Predicting Hearing Aid Outcomes Using Machine Learning.

Abstract

Introduction: The aims of this study were to measure the effectiveness of hearing aid (HA) fitting in improving understanding in quiet and in noise and to investigate the factors that significantly influence these results. This study will be carried out through a retrospective analysis of the results obtained from patients fitted with HAs at Amplifon HA centers between 2018 and 2021. This study explores and classifies the predictive factors of HAs outcomes, looking at the impact of HA technology, personalized adjustments made by the hearing care professional, and patient follow-up and daily use (data logging).

Methods: The study is based on the analysis of a large population of HA users who were fitted in HA centers between 2018 and 2021. It included 77,661 patients. HA outcome is measured through the improvement of intelligibility in quiet and noise. eXtreme Gradient Boosting machine learning method is used to identify predictive factors of HA outcome. SHapley Additive exPlanations Value analysis derived from the game theory is used to evaluate the individual impact of each factor.

Results: HA outcomes are significant in terms of both average improvement per patient of speech intelligibility and the percentage of patients improved. The analysis shows that the level of aided speech perception in quiet and noise is impacted by the choice of technology (category level and manufacturer), fitting parameters (amplification level and binaural loudness balancing) as well as by a high therapy adherence. In particular, binaural loudness balancing was shown to be systematically beneficial to all patients.

Conclusion: Big data analysis is a new relevant method to evaluate predictive factors for HA outcomes. It demonstrates HA efficiency to improve intelligibility in quiet and noise and shows the impact of hearing care professionals in maximizing patient's outcome through the selection of the most appropriate technology, fitting parameters, and a regular follow-up ensuring a high daily usage. However, global results must be interpreted with caution on such a heterogeneous population. They would need to be refined by an approach using clusters of patients with similar audiological profiles.