“Accuracy” is the most reassuring word in diagnostics and the least informative. It sounds like a property of the test, the way weight is a property of an object. It is not. It is a property of a particular measurement, made on a particular group of people, at a particular setting, judged against a particular idea of the truth. Change any of those and the number changes, sometimes enormously, without anything about the test changing at all.

This guide is written for anyone who has to decide what a diagnostic-AI number means: a journalist with a press release, a commissioner with a business case, a clinician with a vendor in the room, a patient with a headline. It assumes no statistics. It works for blood tests, radiology models, digital pathology, AI stethoscopes, symptom checkers, triage tools and wearables — anything that looks at a person and outputs a verdict.

The running example is real and recent, and we audited it in full: an NHS blood test reported in July 2026 as 99% accurate at both detecting and ruling out gynaecological cancer. Every question below caught something in that story. But the questions come first; the case is only the demonstration.

Question one

Ninety-nine percent of what?

Start by deleting the word “accuracy” and replacing it with the metric’s actual name. There are at least eight candidates — sensitivity, specificity, positive and negative predictive value, overall classification accuracy, balanced accuracy, AUC, agreement with clinicians, and positive/negative percent agreement — and a release that says “accurate” often does not say which one it means.

The four that matter most come from a single table. Imagine everyone who took the test, sorted by whether they really had the disease and what the test said:

  • Sensitivity — of the people who had it, how many did the test catch?
  • Specificity — of the people who didn’t, how many did it correctly leave alone?
  • Positive predictive value — of the people it flagged, how many really had it?
  • Negative predictive value — of the people it cleared, how many really were clear?

The first two read down the columns — they are facts about the disease groups. The last two read across the rows — they are facts about the test-result groups. They answer genuinely different questions, and a test can score brilliantly on one while failing on another.

There is also a quantity that is literally called accuracy — the share of everyone the test classified correctly. It is rarely reported, because for most real tools it is either flattering to the point of meaninglessness or brutal.

Illustrative · 1,000 people tested

The same four boxes. Four different questions.

Reading down a column — a fact about a disease group
1,000 people tested, by true disease status and test result
Really has itReally doesn’t
Test says yes
36true positive
96false positive
Test says no
4false negative
864true negative

Of the 40 people who really have it, 36 test positive. Sensitivity = 36/40 = 90%. It is a fact about people with the disease, and it says nothing about the 96 healthy people who were also flagged.

Two of these read down the columns; two read across the rows. Column metrics answer “given the disease, what does the test do?” — unlike the row metrics, they do not change just because the disease gets rarer. (They still move when the mix of patients changes; see question three.) Row metrics answer “given the result, what do I know?” — the question a patient is actually asking, and the one whose answer shifts with how common the disease is.

Ask: what is the metric’s exact technical name — and if the source won’t say, why not?

In the case The two numbers behind “99% accurate” were a sensitivity of 99.1% and a negative predictive value of 99.8% — one about women with cancer, one about women the test cleared. Specificity at that setting was 20.0%. Read as overall classification accuracy, the figure was 23.0%.
Question two

Out of how many?

“99% sensitivity” can mean one case missed out of a hundred, ten out of a thousand, or one out of a hundred and fourteen. It can also rest on seven cancers total, in which case the percentage is theatre.

Percentages hide two things at once: the clinical scale of the error, and the statistical uncertainty around it. A result built on a handful of positive cases has a confidence interval wide enough to reach well below the number in the headline. Always convert back to counts.

Illustrative · the same 99%, four times

Identical headline. Four very different amounts of evidence.

100 cases
99.0% (94.6–99.8)
200 cases
99.0% (96.4–99.7)
400 cases
99.0% (97.5–99.6)
1,000 cases
99.0% (98.2–99.5)
90%95%100%
Every row reports 99.0% sensitivity. Only the case count differs. The bar is the 95% confidence interval — the range of true sensitivities consistent with what was seen. On 100 cases it reaches down to 94.6%; on 1,000 it barely moves. The percentage alone cannot tell you which of these you are reading. For scale, the NHS blood test in our investigation rested on 114 cancers.

