Predictive Values (PPV and NPV)

Definition and Formula

PPV and NPV express how accurately a test result reflects true condition status. The core formulas are: PPV = True Positive / (True Positive + False Positive); NPV = True Negative / (True Negative + False Negative). Both values range from 0 to 1 (or 0 to 100 percent). A higher PPV means a positive result more reliably indicates the condition; a higher NPV means a negative result provides stronger reassurance. The two values should always be interpreted together, not used in isolation as a single index of test quality.

Prevalence Dependence: The Most Overlooked Fact

Sensitivity and specificity are treated as fixed properties of the test and do not change substantially across populations. PPV and NPV, however, are directly tied to prevalence. For example, a test with 95 percent sensitivity and 95 percent specificity achieves a high PPV in a 50 percent prevalence setting, but PPV falls to roughly 16 percent when prevalence drops to 1 percent. This means that in low-prevalence screening programs, most positive results may be false positives. Researchers applying a test across different settings must account for differences in prevalence before drawing conclusions about predictive performance.

Common Misuses and Misconceptions

The most frequent error is reporting sensitivity instead of PPV, which overstates the clinical utility of a positive result. Another mistake is applying a PPV calculated in one population directly to a different population with a different prevalence. Some studies also declare a test near-perfect because its NPV is very high; but NPV rises automatically when the condition is rare, reflecting low prevalence rather than genuine test quality. Whenever predictive values are reported, the prevalence of the condition in the study sample must be explicitly stated so readers can assess generalizability.

Why It Matters and How to Report It

Predictive values are the real-world probabilities that a clinician or researcher needs to interpret a test result in practice. Even when a test has high sensitivity, its PPV will remain low in a low-prevalence setting, so positive results will mostly be false alarms. Recommended reporting includes: the prevalence of the condition in the study population; PPV and NPV each with a 95 percent confidence interval; sensitivity and specificity; and a clear statement of limitations related to prevalence. Altman and Bland (1994) laid out these principles concisely and remain the standard reference for this topic.

Sources