gms | German Medical Science

The care situation of hearing-impaired people with hearing systems, in particular bimodal patients with hearing aids (HA) on the one ear and cochlear implants (CI) on the other, is constantly improving. A study showed that most subjects in the bimodal condition showed improved speech understanding in quiet and in noise compared to the hearing aid or cochlear implant only mode. The bimodal benefit in quiet could be partly explained by the degree of pure-tone loss; subjects with better hearing on the acoustic side benefited significantly from the additional electrical input [1].

Although the hearing systems are generally optimally adjusted and tested under ideal laboratory conditions (‘efficacy’), patients repeatedly report everyday situations in which they are dissatisfied with their hearing (‘effectiveness’). Another complicating factor is that the hearing aid is usually adjusted by a hearing aid acoustician in the shop and the CI is adjusted by an audiologist at the clinic, i.e., at separate times and locations. In addition, the signal processing of the two hearing systems is often not synchronised and each system is optimised to capture the entire acoustic environment. When adjusting both hearing systems, the focus should be on creating an improved binaural perception of hearing for the user. The goal of achieving comprehensive patient satisfaction and quality of life with the hearing systems [2], [3] applies not only to bimodal fitting, but also to conventional fitting with one or two hearing aids, CIs and/or acoustic implants. In order to optimise the benefit for everyday hearing situations and to optimise medical devices and/or their fitting in everyday use under conditions with high ecological validity [4], [5], [6], [7], [8], it is necessary not only to collect objectifiable acoustic and audiological data in the laboratory and in the field, but also to collect subjective data from the patient’s perspective in the sense of “Patient Reported Outcome” (PRO). From the manufacturer’s perspective, it is also necessary to fulfil the requirements of the Food and Drug Administration (FDA) in the United States and the Medical Device Regulation (MDR) in the European Union in the context of post-market clinical follow-up studies (PMCF) by collecting “real-life data” and “real-life evidence” of the medical devices.

Measurement procedures for PRO (PROM) are still predominantly collected in clinical, experimental settings, often using paper-pencil versions, which can be time-consuming and error-prone for healthcare professionals and patients. Although some telemedical applications are available, such as the Remote Check app including a speech audiometric measurement [9], or in the field of Ecological Momentary Assessment (EMA, see e.g. [8], [10]), the recording of outcomes in everyday life with the following areas and focal points is lacking:

1.

Electronic questionnaire collection as PROM, which fulfils national, international, and regulatory aspects as well as national guidelines in its orientation and includes rehabilitation-specific aspects according to ICF (International Classification of Functioning, Disability, and Health, [11])

2.

Simplified measurement of listening effort as a telemedical application outside a laboratory setting to demonstrate cognitive load beyond speech intelligibility

3.

Procedures for EMA that allow a simple classification of situations and their evaluation as well as recording individually significant situations; see “Client Oriented Scale of Improvement” (COSI, [12])

4.

Recording of sound descriptors when the patient’s hearing aid and CI are evaluated differently in defined listening situations to obtain further information for bimodal fitting

The aim of the HearDL app is, firstly, to optimise the individual fitting of bimodal fittings and, secondly, to collect everyday patient data on satisfaction with hearing systems electronically, i.e., paperless, even outside a clinical setting. The app is intended to cover a variety of PROMs relevant to everyday life. For patients, audiologists and hearing care professionals in their daily routine, the application should be easy to use and the data analysis uncomplicated.

Four different areas of hearing evaluation and sound perception are addressed in the HearDL app:

1.

Classic questionnaires with retrospective queries,

2.

Recording of subjective listening effort using ACALES mobile (see [13], [14] for the computerised version of ACALES),

3.

EMA in the corresponding situation, and

4.

“Voting tools” (A/B comparison) as well as a sound and hearing evaluation query for bimodal fitting, embedded in the EMA task.

A dashboard was developed for Windows, which on the one hand organises the complete patient administration, a chronological data overview and the data export, and on the other hand allows the individual tasks, frequency of the query and the times of the measurements to be scheduled for the patients individually and at group level using a study planner.

