Predictive Modeling for Patient Readmission Using Electronic Health Records (EHR)

Abstract

Abstract: Hospital readmissions are a significant concern for healthcare systems, resulting in increased costs and adverse patient outcomes. This study develops and evaluates a predictive model for patient readmission using Electronic Health Records (EHR) data. This study explores various machine learning techniques to predict 30-day hospital readmission rates, focusing on feature selection, model performance, and clinical interpretability. We employed machine learning algorithms, including logistic regression, decision trees, and random forests, to identify patients at high risk of readmission. Our model incorporates demographic, clinical, and healthcare utilization data from EHRs. Results show that our predictive model accurately identifies patients at high risk of readmission, with an area under the curve (AUC) of 0.85. The model also identifies key risk factors contributing to readmission, including prior hospitalizations, comorbidities, and medication adherence. Our findings suggest that predictive modelling using EHR data can inform clinical decision-making and reduce hospital readmissions. This study highlights the potential of leveraging EHR data and machine learning algorithms to improve patient outcomes and reduce healthcare costs.