Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

>Or if all medical data was public, we'd have a medical revolution in understanding.

US medical billing data (which include who, where, what, why and when) is fairly easy to access (ie: 100+ million patients anonymized) and hasn't helped patients much from what I can tell. And many many companies have tried.

>Or if your Facebook likes were public, other competing websites could generate a network easier.

And oppressive governments can hunt down dissidents, hate groups can hunt down minorities, parents can disown their gay kids earlier, etc. Once you add up all the groups that make up "the few" you get rather a lot of people.



It’s probably not good enough to guess patient outcomes. It’s not like you could look up a surgeon and find how often the patient has a problem and then correct that against the risk pool of the patients


I mean you literally can. Billing data included the doctor, the procedures performed, the diagnosis given, patient demographics (age, gender, zip code, etc.). Plus a whole bunch more. Very well structured data actually although there's biases in it. My last company literally did what you said using the data. Not enough money in it so it all goes nowhere and all the big players in the space don't really care.

edit: You also have full medical billing history for each patient in the data set so you can build a historical risk profiles for patients.


That sounds like it would be a trivial to de-anonymize.

Also I’d be curious how good medical experts actually think the data is. In my field i find that data that looks legit is often actually bad enough to the point of near uselessness because the incentives aren’t aligned with collecting accurate data


The incentives are with collecting data that insurance companies will not reject your payments over. If an insurance company noticed you're mis-billing then you lose money. They may request full medical records of patients to verify. Which incentives generally clean data although there are biases such as up billing and spurious diagnosis to get medication covered. Medical records themselves are a mess often but billing data is better since a lot of money is on the line.

>That sounds like it would be a trivial to de-anonymize.

In theory yes if you joined to other identified data sets but in practice doing so makes it a massive HIPAA violation with very large government enforce penalties. There's also enough contractual sanity checking and costs that no one who gets access would bother.


I'd like to add that anonymized billing data has helped a lot of companies make a lot money. That however is separate from helping patients. Helping pharma companies better market medications to doctors makes a lot of money but is arguably a negative for patients.




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

Search: