COVID-19 Scratch Models To Support Local Decisions

Abstract

This article is based on modeling studies conducted in response to requests from Yale University, the Yale New Haven Hospital and the State of Connecticut during the early weeks of the SARS-CoV-2 outbreak. Much of this work relied on scratch modeling, that is, models created from scratch in real time. Applications included recommending event crowd-size restrictions, hospital surge planning, timing decisions (when to stop and possibly restart university activities), and scenario analyses to assess the impacts of alternative interventions, among other problems. This paper documents the problems faced, models developed, and advice offered during real-time response to the COVID-19 crisis at the local level. Results include a simple formula for the maximum size of an event that ensures no infected persons are present with 99% probability; the determination that existing ICU capacity was insufficient for COVID-19 arrivals which led to creating a large dedicated COVID-19 negative pressure ICU; and a new epidemic model that showed the infeasibility of the university hosting normal spring and summer events, that lockdown-like stay-at-home and social distancing restrictions without additional public health action would only delay transmission and enable a rebound after restrictions are lifted, and that aggressive community screening to rapidly detect and isolate infected persons could end the outbreak.