Ask: how many actual cases produced this percentage, and what is the confidence interval?

In the case 113 of 114 diagnosed cancers scored above the rule-out threshold; one fell below it. A good result — and one whose confidence interval runs from 95% to 100%, because a single false negative is all that stands behind it. Clinicians were blinded, so that was a test error, not a clinical one. The study did not report the site or stage.
Question three

Among whom — and how common was the disease?

This is the question that catches the most people, because the effect is so large and so counter-intuitive.

In the predictive-value equations, sensitivity and specificity are held fixed and prevalence enters through the mix of diseased and non-diseased people. Hold them fixed and take a test to a population where almost nobody is sick: its negative predictive value climbs towards 100% almost regardless of how well it discriminates, because “you’re fine” is nearly always the right answer. Take it to a specialist clinic and the predictive values change without a line of code changing. In the real world all three can shift at once, because moving a test to a new population usually changes the spectrum of disease, who gets verified, and how the threshold is applied.

So the reference point for any negative predictive value is not zero. A useful no-information comparator is clearing people at random, whose expected negative predictive value is simply the share of people who don’t have the disease. Drag the sliders and watch the gap between the bar and the marker. That gap shows how much the test enriched its cleared group with people who really were clear — not its total clinical value, which also depends on how many it cleared, what the misses cost, and what anyone does with the result.

Calculator — you supply the inputs

A 99% negative predictive value can be mostly a fact about the population.

0.1%20%40%

Jump to: · ·

Negative predictive value99.8%

Of those told “low risk”, the share who really are clear. Marker: 96.1% — what you would score by clearing people at random.

Positive predictive value4.8%

Of those flagged, the share who really have it. Marker: 3.9% — the disease rate you started from.

Test 1,000 people at this disease rate and 39 of them have it. The test clears 193 people — and misses fewer than one of them. It flags 807 — of whom 769 do not have it.

The sliders hold sensitivity and specificity fixed, to isolate the mathematical effect of prevalence. Real-world transport can change all three. Either way, a headline negative predictive value has to be read next to the disease rate — and next to what guessing would have scored.

Ask: how common was the disease in the study, how common will it be where the tool is actually used, and what would random guessing have scored?

In the case Cancer prevalence in the studied cohort was 3.9%, so clearing women at random would already have produced a “negative predictive value” of 96.1%. The reported 99.8% is a real improvement on that — but a much smaller one than 99.8% sounds. And the claim was then applied to post-menopausal-bleeding referrals, a group with a materially higher cancer rate, where the same test would perform differently.
Question four

At which threshold — and who chose it?

Almost every diagnostic AI outputs a continuous score. “Positive” and “negative” only exist once somebody draws a line. Move the line and sensitivity, specificity, workload, cost and the number of missed cases all move together, in opposite directions. There is no setting that is best at everything.

This creates two distinct problems. The first is honest but easy to miss: a result quoted at a rule-out threshold and a result quoted at a rule-in threshold are not comparable, even from the same model on the same day. The second is not honest: a threshold chosen after seeing the data, because it produced the nicest number, is a form of overfitting.

And watch for the reverse trick — where the setting is reported as though it were a discovery. If a threshold is defined as “clear the lowest-scoring 20%,” then “this test can clear one in five patients” is not a finding. It is the instruction, read back.

Simulated — not PinPoint · AUC 0.86

Move the threshold. Every number in the headline moves with it.

Scores the model gives

The whole curve = AUC 0.86

80%Sensitivity
75%Specificity
25False alarms per 100 well
AUC — a single number for how well the model ranks sick people above healthy ones, across every possible line — is the whole curve. A clinic lives at one point on it. Nothing about the model changes as you move it — only where the line is drawn.

Ask: was the threshold fixed in advance — and is any headline proportion actually just the threshold restated?

