> ML_DATASET // MIMIC-IV-CLINICAL-DATABASE_v1.0
MIMIC-IV (Medical Information Mart for Intensive Care IV)
MIT PhysioNet / Beth Israel Deaconess Medical Center · Healthcare & Clinical Informatics · Hospital stays for >300,000 patients admitted between 2008-2019
Healthcare & Clinical InformaticsPhysioNet Credentialed Health Data License (v1.5.0)Hospital stays for >300,000 patients admitted between 2008-2019credentialed-research-only
Dataset Profile & Characteristics
Label Type:Clinical vitals, ICD-9/10 codes, lab tests, medications, mortality
Languages:en
License Tier:research-only
Modalities:tabular, text, time-series
Intended Use
- In-hospital mortality prediction
- ICU readmission, sepsis onset, length-of-stay modeling
Prohibited / Discouraged Use
- Commercial re-identification or individual patient profiling
- Direct clinical decision automation without trial
Bias, Leakage & Privacy Risk Analysis
Privacy / Sensitive Data Risks:
De-identified electronic health records (EHR) of real hospital ICU patients. Re-identification strictly prohibited.
Known Bias:
Single tertiary academic medical center in Boston, MA; US healthcare billing practices.
Known Benchmark Leakage:
Careful timestamp alignment needed to avoid future-information leakage during admission predictions.
