Retrospective Analyses of Interventions to Epidemics using a Continuously Updated Model, with Application to the COVID-19 Crisis in New York City

Retrospective analyses of interventions to epidemics, in which the effectiveness of strategies implemented are compared to hypothetical alternatives, are valuable for performing the cost-benefit calculations necessary to optimize infection countermeasures. SIR (susceptible-infected-removed) models are useful in this regard but are limited by the challenge of deciding how and when to update the numerous parameters as the epidemic progresses. We present a method that uses a dynamic spread function to systematically capture the continuous variation in the population behavior throughout an epidemic. There is no need to update parameters as the effects of interventions are gradually manifested in the infection dynamics. We use the tool to quantify the reduction in infection rate realizable from the population of New York City adopting different facemask strategies during COVID-19. Assuming a baseline facemask of 67% filtration efficiency, calculations show that increasing the efficiency to 75% could reduce the roughly 5000 new infections per day occurring at the peak of the epidemic by 40%. The turn-around time for the epidemic decreases from around 37 days to 31 days. Mitigation strategies that may not be varied as part of the retrospective analysis, such as social distancing, are automatically captured as part of the calibration of the dynamic spread function.