In the case The three operating points were prespecified, which is good practice. But the rule-out threshold is defined as clearing 20% of the patients without cancer — so “one in five women could be safely ruled out” reports the dial setting. At the study’s other prespecified setting, the same model on the same women missed twelve cancers instead of one.
Question five

What happens to the people it gets wrong?

Every threshold is a decision about which error to prefer, and that decision should be made by thinking about people, not about the shape of a curve.

For a cancer rule-out, it depends entirely on what is done with the result. If a low-risk score is used to stop or defer investigation, a false negative may mean false reassurance and delay. A false positive usually just leaves the patient on the pathway they were already on, though it can trigger extra work. Those are not equivalent harms — but note that low specificity mostly limits how much capacity a rule-out policy can release, rather than proving it creates new work.

Note also what the displaced test was doing. Replacing a scan with a blood test is only equivalent if the scan was only ever answering the one question. Scans tend to see more than the thing they were ordered for.

Ask: what happens to a real patient when this model is wrong in each direction — and what else was the old test catching?

In the case The scan the test would replace measures the womb lining — but it also images the ovaries. Seventeen of the 114 cancers in the study were ovarian (nine) or coded to retroperitoneum and peritoneum (eight). A blood-test rule-out would have to catch those some other way.
Question six

How was “truth” decided?

Every accuracy figure is a comparison against something the researchers treated as the truth: a biopsy, a pathology report, a specialist’s opinion, a code in a medical record, an expert panel, or just what happened over the following year. A model cannot be more reliable than the labels it was graded against.

Record-derived labels are the common weak point. If “had cancer” means “has a cancer code recorded within three months,” then anyone diagnosed in month four is counted as a true negative — and the test gets credit for clearing them.

Ask: how did the researchers decide who truly had the disease, and over what window?

Question seven

Who was left out?

Follow the flow chart. Almost every study loses people between the front door and the final table — incomplete data, unreadable images, unresolved diagnoses, patients whose records could not be linked. Sometimes that is unavoidable. It is still a place where performance can quietly improve.

The specific thing to check: are the enrolment number and the analysis number the same? If a claim quotes the larger one next to statistics computed on the smaller one, the arithmetic on offer is not the arithmetic that was done.

Ask: who entered, who was excluded, who disappeared before the final analysis — and which of those numbers is the denominator of the headline?

In the case 3,313 women were enrolled on the pathway; 2,953 reached the analysis. Every performance figure comes from the 2,953. The Guardian and FemTech World attached the 99% to the 3,313, and the NHS body promoting the test gave 3,313 as the number assessed — each restatement is quoted in the investigation.
Question eight

Compared with what?

The interesting question is almost never whether the AI beats chance. It is whether it beats what the clinic does today — the existing blood test, the existing scan, the existing guideline, or an experienced clinician’s judgement. A model can be genuinely impressive and still be worse than the cheap thing already in use.

Be careful with AUC here. It measures how well a model ranks sick above healthy across every possible threshold, which makes it good for comparing models and poor for telling you what happens to patients. A better AUC does not guarantee better behaviour at the one setting clinicians will actually use.

And establish where any head-to-head comparison came from. Comparisons are the most quotable thing in a press release and the easiest to introduce from outside the study.

Ask: is this replacing an existing test, adding to it, or deciding who gets it — and is the comparison in the paper?

In the case The paper compares the test with CA125, the incumbent marker: AUC 0.81 versus 0.71, which the authors themselves call “an imperfect comparison.” The head-to-head against ultrasound that circulated with the story — 0.832 against 0.792, in 578 women — is one we could not find in the paper or its supplement; we searched the full text, every supplementary table and the figures. It appears on one NHS body’s news page, which cites nothing, and in the company’s own press release, which attributes it to an unpublished sub-analysis held as “data on file.” We could not trace it to any published study, dataset or statistical report.
Question nine

Was it measured, or modelled?

