AI Discharge Summaries Show Low Harm Risk - Summary - MDSpire
From the Journals
Conference News

AI Discharge Summaries Show Low Harm Risk

  • By

  • Kerri Miller

  • May 13, 2026

  • 4 min

Share

A pilot study published in JAMA Network Open evaluated an AI system, MedAgentBrief, for drafting hospital course summaries used in discharge notes at Stanford Health Care. Implemented from August to October 2025, the system generated 1,274 summaries, incorporated into 57% of discharges. Most reviewed summaries (88%) were rated as having no harm potential. Though omissions (25%) and inaccuracies (20%) were noted, the AI's hallucination rate (2%) significantly outperformed previous studies. Physician burnout scores decreased post-implementation, indicating potential cognitive offloading benefits despite modest time savings and other limitations.

Original Source(s)

Related Content