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

The document discussed the issue of collider bias in observational studies of COVID-19, particularly those using non-random samples such as hospital admissions or voluntary testing. It highlighted how such bias could distort understanding of COVID-19 risk factors and disease progression, using data from the UK Biobank as an example. The authors emphasized the importance of recognizing and addressing collider bias in existing studies and suggested that the best way to mitigate this problem was through appropriate sampling strategies during the study design phase. They also provided tools and strategies to help reduce the impact of collider bias and developed a web app for sensitivity analyses.