Dynamic Control of Service Systems with Returns: Application to Design of Postdischarge Hospital Readmission Prevention Programs.

This article focuses on the dynamic control of queueing systems with returns, specifically applied to optimizing postdischarge hospital readmission prevention programs. It models the trade-off between the costs of postservice interventions—which reduce the probability of patient readmission—and the benefits of decreased congestion and service costs in a multiserver queueing system known as the Erlang-R model. Using fluid approximations, the authors characterize optimal long-run average and transient control policies, showing that under piecewise-linear intervention costs, the state space partitions into regions prescribing different intervention intensities based on congestion levels. Simulation experiments demonstrate that dynamically adjusting intervention intensity according to congestion can yield significant cost savings—up to 25.4% in long-run average costs and 33.7% in finite-horizon costs—compared to fixed or simple aggressive policies, with robustness to time-varying arrivals and practical applicability illustrated through a hospital readmission case study.