One Pooled Sensitivity, Two Hospitals: Check Diagnostic Accuracy by Site
October 27, 2026 · 5 min read · Aqrab editorial team
A multicenter diagnostic study reports 80% sensitivity. That pooled number may describe neither hospital particularly well. Before using it to judge a test for your own setting, ask how the test performed at each site and who was tested there.
Same pooled result, different local experience
Imagine two hospitals, each with 100 people who have the target condition and 100 who do not. All receive the same hypothetical index test and an independent reference standard. These invented counts illustrate arithmetic, not the performance of any real test.
| Site | True positive | False negative | False positive | True negative | Sensitivity |
|---|---|---|---|---|---|
| Hospital A | 95 | 5 | 10 | 90 | 95% |
| Hospital B | 65 | 35 | 10 | 90 | 65% |
| Pooled | 160 | 40 | 20 | 180 | 80% |
The pooled sensitivity is 160/200 = 80%, yet the site estimates are 95/100 = 95% and 65/100 = 65%. Specificity is 90/100 = 90% at both sites. The pooled calculation is not wrong; it answers an average-over-these-sites question. It cannot, on its own, promise 80% sensitivity in a third hospital.
The practical takeaway
Request site-specific 2×2 counts and uncertainty intervals alongside the pooled estimate. Then compare each site's patients and testing workflow with the proposed use setting.
Variation is a finding, not yet an explanation
A site difference could reflect chance, differences in disease severity or alternative diagnoses, reader expertise, specimen handling, test execution, or reference-standard procedures. The table cannot distinguish these explanations. A post hoc story about one site being “better” would be unsupported. Nor does a small subgroup with an extreme percentage establish a stable difference; examine its numerator, denominator, and interval.
Site-specific reporting also needs context. If only a handful of target-condition cases were tested at one site, its sensitivity estimate will be imprecise. If the sites enrolled unlike patients, the contrast is not automatically a causal effect of the hospital. The useful next step is to identify what would need to match for the result to travel.
Four questions for the paper
- Counts: Are the four diagnostic-table cells and uncertainty intervals available for each site, not only for the pooled sample?
- Case mix: Were disease severity and alternative diagnoses comparable across sites and to the intended-use population?
- Workflow: Did sites use the same readers, training, specimen handling, index-test protocol, and reference standard?
- Interpretation: Was variation assessed as planned, or is the explanation an exploratory story after seeing the data?
Where Aqrab fits
Use Aqrab to structure a methods critique around site-specific counts, patient selection, measurement procedures, and the intended destination of a diagnostic claim. A critique can flag an unsupported transport assumption; it cannot establish local accuracy without local evidence.
Source anchors
The FDA statistical guidance for diagnostic-test studies recommends reporting results by clinical or specimen-collection site and relevant subgroups, with accuracy estimates and confidence intervals. The STARD 2015 checklist asks authors to distinguish prespecified from exploratory analyses of variation in diagnostic accuracy. Neither source validates the invented counts above.
Keep reading
Don't stop at one method.
Good methods judgment comes from contrast. Read the neighboring guides, see where the assumptions diverge, and avoid treating every observational problem like it needs the same hammer.
Only the Disagreements Were Checked: The Discrepant-Resolution Trap
When a third test checks only disagreements, agreement is not proof that both tests were right. Use this three-question audit before accepting a revised diagnostic-accuracy table.
The Result Arrived After the Decision: Audit Diagnostic Turnaround
A practical timeline shows why diagnostic accuracy alone cannot establish that a result can reach the clinical decision it is meant to support.
Before the Analyzer: Does Specimen Handling Match Diagnostic Test Use?
A practical specimen-path audit for diagnostic studies: collection, transport, storage, and exclusions can separate the reported accuracy from the intended clinical workflow.