How Artificial Intelligence Enhances the Practice of Audiology



Author: Chris Heddon, D.O.

How much time per day could hearing care practices save by incorporating automated pure-tone audiograms into a busy clinical practice? Having this technology in place not only saves time, it also changes the clinic staffing model for hearing assessment. Can we gain insights about changes in overall health by assessing changes in cochlear health over time – even daily? If you believe that hearing thresholds change in response to changes in cochlear perfusion pressure (cardiovascular health marker) and cellular stress (immune response marker), then the answer obviously must be yes. In the next few years, automation of the hearing assessment, like automation of blood pressure assessment before it, will not only save time spent doing sound booth-based audiograms in the clinic, it will also give astute audiologists insights into the general health of the patient that few other healthcare providers have the training and clinical expertise to assess. In the very near future, pharmacies will have kiosks that perform automated hearing assessments, in addition to blood pressure readings, and mobile device operating systems will incorporate hearing assessments into their integrated health apps. Where will these pharmacy kiosks and mobile device health apps send users when there is change other irregularity in the user’s hearing assessment? It is incumbent upon audiologists and hearing care professionals to realize the opportunity that artificial intelligence (AI) and machine learning (ML) present to their profession. While automation technologies like these will do a great job of screening for threshold changes, only clinically trained human beings are suited to assess the ambiguous signal of hearing assessments and diagnostic audiograms. In the coming months you will be seeing a great deal of hype around AI in hearing aids, and it will be important to have a clear-eyed view of what is a business opportunity for you versus a marketing tool intended to sell more hearing aids.

Since the early 2010’s, artificial intelligence AI and ML have been grabbing headlines in the popular media. This has not been just hype. There has been tangible value creation in terms of market capitalization, user adoption, and our understanding of machine intelligence by companies like Google, Amazon, and Facebook who have invested in AI by hiring neuroscientists formally trained to program models of brain function into computers. With big tech companies spending billions on AI and hiring from an incredibly small talent pool of around 25,000 AI researchers worldwide, the market rate for candidates just out of PhD programs is well in excess of $300,000 per year. For experienced AI talent, salaries balloon above seven figures—not including signing bonuses. As high-tech and innovative as hearing aid companies are, this is not a space in which they are well positioned to compete for talent. Nevertheless, before the market for AI researchers became white hot, hearing aid companies had been working on various AI and ML approaches for quite some time. The removal of specific technological constraints, combined with the hearing aid industry’s need to address new and disruptive service delivery models, indicates that the time to bring AI to the hearing care market is now.

Until the widespread adoption of Bluetooth Low Energy (BLE)-enabled hearing aids, there was not a clear use case for either ML or AI in hearing aids. Hearing aids were small islands of computing power limited by their small energy-efficient microprocessors. There was simply no option to perform the computationally intensive processes necessary for AI. Now, BLE connectivity grants hearing aids connection, not just to mobile devices, but also to our greater cloud computing infrastructure.

Given the high cost of premium clinic-based audiology services bundled with today’s hearing aids, the next evolution necessary for the hearing industry to reach additional markets is in providing an intelligent hearing aid that can be programmed outside of the clinic. These next generation intelligent hearing aid systems require three specific elements: (1) hearing aids with energy-efficient wireless connectivity, which gives them access to external computing power; (2) the ability for an audiologist to securely program a hearing aid from a distance, which gives the user access to the highest level of hearing care at all times—even in real-world environments; and (3) mobile phones with sufficient computing power to run AI on-device, which support the dynamically responsive intelligent hearing aids and provide the additional benefits of protecting user privacy and reducing the mobile device power consumption associated with cellular connection to a cloud based server (which would have been needed if the AI was run in the cloud rather than on the user’s mobile device). The first element enables the latter two, while the latter two elements give end-users access to the best combination of human and machine intelligence.

