USMLE Step 1Biostatistics and epidemiologyEP 02 of 08

PPV, NPV and prevalence explained

A test can be 90% sensitive and 90% specific and still be wrong most of the time when it says positive, if the disease is rare. With the test fixed, prevalence moves the predictive values.

  • 5:27 video
  • 3 min read
  • 4 sources
  • Published 16 Nov 2026
Chapters (10)
By the end of this lesson

Calculate positive and negative predictive values from a 2×2 table and predict how they change with disease prevalence.

The short version

  1. Positive predictive value (PPV) = TP / (TP + FP); negative predictive value (NPV) = TN / (TN + FN).
  2. Sensitivity and specificity are read down the columns of the 2×2 table; predictive values are read across the rows.
  3. Prevalence is the fraction of the population that has the disease.
  4. As prevalence rises, PPV rises and NPV falls.
  5. Sensitivity and specificity stay the same; the predictive values change with the population.

A positive result from a good test: is the disease real?

A test is 90% sensitive and 90% specific. Your patient tests positive. Is there a 90% chance the disease is real? Not even close, if the disease is rare.

This lesson covers PPV, NPV and prevalence: how to calculate positive and negative predictive values from a 2×2 table, and how they move as prevalence changes.[1][3]

Columns versus rows: sensitivity and specificity, then PPV and NPV

Sensitivity and specificity are read down the columns of the 2×2 table. They start from the truth, sick or healthy, and ask how the test behaves.[1][4]

Predictive values look from the other side. The person holding the result sees the result first and asks what it means. So predictive values are read across the rows.[1][2]

Put simply: the columns hold the truth (condition present or absent), the rows hold the test result (positive or negative). The same four cells give both pairs of measures, read in different directions.[1]

PPV, NPV and prevalence defined

Positive predictive value (PPV)
PPV = TP / (TP + FP)

Of everyone who tests positive, the fraction who truly have the disease.

Negative predictive value (NPV)
NPV = TN / (TN + FN)

Of everyone who tests negative, the fraction who truly do not have the disease.

Prevalence is the fraction of the population that has the disease. PPV and NPV both depend on it.[1][3]

Notice what is in each denominator. PPV divides by every positive result, true and false. NPV divides by every negative result, true and false. That is why both change when the mix of sick and healthy people changes.

Building the 2×2 table three times with the same test

Keep the test fixed at 90% sensitivity and 90% specificity. Only the population changes: 1000 people, with prevalence 1%, 10% and 50%.[1]

  1. Split the populationMultiply 1000 by the prevalence to get the sick; the rest are healthy.
  2. Fill the sick columnMultiply the sick by the sensitivity to get those who test positive (TP); the remainder are false negatives (FN).
  3. Fill the healthy columnMultiply the healthy by the specificity to get those who test negative (TN); the remainder are false positives (FP).
  4. Read across the rowsPPV from the positive row, NPV from the negative row.
Prevalence 1%: 10 sick, 990 healthy
  1. Sick: 90% of 10 test positive, so TP = 9 and FN = 1
  2. Healthy: 90% of 990 test negative, so TN = 891 and FP = 99
  3. PPV = 9 / (9 + 99) = 9 / 108 ≈ 8.3%
  4. NPV = 891 / (891 + 1) = 891 / 892 ≈ 99.9%
Prevalence 10%: 100 sick, 900 healthy
  1. TP = 90, FN = 10, TN = 810, FP = 90
  2. PPV = 90 / (90 + 90) = 90 / 180 = 50%
  3. NPV = 810 / (810 + 10) = 810 / 820 ≈ 98.8%
Prevalence 50%: 500 sick, 500 healthy
  1. TP = 450, FN = 50, TN = 450, FP = 50
  2. PPV = 450 / 500 = 90%
  3. NPV = 450 / 500 = 90%

With this test at 1% prevalence, a positive result is wrong more than nine times out of ten: 99 of the 108 positives are false.

