[Sniffnoy]

I think you're basically making a bucketing error here. As Scott Alexander points out, the exact same things he wanted to do were already being widely done elsewhere to no ill effect with no such worrying about consent. Why then does it make sense to group it under "research" together with the Tuskegee syphilis experiment and conclude things about it from that experiment, when the two have nothing in common except being research? I think that's a mistake of inference. Like, why should we conclude that we need that sort of consent to administer a questionnaire for experimental purposes (i.e.: gaining generally useful knowledge), but that there are no substantial issues of consent when it’s for diagnostic purposes (i.e.: gaining information about that particular patient) instead? Wouldn't it make more sense to group this case with the case with similar risk profile but a different purpose, rather than a related purpose but an entirely different risk profile? Something being research isn't what causes problems.

[euyyn]

> why should we conclude that we need that sort of consent to administer a questionnaire for experimental purposes [...], but that there are no substantial issues of consent when it’s for diagnostic purposes [...] instead?

A patient consenting to being treated doesn't mean they'd consent for that personal information to be used for any other purpose.

[vacri]

The interesting thing is that HN is usually a place where people are usually quite militant about their data not being used for purposes other than intended.

From the responses in this thread, it seems HNer's are actually fine with their data being used for secret reasons without their consent, as long as it means less bureaucracy. "How dare you use my data for your own purposes... unless it helps you avoid an inconveniencing committee, then go for it!"

[euyyn]

Maybe it's simply a matter of from whose point of view the story is being told. "They used my data without asking" vs "they won't let me use this data that I already have anyway".

It helps that it's painted as the poor resourceless scientist vs the powerful abusive entity (IRB). Rather than the poor resourceless patient vs the powerful abusive entity (hospital and its staff).

[throwawayjava]

> Like, why should we conclude that we need that sort of consent to administer a questionnaire for experimental purposes (i.e.: gaining generally useful knowledge), but that there are no substantial issues of consent when it’s for diagnostic purposes (i.e.: gaining information about that particular patient) instead?

1. Researcher incentives. Data => publications => prestige. If the procedure actually contains some inherent risk (e.g., a surgery) then this mismatch between the priorities of the physician and the priorities of the researcher matters. Even (actually, especially!!!) if the physician and the researcher are the same person.

2. Publication. De-anonymization is a real risk even if you're careful to only talk about patient # N in the publication.

3. Patient preference. Just because I allow my hosting provider to access my servers for the purpose of maintenance doesn't mean I'm OK with them accessing my servers for the purpose of surveillance or to read my personal emails. Purpose matters. Some patients won't want data about them used in the context of a scientific study. Patients should and do have the right to insist on that preference. Frankly, the presumption otherwise is exactly why IRBs exist.

This story sounds stupid because -- aside from de-anonymization or data leaks -- there are actually no substantive risks associated with this study. Change "questionnaire" to "open heart surgery" and all of this process starts to make a lot more sense.

FWIW I'm not saying IRB's aren't completely overbearing when it comes to benign studies. They totally are. But that doesn't mean these processes aren't well-justified in other circumstances.

(Also, a lot of this pain is completely avoidable if you learn and follow the rules, which isn't actually nearly as hard as this post makes it out to be IME.)