Peace TI. **Inflammaging and Oxidative Stress in Human Diseases: From Molecular Mechanisms to Novel Treatments**. *Int J Mol Sci* (2019.0) **20** 4472. DOI: 10.3390/ijms20184472
8. Palackal NT, Burczynski ME, Harvey RG, Penning TM. **The ubiquitous aldehyde reductase (AKR1A1) oxidizes proximate carcinogen trans-dihydrodiols to o-quinones: potential role in polycyclic aromatic hydrocarbon activation**. *Biochemistry* (2001.0) **40** 10901-10910. DOI: 10.1021/bi010872t
9. Liu AL, Lu WQ, Wang ZZ, Chen WH, Lu WH, Yuan J. **Elevated levels of urinary 8-hydroxy-2 -deoxyguanosine, lymphocytic micronuclei, and serum glutathione S-transferase in workers exposed to coke oven emissions**. *Environ Health Perspect* (2006.0) **114** 673-677. DOI: 10.1289/ehp.8562
10. Hadrup N, Mielżyńska-Švach D, Kozłowska A, Campisi M, Pavanello S, Vogel U. **Association between a urinary biomarker for exposure to PAH and blood level of the acute phase protein serum amyloid A in coke oven workers**. *Environ Health* (2019.0) **18** 81. DOI: 10.1186/s12940-019-0523-1
11. Pavanello S, Bollati V, Pesatori AC, Kapka L, Bolognesi C, Bertazzi PA. **Global and gene-specific promoter methylation changes are related to anti-B[a]PDE-DNA adduct levels and influence micronuclei levels in polycyclic aromatic hydrocarbon-exposed individuals**. *Int J Cancer* (2009.0) **125** 1692-1697. DOI: 10.1002/ijc.24492
12. Pavanello S, Pesatori AC, Dioni L, Hoxha M, Bollati V, Siwinska E. **Shorter telomere length in peripheral blood lymphocytes of workers exposed to polycyclic aromatic hydrocarbons**. *Carcinogenesis* (2010.0) **31** 216-221. DOI: 10.1093/carcin/bgp278
13. Pavanello S, Dioni L, Hoxha M, Fedeli U, Mielzynska-Svach D, Baccarelli AA. **Mitochondrial DNA copy number and exposure to polycyclic aromatic hydrocarbons**. *Cancer Epidemiol Biomarkers Prev* (2013.0) **22** 1722-1729. DOI: 10.1158/1055-9965.EPI-13-0118
14. Sahin E, DePinho RA. **Axis of ageing: telomeres, p53 and mitochondria**. *Nat Rev Mol Cell Biol* (2012.0) **13** 397-404. DOI: 10.1038/nrm3352
15. Lowe D, Horvath S, Raj K. **Epigenetic clock analyses of cellular senescence and ageing**. *Oncotarget* (2016.0) **7** 8524-8531. DOI: 10.18632/oncotarget.7383
16. Horvath S, Raj K. **DNA methylation-based biomarkers and the epigenetic clock theory of ageing**. *Nat Rev Genet* (2018.0) **19** 371-384. DOI: 10.1038/s41576-018-0004-3
17. Horvath S. **DNA methylation age of human tissues and cell types**. *Genome Biol* (2013.0) **14** R115. DOI: 10.1186/gb-2013-14-10-r115
18. Hannum G, Guinney J, Zhao L, Zhang L, Hughes G, Sadda S. **Genome-wide methylation profiles reveal quantitative views of human aging rates**. *Mol Cell* (2013.0) **49** 359-367. DOI: 10.1016/j.molcel.2012.10.016
19. Zbieć-Piekarska R, Spólnicka M, Kupiec T, Parys-Proszek A, Makowska Ż, Pałeczka A. **Development of a forensically useful age prediction method based on DNA methylation analysis**. *Forensic Sci Int Genet* (2015.0) **17** 173-179. DOI: 10.1016/j.fsigen.2015.05.001
20. Pavanello S, Campisi M, Tona F, Lin CD, Iliceto S. **Exploring Epigenetic Age in Response to Intensive Relaxing Training: A Pilot Study to Slow Down Biological Age**. *Int J Environ Res Public Health* (2019.0) **16** 3074. DOI: 10.3390/ijerph16173074
21. Dhingra R, Nwanaji-Enwerem JC, Samet M, Ward-Caviness CK. **DNA Methylation Age—Environmental Influences, Health Impacts, and Its Role in Environmental Epidemiology**. *Curr Envir Health Rpt* (2018.0) **5** 317-327. DOI: 10.1007/s40572-018-0203-2
22. Fransquet PD, Wrigglesworth J, Woods RL, Ernst ME, Ryan J. **The epigenetic clock as a predictor of disease and mortality risk: a systematic review and meta-analysis**. *Clin Epigenetics* (2019.0) **11** 62. DOI: 10.1186/s13148-019-0656-7
23. Campisi M, Liviero F, Maestrelli P, Guarnieri G, Pavanello S. **DNA Methylation-Based Age Prediction and Telomere Length Reveal an Accelerated Aging in Induced Sputum Cells Compared to Blood Leukocytes: A Pilot Study in COPD Patients**. *Front Med (Lausanne)* (2021.0) **8** 690312. DOI: 10.3389/fmed.2021.690312
24. 24.Pavanello S, Stendardo M, Mastrangelo G, Bonci M, Bottazzi B, Campisi M, et al. Inflammatory Long Pentraxin 3 is Associated with Leukocyte Telomere Length in Night-Shift Workers. Front Immunol. 2017;8. Available from: https://www.frontiersin.org/articles/,
10.3389/fimmu.2017.00516/full.
