travisgerke.bsky.social
sometimes #dataBS #causalsky #statsky #episky #rstats but mostly yolo skeeting nonsense
459 posts
4,285 followers
2,104 following
Regular Contributor
Active Commenter
comment in response to
post
Would love to see it! In oncology, for example, I very often see that outcomes just aren’t captured in EMRs - you only see things like treatment changes which probably indicate progression but it’s very hard to know for sure
comment in response to
post
🤣 I hear this, same. There was some admirable company stuff in that episode
comment in response to
post
For sure, also spent a lot of time working with Epic/EMR data and it’s not just billing per se, but that is the focus of *why* data is captured and the completeness/missingness that results poses intractable problems. Certainly valuable for some use-cases but not as good as say natl health registry
comment in response to
post
The transcription stuff is legit, that could change a lot of things for the better. That Epic Cosmos stuff on the other hand is silly. Doesn’t matter how many patients it covers, what is effectively billing data will not be sufficient for causal RWE work (with AI or not)
comment in response to
post
Great point and agree. Though I was disappointed to hear them rave at the end about how AI on top of massive aggregated EHR data is going to lead to advances. Hard nope. Garbage in, garbage out. That data is not fit for purpose.
comment in response to
post
I just finished listening to this one too! Insufficient amount of dooming around “EMRs are made for billing and not to improve health, knowledge, or doctor/patient experience” for my taste 😅 but they did excellent research and it was a fascinating listen
comment in response to
post
So is the new HHS
comment in response to
post
Fahrenheit 182, Mark Hoppus
comment in response to
post
Thanks for the shoutout! I stopped active development on shinydag since shiny doesn’t ideally scale to all the desired features. Been working on this instead cstructure.io
comment in response to
post
😣
comment in response to
post
Also fostering right now! 🐈
comment in response to
post
YES
comment in response to
post
every time I review a paper now I will demand ECOG only be presented as a mean
comment in response to
post
Same, my coworkers were all horrified to learn I’ve never had any retirement savings or 401k a couple months ago. But now 🤷♂️ 🫠
comment in response to
post
But some of my original thinking remains: as much as I love working in RStudio (and FAR prefer it to jupyter for example), companies can't drop that IDE into their software because of GPL. But you see jupyter integrations everywhere because it's permissive
comment in response to
post
Well yes, rstudio is also GPL (and open source), so that checks out. I admittedly know almost nothing about licensing history of rstudio -- I wonder if the pandoc bundling happened later? Interesting (and probably a more thoughtful move) that Positron is not GPL (it's Elastic 2.0)
comment in response to
post
Exactly! R is great for academia, but will always be a huge missed opportunity with regards to building financially viable products and companies.
comment in response to
post
So called “copy left” licensing. Incompatible with proprietary software since it requires all downstream embedding of that software to be open sourced as well. R itself is GPL. Python is not copy left (it’s “permissive”)
comment in response to
post
He may or may not have 10,000 person years in this cohort
comment in response to
post
comment in response to
post
All least you got a response from @schiff.senate.gov those offices blanked me
comment in response to
post
Same with @padilla.senate.gov. I called (couldn’t get through), so emailed about Schumer’s inability to meet the moment (I.e. call for new leadership) and got this form response about the Trump administration. Nothing like knowing you’re being heard 🙃
comment in response to
post
And cstructure.io 😀!
comment in response to
post
comment in response to
post
for me, this is one of the two primary (only?) reasons AI is useful (alongside generating regex). Not sure if it's worth burning the planet down tho
comment in response to
post
no DAG, no swag
comment in response to
post
This, but make the guy 35
comment in response to
post
Both. My parents didn’t vote because “they’re all corrupt” and now they (boomer generation) are so privileged they can tune out the news because it’s negative
comment in response to
post
If you are represented by a congressperson on the appropriations committee, please reach out to ask that CDMRP funds be maintained at the previously approved levels. There is time today to add anomaly language to the CR that makes this happen. cdmrp.health.mil
comment in response to
post
Magnetic Fields ref ftw!
comment in response to
post