Physicians Choose AI Over Human-Written Discharge Notes 95% of the Time in New Study
Large language models outperformed traditional documentation in quality, accuracy, and completeness for complex hospitalizations, generating summaries in minutes rather than hours.
Illustration: AI-generated
Hospital discharge summaries—critical documents that bridge inpatient and outpatient care—may soon get a major upgrade thanks to artificial intelligence. A new study from Stony Brook University reveals that physicians overwhelmingly preferred AI-generated discharge summaries over human-authored versions, marking a significant milestone in clinical documentation automation.
AI Summaries Win Physician Approval in Head-to-Head Comparison
The study, published in npj Digital Medicine, compared large language model (LLM)-generated discharge summaries against traditional human-authored summaries for 60 Internal Medicine hospitalizations at Stony Brook University Hospital. The hospitalizations ranged from 7 to 21 days—complex cases that typically generate lengthy, detailed documentation.
The results were striking: physicians preferred the AI-generated summaries in 95% of evaluated cases. The LLM-produced documents scored higher across multiple quality metrics, including overall quality, readability, accuracy, and completeness. Notably, AI summaries contained fewer missing clinical details compared to their human-authored counterparts and were generated within just a few minutes.
"This interdisciplinary team of computer scientists, biomedical informaticists, physicians, and a communications expert demonstrated that AI can significantly improve one of healthcare's most time-consuming documentation tasks," according to the research led by Tyler Osborne, a PhD candidate in Stony Brook University's Department of Computer Science, alongside co-corresponding authors Dr. Rachel Wong, Clinical Associate Professor of Internal Medicine and Biomedical Informatics, and Dr. Fusheng Wang, Professor of Biomedical Informatics and Computer Science.
Primary Care Physicians See Clear Benefits for Patient Follow-Up
The study revealed particularly strong support from primary care physicians, who are responsible for managing patients after hospital discharge. These physicians found AI-generated summaries easier to understand, easier to explain to patients and caregivers, and more useful for planning follow-up care.
This improvement in care transition documentation could have significant implications for patient outcomes. Better discharge summaries facilitate clearer communication between hospital and outpatient providers, potentially reducing readmissions and adverse events during the vulnerable post-discharge period.
However, the research team emphasized an important caveat: both AI-generated and human-authored summaries contained errors, with no statistically significant difference in the estimated severity or likelihood of harm from those errors. This finding underscores that physician review remains essential before any AI-generated summary enters clinical use.
Next Steps: Pilot Programs and Continued Development
The research team is now focusing on continued LLM development with emphasis on accuracy, safety, time savings, and clinical workflow integration. A pilot study at Stony Brook Medicine is under consideration to test real-world implementation.
The potential impact extends beyond documentation quality. By reducing the time physicians spend on discharge summaries—particularly for complex, longer hospitalizations—AI tools could help address physician burnout while improving the information available to care teams managing patient transitions.
For health technology companies, the study demonstrates a viable use case for LLM integration into hospital electronic health record workflows, particularly for high-complexity documentation tasks that consume significant physician time while being critical for care coordination.
Sources
Google Search
Follow Health AI Central on Google
Choose Health AI Central as a preferred source, and Google is more likely to show our reporting in Top Stories.
Reporting & accountability
How we maintain accuracy
Health AI Central separates reported facts, analysis, and public-source data. Readers can review our standards, request a correction, or consult the relevant research methodology before relying on a claim.