Hospital readmissions within 30 days are costly and often preventable, making them a critical focus for healthcare systems. Predictive models can help flag high-risk patients for early intervention. However, if these models perform unevenly across demographic groups, they can unintentionally reinforce health disparities.

This project aims to build an interpretable, accurate, and fair model for predicting 30-day hospital readmissions using the UCI Diabetes dataset. We also tend to question if bias audits should be integrated into the modeling pipeline to ensure equitable performance across racial groups.
In addition to evaluating traditional performance metrics, we will assess fairness using disaggregated metrics, embedding equity checks throughout the development process. The project not only pursues algorithmic efficiency but also aligns with ethical and inclusive health outcomes. Ultimately, we hope to provide a framework responsible for predictive modeling in clinical settings.