a little thread on tips and tricks for working with the IRB when doing health economics and health services research. this is targeted at folks doing observational studies and interventions, especially field experiments w/ health care orgs. but others may find it useful too!

point one: if you don't know what to do, reach out to your IRB. this can head off misunderstandings that cause delays. IRB staff are juggling complex admin rules and they probably aren't experts in your field (they have to wear so many hats), so they will find it helpful too.

next, know the levels of review for human subjects research:
- exempt *
- expedited *
- full review
listed in increasing order of the scrutiny of review & how long it'll take to get approved.
* retrospective+observational studies in health care should fall into these

for observational work, the key question is whether your data is identified or not. analyses of de-identified data are generally exempt. otherwise, expedited.
** to be exempt research, your data does not necessarily have to be HIPAA-deidentified!!!

HIPAA has de-identification standards so data is longer protected health information (PHI). every IRB i've worked w/ has applied a much easier bar than HIPAA. all considered Limited Data Sets w/ dates & scrambled patient IDs de-identified.
re the jargon: https://privacyruleandresearch.nih.gov/pr_08.asp 

so if you're doing analyses of claims data, discharge data, data from RCTs already conducted, etc., if patients are not directly identified, you should explore seeking a "category 4" exempt determination. https://www.hhs.gov/ohrp/regulations-and-policy/decision-charts-2018/index.html#c6

for interventional work, specifically field experiments in the health care delivery system, you are likely going to full review. that said, there are two important times when full review *may* not be appropriate:
1) procedures happening elsewhere
2) quality improvement (QI)

1) is another organization (e.g. gov't agency) running the show, and you're merely helping them out? do they have final say on procedures? are they responsible for intervention procedures like enrollment? tell your IRB before submitting!

your IRB should know that the procedures that normally would trigger full review aren't happening at your institution. you may get more limited + faster review. (NB: this might mean full review should happen at a different IRB.)

2) is the work happening for the purposes of quality improvement (QI)? QI is not considered human subjects research. many orgs have procedures for seeking a "not human subjects research" determination, taking the project out of IRB review.

distinguishing QI vs. research is a tricky area! there's a long history of QI in health care, think work to hasten adoption of best practices, improve patient safety, etc.
in QI, evaluation occurs for the organization, not for generalizable knowledge. (at first, at least)

what many do not realize is that QI and randomization are perfectly compatible. my favorite example of folks doing randomized QI is happening at @nyulangone led by @leorahorwitzmd. i can't recommend learning about that work enough. https://twitter.com/leorahorwitzmd/status/1174487025359249413

the first-order purpose of QI should be helping the organization you're working with. but in practice, QI can also produce generalizable knowledge that gets published. @JAMAInternalMed even put out a guide for publishing QI. https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2664062

if you're doing work to help an org improve, even randomized, your IRB might consider it "not human subjects research". you can still evaluate that work!! the IRB might say that evaluation is research - but lacking an interventional component, it could be exempt or expedited.

circling back to where i started: contact your IRB. especially if you think you might be in the space between QI and research. the rules are complicated but once you find the right approach, approval is often surprisingly quick.