Audiologists provide fantastic quality service to people who need hearing aids. This is the kind of personalized service I prefer as a hearing aid user, but with 15% market penetration among those who need a hearing aid, and an existing shortage of audiologists in the US market, it is clear that these professionals will not be able to serve all who need quality hearing care without a marked change in the technology underpinning the hearing care industry’s service delivery model. In addition to taking the much-needed step of making hearing aids over-the-counter (OTC), the industry needs to find a way to program OTC hearing aids to fit the specific needs and preferences of the end-user. Without the automation of initial and ongoing fitting, the OTC designation for hearing aids will be much ado about nothing. Look no further than Japan’s longstanding OTC market, which was often, and rightfully, referenced as a counterpoint to the OTC model in the United States. With device return rates approaching 50% for some online direct-to-consumer (DTC) channels, OTC, without intelligent connected hearing aids that give the end user some degree of personalization, is an unsustainable proposition. Perhaps more important than the OTC category in the United States, China is the hearing industry’s fastest growing market (between 3-4x current US and EU growth, according to publicly available Big 6 investor reports). According to the World Health Organization (WHO), over 45% of people 65 or older in China have moderate-severe hearing loss. With these numbers in hand, hearing aid manufacturers have an obligation to their shareholders to begin deploying novel service delivery models that respond quickly to the Asian market with economics that compete with commodity Chinese hardware—and that means embracing automation!

How close are we to intelligent connected hearing aids? Widex appears to be first to market with the announcement of their mobile-based ML platform at the 2018 American Academy of Audiology (AAA) conference, which was also described in an overview of the technology in the April 2018 Hearing Review. For clarity, AI and ML are terms often used interchangeably by companies to describe their products. Technically speaking, AI is a broader system of machine intelligence that embodies a set of more narrowly focused ML algorithms. Widex’s ML approach appears to be a method of determining the end-user’s “auditory intention” as it relates to a particular acoustic environment in order to reduce the amount of clinic time spent fine-tuning a hearing aid to the end-user’s preferences. It would be reasonable to assume that Widex will also compare the individual preferences of large sets of anonymized users in order to more efficiently suggest appropriate settings to each user. Over time, this population level approach should reduce the number of end-user interactions to something smaller than the current 20 interactions reported by Widex.

Again, Widex and others have been working on ML and AI solutions for quite some time. GN Hearing and Starkey have also built and published papers on desktop computer-based AI prototypes, However, these solutions have generally remained confined to the research lab. The real challenge with creating AI-driven hearing aids is in taking these novel technologies and deploying them on mobile devices. Without securely pairing with mobile devices capable of performing AI processes, hearing aid performance is held hostage to the connection speed of the mobile device as it attempts to offload AI operations to a server and, as noted above, increases the risk of exposing protected health information.

A connected, intelligent hearing aid requires the support of AI researchers who select appropriate algorithms for optimizing the performance of hearing aids, as well as specialized developers who know how to efficiently program high performance AI into mobile platforms. Right now, this talent predominantly works at places like Apple and Google, where they have recently begun publishing mobile AI tools for developers to incorporate into their apps, including hearing-related apps. However, these general-purpose AI tools have not been specifically tailored to the needs of hearing assessment, environmental detection, and device programming, so there is still a lot of customized development required when creating a mobile-optimized AI architecture for hearing aids, even when using the best available off-the-shelf solutions.

Based on the above, and despite their developmental costs and challenges, AI-driven hearing aids will drive the next wave of growth in the hearing aid market. They will provide better quality experiences for users, reach additional markets, and perform functions that will be seen as essential to users who benefit from the power of these AI-driven tools.
Five-Year Outlook in the Hearing Aid Industry
What will the next five years look like in the hearing aid industry, and how does AI fit into the future of hearing care?

There are several factors that will influence the hearing aid industry over the next five years to drive hearing aid prices down, increase the funds available for addressing hearing-related health issues, increase market penetration, and improve customer outcomes with improved hearing aids and hearing aid functionality.

A first significant factor affecting the hearing aid industry over the next five years is the introduction of over-the-counter (OTC) hearing aid sales and their effect on the price of hearing aids. With the OTC category soon opening in the US, ON Semiconductor, a major US-based manufacturer of chips for the global hearing aid industry, has released a system-on-a-chip platform that is specifically intended to allow startups and second-tier manufacturers to create hearing aids aimed at the OTC category in the US and the direct-to-consumer (DTC) Chinese market. If these new hearing aids, and others like them, gain traction in the mild hearing loss market (which is currently only 10% penetrated and makes up 75% of the overall market) and/or with value-conscious users who otherwise balk at the current average of $2,000 per hearing aid from premium first-tier manufacturers, there will be a sizeable downward shift in the price points of hearing aids. Return rates on OTC hearing aids are expected to present a challenge to retailers; however, embracing the automated personalization that AI offers may be exactly what is needed to reduce device return rates to a sustainable level and allow the hearing aid prices to fall.

