The timely diagnosis of hearing loss in children has significant implications for a child’s social and cognitive development. In young children, hearing loss is often inconspicuous resulting in delayed diagnosis and rehabilitation. As a result, children are at increased risk of delayed speech acquisition and the subsequent long-term sequelae. Speech delay and hearing loss also represents a significant cost to society due to the resources required for their treatment, such as special education, speech therapy, and social services [1, 2]. However, it is clear that early detection is the key to restoring normal speech development and a favorable long-term outcome [3, 4].

Definitive diagnosis of hearing loss is typically made through audiologic assessment of pure-tone air, bone and speech thresholds. Traditionally, pure-tone thresholds are documented by asking the subject if they can hear tones of varying frequency and intensity. While auditory brainstem reflex and otoacoustic emissions testing remain the gold standard of hearing assessment, these investigations pose some difficulty in children. Currently, standard clinical audiometry is performed using non-portable hardware and access is therefore limited in developing countries, where hearing loss is more prevalent [5].

Automation of pure-tone audiometry offers several potential benefits [6]. Particularly, a portable, automated audiometer improves accessibility, creating the possibility of routine pure-tone audiometry in the primary care setting or remote communities [6–11]. Such a device may eventually permit a parent or teacher access to screening audiometry, resulting in earlier detection of hearing impairment. However, there may be significant differences between mean hearing thresholds determined though automated audiometry and manual audiometry [12–14]. As such, clinical validation of any novel automated audiometer is a necessity.

Pure-tone audiometry, regardless of automation or lack thereof, is a mundane task. It is especially challenging to perform in the pediatric population where short attention span and the extent of cognitive development can be limiting factors. Children with hearing impairment may find audiometric testing even more difficult. Several adaptations of pure-tone audiometry are used to evaluate hearing in children. These techniques include behavioral observation, visual re-enforcement, and conditioned play audiometry. While more successful than conventional pure-tone audiometry, these adaptations are resource intensive and typically require two specially trained audiologists to administer. Naturally, an ideal solution would capitalize on the advantages of automation, while maintaining clinical validity and age appropriateness.

With the success of the Apple® iPad® (Apple Inc., Cupertino, CA), it has become evident that touch interface computing enables even the youngest of users to interact intuitively with complex systems. It is not surprising then that portable tablet computers have been widely embraced in healthcare and education [15–19]. Thus, the tablet operating system appears to the optimal environment in which to develop portable, automated diagnostic applications targeted at children. Our group has developed the first such pediatric tablet-based play audiometer [20].

The purpose of this study is to (1) determine the accuracy of audiometric thresholds obtained by tablet audiometry compared to the accepted standard, conditioned play audiometry in children; and (2) describe our experiences with administering an iPad-based hearing assessment in the pediatric population.

We expect children ages 3 and above to be capable of completing a hearing assessment using the tablet audiometer. Further, we hypothesize that the tablet audiometer will be a sensitive screening test and that the audiometric thresholds obtained will closely correlate with conventional audiometric evaluation.

Overall, 55 patients were identified to have normal hearing by conditioned play audiometry, the traditionally accepted standard test. Of these, 52 were found to have normal hearing by tablet audiometry. The 3 remaining children scored slightly outside the defined parameters of normal hearing, namely a 30 dB threshold, in at least one frequency. This appeared to either be due to the child moving through the game too quickly or the presentation timing out before the child could make a decision.

The mean time to complete a binaural hearing test with the tablet was 152 s (SD±114 s) overall. In the group with normal hearing, the time was 109 s (SD±65 s). This increased to 317 s (SD±113 s) in children with abnormal hearing. The difference in test duration was statistically significant (p<0.0001).

For the repeated measures analysis, 56 participants with valid results and binaural testing were available. Patients who were tested with a sound field were excluded from this particular analysis, as ear specific information was not available, making analysis impossible.

The model showed no significant effect of testing modality (compared to audiologists, the mean tablet threshold was 0.21 dB lower (95% CI=0.18 to 0.6 dB, p=0.29). The modeling was repeated for the abnormal cases (n=15), and for the normal cases (n=41). In both cases, the model showed no significant effect of testing modality. Compared to play audiometry, the mean tablet threshold was 1.13 dB lower (95% CI=0.27 to 2.52 dB, p=0.12) and 0.12 dB higher (95% CI=0.02 to 0.27 dB, p=0.10), respectively.

This paper describes the first trial of both a novel play algorithm using interactive audiometry and a new tablet audiometer. It is, to our knowledge, the first tablet-based, semi-automated, play audiometer to be used in a pediatric setting. The purpose of this study was two-fold, to validate the tablet audiometer as a child-friendly application for hearing assessment, as well as to compare tablet thresholds to the traditionally accepted standard play audiometry.

