Sensitivity and specificity
Sensitivity is the proportion of people with a disease that a test correctly identifies as positive, while specificity is the proportion of people without the disease that it correctly identifies as negative. Together they describe a diagnostic test's accuracy.
Sensitivity and specificity are the two core measures of how well a diagnostic test performs. Sensitivity is the true positive rate: of everyone who truly has the disease, the fraction the test flags as positive, calculated as TP / (TP + FN). Specificity is the true negative rate: of everyone free of the disease, the fraction the test correctly calls negative, calculated as TN / (TN + FP).
Two classic mnemonics link the measures to clinical use. SnNout: a highly Sensitive test, when Negative, rules out disease, because a sensitive test misses very few true cases. SpPin: a highly Specific test, when Positive, rules in disease, because a specific test rarely mislabels healthy people. Highly sensitive tests suit screening; highly specific tests suit confirmation.
The two measures usually trade off against each other. Lowering a test's positivity cutoff catches more true cases (higher sensitivity) but mislabels more healthy people (lower specificity). The receiver operating characteristic (ROC) curve plots this trade-off, graphing sensitivity against 1 − specificity (the false positive rate) across every possible cutoff; a curve hugging the upper-left corner indicates a better test. Unlike predictive values, sensitivity and specificity are properties of the test itself and do not change with disease prevalence. Likelihood ratios combine both: the positive likelihood ratio is sensitivity / (1 − specificity), and the negative is (1 − sensitivity) / specificity.
USMLE Step 1 tests these concepts heavily in biostatistics — expect 2×2 table calculations, ROC curve interpretation, and questions about shifting cutoffs. The NPTE-PT and NPTE-PTA apply the same ideas to orthopedic special tests, where a test's sensitivity and specificity determine whether it is better at ruling a pathology out or in.
Key takeaways
- Sensitivity = TP / (TP + FN): the fraction of diseased patients the test correctly detects.
- Specificity = TN / (TN + FP): the fraction of healthy patients the test correctly clears.
- SnNout and SpPin: sensitive tests rule out when negative; specific tests rule in when positive.
- ROC curves plot sensitivity against 1 − specificity across cutoffs; changing the cutoff trades one measure for the other.
- Unlike predictive values, sensitivity and specificity do not depend on disease prevalence.