A single board computer was used as the hardware with software coded using Python (Python Software Foundation) language. App modules were developed for hearing screening for children from birth to 6 years of age. The user interface was designed in both English and the local language with simple texts, images, and symbols such that nontechnical screening personnel can use the software. Three hearing screening test procedures were included: BOA-based screening (for children younger than 1 year), speech spectrum awareness task (SSAT)–based screening (for children older than 1 year and younger than 3 years), and speech recognition task (SRT)–based screening (for children older than 3 years and younger than 6 years).

The test module gets automatically selected based on the age selection in the app by the tester. The test modules include screening frequencies of 500 Hz, 1000 Hz, 2000 Hz, and 4000 Hz at a specified moderate intensity level (selected based on calibration). For each test procedure, age-appropriate stimuli were developed (described under each test procedure in detail) and uploaded to the software. An inbuilt noise monitoring software was included which provides access to the screening module only when the ambient noise levels are less than 50 dB (A). A cloud-based back-end data management was integrated via Google Drive. The screening result is displayed automatically based on the response entered for each frequency.

Suitable commercial transducers appropriate to the test module and age were selected based on frequency response at 500 Hz to 4 kHz, output intensity range, and distortions at intensities up to 80 dB SPL (sound pressure level). At first, 3 commercial grade supra-aural headphones (Sony WHCH510 headphones, Sennheiser HD 450 BT, and Boatrockers 410), and 3 speakers (JBL Go 2, Sony SRSXB12, and Fire Boltt Xplode 1500) were selected for output comparison based on manufacturer information on frequency, intensity range, cost, and ease of availability in the market.

Headphones were connected via Bluetooth to the SRESHT screener at maximum volume setting. The headphone diaphragm was coupled to the calibration microphone with a 6CC coupler connected to the Larsen Davis sound level meter (SLM).

Bluetooth speakers were connected to the SRESHT screener at maximum volume setting and placed at a distance of 1 m and at 0-degree azimuth from the SLM microphone. A subject was seated in a chair and the ear level was marked from the floor. The subject was then removed and the SLM microphone was suspended from a stand approximately at the ear level marked.

For all the transducers, warble tone stimuli were presented at 500 Hz, 1000 Hz, 2000 Hz, and 4000 Hz at different volume settings (%) of the SRESHT screener hardware, and the corresponding dB SPL values were measured in the SLM. The software was initially coded to produce the desired output level for moderate intensity levels in SPL (40 dB, 50 dB, and 60 dB) across the frequencies.

The Sony WHCH510 headphones and JBL Go 2 pro speakers were selected based on their accuracy of output, consistency of output, and output range (described in Development of the SRESHT Screener section).

Both tonal and nontonal stimuli were assessed for BOA [32,36]. Studies suggested that children’s responses to nontonal stimuli were as effective as tonal stimuli [36]. Nontonal stimuli included environmental sounds, animal sounds, and noisemakers that were contextually relevant to screen children. The stimuli were downloaded from an open-source website [37,38]. The acoustic parameters of the stimuli were then modified and corrected or normalized using Audacity (version 3.5.0; Audacity Team) and Apple Logic Pro (version 10.6; Apple Inc). The stimuli and the characteristics are summarized in Table 1.

A sound-treated room was preferred for conducting the screening test at the SRIHER clinic or laboratory. Participants were children with normal hearing and hearing loss between 0 to 1 year of age. Participants’ details are described in Table 2. A standard diagnostic audiometer (Madsen Astera²) with calibrated speakers (Martin Audio C115) was used to conduct the test using the newly developed stimuli (warble, noisemakers, environmental sounds, and animal sounds) at 500 Hz, 1000 Hz, 2000 Hz, and 4000 Hz for children. Behavioral responses such as eyeblink, eye widening, searching, and head turn toward the source were considered acceptable responses. This was noted by an audiology intern. Initially, the stimuli presentation started at 20 dB HL in 500 Hz for children with normal hearing and at a higher intensity level for children with hearing loss. If a response was obtained, the stimuli presentation was presented thrice in the same intensity level to confirm the response and if not, the presentation was increased by 5 dB steps (ascending method). Whenever the child got distracted or bored, tangible or visual reinforcement was given by the audiology assistant. The same procedure was repeated for each stimulus at each frequency. Minimum response levels or thresholds were obtained for each stimulus at each frequency and compared to identify the stimuli that resulted in the lowest thresholds.

For children between 1 and 3 years of age, speech-like animal sounds, such as a cows’ moo, dogs’ bow-wow, goats’ meh, and cats’ meow, and nonsense syllables like mamama and bababa were included.

The spectrum of the animal sounds using PRAAT software (version 6.3.2; Softsonic International) was compared to the spectrum of consonants and vowels to verify that it resembled the properties of speech. The formant frequencies (F1, F2, and F3) of animal sounds were approximately similar to the corresponding formant frequencies of the target vowels. Table 3 shows the formant frequencies of recorded animal sounds and existing normative vowel formant frequencies, as adapted from Raphael et al [39].

