Risk, in the language that health information actually uses

Two ways of describing the same change

A change in risk can be expressed as a proportion of the original risk or as the difference between the two figures, and these sound very different. The proportional form makes small changes to small risks sound substantial, which is why it dominates headlines and promotional material. The difference form conveys what would actually change for a group of people, which is usually the question a reader is trying to answer.

Neither form is dishonest, and using one without the other is where the misleading impression is generated. Asking for the starting figure whenever a proportional change is quoted resolves most of the confusion immediately.

Why the baseline does the work

The same proportional change applied to a common condition and a rare one produces entirely different numbers of affected people. This is why risk factors for rare conditions can be real, well established and almost irrelevant to how an individual should behave. It also explains why the accumulation of many small risk factors is discussed differently from any single one of them.

Baselines vary by age, sex and existing health, which means the same finding has different implications for different readers. General coverage almost never supplies these variations, so the figure quoted describes an average person who may resemble nobody in particular.

Risk applies to groups, not to individuals

A statement that some proportion of a group will experience an outcome does not say which members of that group, and often nothing can identify them in advance. This is why an individual can follow every recommendation and still develop a condition, and why someone else can do the opposite and not. Neither case is evidence against the underlying relationship, though both are used that way constantly in ordinary conversation.

Understanding risk as a description of a distribution rather than a prediction about a person removes a great deal of unnecessary confusion. It also makes clear why guarantees about health outcomes are never available regardless of how confident a claim sounds.

Compounding and time

Risks are usually quoted over a defined period, and a figure over ten years describes something quite different from the same figure over a lifetime. Comparisons that omit the time period are uninterpretable, and the omission is common enough to be worth checking for routinely. Risk also changes over a life, so a figure calculated at one age does not carry forward unchanged to a later one.

Tools that estimate individual risk combine several factors and produce a figure with real uncertainty attached, which is often lost in presentation. Those tools are designed for use within a consultation rather than as something to be run alone and interpreted without support.

What to do with a number that worries you

The useful response to an alarming figure is to establish what the baseline was, over what period, and in which group it was measured. If any of those is missing, the figure cannot support the conclusion attached to it, whatever the accompanying commentary says. Where a risk genuinely applies, what follows from it depends on individual circumstances that only a clinician with the full picture can weigh.

Changing or stopping any prescribed treatment on the basis of a statistic encountered in general reading is specifically a bad idea. The purpose of understanding risk language is to make conversations with clinicians more productive rather than to replace them.