Stayin' Alive, Stayin' Alive: An Optimization Analysis on Reducing Hospital Readmissions for Heart Failure

Abstract

Reducing hospital readmissions, particularly for patients with heart failure, has been a long-standing goal and a crucial challenge in the United States healthcare sector. High readmission rates can indicate a low quality of care for patients and have a negative financial impact on hospitals and providers. This issue has not yet been fully addressed using a data-driven and machine learning approach. In this work, we use real patient data from the Nationwide Readmissions Database and machine learning models to identify factors contributing to readmission. We then propose strategies for reducing these readmission rates. First, we develop a support vector machine model to find key predictors of readmission in heart failure patients. We find significant patient features, such as chronic kidney disease and chronic obstructive pulmonary disease, that increase the likelihood of a patient’s readmission and provide a detailed analysis of these elements. Second, we formulate and evaluate two cost-effective interventions - transferring patients to cardiac rehabilitation centers or extending their hospital stay - for reducing readmissions. We carefully analyze the cost implications of these interventions and propose an optimal solution that balances effectiveness and feasibility. Our findings suggest a push for transferring patients to cardiac rehabilitation facilities, and reveal specific interventions, like increasing a patient’s length of stay, could significantly reduce readmission rates and save a substantial amount healthcare spending. Our findings contribute to the wider research of healthcare management and decision making using machine learning. This work positions the mathematical and computational foundations to assist healthcare professionals, policymakers, and administrative bodies in developing targeted strategies to reduce heart failure readmissions, and improve the quality of care for all patients.