StudyRareStudyRare

Positive and Negative Predictive Value

Log in to star

Last updated 2mo ago

Log in to add personal notes on this page.

Predictive values answer the clinically relevant question: given a test result, what is the probability that the patient actually has (or does not have) the condition? Unlike sensitivity and specificity, predictive values depend on disease prevalence.

  • Positive Predictive Value (PPV) = TP / (TP + FP) = probability of disease given a positive test
  • Negative Predictive Value (NPV) = TN / (TN + FN) = probability of no disease given a negative test
  • PPV increases as prevalence increases (more true positives relative to false positives)
  • PPV decreases as prevalence decreases (false positives overwhelm true positives)
  • NPV is generally high when prevalence is low (most negatives are true negatives)

This is why screening tests perform differently in high-risk vs low-risk populations.

Example 1: NIPT in high-risk vs low-risk

High-risk population (prevalence 1/100):

T21+ (100)T21- (9,900)
Screen+9950
Screen-19,850

PPV = 99/149 = 66%

Low-risk population (prevalence 1/1,000):

T21+ (10)T21- (9,990)
Screen+9.950
Screen-0.19,940

PPV = 9.9/59.9 = 17%

Same test, same sensitivity/specificity, but PPV drops from 66% to 17% because prevalence is lower.

Example 2: Why confirmatory testing is needed

Problem: A 25-year-old with no risk factors has a positive NIPT for trisomy 18. Prevalence at her age is approximately 1/5,000. Test sensitivity is 96%, specificity is 99.5%. What is the PPV?

Solution:

  • In 100,000 pregnancies: 20 with T18, 99,980 without
  • True positives: 20 x 0.96 = 19.2
  • False positives: 99,980 x 0.005 = 499.9
  • PPV = 19.2 / (19.2 + 499.9) = 3.7%
  • Over 96% of positive screens at this prevalence are false positives
  • This is why diagnostic confirmation (amniocentesis/CVS) is essential after a positive screen
  • A positive screening test in a low-prevalence population is more likely to be a false positive than a true positive
  • Diagnostic tests (amnio, CVS) have near-100% sensitivity and specificity, so PPV approaches 100%
  • When counseling a patient about a positive screen, always contextualize the PPV based on their prior risk

"Prevalence Powers PPV": PPV is driven by prevalence. Low prevalence = low PPV, even with an excellent test. Always consider the prior probability (age, family history, ultrasound findings) when interpreting a positive screen.