For your listening pleasure on the PedsCrit Podcast - MelissaJerdonekSacco (UVA Department of Pediatrics/#PedsICU) and Andrew Parsons (UVA Department of Medicine) on clinical reasoning #FOAM #MedEd pedscrit.com/clinical-reaso…

Unsurprisingly Cyril Zakka, MD says it better than I could. These are the exact concerns I have about the way we are integrating LLMs into clinical care. It *might* not matter, but we are all participating in an educational Pascal's wager right now. A 🧵🔽

This looks like a powerful AI tool by ⁦Raja-Elie Abdulnour⁩ and team. Real-time feedback on the reasoning process — can’t wait to try it out! #MedEd medrxiv.org/content/10.110…

Disruptions—new, neglected, and insidious—are reshaping medical education. In this NEJM editorial, we challenge ourselved us to move beyond “business as usual” and reimagine how we prepare future physicians. nejm.org/doi/full/10.10… NEJM

How should one evaluate the usefulness of a question (or test or experiment) e.g. for medical diagnosis? ✅ Information gain, KL Distance, Probability Gain, Impact ⛔️Bayesian Diagnosticity [=Max(LR)] Niche, but cc: Cory Rohlfsen Dr. Shuhan He 🫀🫁

🧠📱 Just published in NEJM AI, we explore the promise and the perils of using #GenAI to prepare students for OSCEs. #AI can simulate patients, but it can’t replace the cognitive work of becoming a clinician. #MedEd Read more here: doi.org/10.1056/AIe250…

Yes. Writing is not a second thing that happens after thinking. The act of writing is an act of thinking. Writing *is* thinking. Students, academics, and anyone else who outsources their writing to LLMs will find their screens full of words and their minds emptied of thought.

MINOCA is not an end point diagnosis. Great example of how diagnosis is not an endpoint but a waypoint. Excited for a new paradigm: ✅ cardiac MRI ✅ coronary functional testing ✅ role of AI in pattern recognition of previously ‘non essential’ data Andrew Parsons

🔔pub on poverty of diagnostic essentialism🔔 Essentialism served us well before AI but how about now? Do end point diagnoses exist? Answers👇 degruyterbrill.com/document/doi/1… Adam Rodman Derya Unutmaz, MD Andrew Sanchez The Phlegmfighter Brian Locke, MD MSCI Adam Cifu Andrew Parsons Medical Education Flamingo, MD, PhD

Reimagining diagnosis in the age of artificial intelligence with ⁦Cory Rohlfsen⁩ degruyterbrill.com/document/doi/1…

The Impact of Definitions of Disease on Overdiagnosis jamanetwork.com/journals/jamai… via Kari Tikkinen et al

Used to recommend openAI as preferred LLM for clinical reasoning support- last week I’m now team gemini until OpenAI OpenAI Developers sort this out and allow us to at least choose o3 by default as legacy model

#AMEE2025

Wow! Thank you ⁦@AMEE_Community⁩ ⁦@AMEEFacDev⁩ #AMEE2025 ⁦UVA School of Medicine⁩ ⁦UVA Research⁩ ⁦UVA SOM Fac Dev⁩ ⁦UVA Department of Medicine⁩

Are A.I. Tools Making Doctors Worse at Their Jobs? - The New York Times nytimes.com/2025/08/28/wel…

Andrew Parsons nice piece. Deskilling, mis-skilling, never-skilling... nejm.org/doi/pdf/10.105…

Niche multi-class reasoning pt 2: I fixed an error in this: when you define a set of "one-vs-rest" LRs (ie. likelihood ratio for dx A vs not A, B vs not B, etc.), you want to ensure the resulting post-test probabilities sum to 1 (coherence)

Proud to represent ⁦UNMC College of Medicine⁩ at AMA ChangeMedEd, introducing novel endpoints of clinical reasoning that weren’t feasible prior to #AI Let’s reimagine how to measure #clinicalreasoning ⁦Brian Locke, MD MSCI⁩ ⁦⁦Dr. Shuhan He 🫀🫁⁩ Andrew Parsons⁩