Limitations of a Mobile Hearing Test Application

Abstract

Figure: iStock/a-imageHearing impairment is one of the most prevalent debilitating illnesses affecting the global populations. The World Health Organization (WHO) estimates that over 360 million people suffer from disabling hearing loss (40 dB loss for adults, 30 dB loss for children; WHO, 2017 http://bit.ly/2gAVBAk). Hearing loss is a high-impact research area because more than half of disabling hearing loss cases can be prevented (WHO, 2017).Figure: Adam W. Pickens, PhD, MPHFigure: Lakshmi Dakuri Robertson, MSPHGovernment agencies in the United States and Europe have identified noise exposures across multiple occupational settings as high priority due to huge population segments reporting hearing loss (NIOSH, 2012 http://bit.ly/2oMOloP; EU-OHSA, 2005 http://bit.ly/2oN2v9d). Regardless of the work setting (e.g., rural or urban, public or private), there is a growing demand for a valid alternative hearing screening methodology. One of the primary driving factors for this demand is the projected shortage of qualified audiologists. This shortage is predicted to be driven by the strain from the increasing number of people who needs hearing tests, particularly the growing elderly population (Gerontologist. 2003;43[5]:661; ASHA, 2013 http://bit.ly/2oMPkFo).Figure: Matthew Lee Smith, PhD, MPH, CHESFigure: Hongwei Zhao, ScDCurrently, pure-tone audiometry is performed by a registered audiologist and considered the industry gold standard (Gerontologist. 2003 http://bit.ly/2oMPkFo). Numerous self-administered tests are used for diagnosing hearing loss among older populations. Additionally, an array of clinician-administered screening tools are available (J Fam Pract. 2003;52([1]:56 http://bit.ly/2oMC2sy; JAMA. 2003; 289[15]:1976 http://bit.ly/2oN05aD). However, the efficacy of these tests is strongly administrator-dependent, which can produce highly variable results with low reproducibility and sensitivity (Gerontologist. 2003 http://bit.ly/2oMPkFo).Figure: Ranjana Mehta, PhDFigure: Sejun Song, PhD, MSPortable microprocessor-based methods to test hearing have recently emerged and shown the ability to accurately reproduce audiologist-quality hearing screenings (Int Arch Otorhinolaryngol. 2013;17[3]:257 http://bit.ly/2oMJW4Y). While not full-spectrum screenings, mobile applications are proliferating and are increasingly embedded in the field of public health for hearing screenings (Hearing Journal. 2017;70[1]:14 http://bit.ly/2oMYGRl). While pure-tone audiologist-administered tests remain the gold standard, the development and implementation of screening alternatives, regardless of their drawbacks, point to the need for portable alternatives that produce reliable results in at-risk populations. To address this screening need, our research team developed the self-administered, pure-tone, full-spectrum “hEAR” mobile screening application, which aims to produce quality screening results while increasing consumers’ access to automated, full-spectrum audiological screenings. STUDY DESIGNTable 1: Summary subject response data statistics for Sound Pressure Level (SPL) in the quiet roomThis was a population prospective cohort pilot study. All recruitment and data collection, analysis, and handling procedures were approved by the Texas A&M University Institutional Review Board (TAMU-IRB). The study aimed to validate the data collection capability and quality of the hEAR application against an audiologist-administered audiometric test. Participants were over 18 years of age and without pre-existing diagnosed hearing loss. Subject trials were randomized and counterbalanced based on initial recording of data. Data for subjects were counterbalanced so that half of the subjects initiated data collection procedures in the laboratory and the other half with the audiologist.Table 2: Summary subject response data statistics for Sound Pressure Level (SPL) in the noisy roomA Samsung Galaxy Tab™ 3.0, an Android device, was chosen to test the hEAR mobile application because it runs on the Android platform architecture, which the hEAR application was built upon. Bose® AE2, readily available off-the-shelf headphones, were used for data collection. An Extech® 600 SPL meter was used to collect background sound pressure levels for the testing rooms. The hEAR application itself was designed based on best practices from a variety of sources, notably the recommendations from the WHO endorsing the Bekesy-style audiometry for self-administered hearing screenings (Franks, 1995 http://bit.ly/2oMTLAa). Additionally, the application complies with Occupational Safety and Health Administration (OSHA)-designated hearing screening specifics as per 29 CFR 1910.95 and the appropriate Appendices (C)(D) for the collection of hearing data in audiometric testing equipment (OSHA, 2015 http://bit.ly/2oMRMLS). For the hEAR laboratory collection procedures, participants were assigned a testing room: noisy (n=15) or quiet (n=15). The data revealed an average ambient background in the noisy environment of 48 dBA and 13 dBA in the quiet environment. Each test involved touching the device screen to begin the testing procedure, after which the application produced randomly assigned frequencies from the previously indicated ranges. Working within the OSHA and WHO guidelines, the application