> ML_DATASET // DIABETES-HOSPITAL-READMISSION_v1.0
Diabetes 130-US Hospitals Readmission (1999-2008)
UCI ML Repository / Strack et al. · Healthcare & Hospital Informatics · 101,766 hospital encounters, 47 features
Healthcare & Hospital InformaticsCC-BY-4.0101,766 hospital encounters, 47 featuresopen
Dataset Profile & Characteristics
Label Type:3-class readmission status (<30 days, >30 days, No readmission)
Languages:en
License Tier:permissive-open-source
Modalities:tabular
Intended Use
- Benchmarking healthcare readmission risk prediction and high-cardinality ICD-9 encoding
Prohibited / Discouraged Use
- Production triage or denying care coverage
Bias, Leakage & Privacy Risk Analysis
Privacy / Sensitive Data Risks:
De-identified clinical encounter records with ICD-9 diagnosis codes and medications.
Known Bias:
Sample from 130 US hospitals; clinical practices from 1999-2008.
Known Benchmark Leakage:
Number of prior inpatient visits correlates heavily with 30-day readmission risk.