Autonomous AI could soon deliver better care than physicians or physician-AI hybrid teams, argues Ezekiel Emanuel, MD, PhD, a University of Pennsylvania bioethicist and former White House health policy adviser, in a recent JAMA Perspective.
He and his co-authors cite five functions where AI already matches or beats physicians: gathering patient information, forming a differential diagnosis, ordering diagnostic tests, prescribing guideline-concordant treatment and managing chronic disease.
In one cited study, Google’s Articulate Medical Intelligence Explorer outperformed physicians on patient complaint elicitation (97% vs 50%) across 159 simulated cases. In another, ChatGPT o3 ranked the correct final diagnosis first in 60% of 377 complex cases, versus 15.9% for 20 internal medicine physicians on a 302-case subset.
The authors also warn that adding physicians back into the loop can worsen outcomes once AI alone already outperforms humans, citing a meta-analysis of 106 human-AI experiments and a 2025 review of 52 clinical studies showing hybrid teams underperforming AI alone.
They conclude autonomous AI could be ready for deployment in select clinical workflows by 2030, requiring shifts in liability, regulation, reimbursement and medical education policy.
Dr. Emanuel co-wrote the Aug. 17 piece with Abe Baker-Butler, BA, also of Philadelphia-based University of Pennsylvania; Neal Khosla, co-founder and CEO of AI virtual care company Curai Health; and his father, Vinod Khosla, founder of venture capital firm Khosla Ventures. Two of the four authors have financial ties to the argument they are making: Curai Health sells AI-powered virtual care, and Khosla Ventures has backed AI health startups, including Curai.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.