The Ledger Beside the Result: Counting the People a Diagnostic Test Never Reached

A diagnostic test can be highly accurate and still miss the people who most need it. That apparent contradiction is not a failure of the assay. It is a failure of the denominator. Sensitivity and specificity describe how a test behaves among people who were tested. They say nothing about the eligible population who were never invited, the households that never returned a kit, the samples that arrived unsuitable, or the patients whose result was issued but never acted upon. In UK diagnostic planning, those gaps are not administrative footnotes. They are the difference between a test that performs well and a service that works.

What the percentage leaves out

Sensitivity asks how often a test detects a condition when it is present. Specificity asks how often it correctly clears people who do not have it. Both are conditional measures. They depend on a defined group of participants, usually those who accepted testing and produced a valid sample. Move from that group to the wider population and another question appears: what proportion of people who could have been tested actually entered the pathway?

This matters across the UK. A community phlebotomy service may report excellent concordance with laboratory reference methods, yet serve mainly people who can travel to a clinic during working hours. A home test kit may show strong clinical validity in a sponsored study, yet be returned less often by people in temporary accommodation, people with low digital confidence, or households where English is not the first language. A screening pathway may have robust laboratory metrics while invitation uptake varies sharply by deprivation. None of these problems is visible in a standalone accuracy figure.

The missing denominator also changes how results should be read. A test with high sensitivity can produce more false positives when used in a low-prevalence population, and a test with high specificity can still miss cases if the people most at risk never take part. Positive predictive value and negative predictive value are useful, but they too are calculated within a tested group. Without a population view, a dashboard can look reassuring while quietly excluding the people whose care is most fragile.

A denominator is a care pathway, not a line in a report

The useful denominator is not a single number. It is a sequence. For any diagnostic service, the population funnel should show: who was eligible; who was invited; who accepted; who received a kit or appointment; who returned a sample; whose sample was valid; whose result was communicated; and who completed the recommended follow-up. Each stage has a different meaning and a different remedy.

A low invitation rate points to identification and outreach. A low acceptance rate points to trust, convenience and communication. A low return rate points to kit design, postal logistics and support. A high invalid-sample rate points to instructions, collection conditions or anatomical and cultural barriers. A low follow-up rate points to result handling, appointment access and clinical ownership. Aggregating all of this into one percentage hides the stage where people are lost.

For commissioners, the practical implication is that service specifications should require a population funnel alongside test performance. Reporting diagnostic yield per 1,000 eligible people is more informative than reporting yield per 1,000 samples received. The former shows reach; the latter shows only what happened after reach had already been achieved. Both belong in the contract, but they answer different questions.

Making equity visible without overclaiming

Equity analysis depends on disaggregation. A national or regional average can conceal divergent experience by age, sex, ethnicity, deprivation, rurality, disability, caring responsibilities and digital access. UK health organisations already hold much of this information, though not always in a form that is complete or safe to analyse. Small numbers require care: subgroup data should be presented with confidence intervals or suppressed where identification risk is real. The aim is not to produce a league table of disadvantage, but to show where a pathway is failing to reach people who should be able to use it.

This is also a research-literacy issue. When a study reports diagnostic accuracy, readers should ask who was recruited, how they were recruited, and who declined. A trial conducted in a well-resourced teaching hospital may not transfer neatly to a community diagnostic centre, a mobile unit or a home-testing service. Local validation is not a bureaucratic extra. It is how a national claim becomes a credible local offer. The same discipline applies to health-tech products entering the NHS: a CE or UKCA mark confirms regulatory compliance, not population reach.

Practical steps for commissioners

Commissioners can make the denominator visible in four ways. First, specify the population funnel in contracts and evaluation frameworks, including invitation, acceptance, return, validity, communication and follow-up. Second, treat non-return and non-attendance as quality indicators rather than patient failures. Third, fund the support that makes participation possible, such as translated instructions, telephone reminders, accessible collection points and reasonable adjustments. Fourth, compare local funnel data with expected population denominators, not only with previous months of activity.

These steps do not require a new national data system. They require a consistent habit: every performance report should state the population it describes and the people it excludes. Where data are missing, that absence should be reported as missing, not smoothed away.

What clinicians can do in everyday practice

Clinicians do not need to become statisticians to use this idea. They can ask a simpler question of any diagnostic result: who is not in this data? In practice, that may mean reviewing non-return lists, checking whether results have been communicated in an accessible format, and noticing when a service is used repeatedly by the same relatively advantaged group. Audit can begin small: a single practice or clinic can track invitation, uptake and follow-up for one pathway over one cycle. The point is to make the invisible stage visible before it becomes an unexplained clinical gap.

What health-tech companies should publish

Companies developing diagnostic products can support the same discipline by publishing the eligibility criteria, recruitment route, return rate, invalid-sample rate and missing-data profile behind their performance claims. A dashboard that shows only sensitivity and specificity is incomplete. A dashboard that shows the population funnel, with appropriate caveats, is more honest and more useful to NHS partners. It also helps companies identify where product design, instructions or support services need to change.

Accuracy still matters. No amount of denominator reporting can rescue a test that does not work. But accuracy alone cannot tell a commissioner whether a service is reaching the population it was commissioned for, or a clinician whether a reassuring result reflects the patient in front of them. The ledger beside the result is not an alternative to diagnostic science. It is the condition for applying that science fairly. The question is not only how well a test performs among the people who used it, but who never had the chance to be counted.