Henderson S. et al., 2026: Validating ChatGPT-4-Generated Patient Information Leaflets for Common Endourological Procedures: A Survey of Urology Health Care Professionals.

Henderson S, Reeves T, Somani B.
J Endourol. 2026 Jan;40(1):117-121. doi: 10.1177/08927790251387371

Abstract

Introduction: Artificial intelligence (AI) is transforming health care, including patient communication. Large language models, like ChatGPT-4, also show great promise for enhancing patient communication and decision-making. This study evaluates ChatGPT-4's ability to produce readable, accurate, evidence-based patient information leaflets (PILs) for three common endourological procedures.

Materials and methods: ChatGPT-4 was given a short prompt to create PILs for extracorporeal shockwave lithotripsy, percutaneous nephrolithotomy, and ureteroscopic removal of stones. A 10-question survey was sent to urology teams in the United Kingdom to assess their accuracy, readability, and up-to-date content.

Results: We received 105 responses (35 per procedure). The AI-generated PILs were rated highly across all domains, including: 85.71% found them evidence-based, 92.40% accurate, 99.05% clear and patient-friendly, and 96.19% would share them with patients despite knowing they were AI-generated.

Conclusions: ChatGPT-4 shows promise as a low-input tool for creating high-quality patient materials. It could support broader health care use with tailored, accessible communication. However, ethical and long-term reliability concerns need further study before wider adoption.

Comment Hans-Göran Tiselius

There is no doubt that artificial intelligence dramatically has altered the methods for patient information. This aspect is clearly emphasized in this report in which ChatGPT-4 was used to produce patient information leaflets for SWL, URS and PCNL.

When this information was exposed to various groups of health professionals, the judgement was that they agreed or strongly agreed to the extremely high level. For SWL 100% of the individuals asked declared that they would use the leaflets to their patients.

One important point is that all leaflets used English as information language and that is also the basic information language found in the literature. I have not had the possibility to use ChatGPT-4 for this purpose with questions in Swedish, but on the other hand it would be easy to translate the leaflets into any language, given that the information is relevant. This approach would also add the possibility to make minor alterations relative to any specific local differences.

Hans-Göran Tiselius