aceComprehension of an AI-generated discharge letter versus a traditional discharge letter: two controlled quasi-experimental parallel-group studies
Abstract: Background Discharge letters are essential for continuity of care, yet patients often misunderstand their content. Poor discharge communication is associated with medication errors, inappropriate healthcare use, and hospital readmissions. Artificial intelligence (AI) language models may improve discharge documentation clarity and accessibility. This study aimed to compare a GPT-4-generated, accessibility-optimised discharge letter with a conventional discharge letter regarding comprehension of key discharge info…
Navy Department BUMED News Letter Vol. 5, No. 3, February 2, 1945 by U.S. Navy. Bureau of Medicine and Surgery. Public domain
In two controlled quasi-experimental studies published 12 September 2026, 341 online-recruited adults and 791 medical and nursing students at the University of Turin were assigned to read either a GPT-4-generated accessibility-optimised discharge letter or a traditional discharge letter. Comprehension was measured with a structured score covering diagnosis, treatment, investigations and follow-up, with secondary measures of readability, clarity and satisfaction.
The results matter because discharge letters are essential for continuity of care and poor understanding is linked to medication errors and readmissions. While the AI letters improved measured comprehension and satisfaction in both populations by the study date, the authors note that validation using real clinical letters and diverse patients is still required to establish generalizability.
- Two controlled quasi-experimental parallel-group studies compared GPT-4-generated versus conventional discharge letters.
- Study 1 analysed 341 adults recruited online; Study 2 analysed 791 medical and nursing students at the University of Turin.
- Primary outcome was structured comprehension score; secondary outcomes included readability, clarity, organisation and satisfaction.
- Health literacy was assessed using the HLS-EU-Q6 and predictors of comprehension were examined with multiple linear regression.
GPT-4-generated, accessibility-optimised discharge letters increased comprehension of diagnosis, treatment, investigations and follow-up compared with conventional letters.
The rundown
Researchers conducted two parallel-group studies allocating participants by age parity to receive either an AI-generated accessibility-optimised letter or a conventional letter, then scored comprehension of diagnosis, treatment, investigations and follow-up.
Both the general adult sample and the higher-literacy student sample at the University of Turin showed higher median comprehension scores with the AI letter, alongside higher satisfaction ratings.
Sources
- Peer-reviewedInternational Journal for Quality in Health Care2026-09-12
How should this claim be treated?
The debate