Communicate Absolute and Relative Risks Together

Begin with the event and time window

State what happened, to whom, and over what period before describing the size of a comparison. “Risk decreased” is incomplete without the event definition and time horizon. Make clear whether the numbers come from observed groups, a model, or an illustrative scenario.

Cochrane's effect-measure chapter distinguishes ratio measures from differences. Those measures can describe the same comparison while emphasizing different aspects. Reporting one as though it were the other changes the meaning, not just the style of the sentence.

Use common denominators

Translate probabilities into a shared denominator when that improves understanding. Two events per hundred and ten per thousand are easier to compare after expressing both per hundred or both per thousand. Retain the underlying observed sample sizes when presenting actual study results so the translation does not imply more observations than were collected.

Choose language that respects the outcome direction. A reduction in an undesirable event and a reduction in a desirable response have different practical meanings. Avoid switching between “success” and “failure” without showing how the counts change.

Compare two fictional settings

In fictional setting one, event risk falls from 20% to 10%. The risk ratio is 0.5, the relative reduction is 50%, and the absolute reduction is ten percentage points. That corresponds to ten fewer events per hundred people over the specified interval.

In fictional setting two, risk falls from 2% to 1%. The ratio and relative reduction are the same, but the absolute reduction is one percentage point, or one fewer event per hundred. These scenarios demonstrate why a relative statement alone cannot communicate the expected absolute difference across different baseline risks.

Preserve uncertainty and study limitations

Use the effect-measure worksheet to record the two risks, comparison order, time horizon, and uncertainty source. The risk-difference tool can illustrate an independent-group difference when its assumptions apply. Do not invent counts from a model's rounded percentages solely to obtain an interval.

Present an appropriate interval for the effect when available, and avoid implying certainty by rounding a small difference to zero. A large relative change in a rare outcome may still have an important consequence, while an apparently large absolute change may reflect bias or confounding. Importance follows context rather than a universal adjective.

Keep causal language proportional to the design. Observational associations should not automatically become promises that changing an exposure will produce the displayed benefit. If translating a study estimate to another population, label the baseline-risk assumption and uncertainty rather than presenting the result as directly observed there.

Finish with a sentence that includes both event frequencies and the named comparison measure. Readers should be able to reconstruct the absolute and relative statements without guessing the denominator, event direction, or period at risk.

Source checked on 3 November 2026; fictional risk comparisons independently recalculated.