The graphical user interfaces (GUI) for the main new outcome measures are shown in Figure 1 [Fig. 1] (A–D); examples of the GUIs for the adaptive listening effort measurement (ACALES mobile), situation definition of the EMA and questionnaire procedures are shown. For the voting tools, an A/B comparison, and the addition of sound descriptors to the EMA were implemented.

Figure 1A [Tab. 1] shows that when users start the app, they can see how many tasks previously defined in the dashboard have already been completed by means of coloured circles that fill up. In the app, this function of the circles is further emphasised by animations. In the ACALES mobile areas, progress bars and percentages are reported back to motivate the user to complete outstanding tasks; for questionnaires: how many questions still need to be answered.

The procedures and their implementation are presented in more detail below.

When selecting suitable questionnaires, care was taken to ensure that they enabled the subjective assessment of hearing over the past two to four weeks to be recorded. The focus is on subjective hearing ability as a functional component, hearing-specific quality of life and long-term satisfaction as well as rehabilitation-specific characteristics. For the selection of the questionnaires, preliminary work by [15] on bimodal care was considered, whereby various questionnaires were analysed as a follow-up. Based on psychometric parameters for change sensitivity Cohen’s d [16] as an effect size measure, the questionnaires “SSQ12” [17] and the Hearing Handicap Inventory “HHI” with the variants –E (Elderly, see [18]) and –A (Adults, see [19]) were identified as sensitive. The new variant of the HHI, the Revised Hearing Handicap Inventory [20], refers to both variants and is also available in a screening version (RHHI-S) and a long version (RHHI).

In contrast to the original scaling, only integer values from 0 to 10 can be entered in the SSQ12, which is due to the small user interface and user requirements, cf. also comparable work in the field of listening effort [21].

Other questionnaire inventories were also implemented. The “Hearing Implant Sound Quality Index HISQUI-19” [22] includes functional components as well as questions on sound quality. The “Nijmegen Cochlear Implant Questionnaire” (NCIQ) [23] has been implemented with all six subscales, a procedure that is mandatory in the CI White Paper on CI provision in Germany [24].

Furthermore, a total of nine rehabilitation-specific questions, which were formulated based on IRES [25] and the ICF concept in audiology [11], were implemented as an in-house development.

In the ACALES listening effort measurement, see [13], [14], sentences from the Oldenburg sentence test (OLSA, [26] are presented in speech-simulating, stationary background noise (Olnoise). The subjects’ task is to rate the subjectively perceived listening effort on a 13-point rating scale (ESCU; Effort Scaling Categorical Unit) from “effortless” (1 ESCU) to “extremely effortful” (13 ESCU) and an additional category “noise only”. During the measurement, the SNR is adaptively changed based on the test subjects’ responses. The aim of this measurement is to determine the individual hearing effort function, in which each SNR value is assigned a rating category [13].

The measurement method was implemented as an independent programme library for use on mobile devices. The standard measurement is carried out “in a quiet domestic environment” with Olnoise. The measurement procedure was shortened for use on mobile devices outside the laboratory. The three phases of the stationary variant with

1.

boundary definition,

2.

estimation of the SNR for the categories and random playback and evaluation of the SNRs, and

3.

recalculation of the limits and SNRs as well as the random playback (repeated twice) were reduced to the first two phases.

In addition, only two of the original three OLSA sets are used. To improve input on a small smartphone screen, the rating scale was reduced from 13/14 rating categories to 7/8 categories (see Figure 1B [Fig. 1]); however, all seven original verbal anchors were retained.

The shortened mobile version vs. the long original version was tested regarding the scale length 13/14 vs. 7/8 and the presentation type PC vs. app with N=15 and normal hearing N=20 CI users. There were no significant differences either descriptively or statistically; a manuscript with detailed results is in preparation.

