Ryan Radecki, MD MS

Putting HFNO on hypoxemic respiratory failure is probably non-inferior to standard therapy, but there are likely clues clinicians rely upon it longer than they ought when patients actually need mechanical ventilation. #FOAMed evidencetriage.com/p/high-flow-ve…

@BrooksWalsh Eh, only if you're being an opportunist attempting to chill its responsible uptake. The idea that individual clinicians hold primary responsibility to evaluate the nuances of, or second-guess, institutionally or professionally-approved AI is precarious.

@emlitofnote Looks like it's a great time for emergency physicians to become expert witnesses who can speak to the *pitfalls* of AI use in medicine.

Ignore AI in #radiology at your peril, unfortunately. Interestingly, however, as this (very small) report notes – *where* AI fits in the workflow might matter as far as a perceived "duty of care" to the a patient harmed by a missed diagnosis. #FOAMed evidencetriage.com/p/more-little-…

@DrToddLee @AppleHelix Literally never a secret that he and his lab was supported by Arnold Ventures.

@AppleHelix And?

Claude was able to find Vinay Prasad's CV from 2023 (for tenure application?). onlinelibrary.wiley.com/pb-assets/CV%2… VP actually received 2 separate grants from the Arnold Foundation. One from 2017-20, the other 20-23. And probably is still supported by the foundation. Total from 2017-23 was ~$4m. Most recently, the foundation was funding VP $800K per year.

@BrooksWalsh I thought 50% was pretty generous! It all depends on the exact wording, but when someone dies because something gets missed ... and AI tells you its there ... I don't see Future Jury being very lenient, at all.

@emlitofnote Yeesh, a few too many layers here to worry about "duty to care " imo. Even the double-read scenario sided with the plaintiff > 50% of the time? Bonkers high rate, doubt external validity.

The SWiFT trial didn't see a difference between whole blood and balanced blood component transfusion when initiated prehospital for major trauma. Always possible the study didn't enroll quite the right population or have the right protocol, but... #FOAMed evidencetriage.com/p/whole-blood-…

A lovely study from Canada doesn't show much favoring rivaroxaban over apixaban for acute VTE – will it be the same for all other indications? #FOAMed evidencetriage.com/p/is-rivaroxab…

It's Google Deepmind's AIME project, deployed as a front-end information-gathering tool for urgent care physicians at BIDMC. It was "fine". Probably some bits left unsaid (particularly around clinical acceptability), but still moving forward. #FOAMed evidencetriage.com/p/deepmind-aim…

Everyone wants a symptom checker, a navigator, a chatbot – you name it – to magically direct patients to the "correct" next step in care. Well, this is what Ada Health is claiming their symptom digester will do: #FOAMed evidencetriage.com/p/how-do-sympt…

Everyone loves a good poop joke – or poop study. Here's a look at potentially using emergency department-specific wastewater monitoring to detect and characterize viruses circulating in the community: #FOAMed evidencetriage.com/p/whats-in-you…

Can we offload some simple, routine radiology to AI? Not quite yet – and this is why uptake is so hard, because it requires workflows to be precisely designed as to augment a human-in-the-loop, not disrupt. #FOAMed evidencetriage.com/p/a-radiology-…

@DrToddLee @venkmurthy No one puts up a fight down here, but we always have to blandly "acknowledge" renal function in context of contrast use – just covering their butts, still.

@venkmurthy I'm fighting this battle every time I'm on service. I feel like I'm in groundhog day

Correlation not causation Fear of contrast nephropathy almost certainly has caused more harm than the thing itself ahajournals.org/doi/10.1161/CI…

Don't trust ChatGPT Health – or likely any general health LLM – to give you medical advice in the context of a serious acute illness. Patient triage is clearly outside its functional requirements. #FOAMed evidencetriage.com/p/the-chatgpt-…

The Zero-Miss Diagnosis, The Final Frontier: aortic dissection. Hundreds of CTs to find a handful of cases, even in Australasia where resource stewardship is Important. Nobel Prize in Medicine for whoever invents A Better D-Dimer, please. #FOAMed evidencetriage.com/p/what-does-ou…

Physicians are good drivers of LLMs, but regular humans who don't know how to tell the "textbook story" can get dangerously divergent results. #FOAMed evidencetriage.com/p/humans-llms-…

A lot of scope creep for thrombolysis in CRAO – but, no, the higher quality data emerging shows harms outweigh the benefit. #FOAMed evidencetriage.com/p/do-not-use-t…

More "tissue not time" evidence – but these data are almost certainly weaker than they appear. Unfortunately, I expect they'll simply be piled upon the ever-expanding indications for stroke thrombolysis, along with corresponding indication creep. #FOAMed evidencetriage.com/p/the-option-f…

The magic wand stethoscope doesn't reliably measure up – yet. Phonographic and single-lead ECG analysis by AI did not reliably improve diagnosis of atrial fibrillation, heart failure, or valvular heart disease. #FOAMed evidencetriage.com/p/not-a-magic-…

More evidence of there not being much of an additive or synergistic effect between different types of analgesia for moderate acute pain – this example in pediatrics: #medsky evidencetriage.com/p/ibuprofen-ch…

These sorts of retrospective analyses are always confounded by indication, but it is probably the case hyPERnatremia can be corrected as rapidly the salient clinical context requires. #FOAMed evidencetriage.com/p/does-it-matt…