Hacker News Reader

Comment by perl4ever | Hacker News Reader

>a false positive here and there

Is that the right characterization? If everyone got a test every year, then there would be 3% false positives, per the article.

Brain cancer cases are around 6-7 per 100,000 per year.

So that means simply screening everyone would produce around 99.8% false positives, wouldn't it?

Or, looking at it another way, there would be 400 - 500 false positives for every real one.

"More data" implies that you use it for something. If there is some kind of intervention, and it almost always is not needed, then it has to be that much safer, in order to not make people worse off in aggregate.

This has long been a controversy with mammograms, I think.

View on HN

If the tests are cheap and can be performed as frequently as necessary, I could imagine a system where a test isn't considered positive until a statistically significant number of them have come back positive. If there is an abundance of data, there is no longer a need to extrapolate from a single data point.

I had a blood test once where my electrolytes were somewhat out of whack. I wasn't called in for an expensive kidney scan or consultation with a urologist. I was called in for another blood test. The second test came back fine. I was likely dehydrated the day of the first test.

The availability of cheaper, more frequent testing means that the response to a single positive signal does not have to be as extreme. Thankfully my doctors office offers walk-in blood testing or my experience could've been a much more anxiety-inducing hassle.

Are false positives always uncorrelated with each other?

I would think it would depend on the nature of what causes them.

True. I've heard of some tests coming back false positive or false negative from specific labs based on the way they did the test. I hear that Lyme disease is a particularly tricky one to get right (or has been in the past).

I've also been told that if you need tests done, it's best to get them dome during the week. Sometimes in some places the weekend team is the "B team" or techs on Friday afternoon just want to get done and make more mistakes. This seems to be similar to the research done about patient outcomes being affected by their surgeon's birthdays being on the procedure date, etc.

Well technically you should've had 2 follow up tests. Or 2-3 initial tests for actual data point.

I agree we need to improve this, yet we are in such an early age of medicine is basically a joke. There are nowhere near enough doctors to even scroll thru your tests, let alone to analyse them. The focus still is mostly on curing the most obvious symptomatic cases.

Also tests require tons of samples. IIRC there's over 2000 different blood tests. Unsure of minimum blood volume needed, but even at 1 ml that's already 30% of your total blood volume! I guess pee is easier lol.

Reply on news.ycombinator.com