1978 study by Casscells showed that physicians were dismal at using PPV test characteristic (positive predictive value, Bayes estimation) for assessing the value of a laboratory test. It was repeated this year with similar disappointing results. Both studies posed a simple question on application of laboratory test results given the test characteristics. Here is the question both studies used. Only 25% got it right. The others were spectacularly wrong.
“If a test with perfect sensitivity to detect a disease whose prevalence is 1/1000 has a false positive rate of 5%, what is the chance that a person found to have a positive result actually has the disease, assuming you know nothing about the person’s symptoms or signs?”
Using PPV and Bayes Formula
There are several ways to solve this problem. Shown below is how calculate the answer using PPV and Bayes formula.
While understanding the math is important, you can develop a more intuitive understanding by visualizing the population (see figure below). If a draw a sample of 1000 patients from the population (all of the dots), only one will have the disease (red dot). Since the test has a 100% sensitivity, it will be a true positive. The remaining 999 patients do not have the disease, but 50 of them will test positive (blue dots), and the rest will test negative (black dots).
Notice that the false positive blue dots far out number the true positives. If this hypothetical example, those patients would be subjected to unnecessary further workup or treatment.