Hacker News Reader

Comment by sigmar | Hacker News Reader

I agree with your comment. Beyond the false premise, I think that this fivethirtyeight piece is extremely dangerous because it seems to prefer shocking headlines over actually depth. i.e. "We Don’t Need More Blood Tests" and the infographic "How accurate tests can be mostly wrong."

View on HN

What false premise? Fivethirtyeight is not wrong. In this case, with seemingly reliable tests (>90% accuracy), only 27% of positive tests are correct. If you're talking about something that has an invasive next step, like a biopsy, and you decide to increase the number of tests done each year, then you are going to also be increasing the number of unnecessary invasive procedures each year as well.

Take mammograms. It's a fairly non-invasive test for breast cancer. Everyone wants to reduce breast cancer, right? Unfortunately, the data is very much like fivethirtyeight's example. In a 10 year period, 1 in 2 women are harmed by a false positive and 1 in 5 were harmed by an unnecessary surgical procedure. What about the lives saved? None.[0] Screening mammograms only end up harming patients.

While it may make sense intuitively that more screening must be better, the data generally fails to back it up.

[0] http://www.thennt.com/nnt/screening-mammography-for-reducing...

If an exploratory test is done for a condition sufficiently rare that a positive is quite likely to be a false positive, then a positive result should not, in the absence of other evidence, lead to an invasive procedure. What action should be taken depends on the particulars of the condition, but just monitoring for actual symptoms of the condition or maybe trying a different test would almost certainly be more sensible options.

The article seems to be decrying additional testing, which is basically just additional information gathering, on the basis that some people might make dumb decisions based on the information. That's probably true, but I don't think it means that more information is a bad thing.

The case of mammography is often brought up for this argument, but is a cherry picked example. Furthermore, it isn't a blood test, so it isn't even what is being argued against by fivethirtyeight.

There is no harm done by extremely accurate testing for HIV. and less accurate tests should be used in screening by medical professionals that know how to proceed from a positive result. Again, blanket statements like "We Don’t Need More Blood Tests" are dangerous

given how 538 seemed to want to be positioned in the market place, I'm pretty disappointed with how similar their model of 'shocking headlines over actual depth' is to what are generally dismissed as low brow competitors

its not actually an informative news source

Reply on news.ycombinator.com