The measurements with CI patients showed that the signals were too soft and should therefore not be presented via the smartphone’s built-in loudspeaker. Other transmission channels are more suitable, such as streaming via a Bluetooth connection or additional active speakers. Although the application is telemedical oriented and is carried out in a private setting, it does not record the listening effort of everyday situations due to reliability aspects and comparability. A subjective survey of listening effort is explicitly included in the EMA surveys.

The EMA approach used here restricts the selection of listening situations to be assessed to individually relevant situations. In the primary concept, patients indicate the five most individually relevant listening situations (e.g., conversation in a restaurant, understanding speech while driving, etc.) at the beginning of the survey, qualify these with a short text (see Figure 1C [Fig. 1]) and then rate them when they are in the corresponding situation (for examples, see Figure 2A–C [Fig. 2]). This enables a before-and-after comparison that relates to identical situations/classes and is intended to improve data quality. The idea of asking about individually significant situations is applied in a similar way in the “Client Oriented Scale of Improvement” questionnaire (COSI; [12] to focus the fitting of hearing systems and the outcome on relevant, individually significant hearing situations.

The “Common Sound Scenarios” scheme from [27] was adapted for the a priori classification of the relevant listening situations, see Table 1 [Tab. 1]. This procedure serves to simplify the subsequent evaluation in clinical routine.

In the first step, the audiologists ask which intention is associated with the respective listening situation: “Conversation”, “Focused listening” or “Monitoring environment” during an activity. Based on the answer, further classification questions are generated. When selecting “Conversation”, the next step asks about the number of conversation partners and the presence of background noise, totalling six situations. If “Focused listening” is selected, the situation is further specified by asking about the presentation of the target signal (live or via a device) and the presence of disturbing background noise (speech, music), again six situations. When selecting the category “Monitoring environment”, the relevance of the ambient noise is asked to determine whether it is background noise during activities that needs to be “monitored” (e.g., traffic noise) or rather casually, incidental noise, such as background music when reading.

As soon as users find themselves in a classified situation, they are asked to answer specific questions relevant to the respective classes. This ensures systematic adaptivity in the selection of items.

Originally, the concept envisaged that users would carry out the classification themselves. User surveys revealed that the classification procedure was too much for patients. The classification is therefore carried out jointly in dialogue with the audiologists in the dashboard; however, classification is still possible in the app for experienced users.

In addition, relevant situations can be assessed spontaneously in the app. Here again, a situation is defined, as shown in Figure 1C [Fig. 1], which is classified by the user based on example situations and queried directly without having to use the expert classification procedure.

Depending on the situation class (see column “No.” in Table 1 [Tab. 1]), various hearing-related aspects are queried by the items via sliders (see column “Items” in Table 1 [Tab. 1]):

Items A, G and F serve as a query on ‘acute’ quality of life/participation, items B and D consider aspects of perception and items C, E and H record functional aspects of everyday life. Left/right sliders were used for the volume, sound, and clarity of speech assessments to provide reference points for fitting the different devices of the bimodal fitting.

The “Mood Scale” distinguishes between five possible answers, while seven scale points are available for the other items.

Once all the hearing situations to be assessed have been entered in the dashboard along with the scheduling, a reminder of the situations is sent via push notification in the HearDL app on the smartphone (Figure 2D [Fig. 2]).

Two “voting tools” (VT) were developed: VT1 for comparing the two hearing systems (for bimodal provision: HA vs. CI) in a current hearing situation and VT2 for directly evaluating different settings or (new) product features in everyday life (A/B comparison). While VT1 is always used in conjunction with a hearing situation as part of the EMA measurement, VT2 is completely stand-alone and is used to compare individual settings of a hearing system (e.g., different hearing programmes) or (new) product features of a hearing system in everyday life (e.g., different parameters of signal pre-processing).

