Forecasting hospital readmission rates in the U.S. using time-series analysis and machine learning

Hospital readmissions pose significant challenges to the healthcare system in the United States, impacting patient well-being and leading to financial penalties for hospitals. This research aims to develop a predictive framework for forecasting hospital readmission rates using a combination of time-series models and machine learning techniques. The study leverages datasets such as the CMS Hospital Readmission Reduction Program and HCUP, focusings on key variables including patient demographics, medical history, length of stay, and comorbidities. Models such as ARIMA, LSTM, and XGBoost are employed to capture temporal patterns and complex relationships within the data. Additionally, SHAP analysis is used to explain the key drivers influencing predictions. Simulations of various healthcare policies, such as follow-up care and reduced hospital stays, are conducted to assess their impact on readmission rates. The results aim to offer actionable insights for healthcare providers, enabling targeted interventions to reduce readmission rates and improve patient outcomes. This study emphasizes the use of data-driven methodologies for effective healthcare delivery, aiming to balance predictive accuracy and model interpretability through the integration of explainable AI techniques.