Signature explainer
Risk is not fate
Health estimates are conditional summaries, not certain forecasts. This short explainer uses invented group examples to show why the baseline, time period, population, and uncertainty matter before a number can support a decision.
Interactive / the crowd
10 in 100 people. Which ones? Nobody can say.
Pick an invented group frequency, then shuffle. The number of filled dots stays fixed, while the filled positions change. The visual emphasizes that a population-derived estimate does not identify which individuals will experience an outcome.
Illustrative example only. This page never calculates anyone's personal risk.
Interactive / the headline trap
"Risk increased by 50 percent"
The same invented comparison, told two ways. A 50 percent relative increase on a 2-in-100 baseline is an absolute increase of 1 in 100. Understanding it requires both the relative change and the baseline.
Baseline: 2 in 100
After a "50 percent increase": 3 in 100
Ask of every dramatic statistic: out of how many, and starting from what baseline?
The four lessons, in full
An estimate summarizes a defined group
A model might estimate that 10 out of 100 people in a defined population will experience an outcome over a stated period. That estimate depends on the population, inputs, outcome definition, time period, and model used.
It does not identify which individuals will experience the outcome. Two people with the same estimate can have different outcomes without making the estimate a personal forecast.
A population-derived estimate is not a prediction of one person's future.
Relative change needs a baseline
A relative increase can sound large while the absolute difference is small. In an invented example, moving from 2 in 100 to 3 in 100 is a 50 percent relative increase and an absolute increase of 1 in 100.
Both descriptions are mathematically correct. The baseline, absolute difference, time period, and uncertainty are needed to understand what the comparison means.
Ask for the starting risk and the absolute difference, not only a relative percentage.
Higher is not certain; lower is not impossible
A higher estimate can support a conversation about priorities, but it does not guarantee an outcome. A lower estimate does not rule one out. Models can also underperform when used in populations unlike the ones in which they were developed or tested.
Some inputs may be modifiable and others may not be. Whether changing an input changes an outcome, and by how much, requires evidence beyond the score itself.
An estimate can inform a decision without becoming a verdict.
Use a number only in a decision context
Useful questions include: which outcome and time period does this estimate cover, which population was the model tested in, how uncertain is it, and would a different result change a real decision?
A clinician can connect those questions to symptoms, family history, local guidance, and the benefits and downsides of the available options. This site does not calculate or interpret a personal score.
The value of an estimate depends on the decision it can responsibly inform.
Put the intuition to work
Clearer probability language can help when discussing screening options and sharing family history with a clinician.
Sources
- Communicating risk and evidence Winton Centre for Risk and Evidence Communication, University of CambridgeResearch and public resources on communicating benefits, harms, uncertainty, and absolute frequencies.
- NHS screening NHSPlain-language framing of screening benefits, risks, and limits.
- Principles of population screening UK National Screening CommitteeCurrent principles for evidence, informed choice, benefit, harm, and uncertainty in screening programmes.
Last reviewed: 2026-07-18