The EarPAD project: a tele-otology study for remote Aboriginal communities

Introduction

Telemedicine continues to be an important part of healthcare delivery to remote and rural populations where geographic distance limits access to in-person consultation with specialist medical practitioners. Australia exemplifies this problem. One potential solution is the use of telehealth to enable real time or asynchronous review by an expert geographically separated from the patient. Otolaryngology has utilised this technique in several countries around the world (1-5) including Australia (6-8).

Aboriginal and Torres Strait Islander children have some of the highest rates of otitis media globally, with twice the likelihood of chronic hearing loss compared to their non-Aboriginal peers. The consequences are varied with influences on speech and language development, numeracy and literacy rates, employment and risk of incarceration (9) resulting in the World Health Organisation declaring it a public health emergency (10).

In 2024, the pilot study for this project was published (11), which critically analysed several potential telemedicine tools that could form a part of a tele-otology platform to assess patients in a rural and remote setting. From the pilot study the best suited telehealth tools were chosen to be incorporated into the suite of diagnostic instruments for the main study, presented here. The pilot also enabled trial of the study protocol with subsequent refinement of the patient health questionnaire and optimisation of user instructions. The platform used to run the software was upgraded from a smartphone-based system in the pilot (EarPHONE project) to a more user-friendly tablet in the main study (EarPAD project). The purpose of this main study was to assess this chosen suite of tele-otology tools in the setting of rural Aboriginal Medical Service (AMS) clinics, with local health care workers operating the devices. Comparison in diagnostic ability was made with the alternative fly-in-fly-out (FIFO) specialist consultation.

Methods

The study is reported according to the STARD reporting guidelines (available at https://www.theajo.com/article/view/10.21037/ajo-2026-1-003/rc). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. It was approved by the New South Wales Aboriginal Health & Medical Research Council (AHMRC) as well as the Sydney Local Health District Ethics Review Committee (project No. X19-0042 & 2019/ETH00242). Written informed consent was obtained from the patients or their parents/guardians prior to participation in the study and participation was entirely voluntary.

The study recruited patients at two remote locations in New South Wales (Bourke and Brewarinna) at their AMS clinics from November 2022 to April 2024 with inclusion criteria being a minimum age of 4 years old and presenting with an ear related complaint. Sequential patients, who were routinely booked into these clinics were invited to participate in the study. Patients were not contacted for recruitment purposes to attend, but rather were already booked into the ear, nose and throat (ENT) clinic and then approached to enroll if they met inclusion criteria. All patients who were invited to enroll, agreed to participate with no abstentions.

The study design algorithm is shown in Figure 1. Patient assessment involved sequential assessment using the telehealth tools followed by a standard clinical assessment with an otolaryngologist (ENT surgeon). The research component was divided into a digital patient history questionnaire (PHQ), video otoscopy, tympanometry and automated audiometry. The clinical component was divided into a clinical assessment with a senior ENT surgeon and audiological assessment with tympanometry. The study then compared how the research tools compared to these gold standard clinical assessments.

Figure 1 Study Algorithm showing patient pathway and data collection. AMS, Aboriginal Medical Service; ENT, ear, nose, and throat.

On-site clinical assessments

The same ENT consultant (A.J.S.) assessed all patients and was blinded to any tele-otology assessments. This surgeon did not perform any of the telehealth assessment and did not have access to any of the telehealth results. A standardised assessment form was completed for each patient, including clinical findings, diagnosis and treatment plan (see Table S1). The clinical audiology component was performed by local health workers at the outreach clinics, qualified in audiological assessment (J.C. and J.G.), following standard audiological protocols. The rooms used for hearing assessments had only limited soundproofing. An Amplivox Model 240 audiometer and Amplivox Otowave 302 tympanometer (Amplivox Ltd, Birmingham, UK) were used for these gold standard audiological assessments.

Tele-otology assessments

Tele-otology assessment included 4 main components for later offsite interpretation:

• PHQ;
• Video otoscopy—HearSCOPE Digital Video Otoscope (HearX Group, Pretoria, South Africa);
• Automated audiometry—HearTEST Audiometer (HearX Group, Pretoria, South Africa);
• Tympanometry—Amplivox Otowave 102 Tympanometer (Amplivox Ltd, Birmingham, UK).

