Hacker News Reader

Comment by xiphias2 | Hacker News Reader

Data availability got much better in the past following years (as a software engineer I can finally do modelling based on data published by medical researchers), but it's still too slow, which costs real lives.

,,Data underlying the study cannot be made publicly available due to ethical concerns about patient confidentiality''

I don't believe that any part of the data couldn't be made available (D-vitamin level - age - covid test results at least, even if location/time would be too sensitive to share). Not sharing any data has ethical concerns as well.

View on HN

I agree! It seems like it could have been anonymized. The government site I check regarding covid news that is local to me has the person’s age, sex, test results, and county. I don’t see how vitamin d levels could be an identifier.

Just venting - data during this pandemic has been one of the most frustrating parts for me. Data should be one of the tools we can use to get through this, but it’s been a struggle knowing what’s accurate and lack of certain information makes it hard to make good decisions.

One interesting thing is that the Russian vaccine tests were criticized for not sharing raw data by lots of researchers from western countries as well.

I'm definately taking vitamin D and not signing up for the Russian vaccine, but it seemed hypocrytical to me that only Russian researchers were shamed for not being transparent.

"One interesting thing is that the Russian vaccine tests were criticized for not sharing raw data"

I'm sure that, if one searched hard enough, one could find such a criticism. However, most of the criticism centered around the fact that the vaccine was "approved" on the basis of nearly zero actual human data.

I saw only 1 open letter signed by lots of researchers which contained many problems with the experiment: the main one for me was that the results that were supposed to be from independent experiments were extremely highly correlated. The reason behind that correlation can be not enough original data (cheating) or bad processing, but the western researchers asked for the original data to find it out.

Anonymizing patient data is quite difficult. Try a Google search:

"reidentifying" patient data

come on, it would be pretty easy to find out where it happened, who was treated and who was not treated.

After you uncloak everything, you just send lawyers to the untreated people (the families of those who died) and start litigating.

or you just let people know their loved ones didn't get the life-saving treatment and let them stew a while.

> It seems like it could have been anonymized.

But maybe the consent was not given for such a usage. Even if anonymized, we must respect the guise under which the data was provided to the researchers.

The really interesting data would be comorbidities that cause low Vitamin D. Those are (independent?) risk factors for severe COVID.

Wouldn't that expose more correlations that aren't proven to be causative?

There's some weird correlation in everything. That's half the numerology. Various correlations can also be useful in meta analysis (if you find some trend across many studies you can run a test for it specifically). Not publishing the raw data just in case someone doesn't know how to interpret the data is not a solution to anything.

> There's some weird correlation in everything.

For anyone who hasn't seen it yet: http://tylervigen.com/spurious-correlations

> Wouldn't that expose more correlations that aren't proven to be causative?

What is wrong with "exposing more correlations"? I don't get how that's an argument against data release.

Reply on news.ycombinator.com