Positive and Negative Predictive Value (PPV & NPV)

01Core Definitions and the 2x2 Contingency Table

Every question on this topic begins with the 2x2 contingency table. You must be able to reconstruct it from memory and populate it rapidly. The four cells represent all possible outcomes when a diagnostic test is applied to a population.

From this table, the four foundational metrics are derived:

Sensitivity (True Positive Rate) is the proportion of truly diseased individuals whom the test correctly identifies as positive. It answers: "Among all sick people, how many tested positive?"

$$\text{Sensitivity}=\frac{TP}{TP+FN}\backslash text\{Sensitivity\}\; =\; \backslash frac\{TP\}\{TP\; +\; FN\}$$

Specificity (True Negative Rate) is the proportion of truly healthy individuals whom the test correctly identifies as negative. It answers: "Among all healthy people, how many tested negative?"

$$\text{Specificity}=\frac{TN}{TN+FP}\backslash text\{Specificity\}\; =\; \backslash frac\{TN\}\{TN\; +\; FP\}$$

These two metrics are intrinsic properties of the test itself. They do not change when you move the test from one population to another. This is the critical distinction that exam questions exploit.

Positive Predictive Value (PPV) is the proportion of all positive test results that are true positives. It answers: "Given that this patient tested positive, what is the probability they actually have the disease?"

$$\text{PPV}=\frac{TP}{TP+FP}\backslash text\{PPV\}\; =\; \backslash frac\{TP\}\{TP\; +\; FP\}$$

Negative Predictive Value (NPV) is the proportion of all negative test results that are true negatives. It answers: "Given that this patient tested negative, what is the probability they are truly disease-free?"

$$\text{NPV}=\frac{TN}{TN+FN}\backslash text\{NPV\}\; =\; \backslash frac\{TN\}\{TN\; +\; FN\}$$

PPV and NPV are not intrinsic to the test. They shift with the prevalence of disease in the population being tested. This is the single most tested concept in this topic.

02Bayes' Theorem and Prevalence Dependency

The relationship between predictive values and prevalence is formalized through Bayes' theorem. This is the mathematical backbone of every exam question that asks you to calculate or reason about PPV and NPV.

$$\text{PPV}=\frac{\text{Sensitivity}\times \text{Prevalence}}{(\text{Sensitivity}\times \text{Prevalence})+((1-\text{Specificity})\times (1-\text{Prevalence}))}\backslash text\{PPV\}\; =\; \backslash frac\{\backslash text\{Sensitivity\}\; \backslash times\; \backslash text\{Prevalence\}\}\{(\backslash text\{Sensitivity\}\; \backslash times\; \backslash text\{Prevalence\})\; +\; ((1\; -\; \backslash text\{Specificity\})\; \backslash times\; (1\; -\; \backslash text\{Prevalence\}))\}$$

$$\text{NPV}=\frac{\text{Specificity}\times (1-\text{Prevalence})}{(\text{Specificity}\times (1-\text{Prevalence}))+((1-\text{Sensitivity})\times \text{Prevalence})}\backslash text\{NPV\}\; =\; \backslash frac\{\backslash text\{Specificity\}\; \backslash times\; (1\; -\; \backslash text\{Prevalence\})\}\{(\backslash text\{Specificity\}\; \backslash times\; (1\; -\; \backslash text\{Prevalence\}))\; +\; ((1\; -\; \backslash text\{Sensitivity\})\; \backslash times\; \backslash text\{Prevalence\})\}$$

The denominator of the PPV formula contains the term $(1-\text{Specificity})\times (1-\text{Prevalence})(1\; -\; \backslash text\{Specificity\})\; \backslash times\; (1\; -\; \backslash text\{Prevalence\})$, which represents the false positive fraction. When prevalence is low, $(1-\text{Prevalence})(1\; -\; \backslash text\{Prevalence\})$ is large, meaning the pool of healthy individuals is enormous. Even a small false positive rate applied to this large pool generates many false positives, which dilute the PPV.

