A prescriptive optimization approach to identification of minimal barriers for surgical patients

Ensuring timely patient discharges is crucial to managing a hospitals patient flow; however, discharges are dependent on the coordination of multiple care teams and thus are highly decentralized in nature. Central capacity teams often lack transparency into how to prioritize scarce resources for patients who need them to leave the hospital (e.g., imaging or post-acute facility placement). Our goal is to identify a small subset of so-called barriers to discharge for hospitalized surgical patients by balancing two aims: a high likelihood that the patient will be discharged from the hospital in the next 24 hours if these barriers are resolved; and a high likelihood that these barriers will indeed be resolved. To do so, we combine discharge and barrier-resolution predictive models with a mixed-integer prescriptive optimization model to identify each patients minimal barriers. We empirically demonstrate the efficacy of the proposed formulation and solution methodology using data from a large academic medical center. Furthermore, we study the influence of uncertainty in discharge prediction estimates on the prescribed barriers and introduce a robust optimization variant that is capable of overcoming the shortcomings of the nominal approach. Our approach can significantly enhance the ability of capacity management teams to identify which barriers are most important to address to expedite a patients discharge while taking into account their inherent heterogeneity.