RP207 - Novel AI-Enabled Low-Cost Tympanometer to Expand Access to Childhood Hearing Screening in Underserved Communities

Abstract: Untreated childhood hearing loss has lifelong consequences and disproportionately affects underserved communities. Most cases are preventable and linked to untreated ear infections. School screening programs offer critical opportunities for early detection, yet most lack tympanometry—a key tool for detecting middle ear disease—due to cost and the need for trained professionals. We developed a low-cost, lay-friendly mHealth tympanometer integrated with a machine learning algorithm for automated tympanogram classification. The tool was validated in underserved South African communities. We present findings and next steps for scaling to expand access globally and close gaps in early identification and management of preventable hearing loss.

Summary: Rationale: By 2050, one in four people worldwide will experience hearing loss. For children, hearing loss has lifelong consequences, affecting language development, academic achievement, and vocational opportunities. In low-resource settings, up to 75% of childhood hearing loss is preventable, most commonly due to untreated ear infections. Tympanometry is an important tool for identifying middle ear pathology, but most screening programs exclude it because commercial tympanometers are expensive and require trained personnel. Schools offer a critical access point for hearing care in underserved communities, and prior work in rural Alaska has shown that incorporating tympanometry improves screening accuracy. However, the high cost and training burden of current devices have limited their adoption. Low-cost tympanometers are not commercially available, in part due to technological complexity, production costs, and a perceived lack of market demand.
To address this gap, we developed a low-cost, lay-friendly mobile health (mHealth) tympanometer using an interdisciplinary team science approach. We conducted a clinical validation study in early childhood development centers in underserved communities in South Africa. Here, we present early findings and implications for scaling.

Design: Children attending Early Childhood Development Centers in Paarl, Khayelitsha, and Mbekweni communities with existing hearing screening programs were eligible. Parental consent and verbal assent were obtained prior to testing. Each child underwent tympanometry twice: once by a community lay screener using the mHealth tympanometer, and once by an audiologist using a commercial device (Maico easyTymp), which served as the gold standard comparison. Otoscopy was also performed.
A validated machine learning algorithm automatically classified mHealth tympanogram results (types A, B, C), while audiologists manually classified commercial tympanogram results in real time. Data were entered directly into a secure research database. Analyses included descriptive statistics and diagnostic accuracy metrics.

Results: Between May and June 2025, 523 children were enrolled (33% female; mean age 4 years, range 2–6 years). Preliminary results indicate an overall agreement of 89% between the mHealth and commercial tympanometers, with a sensitivity of 99% and specificity of 67% for pass/refer classifications.

Conclusions: Early findings suggest that the low-cost, lay-friendly mHealth tympanometer performs comparably to commercial devices in identifying middle ear pathology in young children. These results support the feasibility of expanding tympanometry to school-based community hearing screening programs in underserved settings. Next steps will focus on refining technology and pursuing regulatory approval to enable broad dissemination and improved access to hearing health care globally.

Brief Summary of Clinical Takeaways: The clinical takeaway for this session is an understanding of the diagnostic performance of a low-cost, AI-enabled tympanometer compared to a commercial device. Attendees will learn how this technology can support early identification, extend screening beyond traditional clinical settings, strengthen referral pathways, and expand access to care in underserved communities.

Assumptions: Attendees should have a basic understanding of tympanometry principles, pediatric hearing screening, and common middle ear pathologies.