Hacker News new | past | comments | ask | show | jobs | submit login

"More data points do not hurt a diagnoisis if a proper scientific method is followed."

Excess tests result in a higher number of false positives. This is just math. And those false positives have consequences.




> Excess tests result in a higher number of false positives

True, but this is an engineering problem:

1. Just because some tests have a higher number of false positives does not mean all tests have a higher number of false positives

2. For those tests that do have a higher number of false positives, the well educated (and paid) professionals are aware and should have measures in place

Eg: Bloom filters are fantastic datastructures to test membership. However, they have false positives. We can control the probability of getting a false positive by controlling the size of the Bloom filter.

This is nothing new in engineering.


Medicine is not engineering.

> Just because some tests have a higher number of false positives does not mean all tests have a higher number of false positives

No, but doing more tests leads to more false positives, which can lead to disastrous consequences. For example, someone could think they have aids through a false positive and lose their family, or choose to make an abortion, or start consuming preventive dangerous drugs, etc.

Not only that, you can bankrupt someone. It would surprise many how medicine is practiced differently based on available costs!

> For those tests that do have a higher number of false positives, the well educated (and paid) professionals are aware and should have measures in place

Yes, the measure is not to do them unless the symptoms and the DDX fit, along with other criteria.

I wonder what kind of doctor you've been to.


"1. Just because some tests have a higher number of false positives does not mean all tests have a higher number of false positives"

No - this is a math problem. As you do tests on low prevalence populations (inherently what you're doing when you say 'excess tests'), positive predictive value drops. Sure, it will drop more for worse tests, but this is a trait of all tests.

"2. For those tests that do have a higher number of false positives, the well educated (and paid) professionals are aware and should have measures in place"

One of the problems is distinguishing between "This thing is here" and "This is a clinical problem". For example, if I give you the (highly sensitive) PCR test for C. difficile, you may very well have C. difficile in your guts. But that doesn't mean that's what's making you sick.

Similarly, there are a number of cancer screening tests that will detect cancers that will kill you decades after something else does - including, potentially, the surgery to deal with said finding.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: