AI Tool Improves Hospital Discharge Summaries to Support Follow-Up Care

Discharge summaries guide care after hospitalization but can be difficult to prepare after long or complex stays. Clinicians must condense treatment, medications, follow-up, and unresolved issues into a clear handoff document. Incomplete summaries can disrupt continuity between hospitalists and primary care physicians. A new study shows that artificial intelligence may help produce clearer, more complete discharge summaries.

Researchers at Stony Brook University (Stony Brook, NY, USA) evaluated a large language model designed to support discharge documentation at Stony Brook University Hospital. The system was developed with clinician input to reflect the information physicians need after a patient leaves the hospital. It organizes medical records and other patient information to generate a draft summary within a few minutes.

The study compared large language model-generated summaries with human-authored summaries for 60 internal medicine hospitalizations lasting seven to 21 days. Hospitalists and primary care physicians performed paired assessments of the summaries. Physician reviewers preferred the AI-generated summaries in 95% of evaluated hospitalizations.

The AI-generated summaries also received higher scores for overall quality, readability, accuracy, and completeness. Reviewers found fewer missing clinical details than in the human-authored summaries. Primary care physicians reported that the AI-generated summaries were easier to understand, easier to explain to patients and caregivers, and more useful for follow-up care.

Both AI-generated and human-authored summaries contained errors. Physician reviewers found no statistically significant difference in the estimated severity or likelihood of harm linked to those errors. The researchers emphasized that physician review and verification remain essential before any large language model-generated discharge draft is entered into a patient record.

The findings were published online in npj Digital Medicine. The research team reported that further work will focus on accuracy, safety, physician time savings, and integration into clinical workflows. A pilot study to evaluate the tool in everyday clinical practice at Stony Brook Medicine is being considered.

“These results provide evidence that artificial intelligence can support discharge documentation for complex hospital stays,” said Rachel Wong, co-corresponding author and clinical associate professor in the departments of Internal Medicine and Biomedical Informatics in the Renaissance School of Medicine at Stony Brook University.

Related Links
Stony Brook University