The VT1 particularly addresses the bimodal fitting case: In a current everyday hearing situation in the EMA application, patients should assess whether they hear equally well with CI and HA and, if not, specify more precisely with which system they hear worse and why (predefined hearing assessment classes). To do this, it is essential that the VT1 is used in a clearly described hearing situation. For this reason, the VT1 is an extension of the EMA survey that can be configured (activated/deactivated) by CI audiologists and hearing care professionals in the dashboard. The predefined sound and hearing evaluation classifications (e.g., speech and/or background noise too loud/too soft, own voice too loud, too muffled/too sharp, noise, distorted, reverberation, echo) allow conclusions to be drawn about further individual programming and – if large amounts of data is available – also about general guidelines for fitting in order to optimise the combination of CI and hearing aid. The collection of items is based on qualitative surveys of CI patients as part of the usability studies during the development of the HearDL app and a collection of descriptors for the remote fitting of hearing aids [28].

The dashboard is a separate Windows PC programme and is set up as a multi-user system in which the patients and their hearing systems are created and managed by the audiologists using a patient ID. The corresponding test procedures can be selected and assigned to specific appointments and intervals so that a complete individual programme or a study design for patient groups can be created. The patient profiles and data can be exported for use in scientific analyses. Individually significant hearing situations can be created for the EMA application on a patient-specific basis. The patient data, the test design and the relevant EMA situations can be transferred from the dashboard to the smartphone app.

Data collected from patients can later be imported into the dashboard from a smartphone. Simple analyses of the collected data and visualisation of key results are possible directly in the dashboard. The data can be exported in a table format so that it can be analysed in more detail in other software (e.g., Microsoft Excel, statistics software).

An iterative approach was chosen for the development and implementation of the HearDL app and dashboard, in which partial functionalities could be gradually integrated, tested and adapted as required. Right from the start, the software was developed using a system that allows cross-platform realisation. As the specific platform on which the app runs was less important, development was initially limited to iOS. At a later stage, the app was also further developed for Android systems so that both smartphone operating systems are available. Similarly, a possible translation into other languages was provided for in the software architecture from the outset. The app is currently available in German and English. The dashboard was realised as desktop software for Windows and iOS and is also available in German and English.

For the iterative development process of the app and the dashboard, new interim releases of the respective software were repeatedly created over the course of the project to test them with the specialist staff and patients and subsequently modify them if necessary. In this way, a fully functional prototype was created that can be used for clinical studies and audiological practice. Further releases of the app and dashboard will only be made for Android, as the administration of the iOS app is more complex, and the Android operating system has a significantly higher market share than iOS.

The development of the app tools and the dashboard followed a usability process over the entire period in accordance with [29], on the one hand through expert panels based on the principle of heuristic evaluation and on the other hand through user involvement and patient testing. The patient measurements were carried out in two waves, namely at the beginning of development with nine experienced CI users (four test subjects, aged 51–70 years) and four months before the end of development as an executable prototype with 10 experienced HG users (five test subjects, aged 44–72 years). The tests were used to continuously improve the design, operation, consistency, and user guidance, especially for ACALES mobile and the EMA tasks. The screenshots, see Figure 1 [Fig. 1] and Figure 2 [Fig. 2], are the result of the tests.

The “Mobile App Security Verification Standard” (MASVS) of the Open Worldwide Application Security Project (OWASP, see [30]) was not explicitly considered in the development of the HearDL app. However, the implementation largely complies with the requirements of the two verification levels MASVS-L1 (Standard Security) and MASVS-L2 (Defence-in-Depth). Both programmes, the app on the smartphone and the dashboard for Windows, are not web applications, but run locally on the respective devices, smartphone, or PC. They do not establish any network connections themselves, but only save and transfer data locally on the respective device. Data transfer between the app and dashboard is only possible via cable. Although the app is designed as an EMA tool and there is therefore an interest in additional information about the location and acoustic environment, the app does not record or store position data or acoustic data.

