Artificial intelligence could significantly improve hospital discharge documentation, according to a new study by researchers at Stony Brook University, which found that physicians preferred AI-generated discharge summaries over clinician-written versions in 95% of evaluated hospitalisations.

Published in npj Digital Medicine, the study examined whether large language models (LLMs) could improve the quality, readability, factual accuracy and completeness of discharge summaries, particularly for patients undergoing prolonged and complex hospital treatment.

The researchers analysed 60 internal medicine hospitalisations at Stony Brook University Hospital, involving patients admitted for seven to 21 days. AI-generated summaries were compared with physician-authored documents through paired evaluations conducted by hospitalists and primary care physicians.

The AI system was developed with clinical input to process electronic health records, organise relevant medical information and generate draft discharge summaries within minutes. These documents included treatment details, medication changes, unresolved medical concerns and follow-up care requirements.

The findings revealed that AI-generated summaries received higher physician ratings for overall quality, readability, accuracy and completeness. Reviewers also identified fewer missing clinical details compared with traditional physician-written summaries.

Primary care physicians reported that the AI-generated documents were easier to understand, communicate to patients and caregivers, and use for planning follow-up treatment.

However, the study also highlighted important safety considerations. Both AI-generated and clinician-written summaries contained errors, with no statistically significant difference in the estimated severity or likelihood of harm associated with those errors.

The researchers emphasised that AI-generated documentation should not replace clinical judgement. Physician verification remains essential before such summaries are incorporated into official patient records.

Rachel Wong, co-corresponding author and clinical associate professor at Stony Brook University, said the findings provide evidence that AI can support discharge documentation for complex hospital stays.

The research team plans to investigate the technology’s accuracy, patient safety implications, actual time savings and integration into routine hospital workflows. A pilot evaluation in everyday clinical practice at Stony Brook Medicine is also under consideration.

While the findings demonstrate the potential of AI-assisted documentation, the study’s relatively small sample and single-hospital setting indicate the need for further evaluation before wider clinical implementation.

The research suggests that AI could become a valuable tool for reducing documentation burdens and strengthening continuity of care, provided that clinical oversight and patient safety remain central to its adoption.

Disclaimer: This Cases, Study is intended solely for general information and awareness and does not constitute professional advice, recommendation or endorsement; MedGate Today assumes no liability for decisions made based on its contents.