Choice of tele-health tools was based on the results of the pilot study (11). All devices were approved for use in Australia by the Therapeutic Good Administration (TGA) of the Australian Government Department of Health. A Galaxy TabA10.1 tablet (Samsung Electronics, Suwon, South Korea) was the platform used to run the two HearX applications (camera and automated audiometer). The tympanometer was a separate portable device with printer. All tele-health assessments were performed by a local health worker or research assistant, which occurred prior to and therefore blinded to the onsite clinical assessment.

PHQ

An intuitive digital PHQ was created on an online platform (REDCap Research Electronic Data Capture Program), whereby responses would prompt further questions related to their positive symptoms whereas negative responses would terminate further questioning in that field. The full questionnaire is available in the published pilot study (11). It was designed to simulate an otological history taken by an ENT surgeon.

Video otoscopy

The HearSCOPE Digital Video Otoscope (HearX Group, Pretoria, South Africa) is shown in Figure 2A. A 5–10 second duration video was recorded of each ear canal by an Aboriginal health worker, who had been trained on how to use the device. The device has hardware capability for both focus and light intensity adjustment.

Figure 2 Photographs of the EarPAD tele-instruments. (A) Video-Otoscopy: HearSCOPE Digital Video Otoscope (HearX Group, Pretoria, South Africa) connected to a Galaxy TabA10.1 tablet (Samsung Electronics, Suwon, South Korea). (B) Automated Audiometry: HearTEST Audiometer (HearX Group, Pretoria, South Africa). (C) Tympanometry: Amplivox Otowave 102 Tympanometer (Amplivox Ltd, Birmingham, UK) with printer.

Automated audiometry and tympanometry

The pilot study selected the HearTEST Audiometer (HearX Group, Pretoria, South Africa), shown in Figure 2B, as the most appropriate for this study’s needs (11). We determined that a tablet-based system would be a more suitable platform on which to use the application rather than a phone, as it enabled a larger screen for visualising commands, whilst remaining highly portable. The tablet was connected to calibrated over-the-ear Sennheiser HD 280 Pro headphones (Sennheiser, Hanover, Germany) with high ambient noise attenuation and accurate linear sound reproduction. Air conduction pure tone thresholds were tested at 500, 1,000, 2,000, 4,000 and 8,000 Hz without masking or bone conduction capabilities. The healthcare worker would explain the test to the patient and apply the headphones, then the program would run automatically, cycling through the five pure tone frequencies and adjusting decibel output in line with their responses to determine the patient’s auditory thresholds.

For tele-tympanometry, the Amplivox Otowave 102 Tympanometer (Amplivox Ltd, Birmingham, UK), Figure 2C, was selected as it was easy to use, highly portable and enabled printable results for subsequent uploading online.

Data collection and storage

All data was de-identified and uploaded with a unique participant ID to a secure online database (Sydney University REDCap Research Education Data Capture Program).

Offsite interpretation of tele-otology data

Two senior ENT surgeons (N.J., N.P.) independently reviewed the 4 tele-otology assessments [PHQ, automated audiology, tympanometry and video otoscopy, plus had access to a de-identified general practitioner (GP) referral letter (de-identified)]. From this, they completed a standardised assessment form, similar to the one completed by the onsite ENT surgeon, where they recorded their diagnosis and management plan based on the tele-otology findings. They were blinded to any of the “gold standard” onsite clinical assessments or the onsite ENT surgeon’s findings.

