Per Hundred Thousand What? How Denominators Decide the Shape of a Health Scare

A number can be accurate and still tell you almost nothing. That is the uncomfortable lesson behind a good many UK health stories in 2026, where a doubling of something in a local paper or a percentage-point leap in a national bulletin often turns out to be a story about the base of the sum rather than the health of the population. The technical term for that base is the denominator, and it deserves far more scrutiny than it usually receives.

This is not a argument against health journalism, still less against the underlying data. It is an argument for reading the arithmetic before accepting the conclusion.

Two words that do most of the work

Incidence and prevalence are routinely used as though they were interchangeable. They are not. Incidence counts new cases in a defined period; prevalence counts everyone living with a condition at a point in time. A treatment that keeps people alive longer will raise prevalence while leaving incidence untouched. A screening programme that brings forward diagnosis will raise measured incidence for a few years without any change in how often disease actually occurs. Neither pattern is a failure; both can produce headlines that look like a failure.

The distinction matters most for long-term conditions. Dementia prevalence estimates, for instance, move with the age structure of the population and with diagnostic practice as much as with underlying risk. When a 2026 bulletin reports a rise, the useful question is not whether the figure went up but which of those three things moved.

Testing is not the same as disease

A striking share of the conditions that generate alarm are diagnosed through tests, and the number of tests performed is a policy choice. Sexual health services across England have spent the past few years expanding asymptomatic screening, and diagnoses of gonorrhoea and syphilis have risen accordingly. Some of that rise is genuine transmission; some is simply more people looking. Untangling the two requires data on testing volume and positivity, which is precisely the data that rarely makes it into a short news item.

The same logic applies to referral-driven conditions. When awareness campaigns shorten the time between first symptoms and a GP appointment, recorded cases rise before outcomes improve. A rising count in the year after a campaign is, in the short run, a sign the campaign worked.

The pull of the small number

Local reporting is where denominators do most damage. A coastal town or a rural district with a few thousand residents can produce a genuinely enormous percentage change from a handful of events. Three cases becoming nine is a 200 per cent increase and, in a population that size, entirely within the range of ordinary year-to-year noise. Without a confidence interval, which almost no local press release supplies, the figure invites a conclusion the data cannot support.

National figures have a related problem. Crude rates reflect the age structure of an area as much as its health. That is why published comparisons are usually age-standardised, a statistical adjustment that asks what the rate would be if every area had the same age profile. When a story quotes a crude rate alongside an age-standardised one from another source, the comparison is meaningless even though both numbers are correct.

Who is actually in the base

Perhaps the most consequential denominator question concerns who counts. GP registered lists are a common denominator for prevalence in England, yet they include people who have moved away and have not yet been removed, and exclude people not registered at all — a group that is younger, more mobile and often less healthy than average. Resident population estimates, derived from census and migration data, produce a different picture again. In cities with large student populations or high turnover, the two denominators can differ enough to change the ranking of areas entirely.

Where testing is voluntary, the denominator can be self-selecting in a second way. A condition measured only in people who attend a clinic is measured in a healthier, more engaged population than the one the headline claims to describe.

Five questions worth asking

Before sharing the next alarming figure, it is worth running through a short list.

1. What is the denominator? Is it the whole population, registered patients, tests performed, hospital admissions or something else? A rate per 100,000 people and a rate per 100,000 tests are not comparable.
2. Is this incidence or prevalence? And does the story treat a change in one as evidence about the other?
3. Has testing, screening or referral changed? If the amount of looking has changed, the amount found will change too.
4. Has it been age-standardised? If two areas or two years are being compared, and age structures differ, the raw comparison misleads.
5. How many actual events are we talking about? Ten cases and one hundred cases can both be described as a tenfold increase. Only one of them is likely to be a public health problem.

What better reporting looks like

None of this argues for complacency. Some 2026 signals are real and serious, and the measles resurgence of recent years is a useful contrast: there, rising case counts tracked falling vaccination coverage in identifiable communities, and the denominator problem reinforced rather than obscured the story. The difference is that the measles data came with coverage figures, age breakdowns and geography attached.

That is the standard to ask for. A single number without a base, a period and a population is not a finding; it is an invitation to guess. Readers who know to ask about the denominator are not being sceptical for its own sake. They are doing the small amount of arithmetic that separates a trend from an artefact — and, on a good day, a genuine cause for concern from a headline that got its risk wrong.