Measures of association

A measure of association compares the frequency of an outcome in exposed people with its frequency in unexposed people. The comparison can be a ratio, which describes the strength of the association, or a difference, which describes how much disease the exposure adds in absolute terms. Both are calculated from a 2×2 table with exposure on one axis and disease on the other.

Relative risk (RR) is the incidence in the exposed divided by the incidence in the unexposed, and it requires a design that follows people forward and measures incidence — a cohort study or a randomized trial. Odds ratio (OR) is the odds of exposure among cases divided by the odds among controls, computed as ad/bc from the 2×2 table; it is the measure used in case-control studies, where incidence cannot be calculated, and it approximates the relative risk when the disease is rare. Attributable risk is the difference in incidence between exposed and unexposed, answering how much of the disease burden in exposed people is due to the exposure. The number needed to treat is the reciprocal of the absolute risk reduction in a trial, and number needed to harm the reciprocal of the absolute risk increase.

Interpretation follows a simple rule: a ratio of 1 means no association, greater than 1 means the exposure is associated with increased risk, and less than 1 suggests a protective effect. A confidence interval that includes 1 indicates the result is not statistically significant. None of these measures establishes causation on its own — confounding, selection bias, and information bias can all generate an association where no causal link exists, which is why Bradford Hill criteria and study design quality matter.

USMLE Step 1 asks you to compute these values from a 2×2 table and, just as often, to recognize which measure is appropriate for a described study design. Pair this topic with study designs, bias, and confounding when you review biostatistics.