Hacker News Reader

Comment by mg | Hacker News Reader

I think that aggregating medical research studies could lead to a significant improvement of the worlds health situation.

There are tens of millions of medical studies available online. But no system to look at the data in an aggregated way.

I have been doing data analysis and aggregation in several fields. Every once in a while, I dabble with the concept and implementation of a tool to do it for medical studies. It would need quite a bit of effort to get it right. But the potential is epic.

View on HN

Closely related, but not quite the same: personal genomics. If everybody could get completely personalized nutrition and medication (including preventive) we could be a lot healthier for a lot less expense. AIUI most of the fundamental technologies are here or close, but there needs to be a way to make that information more accessible/usable on the "front line" of patient care and that's where software comes into it.

Even more interesting would be, to automatically link patients medical history with studies, and show the doctor some suggestions: did you consider rare disease X?; there is not enough data, do test Y&Z because then we know much more

You mean like this https://blog.tjcx.me/p/40k-fantastic-startup-idea ?

I remember reading that post a while back.

I was surprised the author gave up after he didn't find doctors who would pay for access. The post says he met 10 doctors in the bay area and when none of them became a paying customer, he shut his project down.

If you really think you are making a dent in the worlds health situation, why give up so fast?

10 seems like such a small number. Was he was getting specific feedback that caused him to shut it down? Like were there some sort of legal liability issues or something that made it something they weren't interested in?

The mistake he made was selling directly to doctors, instead of to pharma, insurers or medical systems (Kaiser). Most individual doctors are not necessarily prioritized to provide better care; the other entities are.

Please remember that we do not have a perfect litmus for every disease, especially rare diseases. And statistics nearly falls apart entirely when assessing a single individual. So although i agree there is potential.. I believe the potential is more along the lines of "You have 15 patients in your practice with X, Y, Z active symptoms - it is incredibly likely that 1 one of them have A, or 3 of them have B"

There is significant reason to be hesitant about having an algorithm suggest every sickness under the sun that hasn't been "ruled out" by testing or statistics yet. Ever use WebMD to try to diagnose yourself?

In other words, I worry about what this approach would do the signal to noise ratio for doctors.

I know one of the programmers that built Origami for ORNL and the DoH

https://www.eurekalert.org/news-releases/913790

seems to go down this path. It is doable and they were able to beat IBM Watson to find some thing or another (_can't find the article off hand_) and won an R&D 100. But as you can tell, no one actually wants to use it.

I agree and for a start since most medical data come from hospital and outpatient systems -there is a very desperate need for something more modern and useable than a painful 'Epic' or 'Cerner' systems. Honestly how has silicon valley and newer tech not replaced those companies and their way of doing things?

I went down this road a few times. Tax laws, contract bidding, and money are the main reasons. Plus, writing such systems requires a large overlap of technology and medicine, which means a lot of high salaries on your payroll. Then you need to find places to buy it. That means getting hospitals and the like to break their usually multi-year contracts with Epic (they are largely subject to Sunken Cost problems as Epic costs a LOT of money).

And so on.

Reply on news.ycombinator.com