“Risk increases by 50%” can sound alarming. But 50% of what starting risk, over what period and in which group? Without those details, the headline leaves out information you need.
The useful habit: ask for the starting risk and the changed risk using the same denominator and time period.
Absolute risk describes the chance of an event
If an event occurs in 2 out of 1,000 people over a specified period, its observed frequency in that group is 0.2%. A risk estimate uses information about a defined group and period; it is not a prediction of exactly what will happen to a particular person.
The time period matters. A one-year risk and a lifetime risk cannot be treated as interchangeable numbers.
Relative risk compares two risks
Relative risk, also called a risk ratio, divides the risk in one group by the risk in a comparison group. A ratio of 1 means the two risks are equal. A ratio of 1.5 corresponds to a 50% relative increase.
Consider these entirely fictional examples. They do not describe a real disease, medicine or exposure.
| Example | Starting risk | Changed risk | Relative increase | Absolute difference |
|---|---|---|---|---|
| A | 2 in 1,000 | 3 in 1,000 | 50% | 1 additional event per 1,000 |
| B | 200 in 1,000 | 300 in 1,000 | 50% | 100 additional events per 1,000 |
The relative increase is identical, but the absolute differences are very different. Neither description should replace the other. Together they show the size of the comparison more clearly.
Keep percentages and percentage points separate
In Example A, risk rises from 0.2% to 0.3%. That is an increase of 0.1 percentage points and a 50% relative increase. It is not an increase of 50 percentage points.
Use the same discipline for reductions. A fictional drop from 10 in 1,000 to 5 in 1,000 is a 50% relative reduction and an absolute reduction of 5 in 1,000, or 0.5 percentage points.
Our percentage-points guide explains the arithmetic in more detail.
Use equal denominators
Comparing “1 in 100” with “5 in 1,000” is easier after rewriting the first as “10 in 1,000.” The changed risk is half the starting risk, rather than five times as large.
Natural frequencies can make a comparison easier to read, but keep the groups and periods comparable. An annual rate based on person-time is not automatically the same as a simple proportion of people over a fixed period.
Ask what the study can establish
An association between an exposure and an outcome does not, by itself, establish that the exposure caused the difference. Differences between groups, measurement problems and other limitations can affect the result.
Read whether the comparison came from an experiment or an observational study, how participants were selected and which factors researchers considered. See correlation versus causation for that distinction.
Read uncertainty and the exact measure
A reported risk ratio is an estimate. Look for uncertainty intervals and the number of events, especially when events are rare. A dramatic ratio based on very few events may be imprecise.
Also check whether the study actually reports a risk ratio, odds ratio or hazard ratio. These measures are related but not interchangeable. Do not automatically translate “twice the odds” into “twice the probability.”
A five-question headline check
- What is the baseline risk?
- What is the risk in the comparison group?
- What population and time period do both numbers cover?
- How uncertain is the estimate, and what did the study design establish?
- What benefits, harms or alternative outcomes are missing from the headline?
A small absolute increase can still matter when an outcome is serious or many people are exposed. A large relative change should not automatically create panic, and a small absolute change should not automatically be dismissed.
Note to keep: risk communication is most useful when it gives both the comparison and its scale. For personal treatment decisions, discuss the evidence and your circumstances with a qualified clinician.