Statistical analysis

• For every patient, the standardised assessment forms were compared; one produced by the onsite surgeon (A.J.S.) based on the clinical assessments and the others produced by the offsite ENT surgeons (N.J., N.P.) based on the tele-otology information. This enabled comparison of the following fields:
  • Canal findings;
  • Diagnosis;
  • Management plan.
• Offsite surgeons were asked to appraise the video otoscopy, in terms of video duration (too short, good length or too long), quality (poor, adequate or good) and visualisation of the tympanic membrane (which quadrants could be seen).
• Confidence in final diagnosis was graded on a 5-point scale (1= no confidence; 5= certain).
• Patient experience and cultural appropriateness was gauged through a feedback questionnaire (See Table S2).
• Field of view comparisons were made between video otoscopy and onsite clinical otoscopy by recording how many quadrants of the tympanic membrane were visible in each case.
• Automated EarPAD audiograms were compared to the standard audiograms performed by a local healthcare worker on their existing local audiometer, which was taken as the gold standard for this rural setting. Bland-Altman plots (12) were used to graphically display deviation from the gold standard result in decibels against the mean of the hearing threshold for the two audiograms at each frequency tested. Sensitivity and specificity were calculated for whether the automated audiometers were able to detect hearing loss, defined as >25 decibels hearing level (dB HL) threshold, compared to the gold standard audiogram, at all frequencies. From this an accuracy calculation was derived using the formula: true positives + true negatives/total number of cases.
• The two tympanometry results (clinical and tele-otology) were assessed using Cohen’s Kappa (k) to gauge agreement.

Statistical calculations were performed using Microsoft Excel for Mac version 16.42. Missing data were excluded from the analysis.

Results

Fifty-one patients (102 ears) were enrolled in the study; 24 females, 27 males with a mean age of 35.4 years [standard deviation (SD) ±27.7; range 4–92 years]. The cohort was balanced between adults and children with 22 participants under the age of 18 years, of which 16 were under 10 years. Complete data sets were achieved for 37 patients (72.5%). Five patients had no gold standard audiogram, due to absence of a local health worker with audiometry training on those outreach visits. Eleven had no clinical tympanometry, while 4 were missing the tele-tympanometry result. Seven patients had no automated audiogram due to patient non-compliance (n=2) or technical failure (n=5). One patient declined to complete the PHQ. Two patients had no tele-otoscopy.

Audiology assessment

The Bland-Altman plots (Figure 3) which visually represent the degree of similarity between audiograms, showed good agreement between the two audiological assessments. The close clustering of data points around the central “zero difference” lines with narrow limits of agreements [the dashed lines representing the 95% limits of agreement (±1.96 SD)], implied good agreement between what the automated audiometer registered and what the clinical assessment found. The breadth of data points across the Y axis confirmed that a good range of hearing ability was assessed (normal, mild, moderate and severe) in the study cohort. Agreement appeared reasonable across the 4 different frequencies examined and independent of patient hearing ability, meaning the automated audiometer gave a similar result to the clinical audiometer regardless of the patient’s hearing status or which frequency was being assessed.

Figure 3 Bland-Altman plots for the 5 tested frequencies comparing the result achieved with the automated audiometer referenced to the gold standard Audiometer. Difference in hearing threshold is plotted against the average threshold to show how performance differed at different patient hearing levels. Each data point represents a different patient. The close clustering of data points around the central “zero difference” lines with narrow limits of agreements (the dashed lines representing the 95% limits of agreement (±1.96 SD), implies good agreement between what the automated audiometer registered and what the clinical assessment found. SD, standard deviation.

The ability of the automated audiometer to successfully diagnose hearing loss (defined as >25 dB HL), using the local clinical audiogram as gold standard and pooling the data across the 5 frequencies, had a sensitivity of 83.7% [95% confidence interval (CI): 79.9–87.5%] and specificity of 56.1% (95% CI: 51.0–61.2%).

Tympanometry

Interobserver agreement between the two offsite surgeons interpreting the tympanometry performed on the Amplivox Otowave tympanometer was deemed “substantial” with a Cohen’s Kappa coefficient of 0.74 and agreement seen in 86.8% of cases. Comparison of these results with those obtained using the gold standard tympanometer, showed agreement in 77.4% of cases, classified as a “moderate” agreement with a Cohen’s Kappa Coefficient of 0.52.

Video quality

The quality of video-otoscopy was subjectively appraised for 161 ears by the two offsite surgeons, being unavailable or unreported in the remainder (Figure 4). It was deemed poor in the majority of cases (93/161, 57.8%) which was highly correlated with those patients who required canal microsuction, which the video preceded (n=22, 21.5%). Patients under 18 years also had a higher proportion of video footage deemed “poor” (Figure 4). Video length was appraised in 157 cases, across the two surgeons. Approximately half of the videos (83/157, 52.9%) were considered too short, the other half adequate (74/157, 47.1%) and none were appraised as too long. Inter-observer agreement in quality assessment was 70%, indicating a significant degree of variability in this subjective measure.

