Doctor, researcher, writer | AI in medicine & #meded | @Princeton | @Penn med | #derm @ Hopkins | rheum/derm fellow @ Harvard
Joined September 2010
- Tweets939
- Following925
- Followers776
- Likes2.3K
Pinned Tweet
#dermresidents: I've been compiling playlists of dermatology-related lectures/grand rounds presentations on YouTube to supplement residency program didactics & provide add'l info/perspectives on various topics in dermatology. You can find them at tinyurl.com/dermlectures
Ramie Fathy retweeted
Our clinical trial of AMIE, a patient-facing AI system, was just published in The Lancet.
I want to explain what *I* think this study ACTUALLY means.
It's about a fundamental question: how do we know that clinical AI works?
🧵⬇️
teapaw.org/Google/status/21083245…
Our medical research system, AMIE, is the first patient-facing conversational diagnostic tool of its kind to be studied prospectively in a real-world clinical setting.
In a study, published today in @TheLancet, we found that when patients chatted with AMIE before an in-person appointment, the interaction built their confidence and helped them organize their thoughts. And physicians who reviewed the information from the patient conversation got more time back to focus on collaborative care and shared decision-making instead of digging through data.
Ramie Fathy retweeted
The clinician's role in the AI medical era
"Clinicians cannot, and should not, resist powerful AI technologies if these systems improve the quality of patient care."
"We contend that the clinician's enduring role is accountability."
[I think the enduring role is well beyond that, but OK]
perspective by @andrewparsonsMD and @AdamRodmanMD @AnnalsofIM
acpjournals.org/doi/10.7326/…
Ramie Fathy retweeted
Out of >1,350 medical AI's cleared or arrived by FDA, a total of 3 (0.2%) were tested for patient outcomes
only 34 (2.5%) linked to prospective, registered clinical trials
@PLOSDigiHealth
journals.plos.org/digitalhea…
Ramie Fathy retweeted
Hi clinician friends - I'm really excited to release Osler Prerounds! Prerounds is a tool I built for hospital medicine, to serve as a "command center" to think, plan, and track everything during busy inpatient service.
I've been using it daily on inpatient medicine service, and it has been extremely useful in helping me better reason about, and stay on top of, patient management.
Some useful features:
- quickly store and track evolving patient data across your entire service
- TODO tracking across all patients + timed reminders for tasks
- export to PDF/printout to take with you during rounds
- all data is stored 100% locally, with nothing leaving your computer!
Available now, for free! any feedback welcome :)
Ramie Fathy retweeted
Dermatology Simulator 2026
As my #derm board exam nears, I find myself becoming increasingly creative in finding ways to avoid studying for it. Here’s my latest solution. Would you play? #dermshift
As my #derm board exam nears, I find myself becoming increasingly creative in finding ways to avoid studying for it. Here’s my latest solution. Would you play? #dermshift
Ramie Fathy retweeted
New research on beneficial RL: models trained on a small amount of beneficial trait data improve on a wide range of alignment and benefits evaluations, even if trained only on health domain data.
We hope it’s a step towards more broadly and persistently beneficial models. 🧵
Ramie Fathy retweeted
Very important study that should energize clinicians and clinical leadership to ask what we want from AI decision-support. What are we maximizing (e.g. accuracy, reproducibility. chain-of-evidence, contextual-specificity etc)? HT @arjunmanrai
For medical information, general AI frontier models (Google, OpenAI, Anthropic) outperformed specialized @openevidence and @UpToDate as assessed by 12 US clinicians, randomized and blinded to which model and extensive testing/benchmarks. This was not anticipated. @NatureMedicine
nature.com/articles/s41591-0…
Ramie Fathy retweeted
This is actually insane. The patient facing tools are performing better than the clinician facing tools.
Direct model access is a big deal.
🤯
Ramie Fathy retweeted
AI in medicine warning! The never skilling effect in medical education! @TomVargheseJr @pferrada1 @SWexner @AmCollSurgeons
Have definitely experienced this with AI scribes. The task shifts from writing to editing, from piecing the story together to supervising the bot’s version thereof.
