A population-based case-control study of COVID-19: methodological considerations on the role of testing bias

Targeting people at risk of COVID-19 infection has been critical to containing the pandemic. Using only differences in cumulative incidence by sociodemographic groups can be misleading, as it reflects both factors related to infection risk and those related to testing for infection. The aim of this analysis was to disentangle the determinants of both mechanisms.

We compared the demographic, socioeconomic, and health characteristics of 813 PCR-confirmed COVID-19 cases over age 20 years with 1630 age- sex- and geography-matched population-based controls, both recruited in 2020-2021 in the RESPIRA study. We used antibody results and previous diagnosis to detect infections in population-based controls.

High socioeconomic status and being older than 60 years old were negatively associated with seropositivity. Obesity and number of people living in the household were positively associated with seropositivity. Among infected (seropositive) people, diagnosis by PCR was more frequent in employees, and in people with asthma or hypertension, and was negatively associated with the number of people living in the household.

Differences between PCR-confirmed cases and non-infected controls reflected differences both in risk of infection, and in PCR-testing in infected people. The possibility of PCR-testing bias in case-control studies of COVID should be considered in future research.