Abstract
OBJECTIVE: This study aimed to evaluate the reliability and usefulness of information generated by Chat Generative Pre-Trained Transformer (ChatGPT) on temporomandibular joint disorders (TMD). METHODS: We asked ChatGPT about the diseases specified in the TMD classification and scored the responses using Likert reliability and usefulness scales, the modified DISCERN (mDISCERN) scale, and the Global Quality Scale (GQS). RESULTS: The highest Likert scores for both reliability and usefulness were for masticatory muscle disorders (mean ± standard deviation [SD]: 6.0 ± 0), and the lowest scores were for inflammatory disorders of the temporomandibular joint (mean ± SD: 4.3 ± 0.6 for reliability, 4.0 ± 0 for usefulness). The median Likert reliability score indicates that the responses are highly reliable. The median Likert usefulness score was 5 (4-6), indicating that the responses were moderately useful. A comparative analysis was performed, and no statistically significant differences were found in any subject for either reliability or usefulness (P = 0.083-1.000). The median mDISCERN score was 4 (3-5) for the two raters. A statistically significant difference was observed in the mean mDISCERN scores between the two raters (P = 0.046). The GQS scores indicated a moderate to high quality (mean ± SD: 3.8 ± 0.8 for rater 1, 4.0 ± 0.5 for rater 2). No statistically significant correlation was found between mDISCERN and GQS scores (r = -0.006, P = 0.980). CONCLUSIONS: Although ChatGPT-4 has significant potential, it can be used as an additional source of information regarding TMD for patients and clinicians.