A Novel and Accurate Method for Estimating Deaths and Cases During Outbreaks of Infectious Diseases Including COVID-19.

Abstract

Epidemiological modelling of infectious diseases plays an important role in driving public health policy. Commonly used models are described, including those based on exponential growth (Laplace and related distributions); susceptible-infected-removed; the Gompertz distribution; and the skew-reflected-Gompertz distribution. These are all sensitive to the timing of peak infection. The development of a novel method for forecasting the number of deaths occurring during epidemics of infectious diseases is described. The mathematical development of the authors' novel asymmetric difference model is detailed in this paper. Its predictions for mortality rates associated with the COVID-19 pandemic for 14 countries were compared with the corresponding published mortality data. Forecasts by the asymmetric difference model of deaths from SARS-CoV-2 in different countries, actual recorded deaths to 30th June 2020, and corresponding errors included UK (42,700; 55,904; -24%); Poland (1490; 1444; +3%); Denmark (580; 605; -4%); Netherlands (6510; 6189; +5%); France (34,280; 29,836; +15%); Canada (1500; 8591; -78%); USA (44,540; 124,734; -64%); and Italy (22,020; 34,980; -37%). The model output was dependent upon forecast date accuracy for the peak of the disease outbreak. For Spain, the forecast date was one day early and for 10 (71%) countries the forecast peak occurred within seven days (inclusive) of the actual date. Mortality prediction by the asymmetric difference model is relatively accurate. Furthermore, this new model does not appear to be as unduly sensitive to the timing of peak infection as other models. Indeed, its prediction of peak infection also appears to be relatively accurate.