25. Pavanello S, Kapka L, Siwinska E, Mielzyñska D, Bolognesi C, Clonfero E. **Micronuclei Related to Anti–B[a]PDE-DNA Adduct in Peripheral Blood Lymphocytes of Heavily Polycyclic Aromatic Hydrocarbon-Exposed Nonsmoking Coke-Oven Workers and Controls**. *Cancer Epidemiol Biomark Prev* (2008.0) **17** 2795-2799. DOI: 10.1158/1055-9965.EPI-08-0346
26. Pavanello S, Siwinska E, Mielzynska D, Clonfero E. **GSTM1 null genotype as a risk factor for anti-BPDE-DNA adduct formation in mononuclear white blood cells of coke-oven workers**. *Mutat Res* (2004.0) **558** 53-62. DOI: 10.1016/j.mrgentox.2003.10.019
27. Pavanello S, Campisi M, Mastrangelo G, Hoxha M, Bollati V. **The effects of everyday-life exposure to polycyclic aromatic hydrocarbons on biological age indicators**. *Environ Health* (2020.0) **19** 128. DOI: 10.1186/s12940-020-00669-9
28. Hou L, Zhu ZZ, Zhang X, Nordio F, Bonzini M, Schwartz J. **Airborne particulate matter and mitochondrial damage: a cross-sectional study**. *Environ Health* (2010.0) **9** 48. DOI: 10.1186/1476-069X-9-48
29. 29.Pavanello S, Angelici L, Hoxha M, Cantone L, Campisi M, Tirelli AS, et al. Sterol 27-Hydroxylase Polymorphism Significantly Associates With Shorter Telomere, Higher Cardiovascular and Type-2 Diabetes Risk in Obese Subjects. Frontiers in Endocrinology. 2018;9. Available from: https://www.frontiersin.org/articles/,10.3389/fendo.2018.00309.
30. Pavanello S, Campisi M, Grassi A, Mastrangelo G, Durante E, Veronesi A. **Longer Leukocytes Telomere Length Predicts a Significant Survival Advantage in the Elderly TRELONG Cohort, with Short Physical Performance Battery Score and Years of Education as Main Determinants for Telomere Elongation**. *J Clin Med* (2021.0) **10** 3700. DOI: 10.3390/jcm10163700
31. Cawthon RM. **Telomere measurement by quantitative PCR**. *Nucleic Acids Res* (2002.0) **30** e47. DOI: 10.1093/nar/30.10.e47
32. Pavanello S, Campisi M, Rigotti P, Bello MD, Nuzzolese E, Neri F. **DNA Methylation - and Telomere - Based Biological Age Estimation as Markers of Biological Aging in Donors Kidneys**. *Front Med (Lausanne)* (2022.0) **9** 832411. DOI: 10.3389/fmed.2022.832411
33. Pavanello S, Campisi M, Fabozzo A, Cibin G, Tarzia V, Toscano G. **The biological age of the heart is consistently younger than chronological age**. *Sci Rep* (2020.0) **10** 10752. DOI: 10.1038/s41598-020-67622-1
34. Cepeda MS, Boston R, Farrar JT, Strom BL. **Optimal matching with a variable number of controls vs. a fixed number of controls for a cohort study. trade-offs**. *J Clin Epidemiol* (2003.0) **56** 230-7. DOI: 10.1016/S0895-4356(02)00583-8
35. Gallo V, Egger M, McCormack V, Farmer PB, Ioannidis JPA, Kirsch-Volders M. **STrengthening the Reporting of OBservational studies in Epidemiology – Molecular Epidemiology (STROBE-ME): An Extension of the STROBE Statement**. *PLoS Med* (2011.0) **8** e1001117. DOI: 10.1371/journal.pmed.1001117
36. 36.Worthington RL, Whittaker TA. Scale Development Research: A Content Analysis and Recommendations for Best Practices. 2022. Available from: https://journals.sagepub.com/doi/abs/10.1177/0011000006288127.