For children between 1 and 3 years of age, SSAT-based screening was conducted with the standard audiometer (Madsen Astera²), with TDH 39 headphones with newly developed stimuli. The participants included children with normal hearing and hearing impairment between the ages of 1 and 3 years. Behavioral responses included head turn, locating the source of the sound, and imitating the sound. Stimuli was presented initially at 1 kHz at 60 dB HL for normal hearing and 80 dB HL for children with hearing impairment. The Hughson-Westlake method of threshold tracking was used to identify the minimum response levels for each stimulus at each frequency. The minimum response levels for these stimuli were compared and the stimuli yielding better responses (thresholds) were finalized for the screening test procedure.

SRT-based screening was developed for children between 3 and 6 years of age. Eighteen picturable spondee words in the local language (Tamil) were excerpted from a standardized spondee list of 50 words [40] that is routinely used in clinical practice. The words were selected based on the relevance of the word in present-day context, and its picturability. Pictures corresponding to the 18 words were downloaded from the open-source web-based repository named Wikimedia Commons [41].

A familiarity test of all the words and their corresponding picture was conducted. Participants were both children with normal hearing and children with hearing impairment above the age of 3 years. For familiarity testing of words, each child was asked to point out the appropriate picture when the word was read out at the conversation level. Similarly, for picture familiarity, each child was asked to name the picture.

These 18 spondee words were further divided into 3 lists of 6 words each. Then, these spondee words were recorded with a female voice centered at 1000 Hz and with loudness balanced at –3 dB peak value, having an interstimulus interval of 5 seconds. A closed-choice response set was developed with 1 target picture and 3 other pictures. The 4 pictures were arranged in a sequential manner with an equal number of repetitions.

Participants were children with normal hearing and children with hearing impairment between the age group of 3 and 9 years. Using the standard audiometer (Madsen Astera²) with TDH 39 headphones, each child’s pure tone thresholds (PTA) were obtained using the Hughson-Westlake method. Then, each child underwent SRT with picture identification as response. This was performed at different sensation levels (0 dB SL, 5 dB SL, 10 dB SL, and 20 dB SL with re-PTA) to identify the level at which 80% correct response (5/6 words) was obtained.

Two groups of children participated in the study. The first group consisted of children under 6 years of age with normal hearing. The other group consisted of children with hearing impairment having a hearing age of less than 6 years and these children were younger than 10 years of chronological age. Children with any other comorbid conditions or any other known disabilities were excluded from the study (eg, children with developmental delay, vision impairment, autism spectrum disorders).

Children were recruited from the outpatient clinic of the tertiary care hospital attached to the institution where the study was conducted, as well as in camps organized in 2 rural districts of Tamil Nadu for the period of 3 months.

Informed written consent was obtained from the parents of the children before the commencement of data collection.

The procedure was conducted in any quiet room near the outpatient department or the campsites with an ambient noise level of less than 50 dB (A) as recorded through the noise monitoring software of the SRESHT screener. SRESHT hearing screening with the finalized stimuli and screening modules for each age group was performed for all children using age-appropriate screening procedures at 60 dB HL.

Then, the minimum response level was estimated for 500 Hz, 1 kHz, 2 kHz, and 4 kHz with the standard diagnostic audiometer (Madsen Astera²) for each test procedure, as a gold standard. Screening results for the SRESHT hearing screener at 60 dB HL (pass or refer) were compared to audiometric thresholds more or less than 60 dB HL.

Nonparametric tests were used to analyze the data as the results of the Shapiro-Wilk test showed that the data were not normally distributed.

The Friedman test was used to compare the children’s responses to different stimuli in BOA. The Wilcoxon signed rank test was used to compare the children’s responses to different stimuli for SSAT. Percentage analysis was used for word familiarity for SRT screening. Agreement statistics were used to beta-validate the screening results.

The summary of output characteristics of various speakers and headphones is shown in Table 4. The Sony WHCH510 headphones and JBL Go 2 pro speakers were selected based on their accuracy of output, consistency of output, output range, ease of use, and cost.

The RETSPL values derived for the Sony WHCH510 headphones are shown in Table 5.

Fifteen children (11 Children with normal hearing and 4 Children with hearing impairment) in the age range of 0 to 1 year of age with a mean age of 6 (SD 1.9) months for children with normal hearing and a mean hearing age of 8 (SD 1.6) months for children with hearing impairment (Table 2), underwent the BOA using the newly developed stimuli. The Friedman test revealed no significant difference in the average minimum response levels of warble tone versus average minimum response levels of noisemakers, animal sounds, and environmental sounds (n=15; χ²₃=15.44; P=.089).

A comparison of the average minimum response levels for different stimuli for children with normal hearing and children with hearing impairment revealed that the average minimum responses to noisemakers for children with hearing impairment were superior to the responses to other stimuli, that is, animal sounds and environmental sounds (Figure 1). Therefore, noisemaker-based stimuli were selected for this age group, even though all 3 categories of non-tonal stimuli were appropriate for hearing screening.