3 ms·
That sounds less like the exams are the problem, and more like the incorrect handling of false positives is the problem.
by JoshTriplett 3y ago
That sounds less like the exams are the problem, and more like the incorrect handling of false positives is the problem.
- gruez 3y ago>and more like the incorrect handling of false positives is the problem. And what would be the "correct" way of handling it? Remember, you don't know that the test returned positive, not that it's a false positive or not. Are you going to tell the patient, "your breast cancer test returned positive, but given your age and other risk factors the probability that you actually have breast cancer is pretty low, so we won't do anything about it"? Even if that's the right thing to do from a statistical perspective, can you imagine the public outrage directed at the bureaucrat/politician that implemented this policy, if it turned out the patient did have breast cancer and died 2 years later?
- JoshTriplett 3y agoDefinitely not "we won't do anything about it". The medical industry (and frankly many other things but especially the medical industry) needs to find an effective way to present the common result "this test returned positive, but given the base rate in the population and the accuracy of this test, this is not necessarily enough information and your probability of having X is still only Y%, so we're going to run another non-invasive test to see if there's even enough information to warrant doing a more detailed check; ZZ% of people get a negative on the second test". Perhaps find a nice graphical presentation of the simplest version of Bayes Rule: "the base rate in the population is this, here's what the test means for the part of the population that doesn't have it, here's what the test means for the part that does, here's how your probability went up but that it's still vastly more likely you don't have it than that you do."