The core exam principle:

Think of it this way: in a population where almost nobody has the disease, a positive result is more likely to be a false alarm. Conversely, in that same low-prevalence population, a negative result is extremely reassuring because there was very little disease to miss in the first place.

Sensitivity and specificity also influence predictive values, but through a different mechanism:

A useful mnemonic: Specificity drives PPV upward (SPin = SPecificity rules In = confirms disease when positive). Sensitivity drives NPV upward (SNout = SeNsitivity rules Out = excludes disease when negative).

03Worked Calculation for Exam Application

This section walks you through a complete example the way exam questions present it. Being able to populate the 2x2 table from a word problem is the skill being tested.

Scenario: A screening test has 90% sensitivity and 95% specificity. It is applied to a population of 100,000 people where the disease prevalence is 1%.

Step 1: Determine the number of truly diseased and truly healthy.

$$\text{Diseased}=\mathrm{100,000}\times 0.01=\mathrm{1,000}\backslash text\{Diseased\}\; =\; 100\{,\}000\; \backslash times\; 0.01\; =\; 1\{,\}000$$

$$\text{Healthy}=\mathrm{100,000}\times 0.99=\mathrm{99,000}\backslash text\{Healthy\}\; =\; 100\{,\}000\; \backslash times\; 0.99\; =\; 99\{,\}000$$

Step 2: Apply sensitivity to the diseased group.

$$TP=\mathrm{1,000}\times 0.90=900TP\; =\; 1\{,\}000\; \backslash times\; 0.90\; =\; 900$$

$$FN=\mathrm{1,000}-900=100FN\; =\; 1\{,\}000\; -\; 900\; =\; 100$$

Step 3: Apply specificity to the healthy group.

$$TN=\mathrm{99,000}\times 0.95=\mathrm{94,050}TN\; =\; 99\{,\}000\; \backslash times\; 0.95\; =\; 94\{,\}050$$

$$FP=\mathrm{99,000}-\mathrm{94,050}=\mathrm{4,950}FP\; =\; 99\{,\}000\; -\; 94\{,\}050\; =\; 4\{,\}950$$

Step 4: Populate the 2x2 table.

Step 5: Calculate PPV and NPV.

$$\text{PPV}=\frac{900}{900+\mathrm{4,950}}=\frac{900}{\mathrm{5,850}}\approx 15.4\mathrm{\%}\backslash text\{PPV\}\; =\; \backslash frac\{900\}\{900\; +\; 4\{,\}950\}\; =\; \backslash frac\{900\}\{5\{,\}850\}\; \backslash approx\; 15.4\backslash \%$$

$$\text{NPV}=\frac{\mathrm{94,050}}{\mathrm{94,050}+100}=\frac{\mathrm{94,050}}{\mathrm{94,150}}\approx 99.9\mathrm{\%}\backslash text\{NPV\}\; =\; \backslash frac\{94\{,\}050\}\{94\{,\}050\; +\; 100\}\; =\; \backslash frac\{94\{,\}050\}\{94\{,\}150\}\; \backslash approx\; 99.9\backslash \%$$

Interpretation: Despite the test having excellent sensitivity (90%) and specificity (95%), the PPV is only about 15%. This means that roughly 5 out of every 6 positive results are false positives. The reason is that the 1% prevalence makes the healthy population overwhelmingly large, and even the 5% false positive rate applied to 99,000 healthy people generates 4,950 false positives, which swamp the 900 true positives.

This is the exact reasoning pattern that exam-writers want you to demonstrate.

04Strategies to Improve PPV in Clinical Practice

When a screening test yields a low PPV in a low-prevalence population, clinicians and public health systems use several strategies to raise the post-test probability before acting on the result. These strategies are frequently tested.

Serial testing is the most commonly examined strategy. Indonesian national guidelines for HIV diagnosis recommend serial testing (using two or three rapid tests sequentially) to reduce the risk of false-positive results. Research from a West Java referral hospital calculated that dual-testing algorithms nearly eliminate false-positive HIV results and are likely as accurate as three-test protocols, even in low-prevalence settings.

