JAMA study finds AI-generated discharge summaries match clinician-authored ones on harm potential
EricTopol · x · 2026-11-18
Eric Topol shares a JAMA study comparing AI-generated hospital discharge summaries against the clinician-authored standard of care.
Reviewers rated the potential for harm across two dimensions:
- Clinical summary: LLM-generated and clinician-authored summaries showed comparable error rates and harm-severity distributions, with most responses clustered in the no-error/no-harm range
- Plan and requested actions: results were similarly close, with no systematic disadvantage for LLM versions at the high-harm end
One of the first head-to-head evaluations of LLM discharge summaries against current clinical practice, suggesting AI-generated documentation is approaching clinician-level safety—though larger real-world validation is still needed.
More from Apps
- Gemini side panel in Chrome leaks model tiers: free vs Pro subscription compared — gaganghotra_ · 2026-11-18
- Tabletop prototype update: auto-linking images to video timestamps looks promising — GrantCuster · 2026-11-18
- Grant Custer shares tabletop prototype progress: autolinking, snapping, touch woes — GrantCuster · 2026-11-18
- Google expands Gemini Tasks free to hundreds of millions, adding agent calls and AP2 payments — sundarpichai · 2026-11-18
- Uber Eats launches drone delivery in Houston suburb at up to 70 mph — Polymarket · 2026-11-18
- The Batman Principle: Why AI Tools Need Knowledge, Experience, and Judgment — afigueiredo · 2026-11-18