The pattern: why PPV rises and NPV falls as prevalence rises

Same test (90% sensitivity, 90% specificity), three populations of 1000
PrevalencePPVNPV
1%8.3%99.9%
10%50%98.8%
50%90%90%

As prevalence rises, PPV rises and NPV falls.[1][3]

Look at the table above. Between 1% and 50% prevalence, PPV climbs from about 8.3% to 90% while NPV slides from 99.9% to 90%. At 10% prevalence, PPV is exactly 50%: half of the positive results are true and half are false.

The reason is simple. The test still wrongly flags the same 10% of healthy people. When almost everyone is healthy, those false alarms swamp the few true cases. When many people are sick, the true cases outnumber the false alarms.

What to keep from PPV, NPV and prevalence

  1. PPV is true positives over all positives, and NPV is true negatives over all negatives.
  2. When prevalence rises, PPV rises and NPV falls.
  3. Sensitivity and specificity stay the same; the predictive values change with the population.

Practice question

Original · USMLE style
Your turnQB-MS-S1-BIOS-002-01

A new blood test for a chronic condition has a sensitivity of 60% and a specificity of 95%. It is used to screen 1000 adults attending a clinic, in which the prevalence of the condition is 20%. A 47-year-old man from this clinic population tests positive.

Which of the following is the probability that this man has the condition?
Pick one option. Your answers stay on this device.

Common questions

What is the difference between PPV and sensitivity?

Sensitivity is read down the column of people who are sick and asks how the test behaves. PPV is read across the positive row and asks what fraction of positive results are truly sick.

How do you calculate PPV and NPV?

PPV = TP / (TP + FP) and NPV = TN / (TN + FN). Build the 2×2 table from sensitivity, specificity and prevalence in a population of 1000, then read across the rows.

Why does PPV fall when prevalence is low?

The same false-positive rate is applied to a large healthy group, so false positives swamp the few true cases. At 1% prevalence with a 90% sensitive, 90% specific test, PPV is about 8.3%.

Do sensitivity and specificity change with prevalence?

No. In this lesson they stay fixed at 90% while PPV and NPV change with the population.

Transcript

Read the full narration, by chapter621 words

0:00The hook: a 90% sensitive, 90% specific test says positive

A test is 90% sensitive and 90% specific. Your patient tests positive. Is there a 90% chance the disease is real? Not even close, if the disease is rare. Today: calculate positive and negative predictive values, and predict how they move with prevalence.

0:19Quick reminder: sensitivity down the columns, predictive values across the rows

Quick reminder from the last lesson. Sensitivity and specificity are read down the columns of the two by two table. They start from the truth, sick or healthy, and ask how the test behaves. A clinician stands on the other side. She sees a result first, and asks what it means. So predictive values are read across the rows.

0:42PPV, NPV and prevalence defined

Positive predictive value: of everyone who tests positive, what fraction truly has the disease? True positives divided by true positives plus false positives. Negative predictive value: of everyone who tests negative, what fraction truly does not? True negatives divided by true negatives plus false negatives. The third ingredient is prevalence: the fraction of the population that has the disease.

1:11Table 1: prevalence 1% (PPV 8.3%, NPV 99.9%)

Now let us build the table, three times, with the same test. The test stays 90% sensitive and 90% specific. Only the population changes. First, a population of 1000, where prevalence is 1%. That is 10 sick people and 990 healthy people. Ninety % of the 10 test positive: nine true positives, and one false negative. Ninety % of the 990 test negative: 891 true negatives, and 99 false positives. Positive predictive value is nine over nine plus 99, nine over 108, about 8.3%. Negative predictive value is 891 over 892, about 99.9%. At 1% prevalence, a positive result is wrong more than nine times out of 10.

2:09Table 2: prevalence 10% (PPV 50%, NPV 98.8%)

Second, prevalence of 10%: a 100 sick, 900 healthy. Ninety true positives, 10 false negatives, 810 true negatives, 90 false positives. Positive predictive value is 90 over 180, exactly 50%. Negative predictive value is 810 over 820, about 98.8%.

