ADS599-Capstone/modeling_data

---

license: other

license_name: physionet-credentialed-health-data-1.5.0

license_link: https://physionet.org/content/mimiciv/view-license/3.1/

pretty_name: ADS599 Capstone Modeling Data — Empty Patient State

dataset_info:

• config_name: all_med_actions

  features:

  • name: ed_stay_id

    dtype: float64

  • name: subject_id

    dtype: int64

  • name: charttime

    dtype: timestamp[ns]

  • name: minutes_into_stay

    dtype: float64

  • name: drug_class

    dtype: string

  • name: action_idx

    dtype: int64

  • name: source

    dtype: string

  splits:

  • name: all_med_actions

    num_bytes: 93408141

    num_examples: 1235609

  download_size: 16415454

  dataset_size: 93408141

• config_name: combined_medication_features

  features:

  • name: ed_stay_id

    dtype: int64

  • name: subject_id

    dtype: int64

  • name: recon_ace_inhibitor

    dtype: uint8

  • name: recon_analgesic___nsaid

    dtype: uint8

  • name: recon_analgesic___opioid_nsaid

    dtype: uint8

  • name: recon_antiarrhythmic

    dtype: uint8

  • name: recon_antibiotic

    dtype: uint8

  • name: recon_anticoagulant

    dtype: uint8

  • name: recon_anticonvulsant

    dtype: uint8

  • name: recon_antiemetic

    dtype: uint8

  • name: recon_antiplatelet

    dtype: uint8

  • name: recon_antipsychotic

    dtype: uint8

  • name: recon_benzodiazepine___sedative_anxiolytic

    dtype: uint8

  • name: recon_beta_blocker

    dtype: uint8

  • name: recon_bronchodilator

    dtype: uint8

  • name: recon_calcium_channel_blocker

    dtype: uint8

  • name: recon_corticosteroid

    dtype: uint8

  • name: recon_diuretic

    dtype: uint8

  • name: recon_gi___acid_suppression

    dtype: uint8

  • name: recon_insulin_glucose

    dtype: uint8

  • name: recon_nitrate

    dtype: uint8

  • name: recon_n_total_meds

    dtype: uint16

  • name: recon_n_drug_classes

    dtype: uint8

  • name: edmeds_ace_inhibitor

    dtype: uint8

  • name: edmeds_analgesic___acetaminophen

    dtype: uint8

  • name: edmeds_analgesic___nsaid

    dtype: uint8

  • name: edmeds_analgesic___opioid_nsaid

    dtype: uint8

  • name: edmeds_antiarrhythmic

    dtype: uint8

  • name: edmeds_antibiotic

    dtype: uint8

  • name: edmeds_anticoagulant

    dtype: uint8

  • name: edmeds_anticonvulsant

    dtype: uint8

  • name: edmeds_antiemetic

    dtype: uint8

  • name: edmeds_antiplatelet

    dtype: uint8

  • name: edmeds_antipsychotic

    dtype: uint8

  • name: edmeds_benzodiazepine___sedative_anxiolytic

    dtype: uint8

  • name: edmeds_beta_blocker

    dtype: uint8

  • name: edmeds_bronchodilator

    dtype: uint8

  • name: edmeds_calcium_channel_blocker

    dtype: uint8

  • name: edmeds_corticosteroid

    dtype: uint8

  • name: edmeds_diuretic

    dtype: uint8

  • name: edmeds_gi___acid_suppression

    dtype: uint8

  • name: edmeds_iv_fluid

    dtype: uint8

  • name: edmeds_insulin_glucose

    dtype: uint8

  • name: edmeds_nitrate

    dtype: uint8

  • name: edmeds_other

    dtype: uint8

  • name: edmeds_n_dispenses

    dtype: uint16

  • name: edmeds_n_drug_classes

    dtype: uint8

  splits:

  • name: med_only

    num_bytes: 25173099

    num_examples: 399573

  download_size: 6049182

  dataset_size: 25173099

• config_name: ecg_features

  features:

