Caution with app-based hearing self-tests with vulnerable populations: preliminarily findings of digits-in-noise (DIN) testing vs. pure tone audiometry (PTA) in a rural, low-resource outreach clinic in Huruma, Tanzania

Introduction

Hearing loss is the fourth leading global cause of disability in the world (Ferrari et al., 2024). Over 5% of the world's population are affected by hearing loss and nearly 80% of people with disabling hearing loss live in low and middle income countries (LMICs) (World Health Organization, 2024). Hearing is fundamental to accessing education, fostering social inclusion, and improving employment prospects. Early hearing screening enables timely detection and prevention of ear disease, advancing equitable outcomes across health, education, and livelihoods. Shortages in Ear and hearing care (EHC) workforce, including audiologists, is one of the leading barriers to accessing EHC in LMICs (Waterworth et al., 2022). App-based hearing assessment tools have been developed to help bridge this gap by providing low-cost access to hearing screening.

Hearing assessment apps use either pure-tone thresholds, measuring a hearing threshold in decibels, or are speech-based, most commonly using digits in noise (DIN) (De Sousa et al., 2021). DIN testing presents the listener with recorded digit triplets and measures a speech recognition threshold (Smits et al., 2013). Speech-based tests can be more easily conducted in real-world settings, unlike pure tone audiometry (PTA) which requires an artificially quiet environment, and have been found to be an easy and reliable screening test (Kwak et al., 2022). However, DIN testing is affected by a user's native language and non-native speakers have been found to perform significantly poorer than native speakers on DIN testing (Potgieter et al., 2018). Furthermore, DIN testing is generally not suitable for young children due to the necessary maturity to use a smartphone and follow instructions. A study of 621 children aged 6–13 years old with normal hearing found age was a significant predictor of DIN SRTs and suggested a cut-off of 7 years for DIN testing to avoid false positives (Wolmarans et al., 2021). Conversely, older age has been found to negatively impact DIN test scores even after accounting for hearing loss (Potgieter et al., 2018).

Despite the large number of smartphone-based hearing assessment apps available relatively few have been validated against gold standard methods (Irace et al., 2021). A review of validated hearing assessment apps reported sensitivity values ranging from 0.982 to 1.00, however, studies focus on adults in high income populations (Bright and Pallawela, 2016). The hearWHO app was developed by the World Health Organization (WHO) and uses DIN to indicate speech recognition thresholds (SRT's) and has been validated for use by people aged 12 and over (World Health Organization, 2025). HearWHO employs antiphasic DIN testing, which has been validated against pure tone audiometry (PTA) and shown to significantly enhance sensitivity, specificity, and reliability compared to diotic presentation across both sensorineural and conductive hearing loss categories (De Sousa et al., 2020). Since its launch in 2019 it has been used extensively around the world, however uptake has been comparatively low in the global south (De Sousa et al., 2022). To account for non-native language and age effects the hearWHO app employs dynamic cut off values for SRTs, which has been found to allow testing in multilanguage populations (Potgieter et al., 2018). Instead of reporting SRTs in decibels the app coverts a user performance into a scaled score from 0 to 100, a score of below 50 indicates hearing loss, 50–74 mild difficulty in noise and above 75 good hearing (World Health Organization, 2025).

Despite the intention of developing app-based hearing assessment tools for use in resource poor contexts there is a paucity of literature exploring their utility in the setting for which they were intended. This study compares the hearWHO app against pure tone audiometry in a low-resource setting in rural Tanzania. This study was conducted in Huruma hospital, a district hospital with a catchment of approximately 300,000 people, that serves low-income and isolated communities near the border with Kenya.

