A Negative Test Is Not a Rule-Out: Calculate the Risk Left Behind
October 6, 2026 · 6 min read · Aqrab editorial team
A paper calls a negative result “reassuring” because its test has 90% sensitivity. But sensitivity alone does not tell you how much risk remains after a negative result. That depends on the chance of the condition before testing, specificity, and whether the study's accuracy estimates apply to the people being tested.
One Negative Result, Two Starting Risks
Consider a hypothetical test with 90% sensitivity and 80% specificity. Its negative likelihood ratio (LR−) is (1 − 0.90) ÷ 0.80 = 0.125. This means a negative result multiplies the odds of the condition by one eighth, assuming those accuracy estimates apply. It does not subtract a fixed number of percentage points from risk.
| Chance before test | Pretest odds | Odds after negative (× 0.125) | Chance after negative |
|---|---|---|---|
| 5% | 0.05 ÷ 0.95 | 0.00658 | 0.65% |
| 50% | 0.50 ÷ 0.50 | 0.125 | 11.1% |
The same negative result leaves roughly 1 in 154 versus 1 in 9 with the condition in these two scenarios. Neither number, by itself, defines what to do next. Whether residual risk is acceptable depends on the condition, the consequences of a missed diagnosis, available follow-up tests, and the care context.
The practical takeaway
When a paper says a negative result “rules out” a condition, ask for the starting risk, the LR−, and the risk remaining after the result—not sensitivity alone.
How to Reproduce the Calculation
- Convert starting probability to odds: p ÷ (1 − p).
- Multiply those odds by LR− = (1 − sensitivity) ÷ specificity.
- Convert back to probability: odds ÷ (1 + odds).
Use a pretest estimate for the relevant clinical setting, not the condition frequency in an unrelated research sample. An LR from a published study is not automatically portable: patient spectrum, disease stage, test execution, the positivity threshold, and handling of indeterminate results can change accuracy. These hypothetical calculations show the consequence if accuracy is stable; they do not validate that assumption.
A Four-Point Rule-Out Audit
- Starting risk: Is the pretest probability defensible for the intended patients and point in the pathway?
- Accuracy: Is LR− available with uncertainty, and were negative results verified against an appropriate reference standard?
- Applicability: Did the study include the difficult, early, or ambiguous presentations in which this test will be used?
- Decision: Is the residual risk low enough for the proposed action, given the consequences of a miss? The study cannot answer this from accuracy alone.
Where Aqrab Fits
Use Aqrab to critique a diagnostic paper's population, reference standard, missing or indeterminate results, and jump from “high sensitivity” to “safe rule-out.” A methods critique can identify the assumptions and missing evidence; it cannot establish a patient-specific risk or a clinical decision threshold.
Source Anchors
The BMJ likelihood-ratio methods note gives the odds-update relationship. AHRQ's diagnostic pathway brief explains pretest and post-test probability and the role of consequences. The STARD explanation describes intended use, clinical role, and applicability. None supplies the hypothetical test values 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.
Ninety Percent Sensitive Based on How Many Cases?
The same 90% sensitivity can rest on 20 or 200 condition-positive participants. A worked confidence-interval comparison shows why the denominator matters before a test is called reliable.
Test Agreement Is Not Diagnostic Accuracy
A worked 97% agreement example shows why matching another imperfect test cannot establish sensitivity or specificity, plus four questions for appraising the comparator and denominators.
The Trial Ended. Who Made It Into the Open-Label Extension?
A denominator-first audit of long-term extension findings: trace who enrolled, who stayed, and what an uncontrolled extension can actually establish.