Age-dependent mean hospitalization and mortality rates in the surrogate-world simulation.

The share f(a) of the population in each age group a is taken from U.S. Census data [69]. The age-dependent death rate in hospitals d′ is obtained from cumulative hospitalization and death rates in NYC by June 1, 2020 [36], under the assumption that 90% of deaths occurred in hospitals. Age-dependent hospitalization rates h(a) and mortality rates d(a) in the community (outside hospitals) are difficult to obtain directly from NYC data because of an age-dependent undercount of infections [37]. We choose hospitalization rates h(a) that approximate data from France [70], adjusting the rates so that the overall hospitalization rate is ∑a f(a)h(a) ≈ 3.1%, which is NYC’s overall hospitalization rate if one assumes a cumulative COVID-19 incidence rate of 23% [71], together with NYC’s actual hospitalization count (52,333 on June 1, 2020) and population (8.34 million) [36]. Similarly, the mortality rate in the community d(a) is chosen such that the overall infection fatality rate is ∑a f(a)[d(a) + h(a)d′(a)] ≈ 1.1%, which is NYC’s overall infection fatality rate if one considers the same cumulative incidence of 23% and the confirmed and probable cumulative death count from COVID-19 (21,607 by June 1, 2020).