This is the question that separates most health-AI claims from most health-AI evidence, and it comes in two parts.

First, what kind of evidence was it? These are five different stages, and press coverage calls them all “trials”. They overlap — a study can be external and prospective at once — but each answers a different question:

  • Internal validation — tested on data from the same source it was built on.
  • External validation — tested somewhere else, on someone else’s patients.
  • Prospective silent validation — run live on new patients, but nobody is shown the output.
  • Live implementation — clinicians see the output and act on it.
  • Impact evaluation — outcomes compared against a credible counterfactual.

A prospective silent validation is a strong design and much better than a retrospective one. But it cannot, even in principle, show that anything improved — because nothing was allowed to change.

Second, benefit claims. “Could save 18,000 procedures,” “could prevent 5,000 deaths,” “could save £100 million” are almost always chains: study performance × national population × assumed uptake × assumed clinician compliance × assumed costs. Ask which links were measured. Usually it is the first one.

Ask: did the tool change a single real decision — and for any headline benefit, which steps were observed and which were assumed?

In the case A prospective silent validation: clinicians were blinded and, in the paper’s words, “the patient’s onward journey was unaffected.” The study provides no evidence that any scan was avoided because of a PinPoint result — no clinician was shown one, and procedure avoidance was never evaluated. The “18,000 spared” figure assumes every unmeasured step in the chain passes at 100%. Not one of the nine restatements we examined mentioned the blinding.
Question ten

Who checked it?

Industry-funded evidence is not worthless — companies fund the studies because nobody else will, and vendor scientists are often the people who understand the model best. Disclosed conflicts are a sign of a functioning system, not a scandal.

But funding changes what still needs to happen, and the question to ask is not “was this conflicted?” so much as “has anyone unconflicted ever repeated it?” Look at the whole evidence base rather than the single paper: if every study of a tool shares authors, funders or data with its manufacturer, then no matter how many papers there are, the result has been produced once.

Regulatory language deserves the same scrutiny. “CE marked,” “UKCA marked” and “fully regulated” can mean an approved body assessed the device’s performance — or that the manufacturer filled in a form. For the lowest device classes, it is the form.

Ask: could an outside team reproduce this — and has one?

In the case All four published evaluations share authors with a financial interest, and one author appears on all four. The conflicts are properly disclosed in every paper. We could find no independent replication. On the MHRA register we found the device listed in the lowest UK in-vitro-diagnostic class, which manufacturers may self-certify.

Evaluate a claim yourself

Take the ten questions to your own AI

Paste a study, press release, model card or news story below. This turns the ten questions into a prompt that makes any assistant show its working — name the metric, find the denominators, check the threshold and the reference standard, and say plainly what is missing. Copy it into ChatGPT, Claude, or Gemini. It does not do the judging for you; it makes the gaps visible so you can.

Evaluation prompt

You are evaluating an accuracy claim about {{SUBJECT}}, using the "Ground Truth" method (groundtruth.health). The material to evaluate is at the end of this message. {{MODE}}

