ABSTRACT
Hearing loss is the second leading impairment worldwide. Childhood hearing loss has lifelong implications and
disproportionately affects individuals in low- and middle-income countries (LMICs). Up to 75% of childhood
hearing loss in LMICs is preventable due to the high prevalence of infection-related hearing loss. School hearing
screening is critical for identification of childhood hearing loss in low resource settings, where newborn screening
is unavailable. However, most screening programs only use pure-tone screening that does not identify middle
ear disease widespread in populations with a high prevalence of infection-related hearing loss. This is because
tympanometry, used to clinically identify middle ear disease, is expensive and designed for trained professionals.
Our goal is to develop and validate an mHealth tympanometer with machine learning diagnostic support
to transform this technology into a low-cost tool that could be broadly disseminated in LMICs, where the
burden of hearing loss is greatest and is not addressed by current hearing screening methodology. Our
study team is comprised of international leaders in hearing loss, audiology, data science, engineering, user-
centered design, and device development in LMICs. We have also partnered with hearX, a University of Pretoria
spinout company that developed the only validated mHealth pure-tone screening device. To test this new device
in an appropriate LMIC setting, we have partnered with the South African site from the Global HEAR
Collaborative, our consortium of collaborators from 28 countries that is the only international research network
dedicated to hearing loss. We documented the need for this device in a large cluster randomized trial recently
completed in rural Alaska, where tympanometry significantly improved the accuracy of school hearing screening
in a population with a high prevalence of infection-related hearing loss. Using data from this trial and pilot funding,
we are developing a machine learning tympanometry algorithm for lay screeners, and early hardware prototype
fabrication is underway. In Aim 1, we will refine the hardware prototype using a user-centered design approach,
cyclically incorporating feedback from South African team members during testing in a lab environment. In Aim
2, we will develop software through user-centered design that integrates the machine learning algorithm and
refined hardware prototype. The resulting mHealth tympanometer will advance to the R33 phase. Technology
development will be completed in Aims 3 and 4 through integration of the mHealth tympanometer with existing
health information technology and an early feasibility study in 15 preschool children in South Africa to optimize
device design for lay users. In Aim 5, we will validate the mHealth tympanometer with lay screeners through a
clinical performance study in 500 preschool children in South Africa. This technology, developed through
partnership and testing in an LMIC setting, will empower teachers and community health workers to identify
children at risk for preventable hearing loss. The Global HEAR Collaborative will provide infrastructure for future
studies with the proposed device across LMICs, directly addressing disparities in childhood hearing loss globally.