  • name: subject_id

    dtype: int64

  • name: stay_id

    dtype: int64

  • name: hadm_id

    dtype: int64

  • name: ecg_time

    dtype: timestamp[us, tz=UTC]

  • name: ecg_acuity

    dtype: int64

  splits:

  • name: ecg_features

    num_bytes: 5892276

    num_examples: 146848

  download_size: 3971052

  dataset_size: 5892276

• config_name: ed_meds_timestep

  features:

  • name: ed_stay_id

    dtype: float64

  • name: subject_id

    dtype: int64

  • name: charttime

    dtype: timestamp[ns]

  • name: minutes_into_stay

    dtype: float64

  • name: drug_class

    dtype: string

  • name: action_idx

    dtype: int64

  • name: source

    dtype: string

  • name: index_level_0

    dtype: int64

  splits:

  • name: ed_meds_timestep

    num_bytes: 27076675

    num_examples: 323190

  download_size: 7130448

  dataset_size: 27076675

• config_name: empty_patient_state

  features:

  • name: subject_id

    dtype: int64

  • name: ed_stay_id

    dtype: int64

  • name: hadm_id

    dtype: float64

  • name: time_steps

    dtype: int64

  splits:

  • name: empty_patient_state

    num_bytes: 4324549602

    num_examples: 134616330

  download_size: 11217914

  dataset_size: 4324549602

• config_name: full_patient_state

  features:

  • name: ed_stay_id

    dtype: int64

  • name: subject_id

    dtype: int64

  • name: hadm_id

    dtype: float64

  • name: time

    dtype: timestamp[us]

  • name: Chemistry-Blood

    dtype: float64

  • name: Hematology-Blood

    dtype: float64

  • name: Hematology-Urine

    dtype: float64

  • name: Blood Gas-Blood

    dtype: float64

  • name: Chemistry-Urine

    dtype: float64

  • name: Chemistry-Other Body Fluid

    dtype: float64

  • name: Hematology-Cerebrospinal Fluid

    dtype: float64

  • name: Chemistry-Cerebrospinal Fluid

    dtype: float64

  • name: Hematology-Ascites

    dtype: float64

  • name: Chemistry-Ascites

    dtype: float64

  • name: Chemistry-Pleural

    dtype: float64

  • name: Hematology-Pleural

    dtype: float64

  • name: Hematology-Joint Fluid

    dtype: float64

  • name: Chemistry-Stool

    dtype: float64

  • name: Hematology-Other Body Fluid

    dtype: float64

  • name: Blood Gas-Other Body Fluid

    dtype: float64

  • name: Hematology-Bone Marrow

    dtype: float64

  • name: Hematology-Stool

    dtype: float64

  • name: Chemistry-Joint Fluid

    dtype: float64

  • name: URINE

    dtype: float64

  • name: Rapid Respiratory Viral Screen & Culture

    dtype: float64

  • name: PERITONEAL FLUID

    dtype: float64

  • name: BLOOD CULTURE

    dtype: float64

  • name: STOOL

    dtype: float64

  • name: OTHER

    dtype: float64

  • name: PLEURAL FLUID

    dtype: float64

  • name: TISSUE

    dtype: float64

  • name: SEROLOGY/BLOOD

    dtype: float64

  • name: IMMUNOLOGY

    dtype: float64

  • name: MRSA SCREEN

    dtype: float64

  • name: SWAB

    dtype: float64

  • name: JOINT FLUID

    dtype: float64

  • name: SPUTUM

    dtype: float64

  • name: FLUID,OTHER

    dtype: float64

  • name: ABSCESS

    dtype: float64

  • name: BRONCHOALVEOLAR LAVAGE

    dtype: float64

  • name: CSF;SPINAL FLUID

    dtype: float64

  • name: Blood (CMV AB)

    dtype: float64

  • name: Blood (EBV)