A second factor is the increased recognition of the economic and emotional burden that the co-morbidities of hearing loss present to our healthcare system, our seniors, and their families. The ongoing Baltimore Longitudinal Study of Aging by Dr. Frank Lin and colleagues at Johns Hopkins demonstrates that hearing loss is independently associated with an increased risk of dementia (2x increased risk dementia for mild hearing loss; 5x increased risk for severe hearing loss) and accelerated loss of brain tissue on magnetic resonance imaging (MRI). In addition, Drs. Lin and Luigi Ferrucci, of the National Institute on Aging (NIA), have demonstrated that mild hearing loss is also associated with a three-fold increased risk of a fall, with a 1.4x increased risk for every additional 10dB of hearing loss. Even if there were no other co-morbidities associated with hearing loss, a reduction in the risk of dementia and falls alone are likely enough to offset the cost of covering a sub-$1000 pair of hearing aids. It is likely that more health insurance payors will follow UnitedHealthcare’s lead in offering a full hearing aid benefit, especially with the lower price points that the OTC category will bring to the US market.

A third factor will be the extent to which modern hearing care practices embrace AI and tele-audiology as tools for increasing throughput and margin for clinic services and the extent to which hearing aid market penetration increases above today’s current 15% penetration rate. There are several reasons to be optimistic that the penetration rate will grow, including: (1) the expansion of DTC service delivery models (including the OTC category in the US); (2) the related reduction in device cost; (3) increasing health insurance coverage; and (4) new entrants into the hearing aid manufacturing market, particularly at the value-conscious end of the market. In today’s market, it takes an average of around ten years for a person who could benefit from a hearing aid to actually purchase one. Lower cost, favorable form factors, and increased accessibility from pharmacies and online channels should reduce this adoption time.

Increased penetration in the value-conscious end of the market is likely to also lead to growth in the premium side of the market. Market data from commercial hearing aid pilots shows that, once a potential hearing aid consumer buys a value-focused hearing aid and sees a benefit, the customer is more likely to trade that model in for a premium hearing aid—the trade-in process ideally involving a referral to a hearing care practice for formal clinical exam. Similarly, if an OTC hearing aid offers a hearing aid screening for initial device fitting, the screening may also include an automatic referral for a full clinic-based hearing evaluation.

Based on these factors, the next five years in the hearing aid industry will see the price of hearing aids come down, an increase in the funds available for hearing-health expenses, an increase in market penetration, and an improvement in customer-patient outcomes.
What Does Artificial Intelligence Mean for the Practice of Audiology?
Patients with hearing loss are being underserved by the current service delivery model, which to date has not had time to fully incorporate technologies like AI and device programming-over-air. Modern audiology practices should embrace these emerging tools as a means of channeling underserved patients to receive appropriate hearing care and increasing their bottom line.

For example, it is evident that many people who should be screened for hearing loss are simply not being screened. Surprisingly, the American Academy of Family Physicians (AAFP) still does not recommend hearing screenings for asymptomatic patients over the age of 50. The low rate of screening contributes to hearing aids’ low market penetration. Factory-calibrated OTC hearing aids that perform an AI-driven, mobile-based, hearing screening for fitting should be positioned, in partnership with hearing aid manufacturers, as a customer acquisition tool for modern audiology practices. In such a role, OTC hearing aid sales will increase the number of people receiving hearing screenings, increase the referrals to audiology practices, and thereby improve the hearing outcomes for a greater percentage of the population.

The belief that the practice of audiology will change significantly with the introduction of intelligent, connected hearing aids is well-founded, but the idea that AI will steal market share from modern audiology practices is not. Look at the recent troubles that Uber and Tesla’s self-driving car programs have had as evidence that complex tasks in ambiguous circumstances are best performed by humans. Audiology is no exception. By embracing the value that AI brings to the field, audiology practices can expand the reach of their services and improve the outcomes for their patients. This is the power of AI and this is what is best for our patients and the growth of the industry.    
Chris Heddon, DO, is the CEO of Resonance Medical, based in Chicago, IL Dr. Heddon can be reached at chris@resonancemed.com.

Portions of this article were published at Hearing Health & Technology Matters (HHTM) blog and are reprinted here with their permission.