Diabetes Patient Records from 130 Hospitals for Fairness Analysis

Diabetes-130-Hospitals_(Fairlearn) is a dataset from OpenML focused on hospital readmission for diabetes patients. It is associated with the Fairlearn toolkit, indicating its use for fairness and bias analysis in machine learning models. The dataset's specific size, creator, and update date are not provided in the input metadata.

Use Cases

• Predicting readmission_status using patient demographic and admission features to assess model fairness.
• Analyzing disparities in predicted outcomes across race or gender columns to detect algorithmic bias.
• Training fairness-aware classifiers to mitigate bias in predictions based on sensitive attribute columns.
• Benchmarking fairness mitigation techniques using clinical outcome labels like time_in_hospital or num_lab_procedures.

Strengths

• Dataset is integrated with the Fairlearn toolkit, providing a direct application for fairness research.
• Originates from 130 hospitals, suggesting a multi-institutional scope for the underlying data.

Limitations

• The exact number of rows, columns, and data completeness are unknown, limiting reproducibility assessments.
• Lack of documented license, author, and update information reduces transparency and trustworthiness.

Provenance

Source

OpenML

Collection Method

null

Time Range

null

Freshness

null

Geography

null