1. Ninety-nine percent of what? Replace the word "accurate" with the metric's exact technical name — sensitivity, specificity, positive or negative predictive value, overall classification accuracy, AUC, agreement, or positive/negative percent agreement (PPA/NPA, which resembles sensitivity and specificity but concedes there is no reference standard). If the source never names it, say so: that is the finding.
2. Out of how many? Give the numerator and denominator behind the headline percentage, quoting the source. If the source reports a confidence interval, quote it. If it does not, write "no interval reported" — and only then, if and only if you have the exact numerator and denominator, compute one, show the arithmetic and name the method. Never state an interval you cannot derive from numbers printed in the source. A percentage resting on a handful of cases is compatible with performance nobody would accept.
3. Among whom, and how common was the disease? Predictive values move with prevalence. Quote the disease rate in the study population. Then say whether the source states where the tool is intended to be used and at what disease rate there — if it does not, write "not stated" and do not supply a figure from your own knowledge. If you have relevant background knowledge, put it on a separate line marked "outside the source, unverified", give a range rather than a point, and do not use it in any calculation. Finally, give the negative predictive value that clearing people at random would have scored in the study population (1 − prevalence), and set the reported NPV against it.
4. At which threshold, and who chose it? Was the operating point fixed before the analysis or picked after? Check whether any headline proportion — "rules out one in five" — is simply the threshold restated.
5. What happens to the people it gets wrong? Take each direction separately, and judge it against what would actually be done with the result. Note anything the replaced test was also catching incidentally.
6. How was "truth" decided? Biopsy, pathology, imaging, a coding record, or just no diagnosis recorded within a window? A model cannot be more reliable than the labels it was graded against.
7. Who was left out? Compare the number enrolled with the number analysed, and check which of the two is the denominator of the headline. Separately, ask how many samples or images returned no usable result at all — invalid, insufficient, indeterminate, QC failure — and whether those people appear in that denominator.
8. Compared with what? Current care, an existing marker, clinician judgement — or nothing? Beware AUC standing in for performance at the threshold clinicians will actually use, and check whether any head-to-head comparison is in the paper at all.
9. Was it measured or modelled? Distinguish internal validation, external validation, prospective silent validation, live implementation and impact evaluation. For any projected benefit — procedures saved, lives saved, money saved — separate the steps that were measured from the steps that were assumed. Then answer plainly: did the study show the tool changing a single real clinical decision, yes or no?
10. Who ran it, who funded it, and has anyone unconflicted repeated it? Treat disclosed conflicts as a signal about what verification is still needed, not as proof of bad faith. Check what "regulated" or "CE/UKCA marked" actually means for this device class.

Quote the exact sentence the source uses for each answer. Where the source is silent, write "not stated" rather than inferring, and never supply a number the source does not give — no invented confidence intervals, denominators or prevalences.

Work only from the material below. If I have given you a link rather than the text and you cannot open it, say so and stop — do not reconstruct the source from memory, and do not produce quotes you have not read. If you recognise the study, keep anything you know from outside the pasted text in a separate section headed "outside the source, unverified", and never render it as a quotation. Before answering question 1, say what you are actually holding: the study itself, an abstract, a press release, or a news story about a study. If it is a restatement rather than the study, say so plainly — your answers then describe the restatement, and the gaps you find may be the reporter's rather than the researchers'.

Finish with: (a) an honest one-line restatement of what the evidence actually supports, (b) the single most important thing that is missing, and (c) what would have to be shown to believe the claim as written.

CLAIM / STUDY TO EVALUATE:
{{CLAIM}}

The method is ours; the judgement stays yours. An assistant can find the denominators and name the metric — it cannot tell you how safe is safe enough for the people in the remaining one percent.

The translation rule

When someone tells you a diagnostic AI is 99% accurate, the useless response is to ask whether 99% is high. Of course it is high. Every number chosen for a headline is high.

The useful response is six questions long, and you now have all six. What exactly was counted? Among whom, and how many of them were ill? At which setting, and who picked it? Against what standard of truth, and compared with what we already do? Did it change a single real decision, or was the benefit modelled? And what happens to the people in the remaining one percent?

A percentage is the end of somebody’s analysis. It should be the beginning of yours.

See these questions applied to a live claim: our investigation of the NHS cancer blood test reported as “99% accurate” — where one diagnosed cancer scored below the rule-out line, and overall classification accuracy is 23%.

Disclosures & provenance

Published
25 Jul 2026 · last updated 27 Jul 2026
Author
The Ground Truth editor. Editorial standard →
Funding
Self-funded. Ground Truth takes no money from, and has no affiliation with, any organization examined here. Independence policy →
Corrections
Corrected 27 Jul 2026 — The 578-woman ultrasound comparison also appears in the company's own press release, which attributes it to an unpublished sub-analysis held as “data on file.” Our finding that it cannot be traced to any published study is unchanged. Corrections log →