Figure 4 Pie chart showing subjective assessment of video otoscopy quality by offsite surgeons, with response options of “Good”, “Acceptable” or “Poor”. Results stratified by age.

The offsite surgeons marked which quadrants could be seen in the video images in 40 cases (Figure 5). There was 100% inter-observer concordance in terms of which quadrants were visible. When compared to the gold standard on-site clinical microscopy (where all 4 quadrants were visible in 95% of cases), the video images equaled the microscopic view in terms of number of quadrants seen in 60% of cases (24/40).

Figure 5 Pie Chart showing how many quadrants of the TM were visible using the video-otoscope (n=40). In comparison, all 4 quadrants were visible with the gold standard microscope in 95% of cases. TM, tympanic membrane.

Where reported, offsite interpretation of tympanic membrane and middle ear status using video otoscopy demonstrated agreement with microscopic findings as the reference standard. For classification of normal vs. abnormal tympanic membranes, diagnostic performance was as follows: Surgeon A demonstrated a sensitivity of 71.4% (95% CI: 54.1–88.7%) and specificity of 88.2% (95% CI: 65.1–100%). Surgeon B demonstrated a sensitivity of 95.2% (95% CI: 66.1–100%) and specificity of 75% (95% CI: 53.3–96.6%). Notable findings such as perforation (n=8; Note that there were 10 perforations within the cohort but only 8 had successful video-otoscopy) were correctly diagnosed on video-otoscopy in 87.5% of cases, with one misdiagnosed as a retraction pocket. In terms of middle ear findings similarly reasonable sensitivity and specificity was seen (Surgeon A: Sensitivity 75% (95% CI: 54.6–95.4%); Specificity 78.9% [95% CI: 57.0–100%), Surgeon B: Sensitivity 100% (95% CI: 62.3–100%), Specificity 71.4% (95% CI: 46.7–96.1%)]. Notable mistakes in this area were middle ear effusions (n=5), misdiagnosed as normal in two cases. There was one case of cholesteatoma, which neither offsite surgeon diagnosed based on video-otoscopy, one labelling it an effusion and the other acute otitis media (AOM).

Clinical diagnosis

An offsite clinical diagnosis was reported in 54 (52.9%) and 59 (57.8%) of the 102 possible diagnostic ear cases by the two respective offsite surgeons. The degree of confidence in diagnosis for each surgeon, using the 5-point Likert Scale had a mean score of 2.13 (SD 1.38) and 2.61 (SD 1.49) respectively (Figure 6).

Figure 6 Bar chart showing the degree of confidence each offsite surgeon had in each diagnosis, graded subjectively from 1 (low confidence) to 5 (high confidence).

The lack of any offsite diagnosis in close to half of the presented cases, indicates that the overall diagnostic ability of the tele-health tool was poor. In those cases where a diagnosis was given, of the pooled 113 cases, the offsite surgeons correctly identified the same diagnosis as the gold standard in 92 cases (81.4%). Analysis of the results by age, showed an improved diagnostic capability within the adult subgroup compared to paediatric, using a cut off of 18 years. This stratification resulted in a correct diagnosis in 36 of the 50 paediatric cases (72.0%) and 56 of the 63 adult cases (88.9%). Calculation of sensitivity and specificity, excluding the cases without a reported diagnosis, compared just whether the ear was normal or abnormal rather than defining a specific clinical diagnosis. For combined adult and paediatric cases (n=113), the overall sensitivity was 88.7% (95% CI: 80.8–96.6%) and specificity was 76.5% (95% CI: 68.6–84.3%). Improved sensitivity and specificity were again seen in the adult subset. For the paediatric subset (n=50) the sensitivity and specificity were 81.5% (95% CI: 66.8–96.1%) and 69.6% (95% CI: 54.9–84.2%) respectively. For the adult subset (n=63) the sensitivity and specificity were 94.3% (95% CI: 86.6–100%) and 82.1% (95% CI: 74.5–89.8%) respectively.

