Non-clinical sciences/Stats & evidence · Test characteristics

PPV & NPV

What it means

PPV = TP/(TP+FP) (of those testing positive, fraction truly diseased); NPV = TN/(TN+FN). Unlike sensitivity/specificity, both depend on prevalence (pre-test probability): even a very specific test has poor PPV in a low-prevalence population (Bayes).

Worked example

A specific screening test applied to a low-risk population yields many false positives → low PPV → confirmatory testing needed.

In the exam

After a positive result in a low-prevalence screening population, most positives turn out to be false because this measure falls as prevalence drops.

What settles it

PPV/NPV move with prevalence; sensitivity/specificity do not. Likelihood ratios are also prevalence-independent.

Classically confused with

Sensitivity

Source: StatPearls — Predictive values