automatically administered a series of “mini-trials” (125; 250; 500; 1,000; 2,000; 4,000; and 8,000 Hz). Every frequency was randomly administered a minimum of four times bilaterally. Each participant underwent at least 28 mini-trials per full trial; each trial ran for 15-20 minutes. In total, there were three complete trials (Trial 1, 2, and 3) for the laboratory data collection and one audiologist-administered pure-tone audiometric test (Trial 4) control variable per participant. With respect to analysis of individual frequencies, the overall mean decibel response values per frequency were calculated for all participants. All statistical analyses were performed using SAS® (Version 13.1). Data were analyzed using a mixed effect model to accommodate the fact that each subject was measured repeatedly at different frequencies and for different trials. Individual subjects were used as random effects. Fixed effects include trial and frequency (seven levels), and their interaction, in which Trial 1 and the frequency of 8,000 Hz were used as the reference levels for analysis (α=0.05). Additionally, analyses were performed in separate testing rooms. STUDY RESULTS Of the 30 subjects, 21 were male and nine were female, with ages ranging from 21 to 67 years old. Most (24) of the subjects were between the ages of 21 and 28 years old. Most (22) of the subjects were undergraduate and graduate students, while eight were staff members at Texas A&M University.Table 3: Estimate outputs for the quiet roomResults from a Repeated Measures ANOVA revealed that Trials 2 and 3 were parallel to Trial 1 throughout the frequencies (no interaction). For the quiet room, when Trial 1 was compared with the audiologist's trial, there were no statistically significant differences for frequencies 250Hz, 500 Hz and 1,000 Hz. Similar results were obtained for the comparisons between Trials 2 and 4 and between Trials 3 and 4 (Table 3).Table 4: Estimate outputs for the noisy roomFor the noisy room (background SPL 48 dBA), there were statistically significant differences between Trial 1 and the audiologist data (Trial 4) at all frequencies. Similar results were obtained for the comparisons between Trials 2 and 4 and between Trials 3 and 4 (i.e., statistically significant differences at all frequencies; Table 4). The app trial measurements in the quiet room (background SPL dB 13 dBA) statistically differed from those of the audiologist trial specifically at certain higher frequencies (i.e., 2,000 Hz; 4,000Hz; and 8,000 Hz). In the noisy room, all frequencies showed a significant difference when compared with audiologist data. Analysis revealed learning curves between trials for numerous frequencies in both the quiet room and the noisy room. Trials at 125 Hz were discarded, as the data from the audiologist were missing for most subjects. DISCUSSION The quiet room, which had lower ambient noise levels, had individual test frequency measurements that were not statistically different from the audiologist test at multiple frequencies. This indicates that the ambient noise levels of the testing environment play a significant role in the quality of the data. When the results for the quiet room were examined alone, mitigating the effect of ambient noise, the statistically different results at the higher frequencies (i.e., 2,000 Hz; 4,000 Hz; and 8,000 Hz) were interesting. These results were most likely due to the headphones that were used. These were over-the-ear headphones, but were not optimized for audiological testing. This is seen in a report published by Inner Fidelity indicating the headphones were capable of good low-frequency sound reproduction, but were not effective at reproduction in the frequency spectrum from 2,000 Hz to 8,000 Hz (Hertsen, 2012). The implications of this work indicate the strong need for a relatively flat frequency reproduction spectrum for headphones being used in mobile hearing screenings. From a health standpoint, the results from this study have potentially significant implications. The results for the quiet room have reproducibility in similar or better test conditions. Though audiometric test rooms are permitted to have a certain level of ambient noise, it may be probable to designate exceptionally quiet areas for hearing tests only. The hEAR mobile application can then be used to ensure workplace compliance with 29 CR 1910.95 for hearing screening. The results collected by the hEAR application are very promising for the quiet environment. Under slightly different conditions, such as better environmental analysis with respect to ambient noise and/or using better headphones, the app can potentially collect audiologist-quality hearing screening data. These results are similar to other studies using portable, software-based audiometers (Int Arch Otorhinolaryngol. 2013 http://bit.ly/2oMJW4Y;Hearing Journal. 2017 http://bit.ly/2oMYGRl). However, these studies indicate that various off-the-shelf headphones are available for testing, while the current study indicates the contrary. Headphone acoustics is of the utmost importance in hearing screenings. The quality of data collected, combined with its portability and the ease of access, makes the application a suitable first line of defense for hearing screening of at-risk populations such as those in high-noise occupations or in vulnerable and underserved areas.