Positive predictive value explained simply

What the positive predictive value answers

After an abnormal test result, the same question almost always comes up: How likely is it that I actually have the condition. The positive predictive value answers this question. It is the proportion of people who truly have the condition among everyone with a positive result. Altman and Bland explained these predictive values clearly in the BMJ in 1994.

The direction of the question matters. Sensitivity and specificity start with the known condition status and ask how the test performs in people with and without the condition. The positive predictive value reverses the direction. It starts with the test result and asks what a positive result represents. For someone who has a test result in hand, this is often the more relevant measure because their true condition status is not yet known.

Why prevalence matters

The key point is that the positive predictive value does not depend on test quality alone. Two tests with exactly the same sensitivity and specificity can have very different positive predictive values depending on how common the condition is in the group being studied. This frequency is the prevalence.

The reason is straightforward. If a condition is rare, there are many people who do not have it. Even if a test incorrectly marks only a small proportion of these people as positive, this large group produces many false-positive results in absolute numbers. At the same time, there are only a few people with the condition who can produce true-positive results. False positives can then make up the majority of all positive results. If the condition is common, the ratio changes, and a positive result becomes more informative.

A simplified calculation example

The following numbers are hypothetical and are only for illustration. They do not come from a study. Suppose a test has a sensitivity of 95 percent and a specificity of 90 percent. These two values remain the same in both cases. Only the frequency of the condition changes.

First, a rare condition. In a group of 10,000 people, 100 actually have the condition, which is 1 percent, and 9,900 do not.

The test identifies 95 of the 100 affected people (95 percent sensitivity) and correctly identifies 8,910 of the 9,900 unaffected people (90 percent specificity). This leaves 990 false-positive results. In total, 95 plus 990, or 1,085 people, test positive. The positive predictive value is 95 divided by 1,085, which is about 9 percent. So only about one in eleven people who test positive actually has the condition, even though the test performs well on both measures.

Now consider the same test quality with a common condition. Among 10,000 people, 2,000 have the condition, which is 20 percent. The test correctly identifies 1,900 of the 2,000 affected people as positive. Of the 8,000 unaffected people, 800 are incorrectly flagged as positive. In total, 1,900 plus 800, or 2,700 people, test positive. The positive predictive value is now 1,900 divided by 2,700, which is about 70 percent. The same test, with the same sensitivity and specificity, has a positive predictive value that rises from about 9 to about 70 percent only because the condition is more common.

Common misinterpretation

The positive predictive value is often confused with sensitivity. They are different measures. A sensitivity of 95 percent means that 95 percent of people with the condition receive a positive result. It says nothing about how many people with a positive result actually have the condition. That is what the positive predictive value measures, and as the example shows, it can be much lower. A test can reliably identify people with the condition and still produce mostly false-positive results in a group where the condition is rare. Because the positive predictive value depends on prevalence, a value applies only to the group in which it was measured. A value from a clinical group with many affected people cannot be transferred directly to the general population.

What this means for self-tests on medtests.net

This relationship is why a positive screening result in the general population is more often a false positive than in an already selected group. If you take a test because you already have relevant symptoms, you belong to a group in which the condition is more common, so a positive result is more informative. If you take the same test without a specific reason in the broader population, where the condition is rare, an abnormal result is more likely to be a false positive.

This is why we word results cautiously. With the Autism Test using the AQ-10 and the Depression Test using the PHQ-9, an abnormal result is a sign that further assessment may be useful, not proof of a condition. With the Anxiety Test using the GAD-7, the same applies. A positive screening result proves nothing. It only indicates that a closer assessment may be worthwhile. We provide specific performance figures for an instrument only in relation to the relevant validation study and the group studied there.

What the term does not mean

The positive predictive value is not a property of the test alone. It always applies to a specific group with a specific prevalence. A high positive predictive value in a study group does not mean that a positive result for you has the same probability of being a true positive. No predictive value turns a screening into a diagnosis. A self-test provides orientation and does not replace medical, psychotherapeutic or diagnostic assessment.