Flattening the curve before it flattens us: hospital critical care capacity limits and mortality from novel coronavirus (SARS-CoV2) cases in US counties

Charles C Branas; Andrew Rundle; Sen Pei; Wan Yang; Brendan G Carr; Sarah Sims; Alexis Zebrowski; Ronan Doorley; Neil Schluger; James W Quinn; Jeffrey Shaman

Transmission dynamics were simulated for all US study counties over the period from February 21, 2020 to March 24, 2020 using an iterated filter-ensemble adjustment Kalman filter framework. 18, 19, 20 This combined model-inference system estimated the trajectories of susceptible, exposed, documented infected, and undocumented infected populations in each county while simultaneously inferring model parameters for the average latent period, the average duration of infection, the transmission reduction factor for undocumented infections, the transmission rate for documented infections, the fraction of documented infections, and the previously mentioned travel multiplicative factor. To account for delays in infection confirmation, a time-to-event observation model using a Gamma distribution with a range of reporting delays and different maximum seeding was employed. Log-likelihood was used to identify the best fitting model-inference posterior. 8, 16 As in prior work 21 , the transmission of SARS-CoV2 under increasing reductions in population physical contact via control measures and behavior change was projected forward in time from March 24,2020 to April 24, 2020, using the optimized model parameter estimates. Control measures included travel restrictions between areas, self-quarantine and contact precautions that were publicly advocated or imposed, and greater availability of rapid testing for infection. Behavior changes in medical care-seeking due to increased awareness of COVID-19 and increased personal protective behavior (e.g., use of facemasks, social distancing, self-isolation when sick) were also considered. Three different contact reduction scenarios were projected, 0% (no contact reduction via controls and behavior change), 25% contact reduction, and 50% contact reduction.