What you saw is what you got? -- Correcting reported incidence data for testing intensity
During the COVID-19 pandemic, different types of non-pharmaceutical interventions played an important role in the efforts to control outbreaks and to limit the spread of the SARS-CoV-2 virus. In certain countries, large-scale voluntary testing of non-symptomatic individuals was done, with the aim of identifying asymptomatic and pre-symptomatic infections as well as gauging the prevalence in the general population. In this work, we present a mathematical model, used to investigate the dynamics of both observed and unobserved infections as a function of the rate of voluntary testing. The model indicate that increasing the rate of testing causes the observed prevalence to increase, despite a decrease in the true prevalence. For large testing rates, the observed prevalence also decrease. The non-monotonicity of observed prevalence explains some of the discrepancies seen when comparing uncorrected case-counts between countries. An example of such discrepancy is the COVID-19 epidemics observed in Denmark and Hungary during winter 2020/2021, for which the reported case-counts were comparable but the true prevalence were very different. The model provides a quantitative measure for the ascertainment rate between observed and true incidence, allowing for test-intensity correction of incidence data. By comparing the model to the country-wide epidemic of the Omicron variant (BA.1 and BA.2) in Denmark during the winter 2021/2022, we find a good agreement between the cumulative incidence as estimated by the model and as suggested by serology-studies. While the model does not capture the full complexity of epidemic outbreaks and the effect of different interventions, it provides a simple way to correct raw case-counts for differences in voluntary testing, making comparison across international borders and testing behaviour possible.