Diabetes130US: Hospital Readmission Records for Diabetes Patients

A 2014 clinical dataset curated for evaluating predictive models on independent and identically distributed tabular data. The data originates from a study analyzing the impact of HbA1c measurement on hospital readmission rates, involving 70,000 patient records. The TabArena team processed the data to avoid target leakage and reversed original ordinal encodings for three features.

Use Cases

• Benchmarking classification models for independent and identically distributed data based on the dataset's stated focus.
• Predicting hospital readmission likelihood for diabetes patients based on the 'readmitted' target column.
• Analyzing the impact of payer code on patient outcomes based on the description mentioning its role in a distribution shift.
• Studying data preprocessing effects on model performance, given the curation comments on handling duplicates and identifiers.

Strengths

• Based on a published study analyzing 70,000 clinical database patient records.
• Curated specifically for evaluating ML models on IID tabular data, with steps taken to avoid target leakage.
• Released under a permissive CC BY 4.0 license.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, which may limit suitability assessment.
• The description notes an original distribution shift based on encounter order, though it vanishes after shuffling.

Provenance

Source

Strack, Beata, et al. 'Impact of HbA1c measurement on hospital readmission rates: analysis of 70,000 clinical database patient records.' BioMed research international 2014.

Collection Method

Clinical database records, curated by the TabArena team for an ML study.

Time Range

2014

Freshness

Dataset year is 2014; last update date is unknown.

Geography

United States (inferred from dataset title)