Improving Access and Automating Fitting for Hearing Aid Devices for the Underserved and Rural Populations

Abstract This proposal responds directly to the purpose of the RFA-MD-22-003 STTR Funding Opportunity Announcement (FOA). Nearly 44 million Americans experience hearing loss, but as much as 86% of adults who could benefit from hearing aid devices (HAD) do not use them. Untreated hearing loss is almost twice more prevalent in rural areas. More than 46 million people live in rural America, but these communities face significant barriers and delays in accessing care and experience a lack of providers for audiology services. The average retail price of HADs was $4,700, making it prohibitively expensive the rural and minority populations who have lower incomes. Hearing loss is also highly prevalent in black and Latino/Hispanic communities and growing, but older black adults were 58% and Mexican American older adults were 78% less likely to use HADs, due to reasons of high cost and lack of insurance coverage. Aside from cost, concerns about comfort, effectiveness and satisfaction are the other significant barriers to HAD usage. A long-term goal of hearing health professionals, the auditory research community, and HAD makers has been to reduce these barriers. The current hearing care delivery in the U.S. follows a model of specialty clinic- based care, through audiologists, which is costly and time-consuming for the rural and minority populations. The introduction of over-the-counter (OTC) HADs is set to disrupt this model and increase accessibility and affordability of hearing care for these populations. Automation of HAD fittings and self-adjustments further enable listeners to pick gain settings according to their individual preferences, which significantly increases listener satisfaction and acceptance. It is common practice today for an audiologist to follow a time-consuming process in which she fine tunes the HAD settings to the users’ individual demands, based on verbal descriptions from the user. This approach is burdensome to both the audiologist and the HAD user and might not provide the satisfaction and comfort the user truly desires. Allowing users to make these changes easily via automation can vastly improve their satisfaction and use of HADs. This project will develop machine learning (ML) based automatic gain adjustment software in a smartphone with a Graphical User Interface (GUI), to minimize the effort required for individualized settings in an OTC HAD. Recent studies of self-adjusted gain have shown consistent within-subject preferences but gain settings that differ greatly from prescribed settings (up to 20dB), and also shown a great deal of between- subject variability in gain preferences. We propose a Deep Neural Network (DNN) based prediction model that learns the non-linear relationships in the data and outcomes, otherwise difficult to ascertain through normal statistics. The DNN will be combined with other ML techniques to refine the final gain settings. With smartphone assistive software and the predictive power of ML, our proposed solution will automate personalized fittings for OTC HADs, significantly improving their accessibility, affordability and acceptance, thus benefitting the rural and minority populations the most. Automating HAD gain settings with the use of machine learning (ML) techniques will increase the ease of use and reduce costs while improving intelligibility and comfort. OTC HADs will improve access and cost and will lead to much higher adoptions of HADs in the rural and minority communities, improving their quality of life.