A demographic scaling model for estimating the total number of COVID-19 infections

Bohk-Ewald, Christina; Dudel, Christian; Myrskyla, Mikko

The total number of COVID-19 infections is critical information for decision makers when assessing the progress of the pandemic, its implications, and policy options. Despite efforts to carefully monitor the COVID-19 pandemic, the reported number of confirmed cases is likely to underestimate the actual number of infections. We aim to estimate the total number of COVID-19 infections in a straightforward manner using a demographic scaling model. This model is broadly applicable as it is based on little input data: deaths attributable to COVID-19, COVID-19 infection fatality rates, and life tables. As many countries lack reliable estimates of age-specific infection fatality rates, we map them from a reference country onto countries of interest based on remaining life expectancy. This scaling accounts for cross-country differences in the age structure, the health status, and the health care system. We also introduce easy to apply formulas to quantify the bias that would be required in death counts and infection fatality rates in order to reproduce a certain estimate of infections. Across the 10 countries with most COVID-19 deaths as of April 17, 2020, our estimates suggest that the total number of infected is approximately 4 times the number of confirmed cases. The uncertainty, however, is high, as the lower bound of the 95% prediction interval suggests on average twice as many infections than confirmed cases, and the upper bound 10 times as many. Comparing our results with findings from local seroprevalence studies and applying our bias formulas shows that some of their infection estimates would only be possible if just a small fraction of COVID-19 related deaths were recorded, indicating that these seroprevalence estimates might not be representative for the total population.