Harnessing language models for streamlined postcolonoscopy patient management: a novel approach

Abstract

Background and aims: ChatGPT, an advanced language model, is increasingly used in diverse fields, including medicine. This study explores using ChatGPT to optimize postcolonoscopy management by providing guideline-based recommendations and addressing low compliance rates and timing issues.

Methods: In this proof-of-concept study, 20 clinical scenarios were prepared as structured reports and free-text notes, and ChatGPT's responses were evaluated by 2 senior gastroenterologists. Compliance with guidelines and accuracy were assessed, and inter-rater agreement was calculated using Fleiss' kappa coefficient.

Results: ChatGPT exhibited 90% compliance with guidelines and 85% accuracy, with a very good inter-rater agreement (Fleiss' kappa coefficient of .84, P < .01). ChatGPT handled multiple variations and descriptions and crafted concise patient letters.

Conclusions: Results suggest that ChatGPT could aid healthcare providers in making informed decisions and improve compliance with postcolonoscopy surveillance guidelines. Future research should investigate integrating ChatGPT into electronic health record systems and evaluating its effectiveness in different healthcare settings and populations.