Authors: Małgorzata Pastucha¹, Anna Ratuszniak¹, Małgorzata Ganc¹, Edyta Piłka¹, Iryna Drohobycka¹, Henryk Skarżyński¹, W. Wiktor Jędrzejczak¹
¹Institute of Physiology and Pathology of Hearing
Background: Hearing loss is a growing global health issue, yet effective long-term monitoring remains challenging due to limited access to specialists and the complexity of interpreting self-test results. With the rise of mobile hearing test apps, there is an increasing need for tools that can interpret these results and guide users on whether specialist consultation is necessary.
Methods: This study assessed the ability of ChatGPT-4.0 to support decision-making in hearing health monitoring. The model was tested on 1,000 synthetic cases simulating monthly hearing threshold data collected over 12 months. For each case, ChatGPT was asked to determine whether a specialist consultation was warranted. Its recommendations were compared with the judgments of five audiology experts. Agreement was evaluated using percent accuracy and Cohen’s Kappa. A multiple-response strategy – selecting the most frequent recommendation from five trials – was also analyzed.
Results: ChatGPT’s recommendations showed strong alignment with expert judgments, achieving agreement rates between 0.80 and 0.84 across individual trials. Accuracy increased to 0.87 when using the most frequent decision across five separate queries. In cases of unanimous expert agreement, ChatGPT reached near-perfect accuracy (0.99). The model also demonstrated the ability to account for random fluctuations in longitudinal data, offering more nuanced recommendations than a fixed-threshold criterion.
Conclusions: This study demonstrates that AI, when integrated with self-monitoring tools, can support accurate and consistent hearing-related decisions. ChatGPT shows strong potential as a decision-support tool for hearing self-monitoring. By linking with mobile apps, it can help users understand when to consult a specialist, bridging the gap between automated testing and clinical expertise. Its ability to adapt to longitudinal patterns makes it a promising asset in scalable, patient-centered hearing healthcare.
