Utility of ChatGPT and Large Language Models in Enhancing Patient Understanding of Urological Conditions

Abstract

Objectives: Large language models such as ChatGPT have been used to generate text in a conversational manner, and may be of use in providing patient information in a urological setting. This study evaluated the accuracy, presence of omissions, and preferability of traditional patient information to the large language models ChatGPT and Bing Chat. Methods: Eight common questions regarding urolithiasis and prostate cancer were selected from traditional patient information and posed to ChatGPT and Bing Chat. Responses from all sources were then evaluated by seven urologists in a blinded fashion for accuracy, omissions, and preferability. Results: We found that 96.43% of ratings of traditional patient information sources were rated accurate, compared to 94.6% for ChatGPT and Bing Chat; 7.1% of ratings of traditional patient information were rated as containing harmful omissions, compared to 10.71% for ChatGPT and 21.4% for Bing Chat; and 55.4% of rater first preferences were given to ChatGPT, compared to 35.7% for traditional patient information and 8.9% for Bing Chat. Conclusions: ChatGPT provided responses of a similar accuracy and preferability to traditional sources, highlighting its potential as a supplementary tool for urological patient information. However, concerns remain regarding omissions and complexity in model-generated responses.

Ähnliche Arbeiten