Define a confounder.

Define a confounder.

• A. A variable related to both exposure and outcome that biases causal estimates.
• B. A variable only related to exposure.
• C. A variable only related to outcome.
• D. A random error term in a regression model.

Answer: (A)

A confounder is a variable that is related to both the exposure and the outcome, and if it’s not accounted for, it biases the estimated effect of the exposure on the outcome. Because it’s linked to both, the observed association can reflect the confounder’s influence rather than a true causal effect of the exposure. For example, if older age makes someone more likely to receive a treatment and also increases the risk of the outcome, age can confound the treatment–outcome relationship. Controlling for confounders—through randomization, stratification, or including the confounder as a covariate in analysis—helps isolate the true effect of the exposure. A variable related only to exposure won’t bias the exposure–outcome estimate, and a variable related only to the outcome won’t either. The random error term in a regression is just unexplained variation, not a confounding factor.

A confounder is a variable that is related to both the exposure and the outcome, and if it’s not accounted for, it biases the estimated effect of the exposure on the outcome. Because it’s linked to both, the observed association can reflect the confounder’s influence rather than a true causal effect of the exposure.

For example, if older age makes someone more likely to receive a treatment and also increases the risk of the outcome, age can confound the treatment–outcome relationship. Controlling for confounders—through randomization, stratification, or including the confounder as a covariate in analysis—helps isolate the true effect of the exposure.

A variable related only to exposure won’t bias the exposure–outcome estimate, and a variable related only to the outcome won’t either. The random error term in a regression is just unexplained variation, not a confounding factor.