S2) yield low discrepancy from observations (table S1). Model estimates for counties with large number of cases and deaths (fig. Further, as many states reopened portions of their economies in early May, the study period was limited to 21 February 2020 to when active control efforts were in place.ĭaily cases and deaths in the United States and the New York metropolitan area are well fit by the transmission model ( Fig. We focus on the national fitting, as well as major metropolitan areas with large populations and abundant data, for which parameter estimates are well informed. This parameter estimation was performed using the ensemble adjustment Kalman filter (EAKF) ( 18) in conjunction with county-level observations of both daily reported cases and deaths in the United States (see Materials and Methods and Supplementary Materials) ( 19). Instead, we inferred changing transmission rates within counties using time series records of COVID-19 activity. Intracounty mobility is not represented as the relationship between mobility and disease transmission is unknown. The metapopulation model explicitly simulates intercounty mobility using observed rates of intercounty visits to points of interest (POI) (e.g., restaurants, stores, etc.) on a county-by-county basis. Other parameters in the model include the ascertainment rate, α, which represents the fraction of infections documented as confirmed cases the average latency period, Z the average duration of infectiousness, D and a travel multiplicative factor, θ. The remaining 2831 counties were apportioned among 16 additional transmission rate parameters depending on cumulative case levels and population density (see Materials and Methods). To reflect heterogeneity in transmission rates across the United States while avoiding a large number of model parameters, we defined a separate β i for counties with greater than 400 cumulative confirmed cases as of ( n = 311) (fig. Here, μ is the relative transmissibility of undocumented infections. We explicitly simulated documented and undocumented infections ( 17), for which separate transmission rates, β and μβ (μ < 1), respectively, are defined. During the study period, 21 February 2020 to, measures to control the virus were actively, but heterogeneously, implemented throughout the United States. S1) and representing SARS-CoV-2 transmission in 3142 U.S.

In this study, we adapted and applied a dynamic metapopulation model ( 13, 14) informed by human mobility data ( 15, 16) (fig.

Post intervention deaths