What is confounding by indication, and is it a concern in observational cohort studies?

• A. It refers to confounding by baseline demographics
• B. It occurs when treatment assignment depends on disease severity; yes, can bias exposure-outcome associations; addressed by adjustment, restriction, or new-user designs
• C. It occurs only in randomized trials
• D. It is not a concern in cohort studies

Answer: (B)

Confounding by indication happens because the reason a patient is given a particular treatment is tied to how sick they are or what their prognosis looks like. In observational cohort studies, treatment isn’t assigned randomly; clinicians base decisions on disease severity and other risk factors. That means the group receiving the treatment may differ from the untreated group in ways that also affect the outcome, so the observed exposure-outcome relationship may reflect these underlying differences rather than a true treatment effect.

This is a real concern in cohort studies because it can bias the estimated effect of the exposure. For example, sicker patients might be more likely to receive a drug and also have worse outcomes regardless of the drug, making the treatment look worse than it is. Conversely, healthier patients might receive a treatment they’re more likely to tolerate, making the treatment look better.

To address it, researchers adjust for measured confounders in analyses, restrict the study to a more uniform subset of patients (same disease severity, for instance), or use a new-user design that includes only patients who are starting the treatment during the study period. These approaches help align groups so differences in outcomes are more likely due to the exposure itself rather than underlying indications.