Bayesian nowcasting with adjustment for delayed and incomplete reporting to estimate COVID-19 infections in the United States
AbstractReal-time estimates of the true size and trajectory of local COVID-19 epidemics are key metrics to guide policy responses. We developed a Bayesian nowcasting approach that explicitly accounts for reporting delays and secular changes in case ascertainment to generate real-time estimates of COVID-19 epidemiology on the basis of reported cases and deaths. Using this approach, we estimate time trends in infections, symptomatic cases, and deaths for all 50 US states and the District of Columbia from early-March through June 11, 2020. At the beginning of June, our best estimates of the effective reproduction number (Rt) are close to 1 in most states, indicating a stabilization of incidence, but there is considerable variability in the level of incidence and the estimated proportion of the population that has already been infected.One Sentence SummaryA new method to track epidemiologic measures of COVID-19, available in the covidestim package for R.