Confounding

Last revised by Joachim Feger on 6 Jan 2024

Confounding is a type of bias that may be seen in observational longitudinal investigations like cohort or case control studies. A confounder is a variable associated with the exposure of and the outcome of interest 1. The confounder distorts the true association between the exposure of interest and the outcome 2,3.

For example, investigators hypothesize that possession of matches is associated with lung cancer and carry out a case control study that shows that study subjects with lung cancer have higher odds of possessing matches than controls without lung cancer. In this case, the confounder is cigarette smoking: smoking is associated both with the possession of matches and with lung cancer. If the study data were “stratified” by smoking status, where the association of matches with lung cancer was considered separately for study subjects who did and did not smoke, the apparent association of matches with lung cancer would disappear.

Perhaps the easiest way of understanding confounding is when the wrong exposure (matches) is blamed for an outcome (lung cancer) by a potentially hidden variable (the confounder, smoking) that is associated with the putative exposure and the health outcome.

Updating… Please wait.

 Unable to process the form. Check for errors and try again.

 Thank you for updating your details.