Sensitivity vs Specificity — Diagnostic Test Metrics
Sensitivity and specificity are the two fundamental properties of any diagnostic or screening test. They describe how well the test correctly identifies people with disease (sensitivity) and people without disease (specificity). Together with disease prevalence, they determine the predictive values that clinicians actually care about at the bedside.
Every diagnostic or screening test has four possible outcomes: true positive (diseased, test positive), false positive (not diseased, test positive), false negative (diseased, test negative), and true negative (not diseased, test negative). These four outcomes form the standard 2x2 table that underpins all diagnostic test evaluation. Sensitivity and specificity are properties of the test itself — they don't change with disease prevalence. Predictive values (PPV, NPV), however, depend on both test properties and disease prevalence.
| Disease + | Disease - | Total | |
|---|---|---|---|
| Test + | TP (True Positive) | FP (False Positive) | TP+FP |
| Test - | FN (False Negative) | TN (True Negative) | FN+TN |
| Total | TP+FN | FP+TN | N |
Key formulas:
- Sensitivity = TP / (TP+FN) = ability to detect disease when present
- Specificity = TN / (TN+FP) = ability to exclude disease when absent
- PPV = TP / (TP+FP) = probability of disease given a positive test
- NPV = TN / (TN+FN) = probability of no disease given a negative test
- Accuracy = (TP+TN) / N = overall correctness
- Prevalence = (TP+FN) / N
Sensitivity is the proportion of truly diseased persons who test positive. A highly sensitive test catches most cases — it has few false negatives. The mnemonic is SnNOut: a Sensitive test, when Negative, rules Out disease.
Use a highly sensitive test when:
- The consequence of missing a case is severe (e.g., HIV screening of blood donors — must not miss any positive unit)
- The disease is highly contagious and contact tracing depends on detection (e.g., TB)
- The initial screening round, followed by a more specific confirmatory test (e.g., HIV ELISA followed by Western blot)
- Rule-out strategy at primary care level (e.g., D-dimer to rule out DVT)
Example: A test with 99% sensitivity misses only 1 in 100 true cases. A test with 80% sensitivity misses 20 in 100 — unacceptable for serious diseases.
Specificity is the proportion of truly non-diseased persons who test negative. A highly specific test correctly identifies healthy people — it has few false positives. The mnemonic is SpPIn: a Specific test, when Positive, rules In disease.
Use a highly specific test when:
- The consequence of a false positive is severe (psychological harm, unnecessary treatment, e.g., HIV confirmation by Western blot)
- The treatment is toxic or expensive (you want to be sure before starting)
- Confirmatory testing after a positive screening test
- Rule-in strategy in patients with suggestive clinical features
Example: Western blot for HIV has >99% specificity — a positive result is essentially diagnostic.
PPV and NPV are what clinicians actually use at the bedside — they answer 'given this test result, what is the probability my patient has the disease?'
- PPV (Positive Predictive Value): Probability that a positive test is a true positive
- NPV (Negative Predictive Value): Probability that a negative test is a true negative
Crucially, predictive values depend on disease prevalence in the tested population:
- In a high-prevalence population, PPV is high (a positive test is more likely to be true)
- In a low-prevalence population, PPV falls dramatically (most positives are false positives)
- Example: An HIV test with 99% sensitivity and 99% specificity has PPV of 50% in a low-prevalence population (1 in 1000) but 99% in a high-prevalence population (1 in 10)
This is why screening tests for rare diseases are problematic — even excellent tests generate too many false positives.
Likelihood ratios (LR) combine sensitivity and specificity into a single metric that is independent of prevalence:
- LR+ (Positive Likelihood Ratio): Sensitivity / (1 - Specificity). Values >10 strongly support diagnosis.
- LR- (Negative Likelihood Ratio): (1 - Sensitivity) / Specificity. Values <0.1 strongly rule out diagnosis.
LRs allow clinicians to combine pre-test probability (clinical suspicion) with test result to get post-test probability using Bayes' theorem. This is the modern way to use diagnostic tests at the bedside, replacing rigid cutoffs.
Sensitivity and specificity are the conceptual core of diagnostic test interpretation. For UPSC CMS aspirants, the 2x2 table, the four formulas, the SnNOut/SpPIn mnemonics, and the effect of prevalence on PPV are among the most frequently tested PSM topics.