The Odds Ratio

Concept and Formula

The odds of an event is the ratio of the probability that it occurs to the probability that it does not: odds = p / (1 - p). The odds ratio is the division of this value between two groups: OR = odds_exposed / odds_unexposed. From a 2x2 table it is calculated as OR = (a x d) / (b x c), where a and d are the concordant cells and b and c are the cross cells. OR = 1 means no association, OR > 1 means higher odds with exposure, and OR < 1 indicates a protective association.

How to Read and Report It

OR should always be reported with a confidence interval and p-value, for example OR = 2.4, 95 CI [1.1, 5.2]. If the confidence interval includes 1, the result is not statistically significant. When interpreting OR, discuss direction and magnitude rather than fixating on the exact number. In logistic regression output, coefficients are in log-odds; apply exp to convert them to OR. Always state the reference category clearly, since the OR is defined relative to that group.

Common Misconceptions

The most common error is treating OR as equivalent to relative risk (RR). OR only approximates RR when the outcome is rare, roughly 10 percent or less. When the outcome is common, OR is always more extreme than RR, exaggerating the apparent effect size. In case-control studies RR cannot be calculated directly, but OR can. Another misconception is reading OR = 2.0 as meaning twice as many people develop the outcome; that statement describes odds, not probability.

When to Use It and Why It Matters

OR emerges naturally from the mathematics of logistic regression and is the standard measure for binary outcomes. In case-control designs, because cases and controls are selected by the researcher, only OR can be meaningfully interpreted. It is also the primary input for meta-analyses in clinical and epidemiological research. When the outcome is common in cohort or experimental studies, RR or risk difference is preferred; however, when the model is logistic regression, correctly interpreting OR is unavoidable.

Sources