This is a method of calculating the predictive value of a single test that gives one of two results, positive or negative, when used in a population with a known incidence of the disease for which the test is done.  A good example is of a test for diabetes that has a 10% rate of giving a false-positive (positive result but person does not have diabetes) result. If the test is given to a population that has only a 1% incidence of diabetes, nine out of 10 of the positive tests will be false-positives and its predictive value will only be 10%.

What is needed is to know the odds that a positive test is a true-positive or that a negative result is a true-negative.  That can be found by calculating the ratio of true results to the sum of true plus false results. In diagnosing HNP, that would be:

Predictive Value of Pos Test = true-pos/true-pos + false-positives


Predictive Value of Neg Test = true-neg/true-neg + false-neg

If the sensitivity of a test is 80% (20% false-neg)  and 70% of those tested are diseased, then the test will return 56% (true) positives.

If the specificity of that test is 90% (10% false-pos) then in that test group, in which 30% are normal, then 10% of 30% = 3% (false) positives

The PVpos will be 56/56+3 = 56/59 = 95%.

The Predictive Value depends mostly on the prevalence (percentage) of disease in the population tested.  One can see how doing MRI scans for HNP with a false-positive rate of 20% in a group of people in which only 10% have a symptomatic herniated disc, twice as many false-positive results will be found than true-positive ones.

    For a full description of the Predictive Value calculations, see ref 17.