When might relative risk still be useful or less misleading?

The short version: relative risk earns its keep when the baseline risk is high enough that the ratio reflects a real change, and when you pair it with the absolute numbers so it cannot mislead.

Relative risk is most useful when the baseline event rate is substantial, because then a large ratio actually means a large absolute difference. The methodological primer in this retrieval shows how the same relative risk can hide wildly different absolute effects depending on the starting risk: a relative risk of 9 can mean a jump from 10% to 90%, while a relative risk of about 41 can mean a jump from 1.2% to 50% [2]. The ratio alone tells you nothing about whether the change matters to a real person.

It becomes less misleading when you present it alongside the absolute risk and the number needed to treat, and when you use percentage formats rather than raw frequencies. One experiment found that the comprehension gap between gain- and loss-framed risk messages disappeared entirely when statistics were shown as percentages instead of frequencies [1]. People understood the numbers better, which is the whole point of communicating risk.

The danger is when relative risk appears alone. People simultaneously overestimate rare event probabilities and underweight them in their choices, a documented pattern [3]. That is exactly the setup for a scary relative risk figure to distort a decision about a rare outcome.

My call: use relative risk only when the baseline risk is high enough to matter, and never without the absolute numbers beside it. Confidence: high.