Methods

A total of 52 patients self-presented to the ear, nose and throat (ENT) outreach clinics with ear-related problems over a 3-day period at Huruma Hospital in the Rombo district of Tanzania. All individuals who attended the clinic with an ear-related complaint were included in the initial clinical assessment, and each participant was assigned an identification number. Demographic data (age, sex, and occupation) and presenting symptoms (hearing loss, tinnitus, otorrhea, itching, and aural fullness; unilateral or bilateral) were recorded. Histories were taken either by a Swahili-speaking clinician or by an English-speaking clinician with Swahili translation. Education level, smartphone experience, and English proficiency were not recorded.

Following clinical assessment, patients were referred for audiological testing only if they met the inclusion criteria for hearing assessment, namely:

(1) A clinician-determined clinical indication for pure tone audiometry (PTA), and

(2) Age ≥12 years for the hearWHO (DIN) component, in line with WHO validation criteria.

Twenty-nine patients met these criteria and completed both pure tone audiometry (PTA) and hearWHO testing. Patients were excluded from pure tone audiometry (PTA) if audiological assessment was not clinically indicated, including cases of isolated wax impaction, acute ear infection requiring treatment first, absence of reported hearing loss, or young age. This reflected real-world constraints of a single audiologist and the need to prioritize patients most likely to benefit from testing.

Nine patients underwent pure tone audiometry (PTA) but did not complete hearWHO (DIN) testing. Five were excluded because they were younger than 12 years, and therefore ineligible for hearWHO testing. The remaining four either declined hearWHO (DIN) testing or were inadvertently missed during clinic flow due to operational constraints.

All patients whom underwent hearing testing completed pure tone audiometry first followed by hearWHO testing. Pure tone audiometry was performed be the same audiologist in a designated clinical room using a calibrated portable audiometer (Interacoustics AD226) with TDH-39 headphones in a designated clinical room. Prior to full testing the audiologist provided an initial explanation of the hearing test and performed a conditioning test at 1,000 Hz. Air conduction thresholds were recorded at 500, 1,000, 2,000, and 4,000 Hz. Pure tone averages were calculated for each ear using four frequencies, followed by a combined bilateral pure tone average.

HearWHO (DIN) testing was performed using the hearWHO app installed on an Apple iPhone XR. JBL 520BT Bluetooth headphones were used because no wired audiometric transducers were available in the outreach setting, and the iPhone XR does not include a 3.5 mm audio port, making Bluetooth the only feasible connection method. Although hearWHO does not specify a required headphone type, WHO guidance indicates that the app is designed for consumer-grade smartphone use rather than calibrated audiometric transducers. The Bluetooth headphones were fully charged at the start of each clinic day and retained sufficient battery throughout testing; no mid-day charging was required, and no connectivity interruptions occurred.

All patients who completed hearWHO (DIN) testing were facilitated by an English speaking clinician and Swahili speaking nurse translator. The clinicians in the HearWho testing room were blind to the pure tone audiometry (PTA) result. Before hearing assessment the app measured the in room noise level and did not allow testing to proceed until a sufficiently low noise reading, which the app sets at 60 dB. Additionally, background noise testing was performed on a Samsung Galaxy S20 FE with Decibel X fast A-weighting soundlevel meter across six separate timeframes with readings between 51 and 56 dB in the testing rooms. Following noise level reading, connectivity with headphones is then confirmed in the app and the patients first language is selected. All patients had Swahili selected as their first language. Before hearWHO (DIN) testing commenced the patient underwent a series of three conditioning tests. During conditioning testing, patients were given the opportunity to ask clinicians questions and clarify their understanding of how to use the app. Upon completion of the test, the score generated by the app was recorded.

The sample was described, and the relationship between the hearWHO score (which represents SRTs) and pure tone audiometry (PTA) was examined using both Pearson's and Spearman's correlation. Hearing loss on pure tone audiometry was defined by using a four-frequency pure tone average of >25 dB hearing loss, consistent with WHO criteria for disabling hearing impairment. Regression modeling and diagnostic accuracy calculations were then performed for the whole group and stratified by age band.

