Collider bias undermines our understanding of COVID-19 disease risk and severity

Abstract

Numerous observational studies have attempted to identify risk factors for infection with SARS-CoV-2 and COVID-19 disease outcomes. Studies have used datasets sampled from patients admitted to hospital, people tested for active infection, or people who volunteered to participate. Here, we highlight the challenge of interpreting observational evidence from such non-representative samples. Collider bias can induce associations between two or more variables which affect the likelihood of an individual being sampled, distorting associations between these variables in the sample. Analysing UK Biobank data, compared to the wider cohort the participants tested for COVID-19 were highly selected for a range of genetic, behavioural, cardiovascular, demographic, and anthropometric traits. We discuss the mechanisms inducing these problems, and approaches that could help mitigate them. While collider bias should be explored in existing studies, the optimal way to mitigate the problem is to use appropriate sampling strategies at the study design stage.

HIV infection and COVID-19 death: a population-based cohort analysis of UK primary care data and linked national death registrations within the OpenSAFELY platform

“… 11 However, any role that HIV might have in increasing the risk of infection or development of severe disease is effectively conditioned out by restriction to hospitalised individuals who are already infected with SARS-COV-2 and who probably have severe disease at the point of inclusion. 25 …”

COVID-19 related health inequality exists even in a city where disease incidence is relatively low: a telephone survey in Hong Kong

“…Second, there might be selection bias because our sample tended to be those who agreed to be followed up from earlier data collection time point and were more educated when compared with our previous sample. 38 Nevertheless, results of our analyses were based on age-weighted and sex-weighted sample to better represent the Hong Kong general adult population, and we were especially interested in examining the difference between the deprived and the non-deprived, as well as the associations of different factors with deprivation. Third, the nature of the analyses was cross-sectional; hence, direct temporality was not established.…”

Population-scale Longitudinal Mapping of COVID-19 Symptoms, Behavior, and Testing Identifies Contributors to Continued Disease Spread in the United States

“…These users are a subset of all the users who reported taking a test in the V3 survey, as some reported test results were outside this time window. To correct for selection bias of receiving a PCR test when studying the risk factors of a positive test result, we incorporated probability of receiving PCR tests as inverse probability weights (IPW) into our logistic model of PCR test result status (+/-) (Methods) 27 . As with the analysis of who received a test, the reported symptoms, loss of taste and/or smell was most strongly associated with a positive test result (OR: 33 .…”