4 ms·
There's also a statistical problem. Let's say 1% of patients have X, and you've got a test for X which is 99% accurate. You screen 10,000 patients; 100 have X
by michaelt 3y ago
There's also a statistical problem.
Let's say 1% of patients have X, and you've got a test for X which is 99% accurate. You screen 10,000 patients; 100 have X and you identify 99 of them. 9,900 don't have X, and you raise 99 false alarms.
But if you expand screening to cover Y which 0.1% of patients suffer from, and your test for Y is equally accurate, you'll identify 10 patients with Y - and get 100 false alarms.
So the rarer the thing you're screening for, the better your test needs to be - unless you want to be buried in false alarms.
As I understand things, this is why we don't generally screen for prostate cancer in the under-50s.
- boxed 3y agoThat's not a statistical problem, it's a problem of doctors not knowing statistics. My point is that you could do these scannings on a broad scale and not look at the result per patient. Don't give it to their doctor to look at by default as that is the problem. Use it for basic science FIRST.
- thfuran 3y agoAnd sometimes the test itself can have rare adverse outcomes even ignoring the issues with false positives.
- zo1 3y agoI may be off base. But it sounds like the accuracy rate of those tests is off if it's based on an assumption that you'll only test people who have risk factors. I would have thought that by now, being the year 2024, we've gotten better than just guessing and statistically inferring stuff. This is where the big disconnect is, and why so many are starting to distrust the medical field. We assume doctors can diagnose stuff and not just go around making educated guesses. No wonder it's a giant government-enforced cartel that restricts competition.