According to local institutional guidelines, this project met criteria for service evaluation/public health practice and therefore did not require formal institutional review board approval. Informed consent was obtained from all participants before testing.

Results

The overall cohort (n = 52) had a mean age of 43.7 years (range 2–85) and was predominantly female (61.5%, Table 1). Within the test population (n = 29), the mean age was higher at 51.3 years (range 12–79), with 58.6% female. Self-reported hearing loss was common, affecting 75% overall and 82.8% in the test group. Tinnitus was reported by around one-third of the test population, while otalgia and otorrhea were less frequent. On otoscopy, approximately half of ears appeared normal, with wax being the most frequent abnormal finding, followed by perforation and other pathologies (Table 1).

Table 1. Patient demographics, clinical history, and examination findings.

^*All patients who completed both PTA and DIN testing.

Exploratory Pearson and Spearman correlations, as well as preliminary regression modeling, were conducted to assess the relationship between hearWHO scores and PTA thresholds; however, these analyses did not yield clinically meaningful associations and were therefore not included in the final Results, which focus on diagnostic accuracy outcomes. In the unstratified analysis (N = 29), the hearWHO (DIN) test showed a sensitivity of 0.45, indicating moderate ability to detect true positives (Table 2). Specificity was very low at 0.11, reflecting poor identification of true negatives. The positive predictive value (0.53) was stronger than the negative predictive value (0.08), suggesting that positive results were more reliable than negative ones. Overall accuracy was 34.5% (Table 2), highlighting limited diagnostic reliability in this sample.

Table 2. Overall diagnostic performance of hearWHO (DIN testing).

Ear-wise diagnostic performance showed that hearWHO correctly identified all ears with pure tone average-defined hearing loss (sensitivity = 1.00 for both ears, Table 3). Specificity was low (left = 0.08; right = 0.05, Table 3), reflecting frequent false-positive classifications. Positive predictive values were modest (left = 0.57; right = 0.36, Table 3), whereas negative predictive values remained perfect (1.00 for both ears, Table 3), indicating that a “normal” hearWHO (DIN) result reliably corresponded to normal pure tone average thresholds. Overall accuracy was 0.59 for the left ear and 0.38 for the right ear (Table 3). Agreement between hearWHO (DIN) and pure tone average was limited: Cohen's κ was 0.08 for the left ear and 0.04 for the right ear, indicating poor agreement beyond chance. Collectively, these findings suggest that hearWHO (DIN) is highly sensitive for detecting hearing loss but demonstrates low specificity and limited concordance with pure tone average thresholds.

Table 3. Ear-wise diagnostic performance of hearWHO (DIN) compared to pure tone audiometry (PTA).

When diagnostic performance was stratified by age, sensitivity declined with increasing age, from 0.67 in participants under 40 to 0.29 in those aged 60 and above, indicating reduced detection in older groups (Table 4). One participant had missing age data and was, therefore, included in the overall diagnostic analysis but excluded from the age-band breakdown, resulting in a total of 28 ears in the stratified table. Specificity was low across all age bands, with only slight improvement in the oldest group. Overall accuracy was modest, highest in the < 40 group (0.50) and lowest in the 60+ group (0.27).

Table 4. Age-stratified diagnostic performance.

^*One participant had missing age data and was included in the overall analysis (Table 2) but excluded from age-band breakdown (Table 4).

Discussion

This study addresses a critical evidence gap by evaluating the hearWHO app in the context for which it was designed and sheds light on why uptake in this setting is poor. Validation of smartphone-based hearing tests is crucial to determine their utility. Ideally such apps should be tested and validated against the gold standard of pure tone audiometry (PTA) in the context for which they are designed. Despite many apps not having undergone formal validation, sensitivity values for those that have been tested range between 0.982 and 1.00 (Bright and Pallawela, 2016), which is a positive indication that such technology can have a role in direct user hearing assessment.

