We are currently losing up to 35,000 lives every year in the U.S. simply because we are afraid to let AI help doctors.
April 16, 2026
Original Paper
Estimating Preventable Patient Mortality Associated With Non-Deployment of Physician-Supervised Artificial Intelligence in United States Healthcare: A Structured Scenario Analysis
SSRN · 6541704
The Takeaway
Most people are worried about AI "hallucinating" in the hospital, but this study quantifies the massive cost of doing nothing. By analyzing existing physician-supervised AI tools, researchers found that not deploying them is leading to between 15,700 and 35,900 preventable deaths annually. That’s like a major airliner crashing every single week because we refused to use the autopilot. It shifts the ethical debate from "is AI safe?" to "is it even ethical to keep using only humans?" For regular people, this means your chances of surviving a hospital stay could go up significantly if your doctor has an AI co-pilot.
From the abstract
Background: The dominant framework for evaluating artificial intelligence (AI) in healthcare focuses on potential harms of deployment. The complementary question - how many patients are harmed because currently available, physician-supervised AI systems are not deployed - has not been systematically addressed. We aimed to estimate the annual preventable mortality burden attributable to AI non-deployment across US healthcare. <div> <br> </div> <div> Methods: We conducted a structured scenario ana