In terms of individual diagnoses (Table 1), the best performance was seen for tympanic membrane perforations, most of which were diagnosed when video-otoscopy was available. Otitis media with effusion was the next most successful diagnosis but was only correctly identified in approximately half of cases. All 4 cases of foreign body were missed by the offsite team, all not recording a diagnosis for those cases. Lack of a recorded diagnosis was very highly correlated with absent or poor-quality video otoscopy which occurred in 90.1% of the “no diagnosis recorded” cases.

Patient feedback

Patients subjectively scored the comfort of the tele-otoscopy and automated audiogram with mean scores of 4.30 (SD 0.81) and 4.14 (SD 1.16) out of 5, respectively, which were comparable to the scores for the gold standard clinical examination and audiogram [4.35 (SD 0.74) and 4.2 (SD 0.97)], respectively. Five patients expressed that they felt the hearing test was too long and one that the tympanometry was painful. None of the 51 patients expressed any cultural concerns about the testing.

Discussion

The EarPAD project tested a set of tele-otology tools designed to assist Aboriginal health workers in a rural setting to diagnose ear conditions. While achieving a precise clinical diagnosis of any otological condition would be desirable, it is more realistic to hope that such a tool could screen for and identify conditions that require further investigation or intervention. This could be in the form of a referral to an ENT either at a central hospital or, if less urgent, at the next outreach clinic. It could also potentially direct the health care worker in terms of what local therapy is appropriate and can be provided by them or a GP.

Initial analysis of the results reveals that a high proportion of offsite assessments did not register a diagnosis. This could be interpreted as demonstrating the suite of tools as failing to meet their primary goal. However, that misses some of the finer conclusions that can be drawn from the study, both in terms of the usefulness of the current model and in designing better tools for the future.

It should be noted that even the “gold standard” (on site) clinical assessments were subject to the challenges of a rural setting. These challenges include the lack of soundproofing for audiological assessments, the lack of surgical equipment and the local skill shortage for useful interventions such as aural toilet. This does, however, represent the best alternative to a tele-based diagnostic tool and the best denominator with which to compare. Given that placement of an ENT surgeon in every remote community is not achievable, it stresses the importance of developing tools such as the EarPAD to address this discrepancy.

Video quality

Correct identification of an otological problem benefits greatly from good otoscopy (13-18). For a telehealth tool to be useful it should incorporate a reliable and accurate video capture of the canal and drum status. That was not achieved in the majority of the study participants, with offsite interpretation of the captured images as “poor” or absent in 66%. While this could imply that the video images did not assist in diagnosis in two thirds of cases, that is not necessarily true as this is a subjective measure. A surgeon might consider the video image poor but still gain important clinical clues from it. More objective measures, such as the number of quadrants visible, demonstrated that the tele-otoscope performed quite well with almost two thirds (60%) achieving parity with the microscope. Other indirect measures of video-otoscopic performance such as tympanic membrane and middle ear appearance, also proved that diagnosis of an abnormality was achieved with reasonable sensitivity, in the absence of the other telehealth tools. However, amongst the suite of tele-tools assessed, the video-otoscope appears the most closely correlated with misdiagnosis or absent diagnosis. It is important to note that video length was deemed insufficient in the majority of cases, and never considered too long. This supports the use of video rather than singular photographs. The length in this study was approximately 10 seconds, which would therefore seem to be a good minimum standard to aim for.

It is disappointing that the chosen video-otoscope, which had performed well in the pilot study, was not more successful in the main study. There are several possible explanations including inadequate user training, poor patient tolerance and device deficiency. Patient feedback indicated adequate comfort levels in the majority, but a small proportion did score comfort as only 3 out of 5, suggesting a degree of intolerance. Insufficient user training could negatively impact on camera positioning and troubleshooting, critical to overcome canal curvature for proper visualisation of the tympanic membrane. Correct image focus is also required which was a manual function in this particular system. Fogging of the lens is also possible, although not described by any of the users in their feedback. All of these issues may have resulted in the low proportion of good quality images. However, if a telehealth tool is to be useful in the wider community it does need to be relatively easy to use with minimal training. Another major factor is canal condition. The presence of cerumen or pus hinders good visualisation. Training in basic ear toilet is a crucial skill for remote clinics to perform tele-otoscopy successfully. In the study, the need for cerumen toilet by the on-site consultant was recorded in 1 in 5 patients, and all of these patients had video-otoscopy graded as poor quality, as one would expect.