The results presented here are contrary to those reported in the literature with a moderately low sensitivity of 0.45 and a very low specificity of 0.11 of the hearWHO app score in relation to pure tone audiometry (PTA). The hearWHO app in this context was more reliable in determining those that had normal hearing (positive predictor value of 0.53) but less reliable in determining those with moderate or severe hearing loss (negative predictive value of 0.08). This differs from the results reported by Potgieter et al. (2018) who reported sensitivity and specificity of 0.962 and 0.903, respectively.

The data reported here, showing an overall test accuracy of 34.5%, indicate that hearWHO (DIN) is not suitable in this setting. The most plausible explanation is the participants' native language. All individuals in this study were native Swahili speakers, and although Swahili and year of birth were selected on the app interface prior to testing, both the test instructions and the hearWHO (DIN) assessment were delivered in South African English. Additionally, smartphone familiarity may have been a factor in poor hearWHO (DIN) performance. While mobile phones are nearly universal across Sub-Saharan Africa, smartphone ownership remains far less common (Poushter, 2016).

However, before completing the full in-app hearing assessment, users of the hearWHO app undertake a short practice version to familiarize themselves with the testing format. In addition, participants in this study received support from a Swahili-speaking clinician, who clarified instructions and answered questions before the full test. It is therefore reasonable to assume that the hearWHO (DIN) results reported here are enhanced compared to what would have been obtained without such support.

Similar to previous studies, older age negatively impacted DIN test scores in this study (Potgieter et al., 2018; Armstrong et al., 2020). However, one study demonstrated DIN testing to be an effective tool for hearing assessment in older aged people with moderate or severe hearing loss, especially when combined with counseling (Koole et al., 2016). Participant familiarity with smartphone technology may have been a factor in age-related performance, as older adults are less likely to be familiar with smartphone technology (Andone et al., 2016).

This study was limited by small sample size, which reduces statistical power and restricts generalizability of the finding to wider populations. Additionally, the very small number of true negative cases substantially constrained the specificity estimates, making them unstable and highly sensitive to single-case variation. Participants prior exposure to smartphone technology was also not explored. Although hearWHO is designed as a self-administered test, participants in this study required assisted administration due to language barriers, which may have artificially improved performance and does not reflect real-world use among individuals with different first-language backgrounds.

Nevertheless, the results presented here raise important questions regarding the utility of DIN testing in non-native language and low resource settings. It can be argued that smartphone technologies intended to serve those in LMICs should be co-designed in partnership with those communities and developed within the contexts in which they are to be used. Further work is needed to provide a broad range of language options or explore alternative non-language-specific testing methods to ensure accessibility and effectiveness across diverse populations. Additionally, the efficacy of DIN testing should be explored in context-relevant settings.

App-based hearing assessment tools have the potential to address critical shortages in EHCs services in LMICs by offering affordable and accessible screening. Hearing screening for children is a WHO priority as early detection of hearing loss is critical to prevent delays in speech, language, cognitive, and social development, and ensures timely intervention that supports educational achievement and lifelong communication (World Health Organization, 2024). This study did not explore the use of the hearWHO app in children younger than 12, though studies suggest that DIN testing could be used for children aged 7 and over (Wolmarans et al., 2021). It can be argued that the development of mobile hearing screening for children should be prioritized as EHC interventions in this population are likely to have a greater impact. Unfortunately, pediatric DIN testing has not been effective at delineating conductive hearing loss (CHL) from normal hearing and is not a suitable screening tool for pediatric CHL (McBride et al., 2025). Further work is needed to develop mobile testing that can differentiate between conductive and sensorineural hearing loss. Future studies that aim to explore the validity of app-based hearing tests designed for low-resource settings should include a larger sample from multiple regions with participants whose primary language is not English.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Katijah Khoza-Shangase, University of the Witwatersrand, South Africa

Reviewed by: B. Deepashree Joshi, Sree Balaji Medical College and Hospital, India

Col Suraj Thapa, Indian Army Med Corps, India