In the current version, it is planned to enter participant’s ID and other data on hearing system care instead of the clear name and date of birth. The app users’ data can only be viewed by the local users of the dashboard, i.e., the audiologists in charge. In addition, only the data requested by the questionnaires, the EMA tool, the ACALES test and the voting tools are collected and stored. All exchange data is stored in the software in an encrypted format. A local/network-compatible Firebird database is used for the dashboard. This uses password-protected access so that only logged-in users have full access to the dashboard via password and username.

Bimodal patients were at the centre of the app development. In general, however, unilateral or bilateral HG and/or CI users or users of other hearing solutions can also use the app. The following fields of application are possible for research and rehabilitation: indications for the individual therapy and rehabilitation corridor, additional data in clinical measurements to increase ecological validity, support for bimodal fitting strategies, development of bimodal signal processing, and documentation of the benefits of aural HA and CI rehabilitation. HearDL is particularly suitable for comparing the everyday data collected with laboratory data. In this way, the question of whether the clinical measurements are meaningful enough to reflect the everyday hearing of patients can be investigated.

This question arises in particular for “bimodal fitting”. It is currently not known exactly how bimodal fitting should be carried out because fitting often takes place in two audiological worlds and separately for each hearing system. With the VT1 voting tool from HearDL, it may be possible to make statements about how “optimal” perception can be achieved, how signal pre-processing must be set for synergy and whether device-specific pre-settings or automatic settings are disadvantageous. It is assumed that the HearDL app will provide data-supported information for bimodal adaptation in the future, which could then lead to adaptation rules.

The app designed here supplements previous telemedical applications, such as Cochlear^TM Remote Check [9], with a shortened hearing effort procedure as ‘ACALES mobile’, an assessment procedure for bimodal fitting and an EMA design that is suitable and optimised for clinical use. It makes sense to implement another audiological measurement procedure, e.g., the digit triple test (DTT, [31]), in HearDL on a terminal device environment to start all essential measurements from one application and have them available for evaluation. This would enable comparability with the Remote Check App [9], in which the DTT is also implemented. In addition, it makes sense to test the implementation of the OLSA, possibly in competition with it, as the OLSA will enable comparison with routine clinical data.

In many EMA measurements, acoustic data of the respective situations were also coupled with the subjective assessments; for an overview, see [8]. In the future, this coupling will also be useful for the HearDL app, but the focus here was initially on the tele-audiological recording of patient data independently of hearing systems from individual manufacturers.

A content management system (CMS) for new questionnaires is currently being created and will be integrated into the dashboard. This will make it possible for audiologists to create further internationally used questionnaire batteries themselves using a GUI and use them in research projects. In general, comparisons of pencil-and-paper questionnaires vs. app-based recording should be examined, as well as the resolution of the SSQ scale (original with fine gradation between integer values vs. the 11-point scale used). An extension is also planned for the list of sound descriptors for the bimodal care case in the context of clinical studies.

The usability of the app and the dashboard was tested by consulting audiological and medical experts as well as potential users. Nevertheless, further extensive pilot tests in different clinical and home settings as well as further validation studies around ACALES mobile and the questionnaires are necessary to ensure comparability with conventional pencil-paper measurements and PC-based

The human studies described here were carried out with the approval of the responsible ethics committee (No.: 2020-071 of the Medical Ethics Committee Oldenburg) in accordance with national law and the Declaration of Helsinki of 1975, in the current, revised version. A declaration of consent has been obtained from all patients.

We would like to thank the patients of the University Clinic for Ear, Nose and Throat Medicine at the Evangelical Hospital in Oldenburg.

Horst Hessel and Markus Meis are employees of Cochlear Deutschland GmbH & Co KG. Rabea Wortmann is an employee of GN Hearing GmbH, Germany. The other co-authors Melanie Krueger, Andreas Radeloff, Mareike Grundmann, Michael Buschermöhle, Inga Holube, and Petra von Gablenz report no conflicts of interest.

The work was carried out as part of the “Hearing in Daily Life (HearDL)” project, funded by the Federal Ministry of Education and Research (BMBF); Grant number: 13GW0266A.