Cognitive offloading and the speedup illusion in human-AI interaction
Sunny Yu, Myra Cheng, Ahmad Jabbar, Ilia Sucholutsky, Katherine M. Collins, Dan Jurafsky, Robert D. Hawkins
arxiv.org/abs/2605.23177 [𝚌𝚜.𝙲𝚈 𝚌𝚜.𝙷𝙲]
ALT Large language models (LLMs) have the potential to boost human productivity by speeding up task completion – provided users know when to offload cognitive work to them. But we do not know if users are well-calibrated in estimating these potential time savings. We conducted a preregistered large-scale behavioral study (N = 1237) to characterize mismatches between expectations and reality, with a focus on simple cognitive tasks. While actual completion times between independent completion and AI-assisted completion did not differ, participants predicted AI to be significantly faster. The same bias was not observed when imagining help from another human participant. We identify a speedup illusion where people have accurate forecasts of independent completion times but significantly underestimate AI-assisted times. Additionally, time and effort dissociate: participants reported lower subjective effort with AI despite equivalent completion times. This suggests that completion time itself is
Ramie Fathy retweeted
I made a summary infographic. The clinical search over trusted sources is a major upgrade. Definitely something I'll be recommending to colleagues.
🤖 Made with AI
Ramie Fathy retweeted
Nothing to see here. Only the AI, Dr Cabot keeping pace with the best of human experts in the most challenging cases @nejm
I’ve been obsessed with the @NEJM CPCs since I was in grad school. Now, with @tabuckley_, it’s surreal to see @NEJM publish the first AI differential diagnosis in the 100+ year history of the series, generated by our AI system Dr. CaBot, alongside the human expert’s.
Ramie Fathy retweeted
A.I. generated clinic notes from ambient out-patient visits helps clinicians in many ways, across 6 health systems jamanetwork.com/journals/jam…
Ramie Fathy retweeted
Thanks to @EricTopol for highlighting our latest study, which looks at some of the failure points in the clinical reasoning performance of reasoning models.
What does this mean for the future of diagnostic decision support? And can LLMs actually medically reason?
A 🧵⬇️
The performance of generative A.I. models for clinical reasoning are not holding up to increased scrutiny
arxiv.org/abs/2509.18234 @MSFTResearch
ai.nejm.org/doi/full/10.1056… @NEJM_AI @AdamRodmanMD @LiamGMcCoy
Ramie Fathy retweeted
#dermresidents: I've been compiling playlists of dermatology-related lectures/grand rounds presentations on YouTube to supplement residency program didactics & provide add'l info/perspectives on various topics in dermatology. You can find them at tinyurl.com/dermlectures
Ramie Fathy retweeted
OpenAI's open source models are bringing a fundamentally new era in medical AI.
We now have an o3-level model that can run locally, HIPAA compliant. All previous medical LLM publications are obsolete.
Ramie Fathy retweeted
Reviewed a Step 1 question today on MEN 2A syndrome, a rare genetic disorder that runs in families. I am glad the boards test students on this because doctors should recognize rare conditions and not just think of common stuff. But the question asked the student to identify *from memorization* the exact gene mutation that causes MEN2A and know its cellular effects. To me, this is ridiculous. This can be easily looked up. Why are we testing students in this way in the modern era? If someone can justify this type of question please let me know. I can’t see the logic of how this helps make better doctors. In fact, it promotes rote memorization which may actually be a bad thing in an era when information changes so rapidly.
Ramie Fathy retweeted
Our first RCT on using an LLM on diagnostic reasoning is out!
And the results are 🔥🌶️... adding ChatGPT did NOT improve diagnostic accuracy or reasoning, and the AI alone outperformed ALL the humans.
What does this mean? A 🧵⬇️
teapaw.org/EricTopol/status/18509…
A small randomized trial of generative #AI for diagnosis again (as seen in a few previous studies) shows higher performance for #AI than physicians + AI. May indicate that physicians need to be trained on how to incorporate AI.
jamanetwork.com/journals/jam…
Ramie Fathy retweeted
96% accuracy in cancer detection by this AI from Harvard.
It predicts tumor molecular profiles from images.
This new model, called CHIEF (Clinical Histopathology Imaging Evaluation Foundation), was trained on 15 million unlabeled images
- Identifies features linked to cancer-related genes
- Detects mutations in 54 common cancer genes
- Performs multiple tasks - detection, genetic profiling, survival prediction
- Outperforms other models by 8% overall
- Validated across 17 institutions
