The PPV and NPV of diagnostic tests are often quoted but what does this really mean?
The sensitivity and specificity are intinsic properties of the test. The sensitity is how good the test is at identifying someone with the disease. The specificity is how good the test is at identifying that someone DOES NOT have the disease.
In contrast the PPV and NPV are dependent on prevalence:
- PPV predicts how confident we are that a positive test has detected disease.
- NPV predicts how confident we are that a negative test can exclude disease.
When thinking about this issue it may be easier to have a picture to work from. For example, you are trying to pick red balls out of a bag and you have a machine that correctly identify a red ball (disease) 90% of the time and a yellow ball (healthy) 80% of the time (it therefore has an 90% sensitivity and 80% specificity).
You draw 2 balls out of this bag and you get:
- 2 red balls
Then you take the machine over to this bag and draw 2 balls but this time you get:
- 1 red ball
- 1 yellow ball
This bag has a disease prevalence of 75%
This bag has a disease prevalence of 25%
So what happened, why is the machine worse at picking red balls out of the second bag? Because it’s not a perfect test, it’s harder to find red balls in the second bag. If the sensitivity and specificity of a test were both 100% then the prevalence wouldn’t matter because the test is never wrong.
Because the test can be wrong, it’s more likely to be wrong when it’s picking something hard to find, when there is a low prevalence.
The red balls represent the people whose disease we are interested in finding, in this case – advanced fibrosis of the liver. The bag on the left are the patients in the hepatology clinics – most of these people have advanced liver disease. But the bag on the right is the general community where most people DO NOT have advanced fibrosis.