In cases where the image quality was poor, the offsite clinicians were still able to correctly identify pathology in some cases, which supports the important additional conclusion that video-otoscopy, while a critical component, is part of a broader set of useful tools in the telehealth suite. Other factors such as the history questionnaire, audiometry and tympanometry have significant value.

Audiometry and tympanometry

Automated audiometers are a cheap and well validated method of hearing assessment in remote settings (19-20). They require no audiological training on the part of the local health worker and are very simple to use and interpret. They represent a cheaper and effective alternative to placement of qualified audiologists in every rural clinic. The advances in noise cancelling technology headphones also presents cost and space saving benefits for a clinic, if installation of a sound-proof booth is not required, although it would still be preferable. Previous studies, including the pilot study for this project, have validated many different commercially available automated audiometers (11,19-20). This study also demonstrated reasonable sensitivity at diagnosing hearing loss [sensitivity of 83.7% (95% CI: 79.9–87.5%)].

The patient feedback highlighted the need to keep automated tests brief, as noted in the feedback forms and informally during the tests. Brevity could be achieved by restricting the tested frequencies to 3 (500, 1,000, 2,000 Hz), which would still give a useful pure tone average, rather than the 5 frequencies used in this study. If hearing loss were identified, a more thorough clinical assessment would be necessary including bone conduction and masking, so keeping the automated test simple and brief seems appropriate. The balance is choosing between obtaining useful audiological data rather than no data, if the patient loses focus and concentration. The automated audiometer lacks the human interaction of an audiologist who can sense if a patient is becoming restless, particularly in the paediatric setting. When methodically running through all five frequencies in the automated setting, the length of the test was counterproductive and resulted in some failures to complete.

Tympanometry would be expected to be fairly consistent across devices but interestingly agreement was only moderate (77.4% agreement, Cohen’s Kappa 0.52). This likely represents subjectivity in trace interpretation, using a more qualitative assessment of the curve rather than objective clinical criteria (e.g., mislabeling a type B for a type As or low rising type C). There is also the possibility of user error, whereby the tympanometer gives an incorrect result due to incorrect positioning in the ear canal or in the presence of canal obstructions (e.g., cerumen), or excessive canal hair, which is likely to be a heightened risk in less experienced or poorly trained users. This again highlights the importance of good training with any tele-health tool.

Clinical diagnosis

The challenge of this tele-otology project was whether experienced ENT surgeons could reach similar diagnostic conclusions when presented with clinical data acquired remotely as opposed to in person. Naturally this is dependent on the quality of the data that is provided. A microscope used in person should provide better interpretation of the ear canal than a 10-second recording taken by a third party. This is why the synthesis of all the available data including the GP referral, objective tests such as tympanometry and audiometry as well as the PHQ were proposed rather than relying solely on video-otoscopy.

Overall, the ability of the offsite ENT surgeons to make a diagnosis was only seen in around half of the cases, but where they felt confident to do so, the accuracy was reasonable and in agreement with the gold standard on-site diagnosis in the majority cases (92 of 113, 81.4%). The EarPAD tool did perform better in assessing adults, which highlights an increased challenge in using telehealth tools in remote paediatric populations. The improved diagnostic capability may relate directly to the quality of the video otoscopy achievable, reflected in the higher proportion of poor-quality recordings in the paediatric subset (Figure 4). Potential causative factors in children could be smaller ear canals, more canal debris or poorer tolerance of the procedure. This correlation again stresses the importance of this particular element within the suite of tele-health tools. Despite a wealth of information being available from audiometry, tympanometry and the history questionnaire, there appears to be a reliance on visual validation for confident offsite diagnosis.

When the tele-health data do not provide adequate information for a complete diagnosis, there is still merit in having a tool that collects a broader clinical assessment to send to a central hub to help guide further management and determine patient follow up requirements. Each element of the tele-health suite gives another perspective on the patient’s clinical condition. The performance of the EarPAD system to delineate normal from abnormal ears had reasonable sensitivity and specificity, especially in the adult subset, suggesting a positive role in this screening process. When diagnosis is possible, it is often in good agreement with the gold standard but when a diagnosis is not possible, it should still distinguish that an abnormality is present and help guide what further investigation or intervention is required.

