Artificial intelligence -assisted tinnitus support with ChatGPT-4.

Abstract

Purpose: To evaluate the effectiveness of ChatGPT-4, an advanced AI language model, in delivering accurate, clear, and relevant information to patients managing tinnitus, especially in settings with limited healthcare access.

Methods: ChatGPT-4 was presented with ten common tinnitus-related Questions, guided by the 2014 American Academy of Otolaryngology guidelines. Ten ENT specialists rated its responses across multiple metrics using a 5-point Likert scale. Statistical analyses assessed accuracy, reliability, and overall response quality.

Results: ChatGPT-4 demonstrated high accuracy (mean = 85.5%, SD = 5.77), relevance (mean = 90.13%, SD = 5.00), and clarity (mean = 83.93%, SD = 5.13) across responses. Response times were consistently efficient (mean = 84.2%, SD = 5.28). Sensitivity and specificity were estimated at 90% and 95%, respectively. Inter-rater reliability analysis yielded moderate to substantial agreement (Cohen's kappa = 0.65-0.78), and effect size analysis indicated a large impact.

Conclusion: ChatGPT-4 provides accurate, timely, and patient-friendly information for tinnitus care, supporting its potential as an adjunct to conventional patient education. This AI tool could bridge gaps in healthcare access, particularly in resource-limited settings, enhancing patient support through accessible and reliable guidance.