The data reveal that the tablet audiometer produces warble-tone thresholds that are in agreement with the accepted standard (traditional play audiometry). This was achieved with narrow confidence intervals, suggesting sufficient statistical power. Audiometric data are acquired in an efficient manner, as demonstrated by a mean test duration of approximately 2.5 minutes. Moreover, the data reveals a high specificity 94.5% with a negative predictive value (NPV) of 98.1%, denoting that tablet audiometry is a robust screening tool. A positive likelihood ratio of 17.1 confirms the tablet audiometer’s capacity to diagnose hearing loss (Figure 3).

As the tablet audiometer is by definition an objective test, if calibrated adequately, it is not likely to be subject to issues with inter-rater reliability. However, it will be prudent in subsequent investigations to ensure strong inter-rater and test-retest reliability.

Conditioned play audiometry can often be employed with children as young as 2 years of age. The supervision and motivation given to the child by a second audiologist in the sound booth allows these very young children to be tested. In the current study, this second audiologist was present for both assessments (tablet and traditional) to maximize the successful completion of assessments, although attaining the appropriate level of support from the audiologists required training and experience.

Our data suggests that using the tablet interactive audiometry method, the majority (82%, 70/85) of children as young as 3 years of age are capable of understanding the concept of the game and completing the hearing assessment. In fact, 4 of the 15 patients were unable to complete the assessment due to a technical issue related to the audiologist. Therefore, 82% is a somewhat conservative assessment of the user-friendliness of the tablet audiometer. Despite the supervision of an audiologist however, some children had difficulties with the tablet audiometer. These difficulties resulted from technical deficiencies of the hardware/software (i.e. attempts to open other software, failure to understand ‘drag and drop’, becoming distracted by visual re-enforcements) and behaviors of the patient (i.e. boredom, poor comprehension of the game). A number of these difficulties eventually lead to the subject being excluded from the statistical analysis. Failure to complete the assessment also appeared to be more prevalent in children with abnormal hearing.

Several technical/gameplay issues were documented during data collection. In particular, children showed signs of fatigue with either test method quite quickly. During standard play audiometry, audiologists often switched games several times during standard play audiometry to keep the child engaged. By contrast, only two games were available when using the tablet, with the current software version. This stresses the importance of maintaining attention in this particular age group.

Furthermore, due to the nature of interactive audiometry, whereby the test is user-directed, action is required at each point in the decision tree. This gives the appearance of more decisions as compared to standard audiometry, where users who did not hear a sound were not required to perform an action. This was exacerbated in children with hearing loss, who were required to sort more objects in order to determine exact thresholds. For example, when testing a normal hearing individual, the minimum number of objects to complete an entire assessment was 16. This number increased to a maximum of 113 when hearing loss was present or unreliable results were being obtained. The average number of objects presented in in hearing loss was 67.1 (SD±19.1) (Table 1). For children with normal hearing, the average was 23.8 (SD±9.45) (p<0.0001). Despite these challenges, the vast majority of children were engaged enough to complete the tablet hearing assessment.

Some younger children were found to have difficulty understanding the concept of sorting. A simplified version of the game was also developed, where the child was only presented with one object and one container. The child placed the object in the container when it produced sound, and a timeout function advanced the game if the child did not. We found this allowed younger children to be accurately tested with the tablet audiometer.

Visual re-enforcements were originally included. Specifically, if the child sorted an object correctly, a pop-up star appeared. However, children with abnormal hearing tended to focus on these visual cues, ignoring the auditory cues (Table 2). Subsequent versions of the tablet game will optimize visual cues to maintain the user’s interest without distracting them.

The authors acknowledge several limitations to this study. Firstly, the majority of patients were normal hearing children. This is simply a reflection of the patient population when conducting sequential recruitment. However, a test population that is predominantly normal hearing will bias the study toward good correlation and successful completion of the relatively shorter hearing test. Second, the methods of our analysis excluded patients who were identified as having clear difficulty with the hearing assessment. This was done to ensure that the results reflect only the performance of the hardware. Although exclusion of ‘difficult’ patients limits the generalizability of our results, this analysis was deliberately used during this hardware validation phase. The proportion of patients excluded from analysis (18%, 15/85) insinuates a high degree of user-friendliness, especially given the potentially difficult patient population. Additionally, this emphasizes the importance of audiologist supervision, as the software is currently unable to determine if a child fully comprehends the game. Further gameplay refinement will likely increase the number of patients who are suitable candidates for tablet audiometry.

The goal of our research is to develop a portable, versatile clinical audiometer designed for children. This study is the first to validate an automated play audiometer designed specifically for the pediatric population. In this study, we demonstrated that air conduction thresholds obtained by tablet audiometer are not significantly different from those obtained by standard play audiometry. The device has a strong predictive value for normal hearing and is highly sensitive for hearing loss. Finally, our results demonstrated that the tablet audiometer is a viable platform for testing children as young as 3 years of age. Future directions will focus on gameplay refinement and further validation testing of air vs. bone conduction thresholds and test-retest reliability.

CHEO AFP Innovation Fund

Department of Surgery Research Fund