What is the most widely accepted conceptual basis for causation in epidemiology?

• A. Bradford Hill criteria
• B. P-value threshold
• C. Counterfactual or potential outcomes model
• D. Post hoc ergo propter hoc

Answer: (C)

Causation in epidemiology is best captured by the counterfactual or potential outcomes framework. The idea is to define the causal effect as the difference between what would happen to the same unit if exposed versus if not exposed. Since we can’t observe both states for the same person, the causal effect is described in terms a population can estimate: the average difference in outcomes if everyone were exposed compared with if everyone were unexposed.

This framework is powerful because it makes the assumptions needed to identify and estimate causal effects explicit. When randomization is present, exposure groups are comparable, and we can interpret differences in outcomes as causal. In observational data, causal inference relies on methods that aim to emulate randomization by adjusting for confounding and carefully handling time and settings—through approaches like regression adjustment, propensity scores, inverse probability weighting, or more advanced g-methods—so that the exposed and unexposed groups resemble each other with respect to confounders.

Historically, the Bradford Hill criteria have served as guidelines for evaluating evidence of causation, but they are not a formal basis for causation themselves. P-value thresholds relate to statistical significance, not causation, and post hoc ergo propter hoc is a logical fallacy. The counterfactual framework provides a precise, testable way to think about causal effects and how to estimate them from data.