Enhancing Hospital Readmission Rate Predictions Using Random Forest and Gradient Boosting Algorithms

Abstract

This research investigates the application of advanced machine learning techniques, specifically Random Forest and Gradient Boosting algorithms, to improve the accuracy of hospital readmission rate predictions. Hospital readmissions pose significant challenges to healthcare systems, both in terms of patient outcomes and financial costs. Traditional prediction models often rely on linear analytical methods that inadequately capture the nonlinear interactions inherent in healthcare data. This study utilizes a comprehensive dataset derived from electronic health records, comprising demographic, clinical, and social factors influencing readmissions. Random Forest and Gradient Boosting, known for their ability to manage high-dimensional data and complex interactions, are employed to develop predictive models. The effectiveness of these models is evaluated against a baseline logistic regression model through metrics such as area under the receiver operating characteristic curve (AUC-ROC), precision, recall, and F1 score. Results demonstrate that both Random Forest and Gradient Boosting significantly outperform the baseline model, with Gradient Boosting achieving the highest predictive accuracy. Additionally, feature importance analysis reveals insights into the determinants of readmissions, underscoring the role of chronic conditions and prior hospitalizations. This study concludes that integrating these machine learning algorithms into predictive modeling frameworks can enhance readmissions management and inform targeted intervention strategies, ultimately improving patient care and reducing healthcare costs.