    dtype: float64

  • name: BILE

    dtype: float64

  • name: vitals_checked

    dtype: float64

  • name: current_temperature

    dtype: float64

  • name: current_heartrate

    dtype: float64

  • name: current_resprate

    dtype: float64

  • name: current_o2sat

    dtype: float64

  • name: current_sbp

    dtype: float64

  • name: current_dbp

    dtype: float64

  • name: current_pain

    dtype: large_string

  • name: current_pulse_pressure

    dtype: float64

  • name: current_map

    dtype: float64

  • name: time_since_last_hrs

    dtype: float64

  • name: temperature_rolling2h

    dtype: float64

  • name: temperature_delta

    dtype: float64

  • name: temperature_rate

    dtype: float64

  • name: heartrate_rolling2h

    dtype: float64

  • name: heartrate_delta

    dtype: float64

  • name: heartrate_rate

    dtype: float64

  • name: resprate_rolling2h

    dtype: float64

  • name: resprate_delta

    dtype: float64

  • name: resprate_rate

    dtype: float64

  • name: o2sat_rolling2h

    dtype: float64

  • name: o2sat_delta

    dtype: float64

  • name: o2sat_rate

    dtype: float64

  • name: sbp_rolling2h

    dtype: float64

  • name: sbp_delta

    dtype: float64

  • name: sbp_rate

    dtype: float64

  • name: dbp_rolling2h

    dtype: float64

  • name: dbp_delta

    dtype: float64

  • name: dbp_rate

    dtype: float64

  • name: Other

    dtype: float64

  • name: Analgesic - Opioid/NSAID

    dtype: float64

  • name: Antiemetic

    dtype: float64

  • name: Analgesic - Acetaminophen

    dtype: float64

  • name: Antibiotic

    dtype: float64

  • name: Benzodiazepine - Sedative/Anxiolytic

    dtype: float64

  • name: Analgesic - NSAID

    dtype: float64

  • name: Bronchodilator

    dtype: float64

  • name: Antiplatelet

    dtype: float64

  • name: GI - Acid Suppression

    dtype: float64

  • name: Corticosteroid

    dtype: float64

  • name: Beta Blocker

    dtype: float64

  • name: Diuretic

    dtype: float64

  • name: Antipsychotic

    dtype: float64

  • name: Anticoagulant

    dtype: float64

  • name: Insulin/Glucose

    dtype: float64

  • name: Calcium Channel Blocker

    dtype: float64

  • name: Nitrate

    dtype: float64

  • name: ACE Inhibitor

    dtype: float64

  • name: IV Fluid

    dtype: float64

  • name: Anticonvulsant

    dtype: float64

  • name: Antiarrhythmic

    dtype: float64

  • name: weight

    dtype: float64

  • name: ecg_status

    dtype: float64

  • name: rad_status

    dtype: float64

  • name: in_ed

    dtype: int64

  • name: in_ward

    dtype: int64

  • name: ed_boarding

    dtype: int64

  • name: terminal_event

    dtype: large_string

  • name: gender

    dtype: large_string

  • name: anchor_age

    dtype: int64

  • name: arrival_transport

    dtype: large_string

  • name: acuity

    dtype: decimal128(10, 9)

  • name: height

    dtype: float64

  • name: recon_ace_inhibitor

    dtype: uint8

  • name: recon_analgesic___nsaid

    dtype: uint8

  • name: recon_analgesic___opioid_nsaid

    dtype: uint8

  • name: recon_antiarrhythmic

    dtype: uint8

  • name: recon_antibiotic

    dtype: uint8

  • name: recon_anticoagulant

    dtype: uint8

  • name: recon_anticonvulsant

    dtype: uint8

  • name: recon_antiemetic

    dtype: uint8

  • name: recon_antiplatelet

    dtype: uint8

  • name: recon_antipsychotic

    dtype: uint8

  • name: recon_benzodiazepine___sedative_anxiolytic

    dtype: uint8

  • name: recon_beta_blocker

    dtype: uint8

  • name: recon_bronchodilator

    dtype: uint8

  • name: recon_calcium_channel_blocker

    dtype: uint8

  • name: recon_corticosteroid

    dtype: uint8

  • name: recon_diuretic

    dtype: uint8

  • name: recon_gi___acid_suppression

    dtype: uint8

  • name: recon_insulin_glucose

    dtype: uint8

  • name: recon_nitrate

    dtype: uint8

  • name: admission_type

    dtype: large_string

  splits:

