aoeverhart.bsky.social
Instructor at Washington University in St. Louis School of Medicine, health economist studying development and adoption of medical technology.
35 posts
432 followers
191 following
Prolific Poster
Conversation Starter
comment in response to
post
Please don't make me post this on the bad site...
comment in response to
post
bsky.app/profile/adri...
comment in response to
post
I just covered this for @thetransmitter.bsky.social — will keep following! www.thetransmitter.org/funding/fede...
comment in response to
post
I used dtable and table for the first time a few weeks ago, so slick. It's like, "What do you mean there's an easy way to export tables that doesn't rely on a 15-year old user-written command?"
comment in response to
post
tl;dr: seems bad.
comment in response to
post
It means having the right data structures to identify emerging quality issues (since we're not screening for these things before market entry) and a responsive FDA that can reduce regulatory uncertainty by clearly communicating to stakeholders whether products are subject to more stringent overview.
comment in response to
post
Look at the main call-to-action in this article. This is fundamentally arguing that we need less regulation, but it still takes a lot of technical know-how to implement this! The responsible approach to reducing regulatory burdens isn't to just slash and burn budgets.
comment in response to
post
Even disagreements about the FDA's recent approach to regulating AI hinge on really technical distinctions. This @scottgottliebmd.bsky.social editorial argues that FDA should regulate AI based on how info is presented to docs, not how docs use that information.
jamanetwork.com/journals/jam...
comment in response to
post
Well at least the reviewers get it.
What do you typically ask for?
comment in response to
post
Agree. In a nice counter anecdote to your experience, I recently reviewed a paper that had an event study similar to Exhibit 1 and every reviewer said, "Uh oh, that doesn't look good," (though to your point, that's another "eye test").
comment in response to
post
I think the larger point is correct though, people aren't think critically about interpreting pre-trend coefficients and ignore the lack of power masking evidence of trends in Ex. 1. But I don't think Ex. 2 would generate significant pre-trends, since the gap seems constant during the pre-period.
comment in response to
post
So even if every pre-treat coef was equal to each other but not zero, you might still interpret that as sort of a "trend", since the gap in outcomes did change at some point during the pre-period. Though in that case I'd probably speculate about an anticipation effect or something, not pre-trends.
comment in response to
post
I'm not sure this is totally right? The coefficients in an event study like Ex. 1 are differences in levels *relative to the difference in t-1*. So observing significant pre-treat coefs means that the gap in outcomes between treat and control got larger/smaller between study start and t-1.
comment in response to
post
The things they don't teach you in grant writing class
comment in response to
post
So if FDA can't fill enough positions (or just takes a big hit to its work force?), then I think they'll have to give a large chunk of the user fees back.
comment in response to
post
I'm not familiar with anything where they can't use the user fees at all (but please share anything if I'm wrong!) But at least on the device side, I think the biggest parts of the user fee agreements with mechanisms for industry claw back are FDA hiring targets.
comment in response to
post
Oh wait, are DMs not a thing here? I guess just comment or shoot me an email.