1. Positive Predictive Value and Negative Predictive Value
Study Outcomes Mini-Presentation
Created by Thomas Hahn, MD
Edited by John Kloke, PhD

2. Objectives Following this presentation, the learner will be able to:
Understand the concepts and calculations for positive predictive value and negative predictive
Apply PPV and NPV to patient care
Communicate PPV and NPV to patients

3. UNDERSTAND the CONCEPT

4. PPV and NPV
Positive Predictive Value: Probability that subjects with a positive test have the disease.
Negative Predictive Value: Probability that subjects with a negative test do not have the disease.

5. Positive Predictive Value
Positive Predictive Value: Proportion of positive test results that are true positives
PPV = True Positives x 100%
All Positives

6. Negative Predictive Value
Negative Predictive Value: Proportion of negative test results that are true negatives
NPV = True Negatives x 100%
All Negatives

7. You need to know these equations for boards!
2x2 Table
You need to know these equations for boards!
Disease Present
No Disease a
(true positive) b
(false positive, Type 1 error) c
(false negative, Type 2 error) d
(true negative)
Positive
Test
PPV = a/(a+b)
Negative
Test
NPV = d/(d+c)
Sensitivity and specificity focus on the DISEASE: total number of people with or without the disease makes up the denominator
Positive and negative predictive values focus on the TEST: total number of people with positive or negative tests makes up the denominator
Sensitivity = a/(a+c) Specificity = d/(d+b)
PPV = true positives x100% NPV = true negatives x100%
(true positives + false positives) (true negatives + false negatives)

8. PPV/NPV Characteristics
Predictive value is determined by the sensitivity and specificity of a test
Sensitivity (low false negatives)  NPV
Specificity (low false positives)  PPV
Predictive value is determined by the prevalence of a disease in the population
Prevalence  PPV, NPV
Prevalence  PPV, NPV
As prevalence of a disease in a population increases, PPV increases and NPV decreases
As prevalence of a disease in a population decreases, PPV decreases and NPV increases
Bayes’ theorem connects PPV/NPV with Sn/Sp

10. Example of Effect of Prevalence on PPV/NPV
Hypothetically test 10 people for Lyme disease in Lyme, Connecticut (high prevalence) and Boise, Idaho (low prevalence)
Lyme, Connecticut
9/10 positive tests. High likelihood that the tests are truly positive (Prevalence  PPV, NPV)
Boise, Idaho
9/10 negative tests. High likelihood that the tests are truly negative (Prevalence  PPV, NPV)

11. Predictive Value and Low Disease Prevalence
How can you increase predictive value of a test when there is a low preclinical disease prevalence?
Example: Breast cancer
Prevalence is low in the population (about 1% in 2012)
To increase PPV, target screening to those at highest risk of developing the disease: women > 50 years old
Prevalence data based on about 3 million cases of breast cancer in 2012, when US population was around 314 million.

12. PPV/NPV vs Sensitivity/Specificity
Sensitivity and specificity
Focus on the disease
Important to epidemiologists and clinicians
Positive and negative predictive values
Focus on the test
Important to patients
The denominator in sens/spec is the total number of people w/ and w/out disease
The denominator in PPV/NPV is total number of positive and negative tests

13. 2x2 Table
Disease Present
No Disease a
(true positive) b
(false positive, Type 1 error) c
(false negative, Type 2 error) d
(true negative)
Positive
Test
PPV = true positives
all positives
Negative
Test
NPV = true negatives
all negatives
Sensitivity and specificity focus on the DISEASE: total number of people with or without the disease makes up the denominator
Positive and negative predictive values focus on the TEST: total number of people with positive or negative tests makes up the denominator
Sensitivity = true positives Specificity = true negatives
all with disease all without disease

14. APPLY the OUTCOME

15. Using Predictive Value
A PPV helps to answer the question: “If the patient’s test is positive, what are the chances that the patient has the disease?”
Predictive values may make more sense to the patient than sensitivity/specificity
Explains how many people with a positive test are truly positive vs how many people with the disease will test positive.

16. Application Example
The negative predictive value of a screening test for Lyme disease is 50%? What does this mean?
This is not a good test for ruling out Lyme disease. 50% of people with a negative test will actually have Lyme disease!!!

17. COMMUNICATE the DATA

18. Communicating to Patients
A patient presents with a tick bite, and you decide to test for Lyme disease using a new serum test. The patient asks you, “How good is this test?” Knowing that the PPV is 85%, how do you answer this question?
“If your test is positive, there is an 85% chance that you have the disease.” (test is 85% accurate)
“Out of all the positive tests, 85% will be correct.”
“If 100 people have a positive test, 85 of those will have the disease and 15 will not have the disease.”

19. Practice Question
A study evaluating a two-tier screening test for Lyme disease in a low prevalence area tested 4723 patients for Lyme disease. Of the patients who had positive tests, 12 had Lyme disease and 58 did not have Lyme disease. What is the positive predictive value of the test in this low prevalence area?
A. 17%
B. 20%
C. 1%
D. 15%
Disease
No Disease
Positive Test 12 58
Negative Test
4653
Lantos, P.M. et al. (2015). Poor positive predictive value of lyme disease serologic testing in an area of low disease incidence. Clinical Infectious Diseases, 584.

20. Practice Question
A study evaluating a two-tier screening test for Lyme disease in a low prevalence area tested 4723 patients for Lyme disease. Of the patients who had positive tests, 12 had Lyme disease and 58 did not have Lyme disease. What is the positive predictive value of the test in this low prevalence area?
A. 17% (12/70 = PPV)
B. 20% (12/58)
C. 1% (70/4653)
D. 15%
Disease
No Disease
Positive Test 12 58
Negative Test
4653

21. The worst predictive value?

22. References AAFP Family Medicine Board Review Course