  • name: full_patient_state

    num_bytes: 6238206732

    num_examples: 6551723

  download_size: 143234905

  dataset_size: 6238206732

• config_name: meds_admitted_er_actions

  features:

  • name: subject_id

    dtype: int64

  • name: stay_id

    dtype: int64

  • name: hadm_id

    dtype: float64

  • name: pyxis_charttime

    dtype: timestamp[ns, tz=UTC]

  • name: pyxis_med_rn

    dtype: float64

  • name: pyxis_medication

    dtype: string

  • name: emar_id

    dtype: 'null'

  • name: emar_charttime

    dtype: timestamp[ns, tz=UTC]

  • name: emar_medication

    dtype: 'null'

  • name: event_txt

    dtype: 'null'

  • name: scheduletime

    dtype: 'null'

  • name: in_er

    dtype: bool

  • name: medication

    dtype: string

  • name: drug_class

    dtype: string

  • name: action_time

    dtype: timestamp[ns]

  • name: ed_stay_id

    dtype: float64

  • name: ed_intime

    dtype: timestamp[ns]

  • name: minutes_into_stay

    dtype: float64

  • name: action_idx

    dtype: int64

  • name: index_level_0

    dtype: int64

  splits:

  • name: adm_er_clean

    num_bytes: 148079076

    num_examples: 912419

  download_size: 29150917

  dataset_size: 148079076

• config_name: radiology_features

  features:

  • name: subject_id

    dtype: int64

  • name: stay_id

    dtype: int64

  • name: hadm_id

    dtype: int64

  • name: exam_time

    dtype: timestamp[us, tz=UTC]

  • name: rad_acuity_level

    dtype: int64

  splits:

  • name: radiology_features

    num_bytes: 10408947

    num_examples: 259413

  download_size: 6821015

  dataset_size: 10408947

• config_name: vitals

  features:

  • name: ed_stay_id

    dtype: int64

  • name: subject_id

    dtype: int64

  • name: charttime

    dtype: timestamp[us]

  • name: temperature

    dtype: float64

  • name: heartrate

    dtype: float64

  • name: resprate

    dtype: float64

  • name: o2sat

    dtype: float64

  • name: sbp

    dtype: float64

  • name: dbp

    dtype: float64

  • name: pain

    dtype: large_string

  • name: time_since_last_hrs

    dtype: float64

  • name: temperature_rolling2h

    dtype: float64

  • name: temperature_delta

    dtype: float64

  • name: temperature_rate

    dtype: float64

  • name: heartrate_rolling2h

    dtype: float64

  • name: heartrate_delta

    dtype: float64

  • name: heartrate_rate

    dtype: float64

  • name: resprate_rolling2h

    dtype: float64

  • name: resprate_delta

    dtype: float64

  • name: resprate_rate

    dtype: float64

  • name: o2sat_rolling2h

    dtype: float64

  • name: o2sat_delta

    dtype: float64

  • name: o2sat_rate

    dtype: float64

  • name: sbp_rolling2h

    dtype: float64

  • name: sbp_delta

    dtype: float64

  • name: sbp_rate

    dtype: float64

  • name: dbp_rolling2h

    dtype: float64

  • name: dbp_delta

    dtype: float64

  • name: dbp_rate

    dtype: float64

  • name: pulse_pressure

    dtype: float64

  • name: map

    dtype: float64

  splits:

  • name: vitals

    num_bytes: 479613448

    num_examples: 1893713

  download_size: 84005509

  dataset_size: 479613448

configs:

• config_name: all_med_actions

  data_files:

  • split: all_med_actions

    path: modeling_data/all_med_actions-*

• config_name: combined_medication_features

  data_files:

  • split: med_only

    path: modeling_data/med_only-*

• config_name: ecg_features

  data_files:

