A model to forecast regional demand for COVID-19 related hospital beds

Johannes Opsahl Ferstad; Angela Jessica Gu; Raymond Ye Lee; Isha Thapa; Andrew Y Shin; Joshua A Salomon; Peter Glynn; Nigam H Shah; Arnold Milstein; Kevin Schulman; David Scheinker

The model accepts as an input the number of COVID-19 hospitalizations and the associated doubling time, if these are available. If these are not available, for each US county, the model imports the latest total number of confirmed cases from the New York Times online repository and accepts user-entered parameters of the ratio of total cases to confirmed cases (e.g., 10:1) [1, 9] and the COVID-19 populationlevel doubling time (e.g., 7 days) [2] . Following a simple exponential model, the total number of COVID-19+ people on each subsequent day N is the product of the initial number of total cases and 2 to the power given by N divided by the doubling time input parameter. Users can simulate the effects of interventions that mitigate the spread of infection (such as social distancing) by entering new doubling times and dates for these interventions to take effect. If these are input, the number of COVID-19+ people on subsequent days is calculated using the new doubling time with the previous formula. Users are referred to online tools to estimate changes in doubling time associated with the impact of socialdistancing interventions. [3] Calculation of Hospitalization Rates County-specific age distributions are derived from the US census [4] and age-group specific estimates of the case rates of severe symptoms, critical symptoms, and mortality derived from Imperial College COVID-19 Response Team [5] . For each county, the proportion of each age group relative to the total population is calculated. Severe and critical symptom case-rates for each age group are weighted by these proportions, and the hospitalization rate is calculated as the sum of these weighted severe and critical symptom case-rates.