Which statement best describes a confounding variable in evaluation?

• A. A third variable related to both exposure and outcome that distorts the observed effect.
• B. A variable that only affects outcome but not exposure.
• C. A random measurement error.
• D. An interaction term in a model.

Answer: (B)

A confounding variable is a third factor that is related to both the exposure and the outcome and, because of that relationship, distorts the observed association between exposure and outcome. This is why it can make an effect look larger or smaller than it truly is if you don’t account for it. For example, suppose you’re evaluating whether a program increases physical activity. Age could be a confounder if younger people are more likely to join the program and also more likely to be active regardless of the program; without adjusting for age, you might misattribute higher activity to the program itself.

The other descriptions don’t capture confounding. A variable that affects the outcome but not the exposure isn’t confounding—it's an independent risk factor. Random measurement error isn’t a confounder either; it introduces bias through misclassification or imprecision. An interaction term describes effect modification, where the exposure effect differs across levels of another variable, not a confounding relationship.