  • split: ecg_features

    path: modeling_data/ecg_features-*

• config_name: ed_meds_timestep

  data_files:

  • split: ed_meds_timestep

    path: modeling_data/ed_meds_timestep-*

• config_name: empty_patient_state

  data_files:

  • split: empty_patient_state

    path: patient_state/*.parquet

• config_name: meds_admitted_er_actions

  data_files:

  • split: adm_er_clean

    path: modeling_data/adm_er_clean-*

• config_name: radiology_features

  data_files:

  • split: radiology_features

    path: modeling_data/radiology_features-*

• config_name: vitals

  data_files:

  • split: vitals

    path: vitals_triage/*.parquet

• config_name: full_patient_state

  data_files:

  • split: full_patient_state

    path: full_patient_state/*.parquet

---

empty_patient_state

Empty data frame for patient state. Subject_id, ed_stay_id and hadm_id are retained for identification. A row is added for each 10 min time block for that unique stay

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

vitals

Vitals cleaned using the same pattern and methods as the triage stats. Rhythm was dropped. Triage stats were added for each ed stay Missing values were handled using forward fill then backward fill for the groups that did not have an initial value. Remaining NaN values were filled using Iterative Imputer from sci-kit learn. Features were engineered for each vital to get a rolling 2h average, the delta from the most recent value, a time delta, rate of change, a general time delta, pulse pressure, and average perfusion pressure

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

full_patient_state

Full patient state dataframe for offline RL training, derived from MIMIC-IV ED data. One row per 30-minute time block per ED stay (ed_stay_id). Built by combining: lab order/result statuses (0=not ordered, 1=pending, 2=normal, 3=abnormal), microbiology order/result statuses (0-4), vital signs with fill-forward and delta/rate features, medication administration flags (binary, first admin onward), ECG status, weight, and transfer location flags (in_ed, in_ward, ed_boarding). Static features merged per stay: gender, age, arrival transport, triage acuity, admission type (ward rows only), height, and pre-ED medication reconciliation (recon_* drug classes). terminal_event column encodes the outcome: ICU transfer, ward discharge, or ED discharge. Intended as the primary input for offline RL state construction.

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

part1_train

Preprocessed classification and time-series escalation datasets derived from full_patient_state. Part 1: one row per ED stay, triage-time features only (no leakage), 4-class cohort label target plus 2-class rare holdout. Part 2: hourly timestep rows (resampled from 30-min), binary escalation_needed label, filtered to stays <= 48 hours, cumulative vital aggregation features.

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

part1_test

Preprocessed classification and time-series escalation datasets derived from full_patient_state. Part 1: one row per ED stay, triage-time features only (no leakage), 4-class cohort label target plus 2-class rare holdout. Part 2: hourly timestep rows (resampled from 30-min), binary escalation_needed label, filtered to stays <= 48 hours, cumulative vital aggregation features.

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

part1_holdout

Preprocessed classification and time-series escalation datasets derived from full_patient_state. Part 1: one row per ED stay, triage-time features only (no leakage), 4-class cohort label target plus 2-class rare holdout. Part 2: hourly timestep rows (resampled from 30-min), binary escalation_needed label, filtered to stays <= 48 hours, cumulative vital aggregation features.

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

part2_train

Preprocessed classification and time-series escalation datasets derived from full_patient_state. Part 1: one row per ED stay, triage-time features only (no leakage), 4-class cohort label target plus 2-class rare holdout. Part 2: hourly timestep rows (resampled from 30-min), binary escalation_needed label, filtered to stays <= 48 hours, cumulative vital aggregation features.

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/

part2_test

Preprocessed classification and time-series escalation datasets derived from full_patient_state. Part 1: one row per ED stay, triage-time features only (no leakage), 4-class cohort label target plus 2-class rare holdout. Part 2: hourly timestep rows (resampled from 30-min), binary escalation_needed label, filtered to stays <= 48 hours, cumulative vital aggregation features.

License: PhysioNet Credentialed Health Data License 1.5.0 — https://physionet.org/content/mimiciv/view-license/3.1/