
Harvard study: #UrgentCare may be an appropriate first setting for imaging low-acuity #lowbackpain. @MarcSucciMD et al. @JACRJournal ow.ly/jXAT5106qYe

Inventor, Radiologist, Associate Chair Innovation, Associate Prof Harvard, Founder, @harvardmed @MGBInnovation @2minmed @meshincubator @forbes30Under30

Harvard study: #UrgentCare may be an appropriate first setting for imaging low-acuity #lowbackpain. @MarcSucciMD et al. @JACRJournal ow.ly/jXAT5106qYe

Given a complete case, 21 frontier models named the correct diagnosis more than 90% of the time. Given only what a clinician actually has at the start of a visit, they failed to produce a comprehensive differential more than 80% of the time. That single pair of numbers, from a JAMA Network Open study out of Mass General Brigham, explains most of the confusion in this field. Here is what it looks like from clinic. A patient comes in with fatigue and some bruising. There is no vignette. There is a story told out of order, a spouse adding details, a CBC from three months ago sitting in an outside system, and my own quiet running list of what I am afraid of. The hard part of oncology has never been naming the disease once the workup is done. It is deciding what to work up. The authors put it better than I can: the model is being asked to do precisely the part of medicine it does worst. And it is being asked that most often at 2am, by a frightened person with no training in how to prompt or how to appraise what comes back. Forty million health questions a day now go to ChatGPT. I build clinical AI. I am not arguing these tools are useless. I am arguing we are grading the wrong half of the encounter. On the other hand, I believe that practicing medicine without using AI is considered a service for the patients. AI gives me eyes and insight that I will never gain on my own, but the current benchmarks don't make sense. Everyone is trying to either prove AI is better than doctors or doctors are better than AI. Believe me, it doesn't work like that. Healthcare is so complex that AI is a so powerful tool. However, it still needs a human in the loop. You can't run a hospital by AI agents without humans. Mark my words! It's not because AI is retarded. It's because patients need humans, and this will not change. jamanetwork.com/journals/jam…

More evidence of #AI limitations in guiding clinical care. Exploration of its use in clinical reasoning found not bad in final #diagnosis and management but error rates in differential diagnosis of 80%. jamanetwork.com/journals/jam…

Honored and thrilled to share that I've been named a 2026 Paul and Daisy Soros Fellow! pdsoros.org/fellows/arya-rao… @PDSoros @harvardmed

Our commentary on this very interesting study by Rao AS et al demonstrating the Limitations of Large Language Models in Clinical Diagnostic Reasoning was just published in JAMA Network Open: jamanetwork.com/journals/jam…
Assessment of 21 LLMs for generating a differential diagnosis "Off-the-shelf LLMs have not yet achieved the intelligence required for safe deployment and remain limited in demonstrating advanced clinical reasoning." jamanetwork.com/journals/jam…

Read about our latest study in @FinancialTimes @MassGenBrigham @meshincubator
While chatbots can identify likely conditions once a case is fully specified, they are less reliable at the earlier, more uncertain stages of clinical reasoning. ft.trib.al/Fo6IrBR

Thanks for highlighting our study released today @EricTopol @meshincubator @MassGenBrigham @harvardmed
Assessment of 21 LLMs for generating a differential diagnosis "Off-the-shelf LLMs have not yet achieved the intelligence required for safe deployment and remain limited in demonstrating advanced clinical reasoning." jamanetwork.com/journals/jam…

AI remains lacking in clinical reasoning abilities, according to @JAMANetworkOpen @MarcSucciMD et al study of 21 large language models medicalxpress.com/news/2026-… @RichardLehman1 @CoryRohlfsen @raj_mehta @DevanSinha

Join us for MESH Core 2026: The Official Healthcare Innovation Bootcamp of @MassGenBrigham. May 4-5, 2026. Themes include AI, longevity, venture, and more. Featuring CEOs, founders, physicians, and more. Open to the Public. Learn more: meshcore.meshincubator.org/

Wonderful work with a huge contribution from @harvardmed students working with the incomparable @MarcSucciMD and other @MassGeneralMDs from several departments. #LLMs #AI #Radiology @meshincubator @NeimanHPI

In a recent study, researchers from @meshincubator and colleagues investigated the use of CTA in emergency departments. The study was published in @JAMANetworkOpen. Read more: doi.org/10.1001/jamanetworko… @MarcSucciMD

🚨 CT angiography use for suspected #GIbleed jumped 7-fold in the ED, but diagnostic yield is dropping. @MarcSucciMD of @meshincubator says: “We should avoid defaulting to #CTA for borderline scenarios — it’s about balancing risk, time and throughput.” bit.ly/4noDDT3

At #MESHCore2025, Dr. Isaac Kohane @zakkohane shared how AI tools can deliver vastly different care recommendations depending on the prompt. His takeaway: AI in healthcare must be guided by ethical frameworks that prioritize patients. Read more: harvardmagazine.com/2025/05/…

At @meshincubator 2025, @harvardmed’s @zakkohane shared a striking case: GPT-4 recommended treatment to a patient when assessing from the perspective of a doctor—but denied treatment as an insurer. Same patient, same facts, different values. #Harvard harvardmagazine.com/2025/05/…

Key takeaways in @JAMA_current 1. AI goes beyond performance: Medical schools must prepare students to use AI not just as a diagnostic tool, but to augment core skills. 2. Transformation needed: Medical education must evolve to teach both technical AI skills and human-centered care 3. LLMs show promise: Large language models demonstrate early success in clinical tasks @MarcSucciMD @AryaRao_ @harvardmed
"As AI takes on tasks such as routine documentation and medical record review, the physicians of the future must serve as responsible stewards of these technologies." Read this Viewpoint about preparing students to succeed in medical school. ja.ma/41RtkyV

And that’s a wrap! Just finished #MESHCore2025! Inspired by clinicians making the world better. Excited to continue advancing tech for efficiency, accuracy & better decisions in healthcare. Looking forward to building the future together 💪🏽 @MGBInnovation @meshincubator