2:36Table 3: prevalence 50% (PPV 90%, NPV 90%)

Third, prevalence of 50%: 500 sick, 500 healthy. Four 150 true positives, 50 false negatives, 450 true negatives, 50 false positives. Now both predictive values are 450 over 500, 90%.

2:57The pattern: why PPV rises and NPV falls with prevalence

Look at the pattern. As prevalence rises, positive predictive value rises, and negative predictive value falls. The reason is simple. The test still wrongly flags the same 10% of healthy people. When almost everyone is healthy, those false alarms swamp the few true cases. When many people are sick, the true cases outnumber the false alarms. Sensitivity and specificity did not move at all. Prevalence moved the predictive values.

3:29Practice question

Your turn. A clinic population of 1000 people is screened with a new test. The test has a sensitivity of 60% and a specificity of 95%. Twenty % of the clinic population has the condition. One patient tests positive. What is the probability that he has the condition? A, 20%. B, 60. C, 75. D, 90. E, 95. Pause here and decide.

4:04Answer and explanation

A is out: 20% is the prevalence, the chance before any test result. B is out: 60% is the sensitivity, which answers a different question. D is out: 760 over 840 is about 90%, the negative predictive value, which answers the question for a negative result. E is out: 95% is the specificity, again a different question. So the answer is C. Two 100 have the condition, a 120 are true positives; of 800 healthy, 40 are false positives; positive predictive value is 120 over 160, 75%.

4:51Recap and next lesson: relative risk and odds ratio

Three things to keep. One: positive predictive value is true positives over all positives, and negative predictive value is true negatives over all negatives. Two: when prevalence rises, positive predictive value rises and negative predictive value falls. Three: sensitivity and specificity stay the same, and the predictive values change with the population. Next, relative risk and odds ratio: two ways to compare risk between groups.

Sources

Numbers in the text, e.g. [1], point here
  1. Parikh R, Mathai A, Parikh S, Chandra Sekhar G, Thomas R. Understanding and using sensitivity, specificity and predictive values. Indian J Ophthalmol. 2008;56(1):45-50. pmc.ncbi.nlm.nih.gov/articles/PMC2636062/
  2. Trevethan R. Sensitivity, Specificity, and Predictive Values: Foundations, Pliabilities, and Pitfalls in Research and Practice. Front Public Health. 2017;5:307. pmc.ncbi.nlm.nih.gov/articles/PMC5701930/
  3. Positive and negative predictive values (Wikipedia, secondary). en.wikipedia.org/wiki/Positive_and_negative_predictive_values
  4. Sensitivity and specificity (Wikipedia, secondary). en.wikipedia.org/wiki/Sensitivity_and_specificity

All numbers on screen are teaching values: 90% sensitivity and 90% specificity at prevalence 1%, 10% and 50% in a population of 1000; the practice question uses its own values.

The practice question is original and its patient is synthetic. Every fact is checked against the sources above in independent accuracy reviews. Always follow your local guidelines and your own laboratory’s ranges.

Image credits. Illustrations: Servier Medical Art (smart.servier.com), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), adapted (cropped, recoloured, outlined): adult man. Health Icons (healthicons.org), CC0 1.0: medical sample tube. The result slip, tables, dot grid, graph, question card and all handwriting are drawn in code (original). No AI-generated teaching images.

Corrections. None so far. Spotted an error? Email info@medservatory.com with the lesson code MS-S1-BIOS-002.

For exam study only, not medical advice. Read the medical disclaimer.

  1. EP 01Sensitivity vs specificity: the 2×2 table, SnNout and SpPin5:34
  2. EP 02PPV, NPV and prevalence explained5:27
  3. EP 03Relative risk vs odds ratio: the 2x2 table explained5:15
  4. EP 04Number needed to treat: ARR, RRR and NNH from trial data5:00
  5. EP 05Study designs explained: case-control, cohort, cross-sectional and RCT5:22
  6. EP 06Types of bias: selection, recall, lead-time and confounding5:30
  7. EP 07Type I and type II errors: alpha, beta, power and the p-value5:16
  8. EP 08Confidence intervals and significance: is the null value inside?5:12