Maintaining Healthcare Capacity in Rural America by Replenishing Personal Protective Equipment: The Case from West Virginia

In the face of a worldwide shortage of personal protective equipment (PPE) at the onset of the COVID-19 pandemic, we report on the challenge of supplying hard-to-forecast PPE demand in the state of West Virginia under challenging replenishment conditions. Part of the challenge included the lack of appropriate healthcare data in a largely rural state such as West Virginia. We describe the work of a joint interagency task force (JIATF) that assembled the data to implement a novel agent-based epidemiological model for forecasting PPE demand of patients with confirmed and suspected COVID-19, extending the capability of popular epidemiological models. On the supply side, we describe our extension of nonparametric approaches for setting inventory parameters necessitated by the combination of nonstandard lead-time distributions and autocorrelated demand. Our simulation studies, grounded in the practical need to replenish PPE stocks, demonstrate considerable improvements in both (a) forecast accuracy possible from our approach over extant epidemiological models, and (b) estimation improvement of our nonparametric inventory estimation approach over the extant approaches. We describe the detailed implementation of these methods under the aegis of the JIATF and discuss the relevance of our results beyond the COVID-19 pandemic to other public health emergencies and supply chain settings.