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

By G. Griffith, T. Morris, Matthew J. Tudball, Annie Herbert, Giulia Mancano, Lindsey Pike, G. Sharp, J. Sterne, T. Palmer, George Davey Smith and 4 more. 2020

742 citations recorded by the source corpus.

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. Many published studies of the current SARS-CoV-2 pandemic have analysed data from non-representative samples from populations. Here, using UK BioBank samples, Gibran Hemani and colleagues discuss the potential for such studies to suffer from collider bias, and provide suggestions for optimising study design to account for this.

Subfields: Infectious Diseases, Modeling and Simulation

Citations in the corpus

• Simpson's Paradox in COVID-19 Case Fatality Rates: A Mediation Analysis of Age-Related Causal Effects (2020, neutral): “differences), as done in some studies [58]–[60].”
• Vaccine side-effects and SARS-CoV-2 infection after vaccination in users of the COVID Symptom Study app in the UK: a prospective observational study (2021, neutral): “2 Moreover, our results might have been affected by collider bias (ie, when a risk factor and an outcome both affect the likelihood of being sampled) 29 if both vaccination status and COVID positivity influenced the probability of participation in the app.”
• Mapping the human genetic architecture of COVID-19 (2021, neutral): “This can result in collider bias which distorts the relationship between genetic variation and the phenotypes being examined 52 .”
• Multimodal biomedical AI (2022, neutral): “159 ).”