Data-driven Adaptive Robust Optimization for Resource Sharing During a Pandemic

Outbreaks of COVID-19 in local communities can lead to a sharp increase in demand for limited resources such as mechanical ventilators. To cope with such demand surges, many hospitals in the US canceled or postponed elective procedures to preserve care capacity (including life-saving ventilators) for patients with COVID-19. These cancellations and delays resulted in a financial loss for the hospitals and poor outcomes for non-COVID-19 patients, a problem our methods can mitigate. Given that COVID-19 spreads at varying rates across different regions, there is an opportunity for sharing scarce portable resources to alleviate capacity shortfalls caused by local outbreaks.This paper develops a novel data-driven adaptive robust optimization methodology for allocating and relocating mechanical ventilators among different regions and states. Our method proposes an effective affine policy and considers a new type of dynamic uncertainty sets for ventilator demand by leveraging a powerful microsimulation model of COVID-19 spread and intervention. Our main methodological contributions lie in a new policy-guided model and an efficient algorithmic framework that mitigate critical limitations of current robust and stochastic models and make resource-sharing decisions implementable in real-time. Proof of concept is given for sharing ventilators among regions in Ohio and Michigan. Our numerical results suggest that the optimal policy derived from our data-driven adaptive robust model could satisfy ventilator demand during the first pandemic’s peak in Ohio and Michigan with 22% (limited sharing) to 65% (full sharing) fewer ventilators compared to a no sharing strategy (status quo), thereby allowing hospitals to preserve more elective procedures