Cultural appropriateness and local engagement

Patient feedback was overwhelmingly positive, suggesting that the use of automated technologies is both culturally appropriate and well received in these communities. Engagement by local health workers will be key to the success of any tele-health initiative.

Limitations

Missing data from the telehealth data acquisition was relatively small but would have had some impact on study results. More significant was the inability of the offsite surgeons to formulate a clinical diagnosis in a large proportion of cases (pooled 91 of 204 ear cases, 44.6%). This affected what data could be used to calculate sensitivity and specificity of the diagnostic model as non-reported cases were excluded. This may have resulted in an overestimation the diagnostic accuracy of the system. Offsite surgeons may have avoided giving a diagnosis due to uncertainty which could have resulted in selection bias. Future studies would benefit from insisting on an offsite diagnosis, ensuring it is linked to a measure of confidence in that decision.

Conclusions

The presented collection of tele-otology tools has features that clearly have merit but also presents several challenges that need to be overcome before successful adoption into the health system. Many refinements are necessary before potential roll out more widely but in principle, the use of a multi-faceted approach to tele-otology seems valid and useful moving forward. The following recommendations are presented to assist in the design of prospective tele-otology screening tools and future research:

• Video-otoscopy is fundamental to the success of any tele-otology tool.
• Video-otoscopy alone is not sufficient and needs to be in combination with the diagnostic contributions of audiometry, tympanometry and history questionnaires.
• Imaging should use video rather than still photography.
• Video length should be at least 10 seconds duration. Longer is more useful than shorter.
• Automated audiometry is easy to perform and well validated but should be kept as brief as possible (e.g., restrict to 3 frequencies: 500,1,000, 2,000 Hz).
• Adequate training of local health workers is crucial with particular attention to:
  • Positioning and focussing the video-otoscope effectively to get successful images;
  • Performing basic ear toilet to avoid obstruction to the video image;
  • Positioning the tympanometer for effective recording.

Acknowledgments

The authors would like to thank the following people who were involved in (a) protocol preparation; (b) clinical implementation; or (c) advisory capacity: Abbott K^b, Annan D^b, Bhutta M^a,c, Bird M^a,b Bruce J^a, Cochrane-Owers B^b, Christie V^c, Dixon B^b, Ferguson H^c, Finlayson H^c, Gordon A^b, Gwynne K^a, Harrop E^a, Jones M^b,c, Kelly D^c, Knight J^c, Kong K^a,c, Rambaldini B^c.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://www.theajo.com/article/view/10.21037/ajo-2026-1-0003/rc

Data Sharing Statement: Available at https://www.theajo.com/article/view/10.21037/ajo-2026-1-0003/dss

Peer Review File: Available at https://www.theajo.com/article/view/10.21037/ajo-2026-1-0003/prf