37. 37.Hair J. Multivariate Data Analysis. Faculty Publications. 2009 Feb 23; Available from: https://digitalcommons.kennesaw.edu/facpubs/2925
38. Li J, Zhu X, Yu K, Jiang H, Zhang Y, Wang B. **Exposure to Polycyclic Aromatic Hydrocarbons and Accelerated DNA Methylation Aging**. *Environ Health Perspect.* (2018.0) **126** 067005. DOI: 10.1289/EHP2773
39. Bell CG, Lowe R, Adams PD, Baccarelli AA, Beck S, Bell JT. **DNA methylation aging clocks: challenges and recommendations**. *Genome Biol* (2019.0) **20** 249. DOI: 10.1186/s13059-019-1824-y
40. Herbstman JB, Wang S, Perera FP, Lederman SA, Vishnevetsky J, Rundle AG. **Predictors and consequences of global DNA methylation in cord blood and at three years**. *PLoS ONE* (2013.0) **8** e72824. DOI: 10.1371/journal.pone.0072824
41. Martino DJ, Tulic MK, Gordon L, Hodder M, Richman TR, Metcalfe J. **Evidence for age-related and individual-specific changes in DNA methylation profile of mononuclear cells during early immune development in humans**. *Epigenetics* (2011.0) **6** 1085-1094. DOI: 10.4161/epi.6.9.16401
42. Fraga MF, Ballestar E, Paz MF, Ropero S, Setien F, Ballestar ML. **Epigenetic differences arise during the lifetime of monozygotic twins**. *Proc Natl Acad Sci U S A* (2005.0) **102** 10604-10609. DOI: 10.1073/pnas.0500398102
43. Tan Q, Heijmans BT, Hjelmborg JVB, Soerensen M, Christensen K, Christiansen L. **Epigenetic drift in the aging genome: a ten-year follow-up in an elderly twin cohort**. *Int J Epidemiol* (2016.0) **45** 1146-1158. PMID: 27498152
44. Gale CR, Marioni RE, Harris SE, Starr JM, Deary IJ. **DNA methylation and the epigenetic clock in relation to physical frailty in older people: the Lothian Birth Cohort 1936**. *Clin Epigenetics* (2018.0) **10** 101. DOI: 10.1186/s13148-018-0538-4
45. Dugué PA, Bassett JK, Joo JE, Jung CH, Ming Wong E, Moreno-Betancur M. **DNA methylation-based biological aging and cancer risk and survival: Pooled analysis of seven prospective studies**. *Int J Cancer* (2018.0) **142** 1611-1619. DOI: 10.1002/ijc.31189
46. Grant CD, Jafari N, Hou L, Li Y, Stewart JD, Zhang G. **A longitudinal study of DNA methylation as a potential mediator of age-related diabetes risk**. *Geroscience* (2017.0) **39** 475-489. DOI: 10.1007/s11357-017-0001-z
47. Roetker NS, Pankow JS, Bressler J, Morrison AC, Boerwinkle E. **Prospective Study of Epigenetic Age Acceleration and Incidence of Cardiovascular Disease Outcomes in the ARIC Study (Atherosclerosis Risk in Communities)**. *Circ Genom Precis Med* (2018.0) **11** e001937. DOI: 10.1161/CIRCGEN.117.001937
48. Horvath S, Ritz BR. **Increased epigenetic age and granulocyte counts in the blood of Parkinson’s disease patients**. *Aging* (2015.0) **7** 1130-1142. DOI: 10.18632/aging.100859
49. Jongeneelen FJ. **A guidance value of 1-hydroxypyrene in urine in view of acceptable occupational exposure to polycyclic aromatic hydrocarbons**. *Toxicol Lett* (2014.0) **231** 239-248. DOI: 10.1016/j.toxlet.2014.05.001
50. Blackburn EH, Epel ES, Lin J. **Human telomere biology: A contributory and interactive factor in aging, disease risks, and protection**. *Science* (2015.0) **350** 1193-1198. DOI: 10.1126/science.aab3389
51. Yeh JK, Wang CY. **Telomeres and Telomerase in Cardiovascular Diseases**. *Genes (Basel)* (2016.0) **7** 58. DOI: 10.3390/genes7090058
52. Herrmann W, Herrmann M. **The Importance of Telomere Shortening for Atherosclerosis and Mortality**. *J Cardiovasc Dev Dis* (2020.0) **7** 29. DOI: 10.3390/jcdd7030029
53. Ruiz A, Flores-Gonzalez J, Buendia-Roldan I, Chavez-Galan L. **Telomere Shortening and Its Association with Cell Dysfunction in Lung Diseases**. *Int J Mol Sci* (2021.0) **23** 425. DOI: 10.3390/ijms23010425
54. Holme JA, Brinchmann BC, Refsnes M, Låg M, Øvrevik J. **Potential role of polycyclic aromatic hydrocarbons as mediators of cardiovascular effects from combustion particles**. *Environ Health* (2019.0) **18** 74. DOI: 10.1186/s12940-019-0514-2
55. Sahin E, Colla S, Liesa M, Moslehi J, Müller FL, Guo M. **Telomere dysfunction induces metabolic and mitochondrial compromise**. *Nature* (2011.0) **470** 359-365. DOI: 10.1038/nature09787
56. Victora CG, Huttly SR, Fuchs SC, Olinto MT. **The role of conceptual frameworks in epidemiological analysis: a hierarchical approach**. *Int J Epidemiol* (1997.0) **26** 224-227. DOI: 10.1093/ije/26.1.224
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.