Informing patients with otologic balance disorders: A performance review of ChatGPT-4

Abstract

This cross-sectional evaluation aimed to evaluate the quality of patient education materials provided by ChatGPT regarding otologic balance disorders. A total of 126 patient-oriented questions covering 9 common vestibular conditions – including benign paroxysmal positional vertigo, vestibular neuritis, labyrinthitis, Meniere disease, superior semicircular canal dehiscence, persistent postural perceptual dizziness, perilymph fistula, presbyvestibulopathy, and acoustic neuroma – were submitted to ChatGPT version 4o. The responses were independently evaluated by 2 otolaryngologists using the DISCERN tool to assess information quality and the PEMAT-P tool to evaluate understandability and actionability. The mean DISCERN score was 48.06 (range: 44.0–53.0), indicating moderate quality. PEMAT scores averaged 80% for understandability (range: 75%–88%) and 43% for actionability (range: 40%–60%). While the outputs were generally easy to understand, many lacked actionable guidance. In terms of information quality, the responses were generally acceptable for patient education purposes, though they occasionally included inaccuracies or omissions. ChatGPT may serve as a supportive tool for patient education on vestibular disorders but should be used with professional oversight to ensure safe and accurate communication.