Explainable AI for Predicting Patient Readmission in Hospitals

Abstract

Hospital readmissions within 30 days of discharge remain a major challenge for healthcare systems, contributing to increased clinical workload and operational costs. Early identification of patients at high risk of readmission can support preventive interventions and improve patient management. This study proposes an explainable artificial intelligence (XAI) framework for predicting 30-day hospital readmissions using a synthetic healthcare dataset consisting of 5,000 patient records and eight clinical and demographic features. Multiple machine learning models, including Logistic Regression, Support Vector Classifier, Random Forest, Gradient Boosting, AdaBoost, and Extreme Gradient Boosting (XGBoost), were trained and evaluated under a unified preprocessing pipeline. Weighted F1-score was selected as the primary evaluation metric to address class imbalance, with XGBoost achieving the best performance (F1-score = 0.742, accuracy = 0.797). Model robustness was validated using 5-fold stratified cross-validation. To enhance interpretability, global explainability techniques such as SHAP and permutation importance were applied, while local interpretability was achieved using LIME to explain individual predictions. The proposed framework demonstrates that integrating predictive modeling with explainable AI improves transparency, reliability, and decision support capability in hospital readmission prediction systems.

Ähnliche Arbeiten