Timing It Right: Balancing Inpatient Congestion vs. Readmission Risk at Discharge
Pengyi Shi, Jonathan E. Helm, Jivan Deglise‐Hawkinson, Julian Pan
Abstract
One of the most important decisions a hospitalist makes at the intersection of cost and quality of care is when to discharge a patient from the hospital. Keeping patients longer (shorter) increases (decreases) overcrowding and hospital costs but also decreases (increases) readmission risk. Here a long-run average cost optimization problem for determining on each day who and how many patients to discharge is developed. The authors combined structural properties of the model with an analytical solution for a special cost structure to approximately solve the high-dimensional Markov decision process. This transformed the originally intractable problem into a simple univariate optimization problem that can be solved efficiently yet allowed capture of time nonstationarity and fully heterogeneous inpatient populations, where each patient has a personalized risk trajectory. Moreover, the authors took one step beyond theory and implemented their discharge decision support tool in a partner hospital. For the tool to be properly parametrized and implementable, the authors developed a model to predict readmission risk as a function of length of stay that integrated several statistical methods in a novel manner. The resulting implementation was described as a showcase, demonstrating the tool’s applicability for integration with general hospital data systems and workflows.