Testing Error

To the heart and mind
Ignorance is kind
There's no comfort in the truth
Pain is all that you'll find
--Wham

Instructive article on CV19 testing error. Any measurement system will have at least some Type I and Type II error. Type I error creates 'false positives' which, in the CV19 context, leads to the conclusion that people have the virus when they really don't. Type II errors, or 'false negatives,' leads to the conclusion that people aren't infected when they really are.

The author pulls an estimate from a recent article in the New England Journal of Medicine for the sensitivity of CV19 tests of about 70%. Sensitivity is the likelihood that a test will detect the disease. Why is the test only 70% sensitive? Part of the reason is that current methods test for only fragments of the virus rather than the complete virus. A test for parts is not as accurate as a test for the entire thing.

Suppose 100 people who actually have CV19 are evaluated with a test with 70% sensitivity. Only 70 infections will be detected. There will be false negatives with 30 people (i.e., they had the disease but the test didn't pick the infection up).

The author goes on to assume a specificity (i.e., the capacity of a test to correctly identify those without the disease of 70%). Thus, if 100 people without the disease are tested, 30 will be detected as being infected even thought they weren't.

It is easy to see the error of assuming that a positive test equates to a true infection. Yet, that is precisely what we are doing. He demonstrates the folly in this approach using current testing protocols of professional sports teams. If they are using a test that is no better than 70% sensitive and specific, then eventually all players will test positive--even if they are not truly infected. After quarantine, they could again falsely test positive.

The author's point is that tests with low sensitivity and specificity are not practically useful, especially at low virus prevalence. Given current state of the art, he suggests that diagnostic testing has done more harm than good due to the false alarms that it creates.

In his view, a better indicator are in the hospital--in ICUs specifically--because of a clinical diagnosis of pneumonia under the assumption that any case of viral pneumonia is coronavirus-related.

That measure would certainly make the virus look much less alarming--something many people do not want. For them, it would seem that the more testing error, the better.