Funding: This work was supported by a financial grant from the Sydney Local Health District, NSW Government, Australia awarded at the 2019 “The Pitch” program. The grant was used solely to purchase the equipment used to run the study. No money was paid to any participants or authors.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://www.theajo.com/article/view/10.21037/ajo-2026-1-0003/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. It was approved by the New South Wales Aboriginal Health & Medical Research Council (AHMRC) as well as the Sydney Local Health District Ethics Review Committee (project No. X19-0042 & 2019/ETH00242). Written informed consent was obtained from the patients or their parents/guardians prior to participation in the study and participation was entirely voluntary.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Robler SK, Platt A, Turner EL, et al. Telemedicine Referral to Improve Access to Specialty Care for Preschool Children in Rural Alaska: A Cluster-Randomized Controlled Trial. Ear Hear 2023;44:1311-21. [Crossref] [PubMed]
2. Habib AR, Perry C, Crossland G, et al. Inter-rater agreement between 13 otolaryngologists to diagnose otitis media in Aboriginal and Torres Strait Islander children using a telehealth approach. Int J Pediatr Otorhinolaryngol 2023;168:111494. [Crossref] [PubMed]
3. Alenezi EM, Veselinović T, Tao KF, et al. Ear Portal: An urban-based ear, nose, and throat, and audiology referral telehealth portal to improve access to specialist ear health services for children. J Telemed Telecare 2024;30:1581-1589. [Crossref] [PubMed]
4. Schuster-Bruce J, Shetty P, O'Donovan J, et al. Comparative performance of prediction model, non-expert and telediagnosis of common external and middle ear disease using a patient cohort from Cambodia that included one hundred and thirty-eight ears. Clin Otolaryngol 2021;46:635-41. [Crossref] [PubMed]
5. Ravi P, Ramkumar V, Rajendran A, et al. Tele-Audiological Surveillance of Middle Ear Status among Individuals with Cleft Lip and/or Palate in Rural South India. J Am Acad Audiol 2020;31:185-94. [Crossref] [PubMed]
6. Elliott G, Smith AC, Bensink ME, et al. The feasibility of a community-based mobile telehealth screening service for Aboriginal and Torres Strait Islander children in Australia. Telemed J E Health 2010;16:950-6. [Crossref] [PubMed]
7. Nguyen KH, Smith AC, Armfield NR, et al. Cost-Effectiveness Analysis of a Mobile Ear Screening and Surveillance Service versus an Outreach Screening, Surveillance and Surgical Service for Indigenous Children in Australia. PLoS One 2015;10:e0138369. [Crossref] [PubMed]
8. Reeve C, Thomas A, Mossenson A, et al. Evaluation of an ear health pathway in remote communities: improvements in ear health access. Aust J Rural Health 2014;22:127-32. [Crossref] [PubMed]
9. Leach A, Morris P. Otitis media and hearing loss among Aboriginal and Torres Strait Islander children: a research summary. Australian Parliament’s Standing Committee on Health, Aged Care and Sport public hearing in reference to the Inquiry into the Hearing Health and Wellbeing of Australia. Submission 108, 2017 Jun.
10. Acuin J. Chronic suppurative otitis media: Burden of illness and management options. World Health Organization; 2004.
11. Saxby AJ, Schofield D, Tout F, et al. The earphone project pilot: a tele-otology study for remote Aboriginal communities. Aust J Otolaryngol 2024;7.
12. Thompson GP, Sladen DP, Borst BJ, et al. Accuracy of a Tablet Audiometer for Measuring Behavioral Hearing Thresholds in a Clinical Population. Otolaryngol Head Neck Surg 2015;153:838-42. [Crossref] [PubMed]
13. Biagio L. Asynchronous video-otoscopy with a telehealth facilitator. Telemed J E Health 2013;19:252-8. [Crossref] [PubMed]
14. Biagio L. Video-otoscopy recordings for diagnosis of childhood ear disease using telehealth at primary health care level. J Telemed Telecare 2014;20:300-6. [Crossref] [PubMed]
15. Mandavia R, Lapa T, Smith M, et al. A cross-sectional evaluation of the validity of a smartphone otoscopy device in screening for ear disease in Nepal. Clin Otolaryngol 2018;43:31-8. [Crossref] [PubMed]
16. Mousseau S, Lapointe A, Gravel J. Diagnosing acute otitis media using a smartphone otoscope; a randomized controlled trial. Am J Emerg Med 2018;36:1796-801. [Crossref] [PubMed]
17. Rappaport KM, McCracken CC, Beniflah J, et al. Assessment of a Smartphone Otoscope Device for the Diagnosis and Management of Otitis Media. Clin Pediatr (Phila) 2016;55:800-10. [Crossref] [PubMed]
18. Moshtaghi O, Sahyouni R, Haidar YM, et al. Smartphone-Enabled Otoscopy in Neurotology/Otology. Otolaryngol Head Neck Surg 2017;156:554-8. [Crossref] [PubMed]
19. Mahomed F. Validity of automated threshold audiometry: a systematic review and meta-analysis. Ear Hear 2013;34:745-52. [Crossref] [PubMed]
20. Pedersen C, Pedersen ER, Sørensen CB, et al. Accuracy of automated and non-audiologist-operated audiometry compared to gold-standard testing. Int J Audiol 2026;65:71-9. [Crossref] [PubMed]