Advanced applications in digits-in-noise testing to detect and differentiate hearing loss

Abstract

More than half a billion people have disabling degrees of hearing loss, which, left untreated, has debilitating consequences to the individual and society. Prevalence is expected to increase rapidly within the next thirty years, making hearing loss a significant public health matter. Thus, increasing efforts should be made towards detection and treatment. Many people with hearing loss reside in low- and middle-income countries, where the capacity to provide care, especially clinic-based models of care, is limited. Furthermore, the inaccessibility has been exacerbated by the COVID-19 pandemic. The digits-in-noise (DIN) test has been a more accessible screening tool over the past two decades, measuring a speech recognition threshold (SRT) that has a high association with standard pure tone audiometry. The test has the benefit of being provided directly to the public over digital modes like smartphones, using familiar stimuli and a simple procedure that does not require calibration. One example is the hearWHO DIN test that has been widely used and promoted as a free hearing screening test to the public. Aside from detecting hearing loss, no studies have developed methods to differentiate and classify hearing loss further. Therefore, this study investigated more advanced DIN test methods that could serve this purpose. Study I evaluated if a combination of two DIN test paradigms (antiphasic and diotic) could accurately categorise hearing into (a) normal hearing (pure tone average [PTA] ≤ 25 dB HL), (b) bilateral sensorineural hearing loss (SNHL; PTA > 25 dB HL), or (c) unilateral SNHL (PTA > 25 dB HL in the poorer ear and ≥ 20 dB interaural PTA difference) or conductive hearing loss (CHL; air conduction PTA > 25 dB HL and ≥ 20 dB air-bone gap). After establishing normative antiphasic SRTs across a sample of 489 adults with varying types and degrees of hearing, 393 participants completed a second diotic DIN test. The antiphasic DIN test had sensitivity and specificity of 90% and 84% to detect hearing loss. Furthermore, the combined antiphasic and diotic DIN test approach with fixed SRT cut-offs could correctly categorize 75% of the sample. Using a fixed antiphasic and sloping diotic SRT cut-off (varying slope and offset) could increase classification to 79%. False-negative rates for both procedures were below 10%. Study II investigated a different approach to determine if CHL could be accurately distinguished from bilateral SNHL using a combination of pure tone audiometry and a diotic DIN test. An analyses of 122 adults with bilateral SNHL and 36 with CHL was conducted. Binomial logistic regression determined the effect of pure tone thresholds, SRT and age on the likelihood of having CHL or bilateral SNHL. A model including low-frequency PTA (0.5 & 1 kHz), diotic DIN SRT, and age had sensitivity and specificity of 97.2% and 93.4%, respectively, to distinguish CHL from bilateral SNHL. Instead of establishing a hearing loss type, Study III aimed to determine if a low-pass (LP) and high-pass (HP) speech filtering technique could estimate pure tone audiometry in separate low and high-frequency bands. Previous work has used LP filtered masking noise to increase the sensitivity of the DIN test to high-frequency hearing loss. However, this study filtered speech at 1.5 kHz to ensure minimal speech information presented above or below the filter cut-off frequency. Results indicated better test-retest reliability (Intraclass correlation coefficient [ICC] = 0.71; 95% confidence interval [CI] 0.52 to 0.82) of the HP DIN test than the LP DIN test (ICC = 0.39; 95% CI -0.01 to 0.63). The HP DIN SRT was more strongly correlated to all the PTA averages (four frequency, low-frequency and high-frequency) than the unfiltered, broadband (BB) or LP DIN test. Subsequently, the HP DIN test showed increased sensitivity and specificity to detect hearing loss in any PTA average, compared to the BB or LP DIN test. The LP DIN test had a weaker correlation to low-frequency thresholds than the BB DIN test. As a result, a combined LP and HP DIN test approach could not accurately predict an audiometric slope or configuration. For ears with normal hearing (PTA ≤ 15 dB HL), the HP DIN showed a stronger correlation (rs = 0.36) to extended high frequencies (8 to 16 kHz) than the BB DIN (rs = 0.26). As an implementation research approach, study IV investigated the global use and uptake, test characteristics and performance of an antiphasic DIN test as provided on the free World Health Organization smartphone hearing screening test (hearWHO). The data of 242 626 tests conducted by adults (> 18 years) conducted between February 2019 and May 2021 were evaluated. The test was completed in nearly every country globally (n = 179/195), with the greatest uptake seen in China and India. Uptake was most significant in the Western Pacific (32.9 %) and European (24.8 %) WHO regions. As expected, referral rates were typically higher for older age groups in most WHO regions, except for the African and Eastern Mediterranean regions, where overall hearWHO test uptake was lowest. There was a high uptake of tests (44%) by young adults under 30 years. The sequential antiphasic-diotic DIN test approach to classify hearing loss has the potential to optimise care pathways using remote and contactless testing by identifying unilateral SNHL and CHL as cases requiring medical referral. In contrast, bilateral SNHL cases could be referred directly to a hearing care professional or be served using non-traditional models. Furthermore, considering restrictions on traditional audiological assessments due to an infectious disease like COVID-19 and under-resourced settings, alternative methods that enable audiological care with minimal physical contact may reduce mortality and infection risk whilst optimising care pathways and resource allocation. This DIN test approach and the combined pure tone audiometry and diotic DIN test method could allow accurate detection of CHL without the use of bone conduction testing conducted in sound-proof booths. The DIN test could further sensitively discriminate hearing, especially occurring in high-frequencies when using HP speech filtered stimuli, and shows potential to detect early signs of hearing loss occurring in the extended high frequency (≥ 8 kHz) range. As an applied public health practice, the test reaches an important target audience of younger adults positioning it as an important measure for public health advocacy to prevent hearing loss due to unsafe listening practices. This study project provides empirical evidence that DIN test methods can support improved detection and classification of different hearing loss types. These advances contribute to growing research to optimise the DIN test efficiency and sensitivity as a screening and potential triaging tool. Furthermore, these classification methods can provide simple, applied solutions that support alternative service delivery models like over-the-counter or direct-to-consumer pathways.