Negative Predictive Value

The negative predictive value (NPV) is the probability that a person who receives a negative test result truly does not have the condition being tested for.^[1][2][3] It is calculated as the number of true negatives divided by the total number of negative test results (true negatives plus false negatives) and is expressed as a proportion or percentage.^[2]

When a diagnostic test returns a negative result, the clinician and the patient need to know how much they can trust that result. The NPV directly answers this question: what is the chance that the patient is truly disease free given a negative test? This information guides clinical decisions about whether additional testing is needed, whether treatment can be withheld, and how to counsel patients about their status. The NPV is essential for evaluating screening programs, ruling out diagnoses, and assessing the clinical utility of laboratory tests, imaging studies, and rapid diagnostic devices.

NPV is calculated from a two by two table of test results and disease status. The columns represent disease present and disease absent; the rows represent test positive and test negative. NPV = true negatives / (true negatives + false negatives). It represents the post test probability of being disease free given a negative result.

A critical feature of NPV is that it depends not only on the test's intrinsic accuracy (its sensitivity and specificity) but also on the prevalence of the disease in the population being tested.^[1][2][3] When prevalence is low (most people being tested do not have the disease), there are many true negatives and few false negatives, so NPV is high. When prevalence is high, the proportion of false negatives among all negative results increases, and NPV decreases. This means that the same test can have a very different NPV depending on the clinical context.

The relationship between NPV, sensitivity, specificity, and prevalence can be expressed using Bayes' theorem: NPV = (specificity times (1 minus prevalence)) divided by ((specificity times (1 minus prevalence)) plus ((1 minus sensitivity) times prevalence)).^[2] This formula makes explicit how each factor contributes to the post test probability.

For screening purposes, where the goal is to rule out disease, a high NPV is particularly important. A negative result from a test with a high NPV provides strong reassurance that the disease is absent. The acronym "SnNOut" (Sensitivity, Negative, rule Out) captures the idea that a highly sensitive test with a negative result is good at ruling out disease, because high sensitivity raises the NPV, although the NPV still depends on prevalence.^[4]

Report NPV when evaluating the performance of a diagnostic or screening test, alongside sensitivity, specificity, and positive predictive value (PPV). Use it to assess whether a negative test result is reliable enough to guide clinical decisions. NPV is particularly important in settings where missing a true case (a false negative) has serious consequences, such as screening for cancer, infectious diseases, or life threatening conditions. Always report the prevalence in the study population because NPV is not generalizable across populations with different disease prevalence.^[2][3]

A rapid antigen test for influenza is evaluated in 2,000 patients presenting with respiratory symptoms during flu season (prevalence 20%). The test has a sensitivity of 85% and a specificity of 95%. Among the test results, there are 340 true positives (85% of 400 diseased), 60 false negatives, 1,520 true negatives (95% of 1,600 nondiseased), and 80 false positives. The NPV is 1,520 divided by (1,520 + 60) = 96.2%, meaning that 96.2% of patients with a negative result are truly influenza free. If the same test is used in a low prevalence setting (5% prevalence), the NPV rises to 99.2%, while in a high prevalence setting (40% prevalence), it drops to 90.5%.