Selection of patients based on the presence of specific pre‐set symptoms, such as fever and cough, lead to biased associations between these symptoms and disease, and sensitivity and specificity estimates that differ from their true values.