The logic is straightforward: after the first test identifies a pool of positives (which includes many false positives if prevalence is low), applying a second independent test to only that pool effectively raises the "prevalence" within the tested subgroup. The individuals entering the second test now have a much higher pre-test probability than the general population, so the second test's PPV climbs.

Indonesia's National Screening Programs and PPV Implications

Indonesia's Ministry of Health launched a free health screening program starting in 2025, offering age-appropriate medical check-ups at Puskesmas (community health centers) on each citizen's birthday. The program covers early detection of congenital diseases in infants, obesity and diabetes screening for adolescents, cancer screening for adults, and Alzheimer's screening for elderly patients.

From a biostatistics perspective, this is a textbook case of population-level screening. When you screen the entire population rather than high-risk subgroups, the prevalence in the tested population is lower, which means the PPV for any given test will be lower. This is why confirmatory testing protocols and risk-stratified screening algorithms are built into such programs.

A validation study of the NS1 rapid diagnostic test for dengue in Batam, Indonesia (2022 to 2024), found a PPV of 62.7% and an NPV of 88.9% when compared against RT-PCR as the gold standard. This illustrates a real-world example where a screening test in an endemic region yields a moderate PPV because not every febrile patient with a positive NS1 actually has dengue, and the specificity of the rapid test (81.2%) allows a meaningful number of false positives.

05Relationship to Other Diagnostic Metrics

PPV and NPV do not exist in isolation. Exam questions often require you to distinguish them from related concepts or to understand how they connect.

Post-test probability is conceptually identical to PPV when the test is positive, and conceptually identical to $(1-\text{NPV})(1\; -\; \backslash text\{NPV\})$ when the test is negative. Questions may use the term "post-test probability" instead of "PPV" to test whether you recognize they are the same concept.

Likelihood ratios are prevalence-independent and are used to convert pre-test probability into post-test probability via Fagan's nomogram or the odds form of Bayes' theorem:

$$\text{Post-test odds}=\text{Pre-test odds}\times \text{Likelihood Ratio}\backslash text\{Post-test\; odds\}\; =\; \backslash text\{Pre-test\; odds\}\; \backslash times\; \backslash text\{Likelihood\; Ratio\}$$

Where:

$$\text{Pre-test odds}=\frac{\text{Prevalence}}{1-\text{Prevalence}}\backslash text\{Pre-test\; odds\}\; =\; \backslash frac\{\backslash text\{Prevalence\}\}\{1\; -\; \backslash text\{Prevalence\}\}$$

06Differential Concepts and Distractors

07Traps and High-Yield Pearls

The single most common way students lose points on PPV/NPV questions is by assuming that a test with high sensitivity and high specificity will always produce reliable positive results. The exam exploits this by placing a good test in a low-prevalence screening scenario and asking "what is the PPV?" or "what proportion of positive results are false positives?" Students who do not account for prevalence will pick an answer close to 90% or 95% and be wrong.

The second trap involves confusing the direction of the conditional probability. Sensitivity asks: "Given disease, what is the probability of a positive test?" PPV asks the reverse: "Given a positive test, what is the probability of disease?" These are not the same question. Reversing the conditional without applying Bayes' theorem is known as the prosecutor's fallacy or the base rate fallacy, and it is a concept exam-writers deliberately exploit.

A third trap appears in questions about serial versus parallel testing. Serial (sequential) testing increases overall specificity and therefore raises PPV, but it decreases overall sensitivity because each additional test adds another chance to miss a true positive. Parallel testing (running two tests simultaneously) increases overall sensitivity and NPV but decreases specificity and PPV. Students who reverse these effects lose the point.

Finally, remember the directional rules as anchors for rapid elimination of wrong answers:

• When prevalence goes up, PPV goes up and NPV goes down.

• When prevalence goes down, PPV goes down and NPV goes up.

• Sensitivity and specificity do not change with prevalence.

• A "screening" test prioritizes sensitivity (you want to catch everyone, accepting more false positives).

• A "confirmatory" test prioritizes specificity (you want to be sure a positive is real, accepting that you might miss a few).

The core competency being tested is whether you understand that the clinical meaning of a test result depends on the context in which the test is applied, not just on the test's technical performance.